Basic Engineering for Medics

and Biologists
An ESEM Primer

Edited by
T. Clive Lee
Royal College of Surgeons in Ireland
and Trinity Centre for Bioengineering, Dublin, Ireland
and
Peter F. Niederer
Institute for Biomedical Engineering
Swiss Federal Institute of Technology (ETH)
and University of Zurich, Zurich, Switzerland

Amsterdam • Berlin • Tokyo • Washington, DC

© 2010 The authors and IOS Press.

All rights reserved. No part of this book may be reproduced, stored in a retrieval system,
or transmitted, in any form or by any means, without prior written permission from the publisher.

ISBN 978-1-60750-526-6 (print)

ISBN 978-1-60750-527-3 (online)
Library of Congress Control Number: 2010923280

Publisher
IOS Press BV
Nieuwe Hemweg 6B
1013 BG Amsterdam
Netherlands
fax: +31 20 687 0019
e-mail: [email protected]

Distributor in the USA and Canada

IOS Press, Inc.
4502 Rachael Manor Drive
Fairfax, VA 22032
USA
fax: +1 703 323 3668
e-mail: [email protected]

LEGAL NOTICE
The publisher is not responsible for the use which might be made of the following information.

PRINTED IN THE NETHERLANDS

Basic Engineering for Medics and Biologists v
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.

Preface
Making Engineering Accessible to
Medics and Biologists
Research and development in bioengineering and medical technology, conducted dur-
ing recent decades, have led to spectacular progress in clinical medicine. These
achievements have triggered an enormous increase in the number of courses offered in
the areas of bioengineering, clinical technology and medical informatics; nowadays,
most major universities offer curricula oriented towards these fields. The majority of
participants however come from engineering backgrounds and so modules dealing with
basic biological and medical sciences have been included. These have been facilitated
by the ready availability of textbooks in this area, many of which were specifically
written for nursing A & P (Anatomy & Physiology) programmes.
In contrast, relatively few participants from medicine and biology have taken up
courses in biomedical engineering, to the detriment of scientific exchanges between
engineers and medical doctors. The reasons for this imbalance are many and may vary
from country to country, but a consistent finding is the difficulty (or courage) in taking
the first step. ‘Introductory’ texts in bioengineering tend to involve vector algebra and
calculus early in Chapter 1. While in most countries entry to medical school is very
competitive and requires, among others, high grades in mathematics, little more than
arithmetic is required during the course itself, so numeracy undergoes disuse atrophy.
Furthermore, to paraphrase George Bernard Shaw, medics and engineers are sepa-
rated by a common language. To the medic, stress and strain are both symptoms of
anxiety, while to the engineer they are defined by equations involving symbols such as
σ and ε and are related to forces and deformations. To the average medic this is, liter-
ally, all Greek.
ESEM is a bridge between medicine and engineering. It promotes cultural and sci-
entific exchanges between engineers and medics and training courses in biomedical
engineering. What better way to achieve these objectives than to address this problem
and help medics and biologists to take this first step.
To this end, we herewith present a series of First Step chapters in book form aimed
at medics and biologists to help bring them to the level where they can begin an MSc in
biomedical engineering, or at medics who “simply” wish to understand a particular
medical technology. Written by engineers who are leaders in their field, with input
from medical colleagues, they will cover the basic engineering principles underpinning
biomechanics, bioelectronics, medical informatics, biomaterials, tissue engineering,
bioimaging and rehabilitation engineering, and will include clinically relevant exam-
ples.
As Albert Einstein may have said ‘Everything should be made as simple as possi-
ble, but not simpler’.

Dublin & Zurich, December 2009 Clive Lee & Peter Niederer
Cover Illustration

Stress trajectories in a curved Culmann crane (left) compared with a schematic repre-
sentation of the trabecular pattern in the proximal human femur (right). From: Wolff J.
Ueber die innere Architectur der Knochen und ihre Bedeutung für die Frage vom
Knochenwachsthum (On the intrinsic architecture of bone and its significance with
respect to the question of bone growth). Virchow’s Arch. 1870; 50: 389–450.
 vii

Contents
Preface: Making Engineering Accessible to Medics and Biologists v
Clive Lee and Peter Niederer

Introduction to Chapter I: Biomechanics 1

Gijsbertus J. Verkerke, Pascal Verdonck and T. Clive Lee

I.1. Statics 3
Gijsbertus J. Verkerke and T. Clive Lee
I.2. Mechanics of Materials 13
Prashant K. Sharma
I.3. Dynamics of Human Movement 27
Bart H.F.J.M. Koopman
I.4. Biofluid Mechanics & the Circulatory System 45
Pascal Verdonck and Kris Dumont
I.5. Biomechanics of Implants 58
Jan G. Hazenberg, Johannes Schmid, T. Clive Lee and
Gijsbertus J. Verkerke

Introduction to Chapter II: Bioelectronics 67

Richard B. Reilly and T. Clive Lee

II.1. Elementary Electrodynamics 69

Jacques Jossinet
II.2. Electrical Safety 81
Jacques Jossinet
II.3. Electrograms (ECG, EEG, EMG, EOG) 90
Richard B. Reilly and T. Clive Lee
II.4. Biosensors 109
Richard B. Reilly and T. Clive Lee

Introduction to Chapter III: Medical Informatics for

Biomedical Engineering 119
Paul McCullagh and T. Clive Lee

III.1. Medical Informatics and eHealth 121

Paul J. McCullagh, Huiru Zheng, Norman D. Black,
Richard Davies, Sue Mawson and Kieran McGlade
III.2. Data Structures, Coding and Classification 140
Huiru Zheng, Haiying Wang, Norman D. Black and John Winder
viii

III.3. Mining, Knowledge and Decision Support 158

Dewar D. Finlay, Chris D. Nugent, Haiying Wang, Mark P. Donnelly
and Paul J. McCullagh
III.4. Remote Healthcare Monitoring and Assessment 172
Chris D. Nugent, Dewar Finlay, Richard Davies, Mark Donnelly,
Josef Hallberg, Norman D. Black and David Craig

Introduction to Chapter IV: Biomaterials and Tissue Engineering 185

Fergal J. O’Brien and Brian O’Connell

IV.1. Scaffolds & Surfaces 187

Sonia Partap, Frank Lyons and Fergal J. O’Brien
IV.2. Cellular & Molecular Biomechanics 202
Veronica A. Campbell and Brian O’Connell
IV.3. Bioreactors in Tissue Engineering 214
Niamh Plunkett and Fergal J. O’Brien
IV.4. Characterisation and Testing of Biomaterials 231
Sebastian Dendorfer, Joachim Hammer and Andreas Lenich

Introduction to Chapter V: Medical Imaging 247

Peter Niederer and T. Clive Lee

V.1. Ultrasound Imaging and Doppler Flow Velocity Measurement 249

Peter F. Niederer
V.2. X-Ray-Based Medical Imaging: X-Ray Projection Technique, Image
Subtraction Method, Direct Digital X-Ray Imaging, Computed Tomography (CT) 274
Peter F. Niederer
V.3. Basic Elements of Nuclear Magnetic Resonance for Use in Medical
Diagnostics: Magnetic Resonance Imaging (MRI), Magnetic Resonance
Spectroscopy (MRS) 302
Peter F. Niederer

Introduction to Chapter VI: Rehabilitation Engineering 321

Tadej Bajd and T. Clive Lee

VI.1. Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 323

Zlatko Matjačić
VI.2. Basic Functional Electrical Stimulation (FES) of Extremites –
An Engineer’s View 343
Tadej Bajd and Marko Munih
VI.3. Rehabilitation Robotics 353
Marko Munih and Tadej Bajd

Subject Index 367

Author Index 369
Basic Engineering for Medics and Biologists 1
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-1

Introduction to
Chapter I: Biomechanics
Gijsbertus J. VERKERKE, Pascal VERDONCK and T. Clive LEE (eds.)

Biomechanics is the science that examines forces acting upon and within a biological
structure and effects produced by such forces (B.M. Nigg, Biomechanics of the
musculoskeletal system, 1995). As such, biomechanics is part of the world of physics
and concentrates on forces that act on biological structures. These forces either
originate from inside the human body, like muscle forces, or forces, created externally,
like an external load or impact. The structures they act upon can be classified as solids,
like skeletal parts or organs and by fluids, like blood and air.
The study on the effect of these forces can be classified as deformation and
displacement.
Biomechanics concentrates on various situations. To apply biomechanics four
subspecialisations are created, each focusing on a specific situation and simplifying the
real world in a specific way to allow a practical model that can be analysed
theoretically.
Statics deals with solid structures that will not move. In this way the forces that act
upon a structure can be calculated. All structures are assumed to be undeformable.
It is based on Newton’s third law, F12 = - F21: when a structure exerts a force on
another structure (for instance by its weight) that last body will exert an equal force in
opposite direction to the first body. It is also based on Newton’s first law:
F = 0 ⇔ v = constant: If the resulting force F on a structure is zero, then that
structure will not change its velocity, so non-moving structures will remain motionless.
Mechanics of materials deals with solid structures that will not move, but are
deformable.
Dynamics deals with solid structures that can move, but cannot deform. It is based
on the second law of Newton, F = m a: when a force F acts on a body with mass m, it
will undergo an acceleration a.
And fluid mechanics deals with fluids that can move and that can deform.
These four different biomechanical subspecialisations and their applications will
be discussed in this chapter.
This page intentionally left blank
Basic Engineering for Medics and Biologists 3
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-3

I.1. Statics
Gijsbertus J. VERKERKEa,b and T. Clive LEEc
a
University Medical Center Groningen, University of Groningen
Dept of Biomedical Engineering, Groningen, The Netherlands
b
University of Twente, Dept of Biomechanical Engineering, Enschede,
The Netherlands
c
Royal College of Surgeons in Ireland, Department of Anatomy, Dublin, Ireland

Abstract. The forces that act on an object determine its dynamic behaviour and
defromation. Analysis of all forces and moments is essential. A free-body diagram
summarizes all forces and moments that act on an object. To calculate the
magnitude of the forces we can use the static equilibrium of forces and moments.

Keywords. Force, moment, couple, static equilibrium, free-body diagram, internal

force

Introduction

If we want to know, how an object will move (kinetics) or deform (mechanics of

materials) we first have to know, which forces are acting upon those objects. Statics is
the field of mechanics that studies forces on objects that don’t move, so the forces
acting upon them balance each other.
Since all objects on earth are subjected to gravity, forces are everywhere. Apart
from gravity, we can distinguish spring forces, magnetic forces, electric forces,
hydraulic and pneumatic forces and inertia forces. A force is defined by a magnitude, a
direction and a point of action. A muscle force acts on its origin and on its insertion,
has a distinct magnitude that can be derived (for isometric contractions) from the EMG
and has a clear direction, the line between origin and insertion. The force of a cue acts
on the surface of a billiard ball, in the direction of the cue and with a magnitude that is
determined by the velocity of the cue. When you are sitting on a thumb tack, you
experience the point of action, magnitude, and direction of a force.
A force creates either movement (of the billiard ball) or deformation (the muscle
force you apply to grab a book will deform that book). By definition, force (F) is equal
to mass (m) times acceleration (a):

F=m•a (1)

Acceleration is a change in velocity (Δv) over time (Δt):

a = Δv / Δt (2)
4 G.J. Verkerke and T.C. Lee / Statics

Velocity (v) is a change in distance (Δs) over time (Δt):

v = Δs / Δt (3)

So velocity is expressed in m/s; acceleration, a change in velocity, is expressed in

(m/s)/s or m/s2. So force is expressed in kg.m/s2, often abbreviated as Newton (N).
Assumptions: To use forces, we have to do some modelling: we assume that a
force is acting in a single point with infinitely small dimensions. In reality we study
only forces that are acting on a surface, called pressure.

1. Resultant Force

When two forces with different lines of action are acting on the same point of an object
(for instance two muscles that are active with the same insertion point but different line
of action), then we can combine them into a single resultant force. The magnitude and
line of action of this force can be found using the rules for adding vectors (Figure 1).

Figure 1. Two vectors A and B can be replaced by a resultant force R using the parallelogram
construction (left) or putting them head-to-tail (middle and right).

A force can also be resolved into two different forces using the same rule. Often
this is performed to simplify the effect of a force. If the seat of your chair is not
horizontal, but inclined under an angle β, then the force F that is acting on this seat due
to gravity can be replaced by two forces: a force N that is acting perpendicular to the
seat (also called the normal direction) and a force V acting parallel to the seat (also
called the shear direction). The force V lets you slide off the seat, while the force N
determines whether or not the chair will be able to carry you.
You can find the force V by multiplying force F with sin β (sinus β). Force N is
found by multiplying force F with cos β (cosinus β). The value of sin β and cos β can
be found on most pocket calculators and using your Windows calculator (scientific
view). The angle β is expressed in degrees. So:

V = F • sin β

N = F • cos β (4)
 G.J. Verkerke and T.C. Lee / Statics 5

1.1. Friction

Four main types of forces exist in and on the human body: gravity forces, muscle
forces, friction forces and forces exerted by others or other objects, like a boxing glove.
A friction force always acts perpendicular to a surface and prevents an object moving
along that surface. So a friction force prevents you from sliding off the inclined seat.
Friction has some strange features:
First of all its magnitude is variable and only as large as is necessary to prevent
movement. If a force V is acting upon you, because you are sitting on an inclined seat,
then the friction force has the same size as V, but acts in the opposite direction.
Secondly there is a limit to the friction force, of course, because you will slide off
a steeply inclined seat. The friction force, acting during motion is slightly smaller than
the static friction force.
Thirdly, the friction force is linearly related to the normal force:

Ff = μs • FN (5)

with Ff = static friction force (N), μs = coefficient of friction and FN = normal force (N).
As you can derive from formula (5), μs is dimensionless. Its value depends on the two
surfaces that are in contact with each other and can be found in handbooks.
Fourthly, the friction force is independent of the area of contact between the two
objects. You can check this by placing a book on a board, start to incline the board and
remember at which angle it starts to move. Then place the book on its side on the board
and repeat the experiment. The angle will be the same. An explanation for this is that
both surfaces are rough and in theory to keep the book stable it rests only at three
contact points on the board. The coefficient of friction is determined by the nature of
these three contact points and thus is not determined by the contact surface.

1.2. Moment and couple

Apart from the tendency to move an object along its line of action a force also has a
second effect: think of a billiard ball that is hit by your cue right in the middle: it will
move straight forward. Now hit it slightly off-centre. It will again move straight
forward, but also start to rotate around its axis (Figure 2). This rotating effect of a force
is called a moment. The effect is expressed by:

M=F•d (6)

with F the force (N), d the distance between the centre of the billiard ball and the line
of action of the cue (m) and M the moment. The distance d, also called the lever arm of
the force, must be perpendicular to the line of action of the force. Note that a moment
always is related to the point of which you would like to know the rotating effect of the
force, in this case the centre of the billiard ball.
Using formula (6), the unit of M is Nm. (Note that mN is not the correct notation
of this unit, since it means milli-Newton)
A moment allows you to open a door, to unscrew bolts, and to spin a ball. To
increase the effect of a moment you can either increase the force or the distance. Our
body uses this effect: since the muscle force is limited, enlarging d is an additional
strategy to increase the rotating effect, that is why you have a patella; it increases the
6 G.J. Verkerke and T.C. Lee / Statics

effect of your quadriceps muscle and makes it possible to extent your knee more
powerfully.

Figure 2. When a force is acting eccentrically over a distance d, a moment appears, resulting in a
rotation (left). Two parallel forces in opposite direction are called a couple (middle), also denoted by a
curved arrow (right).

A specific situation appears when two parallel and equal forces with an opposite
line of action are acting on an object (Figure 2, middle). The net action of the forces in
horizontal movement is zero, they only have a rotating action caused by their moment,
which is twice the size of the moment when one force is acting. Such a pair of forces is
called a couple and denoted by a curved arrow (Fig. 2, right), since its line of action
(the rotation line) is pointing out of the plane. A couple has a magnitude and
orientation and therefore is a vector.
In statics, moving a force along its line of action - translation - does not have any
influence. It does not matter whether you push an object or pull it. However, moving a
force perpendicular to its line of action does have an influence. Rotation is affected by
it, since its arm is changing. Moving a couple over an object is allowed. Although the
arm of one force is increasing, the arm of the other force is decreasing by the same
amount, leaving no net change. Therefore a couple is called a free vector.
In practice the terms couple and moment are used inaccurately and often
exchanged, which is not logical, since a couple is a free vector, but a moment is related
to a point on an object. The term couple moment is often used instead of couple, which
is also confusing.

2. Static Equilibrium

Now we know the effects of forces, either translation (movement along the line of a
force) or rotation, we can apply Newton’s First Law. According to this law, if an object
is not subjected to a force, this object will not start to move or change its velocity.
Since, in statics, we consider all objects to be at rest, no net forces should act on them.
Also the sum of their moments should also be zero to avoid rotation. So in statics the
sum of all forces should be zero and the sum of all moments in relation to a certain
point on the object is also zero. This means that the object has no degree of freedom to
move.
 G.J. Verkerke and T.C. Lee / Statics 7

When we consider a two-dimensional situation, each force can be resolved into

two components, a horizontal one, Fx, and a vertical one Fy. For static equilibrium all
horizontal forces should be zero and all vertical ones should be zero.
This leads to three equations of equilibrium:

Σ Fx = 0
Σ Fy = 0 (7)
Σ Mp = 0
where Σ is capital sigma and means ‘the sum of’.
Note that when the sum of all moments is zero in relation to a certain point on the
object, it will also be zero to another point on that object. So the reference point is a
free choice, but must be used subsequently for all forces.
In the three-dimensional world, which we will not consider in this paper, six
equations of equilibrium exist:

Σ Fx = 0
Σ Fy = 0
Σ Fz = 0 (8)
Σ Mx = 0
Σ My = 0
Σ Mz = 0

3. Free-Body Diagram

In 2D, in order to apply the three equations of equilibrium, we have to know which
forces are acting upon an object. We already found out, that not only obvious external
forces are active, like gravity, but also hidden ones, like the force of a seat acting upon
your bottom. To make this last type of force visible, we have to free the object from its
surroundings. After that we have to draw all forces of the surroundings that act upon
this object. These forces are unknown, but by applying the three equations of
equilibrium we can calculate them and so complete our static analysis. Thus, all forces
that are acting upon an object are known.
Then we can proceed to dynamics or mechanics of materials to find out the
movement of this object or its deformation.
The main problem in making an object free from its surroundings is to determine
the reaction forces from the surroundings. In Figue 3 a list of connections between the
object and its surroundings (the supports) is shown, along with the replacing reaction
forces and couples. As a general rule, if a support prevents translation of an object in a
given direction, then a force is present on the object in the opposite direction. Similarly,
if a rotation is prevented, a couple is exerted on the object.
8 G.J. Verkerke and T.C. Lee / Statics

F2

F1

M
F2

F1

Figure 3. Overview of possible supports and their replacing forces and couples.

Making an object free from its surroundings is a very delicate action, since we
have to decide which object we select. If we select a whole human body as our object,
and we are interested in the forces acting in our shoulder joint when lifting a bag, we
will not be able to calculate these forces, because they are hidden inside the object. So
in that case we have to select the upper limb as our object and free it from the rest of
the body. We again have to determine the reaction forces of the rest of the body. Most
supports consist of joints, which in general cannot be characterised as regular hinge-
type connections. We can distinguish 6 types of joints (Figure 4):
1. Ball and socket joints, like the hip joint, allow three rotations and no translations.
2. Condyloid (ellipsoid) joints, like the knee, allow in general rotation around one
axis and some rotation around another axis. When the knee is extended there is no
axial rotation possible, when it is flexed some axial rotation is possible.
3. Saddle joints, like the thumb (between the 1st metacarpal and trapezium), allow
rotation around two axes.
4. Hinge joints, like the elbow (between the humerus and the ulna), act like a door
hinge, allowing only flexion and extension around one axis
 G.J. Verkerke and T.C. Lee / Statics 9

5. Pivot joints, like the elbow (between radius and ulna). This is where one bone
rotates about another.
6. Gliding joints, such as in the carpals of the wrist (not shown in Figure 4). These
joints allow a wide variety of movement, but not over a large distance.
A moment allows you to open a door, to unscrew bolts, and to spin a ball. To
increase the effect of a moment you can either increase the force or the distance. Our
body uses this effect: since the muscle force is limited, enlarging d is an additional
strategy to increase the rotating effect, that is why you have a patella; it increases the
effect of your quadriceps muscle and makes it possible to extent your knee more
powerfully.
A moment allows you to open a door, to unscrew bolts, and to spin a ball. To
increase the effect of a moment you can either increase the force or the distance. Our
body uses this effect: since the muscle force is limited, enlarging d is an additional
strategy to increase the rotating effect, that is why you have a patella; it increases the
effect of your quadriceps muscle and makes it possible to extend your knee more
powerfully.

Figure 4. Schematic drawing of 5 joint types, ball and socket (1), condyloid (2), saddle (3), hinge (4)
and pivot joints (5).

Apart from separating the joint surfaces we also have to cut through muscles to
free our upper limb from the rest of our body. A muscle can be categorised as a cable
and must be replaced by a single force with an orientation from origin to insertion.
Let us now summarize the procedure for determining the shoulder joint and muscle
forces in case we think we need to lift some weights (of in total 100 kg, see Figure 5,
left). First we must define a positive x-axis, y-axis and direction of moments (Figure 5,
right). Then we need to free the object of interest, in this case the upper limb.
10 G.J. Verkerke and T.C. Lee / Statics

The upper limb is connected to the rest of the body by the shoulder joint, which
can characterised as a ball and socket joint in 3D, and in 2D as a hinge. So the
shoulder joint can be replaced by two forces, F1 and F2. The upper limb is also
connected with muscles to the rest of the body. In this situation the deltoid muscle will
be (very) active. A muscle can be replaced by a single force, pointing along its line of
action, Fs. Now we can make the free-body diagram (Fig. 5, middle), including all
forces acting on the upper limb: half of the weight of the dumb-bell (50 kg x 10 m/s 2 =
500 N), the weight of the upper limb (assumed to be 2.5 kg, resulting in 25 N), the
unknown muscle force Fs, and the unknown joint reaction forces F1 and F2. To be able
to use our equations of motion, we first have to split F1, F2 and Fs in a force, acting in
x-direction and a force, acting in y-direction using formula (4).

600
M+
Fg Fw
Fs F1 x
F2

0.05 m
0.2 m 0.25 m

Figure 5. How to calculate shoulder joint and deltoid forces when lifting a weight (left): a free-body
diagram is made of the arm (middle), including the orientation of the x- and y-axis and the positive direction
of moments (right)

Finally we can calculate the unknown forces using the three equations of
equilibrium:

Σ Mp = 0 (p = shoulder joint centre) - Fg • 0.2 – Fw • 0.45 + Fs • 0.05 = 0

Fs = 4600 N (9)

Σ Fx = 0 - Fs • cos 600 – F1 • cos 600 + F2 • cos 300 = 0

F1 = √3 F2 – Fs (10)

Σ Fy = 0 → - Fs • sin 600 – F1 • sin 600 – F2 • sin 300 – Fg – Fw = 0

→ F1 = - Fs – √3/3 (F2 + 1050) (11)
 G.J. Verkerke and T.C. Lee / Statics 11

(9), (10), (11) → F2 = - 262.5 N ( 12)

(9), (10), (12) → F1 = - 5055 N (13)

Obviously we have chosen both F1 and F2 in the wrong direction, because the
outcome is negative for both forces. So the direction of the unknown forces can be
chosen arbitrarily, after the calculation will become clear, if the assumed direction is
the proper one.
The magnitude of Fs and F1 are considerable, corresponding with a mass of 460
and 505 kg, respectively! The main reason for this is the small arm of Fs compared to
the arm of Fw; muscle arms are small due to geometric considerations. So muscles must
be very strong to compensate for this short arm.

4. Internal Loads

Now we are able to determine the forces, acting on an object and to calculate the
magnitude of these forces, next we can analyse two situations:
Firstly we can find out what will happen if an extra force or couple is applied and
thus the equilibrium of forces and moments is disrupted. In the ‘Dynamics’ chapter we
will calculate the resulting accelerations.
Secondly we can consider our object to be not rigid, so deformable and calculate
the deformations that result from the forces and determine, if our object can resist the
forces, and thus if it is strong enough. For this last action we need to do an extra step in
statics, we have to calculate the internal loads and determine what location is the
weakest point.

d
y

600
M+
M
Fg Fw
F2 x
F1
0.25 m

Figure 6. For the body builder on the left the internal forces of his arm are calculated (middle), using a
co-ordinate system (right).

The first step we have to do is to isolate our object (Figure 6, left), define a
positive x- and y-axis and direction of moments (Fig. 6, right) and calculate the external
12 G.J. Verkerke and T.C. Lee / Statics

reaction forces (in our case there are none). Then we have to cut our object in two parts
and replace one part of our object (in our case the lower left part) by its reaction forces
F1, F2 and M at the cutting edge (Fig. 6, middle).
The resulting free-body diagram (Fig. 6, middle) contains three unknowns, F1, F2
and M that can be calculated using the three equations of equilibrium. Note that the
position of the cut is defined by the distance d between the cut and the upper right end
of our object. By varying this distance we can scan the entire arm and thus calculate the
internal forces at any spot. It then is possible to determine the site at which the highest
loads will occur and thus the site where failure will occur first:

Σ Mq = 0 (q = cutting edge site)

→ - Fg • d • cos 600 – Fw • (d • cos 600 – 0.25) + M = M = 262.5 • d – 125 Nm (14)

Σ Fx = 0 → - F1 • cos 600 + F2 • cos 300 = 0

→ F1 = √3 F2 (15)

Σ Fy = 0 → - F1 • sin 600 – F2 • sin 300 – Fg – Fw = 0

→ F1 = - √3/3 (F2 + 1050) (16)

(15), (16) → F2 = - 262.5 N (17)

(15), (17) → F1 = - 4547 N (18)

Again, both F1 and F2 are acting in the opposite direction as was presumed.
Now we can find out, where the most critical spot in the upper arm is. Both F1 and
F2 are constant, M is maximal for d = maximal. So the highest load will act where d is
maximal, which is the lower left end of the upper arm.

References

[1] G.L. Lucas, F.W. Cooke, E.A. Friis, A Primer of Biomechanics, Springer-Verlag,
1998
[2] R.C. Hibbeler, Statics and Mechanics of Materials (Second edition), Pearson –
Prentice Hall, 2004
[3] N. Özkaya and M. Nordin, Fundamentals of Biomechanics; Equilibrium, Motion
and Deformation (Second edition), Springer Science+Business Media, 1999
Basic Engineering for Medics and Biologists 13
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-13

I.2. Mechanics of Materials

Prashant K. SHARMAa
a
Department of BioMedical Engineering, University Medical
Center Groningen, The Netherlands

Abstract. Mechanics of materials is the science of forces applied on a body and

response of the body in terms of deformation. Different type of loadings on bodies
with different geometries or made of different material give rise to different
deformations. Last but not the least, this science allows to predict the failure of a
body under certain loading condition hence makes it possible to optimize the
design for that particular condition.

Keywords. Stress, strain, stiffness, torsion, bending, ductile, brittle

Introduction

This field of science deals with the relation of external loads acting on a body to the
induced internal forces experienced by the material of that body. Internal forces either
result in deformation i.e. gross changes in the external shape or failure.

1. Force and Deformation

Assume a bucket of mass 1 kg hanging on a rope, then from intuition we know that the
bucket is pulling the rope downwards i.e. exerting a force due to gravity. The
magnitude of the force is equal to the mass of the bucket times the acceleration due to
gravity, g, with a value of 9.8 m s-2. Thus the force experienced by the rope (F = m x g)
is 9.8 kg m s-2, or 9.8 N (≈ 10 N). In other words, one Newton of force will act if 102
(≈100) g of mass is hanging from the rope.
Let us now analyze a simple situation where mass m is hanging on a cylinder with
radius of r (Figure 1, left) and a mass of 2m on a cylinder with radius 2r (Figure 1,
right). Can we say which cylinder will fail first? A force of mg is acting on the left
cylinder and twice the force 2mg on the right cylinder. Cross-sectional area of the
cylinder on which this force is active is also different, the left cylinder has an area of
πr2 (pi, a constant = 3.142 x radius squared) whereas the right cylinder has 4 times the
area i.e. 4πr2. We have to invoke the concept of Stress, denoted by σ, (sigma) in order
to answer this question.
14 P.K. Sharma / Mechanics of Materials

F
σ= (1)
A

where F is the force and A is the cross-sectional area on which the force is active. Unit
of σ is N/m2, also called Pascal (Pa)

r 2r

m 2m

Figure 1. Loading situation of two cylinders

The left cylinder is experiencing a stress of mg/πr2 whereas the right cylinder
experiences half that much (mg/2πr2), therefore we can now say that the left cylinder
will fail first.
The situation we analyzed is called uniaxial state of stress because the cylinders
are experiencing stress along one axis and the stress is called normal stress because
the force is acting perpendicular to the cross-sectional area (Figure 2A). Normal stress
always tries to either pull, called tension, or push, called compression, a body (Figure
2B).

A B

90o

F Tension Compression

Figure 2. Normal stresses

Material inside the cylinders as shown in Fig 1 responds to this tensile stress and
deforms by stretching itself resulting in increase in the length of the cylinder at the cost
of a decrease in the diameter. This deformation is expressed in terms of normal strain,
 P.K. Sharma / Mechanics of Materials 15

denoted by ε (epsilon). Strain is a concept where the deformation due to the applied
stress is normalized by the initial dimensions of the body.
ΔL
ε= (2)
L

where L is the original length and ΔL is the increase in length (Figure 3A). Strain
is a dimension-less quantity since both ΔL and L are expressed in the same unit.

L L

ΔL

A. Tension

ΔL
L

B. Compression

Figure 3. Normal strain

In case of compression (Figure 3B) the formula to calculate the strain remains the
same but the meaning of ΔL changes into a decrease in height thus takes a negative
sign.
Why can’t we simply measure the applied force, in Newtons and deformation in
centimeters instead of using stress and strain? The concept of stress and strain helps us
to compare the mechanical properties of say stainless steel and aluminum cylinders
without having to make them exactly of the same size and shape and applying exactly
the same force. It helps us to compare our results with laboratories from the other side
of the world and most importantly by measuring the mechanical properties of a small
piece of metal in the laboratory we can predict the performance of a long beam of the
same metal placed in a bridge.

2. Mechanical Properties of Materials

In day to day life we talk about the strength of different materials and compare them by
saying that one is stronger, tougher or harder than the other.
16 P.K. Sharma / Mechanics of Materials

Elastic Plastic

I II III IV
Strain to
failure

Stress, σ (Pa)
Elastic
limit

Tensile Failure
strength stress
Yield
stress

Strain ε

Figure 4. Stress – strain diagram

What do we mean by all these words- strength, toughness and hardness? Strength,
toughness and hardness depend on the ability of a material to sustain load without
undue deformation. If we keep on increasing the force on the left cylinder in Figure 1
then we know that the cylinder will deform, but do we know at what applied force it
will fail and how long it will become? The answer is that it depends on the strength of
material the cylinder is made of. Strength of a material can be determined by
performing either a tensile or compression test and record the results in terms of stress
and strain. During a tensile test, an increasing amount of stress is applied to a standard
specimen and the resulting strain is measured, these stress and strain data are plotted on
a so called stress-strain diagram (Figure 4). Starting from zero, as the applied stress is
increased the strain increases linearly up to a limit called the elastic limit. In this region
I the material shows elastic behavior and follows Hooke’s Law (Eq. 3), where the strain
is linearly proportional to the applied stress and the constant of proportionality E is the
Young’s modulus having the same units as stress i.e. Pascal

σ = Eε (3)

If the applied stress is removed within the elastic limit then the material regains its
original size meaning that the strain reduces to zero and there is no permanent
deformation in the specimen (Figure 5A). At the end of the elastic region, the plastic
region starts where a slight increase in stress above the elastic limit i.e. yield stress
results in deformation almost without an increase in stress, this process is called
yielding (region II). In region III the applied stress again increases, resulting in an
increase in strain. The point where the maximum amount of stress is supported by the
specimen is called the tensile strength. In region IV the supported stress continuously
decreases leading finally to failure. If the applied stress is removed in the plastic region,
the strain does not reduce to zero but some residual strain remains in the form of
permanent deformation (Figure 5B). The start of the plastic region is difficult to
determine, since the linear increase of the stress is gradually changing into the yielding
phase. For practical reasons the start of the plastic region is defined by the point where
 P.K. Sharma / Mechanics of Materials 17

if the applied stress is removed the specimen keeps a small amount of permanent
deformation (ε = 0.002 = 0.2 %); this stress is therefore called the yield stress σ0.2.

Figure 5. Elastic and plastic deformation

Figure 6. Different materials and their failure

Figure 6 shows two different type of materials; material A (Figure 6A) is strong
because it can support very high stresses but this material is also brittle, meaning that it
cannot be given any permanent (plastic) deformation, just after reaching the maximum
stress the material fails giving rise to a flat fracture surface (Figure 6C). Material B
18 P.K. Sharma / Mechanics of Materials

(Figure 6B) on the other hand is less strong, but tough and ductile because it deforms a
lot before failure meaning that it can absorb a lot of energy by deforming permanently
before failure. This amount of energy is represented by the area under the curve,
Figure 6B inset.
For a tough material the cross sectional area reduces at a localized area and causes
a large plastic deformation, called necking, before the material fails (Figure 6D).
While constructing a bridge, engineers choose a tougher material because then they
can see their bridge change its shape before it finally fails. If they use a strong and
brittle material then the bridge will fail catastrophically without any prior warning.
Hardness tells us the resistance of a material to scratching by another material, this
property is important if we do not want our material to wear out during rubbing with
other surfaces. This is the reason for use of hard materials as artificial joint surfaces and
drill bits. Similarly cortical bone is stronger and harder to avoid wear and tear whereas
trabecular bone is tougher to absorb energy during jumping and falling.
Till now we have talked about the tensile strength of a material, but the normal
stress can also be compressive in nature. Materials like rubber can deform to up to 10
times their original size under tension where as concrete cannot (Figure 7). Concrete
has a very low tensile strength but have very high compressive strength; that is why
steel rods are embedded in concrete to bear the tensile part of the stress in a building
column. In the human body most of the bone structure and teeth are good at bearing
compressive stresses whereas ligaments are specialized structures meant to bear tensile
stresses.

Rubber

ε
1 10

σ
Cement
ε
-0.06 0.01

Figure 7. Materials to bear tensile and compressive stresses

 P.K. Sharma / Mechanics of Materials 19

A B

F
Figure 8. Shear Stress

3. Shear Stress and Strain

We rarely feel the presence of yet another type of stress, which is called shear stress.
This occurs when the force is active parallel to the cross-sectional area (Figure 8A) and
tries to slide one part of the body in relation to the other part (Figure 8B). Shear stress
is denoted by τ (tau) and also equal to the force acting per unit cross-sectional area
with the unit Newton (N) (Eq.4).
F
τ= (4)
A

A B

L
d
Figure 9. Shear Strain

The material reacts to this shear stress as shown in Figure 9A and the shear strain,
denoted by γ (gamma), is calculated using equation 5 (Figure 9B) and again it is a
dimension-less number
d
γ= (5)
L
A material experiencing shear stress also demonstrates a shear stress-strain curve
very similar to a normal stress strain curve, with an elastic region where it follows
Hooke’s law and shear strain is linearly proportional to shear stress and having a
modulus of rigidity, G, (Eq.6). Above the elastic limit the material deforms
permanently
20 P.K. Sharma / Mechanics of Materials

τ = Gγ (6)
Normal stresses (both tensile and compressive) and shear stress are the only two
types of stresses which can exist inside the material and hence the material manifests
only normal and shear strain. All the different types of external forces acting on the
body induce normal or shear stresses, although active in more than one direction.
There are two more ways in which external forces can be active on a body; one in
which the cylinder is twisted called torsion (Figure 10) and second where the cylinder
is bent called bending (Figure 14A). We shall now analyze how these external forces
translate into internal stresses.

T T

τ τ

Cylinder Tube

Figure 12. Shear stress distribution

Figure 10. Torsion

If the cylinder surface is marked with a black square (Figure 11A) and external
torsional moment, T, is applied then the cylinder deforms in such a way that the square
is transformed to a rhombus shape (Figure 11B&C). We have seen that this
transformation takes place only when the forces are active parallel to the cross-
sectional area i.e. shear stress (Figure 8, 9). Thus externally applied torsional forces
induce shear stresses in the cylinder hence the cylinder shows shear strain (γ=d/L,
Figure11B,C,D).
If we know the magnitude of the externally applied torsional moment, also called
torque, on a cylinder of radius r then we can calculate the induced maximum shear
stress at the cylinder surface using Eq.7.
2T
τmax = (7)
πr 3
If we look at the cross section of the cylinder then we will see that the induced
shear stresses vary linearly from zero in the center to a maximum value at the surface,
whereas in a hollow tube the stress is a finite value at the inside diameter with
maximum at the outer diameter (Figure 12). The maximum shear stress at the hollow
tube surface with outside radius, ro, and inside radius, ri, is calculated using Eq. 8. For
the same applied torque and outside tube diameter, as the inside tube diameter increases
the maximum induces shear stress at the outside surface increases.
 P.K. Sharma / Mechanics of Materials 21

2Tro
τmax = (8)
π( ro4 − ri4 )

A A B
h
b
F

L
B C D D
C
d

L E σ
+ τ

- 0 0

Figure 11. Internal shear stress Figure 14. Normal and shear stresses induced by the external
induced due to external torsional bending moment
forces.

If we know the material properties then we can calculate the shear strain at each
point, this shear strain cumulatively results in twisting of the cylinder. The angle of
twist, θ (theta), (Figure 13) for a cylinder made of material with modulus of rigidity G
and length L experiencing an external torque of, T, can be calculated using Eq. 9. The
longer and thinner a cylinder greater is the angle of twist.
2TL
θ= (9)
πGr 4
If a square cross-section beam is fixed at one end and a force is applied at the
other end perpendicular to its axis then there is a induced bending moment M in the
bar. A bending moment is the applied force multiplied by the distance of the force
from the fixed end (M=F.L) and is thus reported in N.m (Figure 14A).
22 P.K. Sharma / Mechanics of Materials

T
Figure 13. Angle of twist due to torsion

This bending moment induces both normal and shear stresses in the beam; Figure
14A shows how a rectangle drawn on the side of a beam changes to a trapezoid due to
stretching of the beam above the central axis and compression below it. This means that
tensile stresses are induces on the beam above and compressive stresses below the
central axis (Figure 14C&E). The maximum tensile or compressive stress at the top or
bottom surface can be calculated by Eq. 10.

6M
σmax = (10)
Tensile / Compressive
Top / bottom bh 2
If we imagine the beam to be constructed with a pile of much thinner beams
(Figure 14B) then by intuition, the same as bending a thick book along its binding axis,
we know that these thin beams will slide and move relative to each other when a
bending moment is applied (Figure 14D). This can only happen if there is a force active
along the interface between the thin beams i.e. shear stress is being active. Even when
the beam is not composed of thinner beams we can see that these shear stresses will be
induced, the proof for this can be seen from the longitudinal fracture through the center
visible in old wooden beams used to support the roof. Induced shear stresses are
maximum at the central axis and drop to zero at the top and bottom surfaces (Figure
14E). Maximum shear stress at the central axis can be calculated by equation 11 where
V is the internal shear force resulting from the applied bending moment.

3V
τmax
centralaxis = (11)
2bh
The deformation caused by a bending moment is deflection “v” and slope of the
beam. Both parameters can be calculated at any point on the beam defined by distance
x from the fixed end by solving a differential equation (Eq. 12), where the Young’s
modulus (E), moment of inertia (I) and bending moment at the point (M(x)) is known.

d 2v
EI = M ( x) (12)
dx 2
 P.K. Sharma / Mechanics of Materials 23

Some typical solutions to the differential equation for a beam loaded with force, F
(N), bending moment, M (Nm) and distributed load, Q (N/m) is presented in Figure 15.

Figure 15. Deflection and slope of a beam under different loading situations

4. Failure

Failure can now be decided in terms of the maximum stress above which we consider
our material to be rendered useless. While constructing a bridge or making a load
bearing implant, we would never like that our structure deforms permanently thus we
can say that the upper limit of any stress induced in the material should be the yield
stress. In building and bridge construction a margin is given to be always on the safe
side i.e. the calculated induced stresses in the material should be always, say, less than
70% of the yield stress and this is called the maximum allowed stress (σallowed , τallowed).
In simple external loading situation like tension, compression, torsion or bending
we can calculate the maximum internal stresses using equations 1, 4, 7, 8, 10 and 11
and if these stresses are higher than the allowed stresses then we know that the
structure will fail. In real situation two or more of these external loading takes place
simultaneously, what happens then? One such simple example of combined loading is
a spherical liquid storage tank, for example the heart. The tank is pressurized and if we
look at a small square piece of material on the tank surface we can see that the square
experiences tensile stresses on both sides as shown in Figure 16. To handle these
situations with combined loading we use Mohr’s circle. Most of the materials have to
be analysed 3-dimensionally but since they are geometric in nature, using the
symmetry we can convert them to 2 dimensions like we did for the spherical storage
tank (Figure 16).
24 P.K. Sharma / Mechanics of Materials

−
σy
d c b σ
σx − σ x +σ y +
2θ p
σ avg =
2 2θ s τ xy

a
e
σx
τ +

Figure 16. Spherical storage tank Figure 17. Mohr’s circle

Mohr’s circle is a plot between the normal stress (σ) on the horizontal or x-axis
and the shear stress (τ) on the vertical or y-axis (Figure 17). Tensile normal stresses are
positive and compressive are negative, for the shear stress we follow a sign convention
shown in Figure 18 where a very small volume element inside the bulk of the material
is shown. Coming back to Figure 17, the position of the center of the circle c is the
average of the two normal stresses (σx and σy and reference point a can be plotted with
τxy on the y-axis and σx on the x-axis. Now we can join points c and a to get the radius
and then the circle can be drawn with a radius given by Eq. 13. From this Mohr’s
circle for a given situation we can find out the maximum normal stress, point b, and
shear stress, point e, active in the material.

2
⎛ σx − σ y ⎞
R = ⎜⎜ ⎟ + τ 2xy (13)
2 ⎟
⎝ ⎠
If we take a simple situation where a tensile stress of 100 Pa is active on a square
cross section beam then the state of stress at a very small volume element will

+σy -σy

+τxy -τxy

+σx -σx

Figure 18. Sign convention

be as shown in Figure 19A. Surprisingly, the Mohr’s circle (Figure 19B) of this volume
element shows that there exists an orientation (45o clockwise, Fig19C) where
 P.K. Sharma / Mechanics of Materials 25

maximum shear stress of 50 Pa is active. Therefore for supporting 100 Pa of stress we

have to choose a material for which the σallowed is at least 100 Pa and τallowed is at least
50 Pa, all the material with τallowed<50Pa cannot be used even when they can support the
normal stress of 100 Pa and will deform (Figure 19D).
Looking at a more complex combined loading of compressive and torsional nature,
Figure 20 shows a cylindrical beam loaded with 100 Pa compressive stress and a
torque giving rise to a shear stress of -50 Pa on its surface. From point d in Mohr’s
circle we can see that the maximum normal stress induced is compressive and 120 Pa
and point e tells us that the maximum shear stress is about 71 Pa. Therefore a material
for this application should be capable of bearing a compressive stress > 120 Pa and a
shear stress > 71 Pa.
From simple loading conditions Mohr’s circle can be constructed and the highest
normal and the highest shear stress can be determined and thus prediction of whether
the construction will deform or totally collapse can be made.

A B σ x −σ y
= 50
2

100 Pa
a σ
+σx σ avg = 50 90O τ xy = 0

σx = 100, σy = 0 e
τ = 50 Pa
τxy = 0
τ
C D

100 Pa 50 Pa
100 Pa

Figure 19. Mohr’s circle for uniaxial tension

e(-50,-70.7)
100 Pa
a(-100,-50)
T
σy=0 σ c(-50,0) b(20.7,0)
d(-120.7,0)
σx=-100 Pa

τxy = -50 Pa
τ

Figure 20. Mohr’s circle for combined compressive and torsional loading
26 P.K. Sharma / Mechanics of Materials

References

[1] R.C. Hibbeler, Mechanics of Materials. Printice Hall International, New Jersey, 2003
[2] N. Özkaya and M. Nordin, Fundamentals of Biomechanics: Equilibrium, Motion, and Deformation.
Springer, 1999
Basic Engineering for Medics and Biologists 27
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-27

I.3. Dynamics of Human Movement

Bart (H.F.J.M.) KOOPMANa
a
University of Twente
Dept of Biomechanical Engineering
Enschede
The Netherlands

Abstract. The part of (bio)mechanics that studies the interaction of forces on the
human skeletal system and its effect on the resulting movement is called rigid
body dynamics. Some basic concepts are presented: A mathematical formulation to
describe human movement and how this relates on the mechanical loads acting on
the skeletal system. These equations of motion depend on the mechanical
properties of the skeletal system, such as dimensions and mass distribution. It is
applied to describe and analyze human gait.

Keywords. Human movement, rigid body dynamics, gait analysis

Introduction

Human movement dynamics relates the applied forces or moments of force (i.e. the
kinetics) to the movement (i.e. kinematics) of the skeletal system. In the control loop to
displace body parts (Figure 1), the Central Nervous System (CNS) is the controller that
activates the motors (muscles). The movement dynamics are represented by the skeletal
system, and the resulting movement is measured and fed back into the CNS.
In this chapter we focus on the skeletal system only. The part of (bio)mechanics
that studies this is called rigid body dynamics. Some basic concepts are presented: A
mathematical description of human movement; how this depends on the mechanical
loads acting on the skeletal system and how this depends on the mechanical properties
of the skeletal system, such as dimensions and mass distribution. It will be applied to
describe human walking.

1. Rigid Body Dynamics

The human body consists of a more or less rigid structure (the bones of the skeleton) to
which soft tissues (muscles, fat, organs etc.) attach. A structure is considered rigid
when under practical loading situations, the structure does not noticeably deform, i.e.
the mechanical stiffness seems to be infinite. Since the soft tissues are not rigid and do
deform in practical loading situations, an accurate mechanical description of human
28 H.F.J.M. Koopman / Dynamics of Human Movement

movement involves the description of both soft tissues and skeleton. The description of
the dynamics of the human body would involve very complex calculations, since not
only movement, but also deformations have to be considered. This approach is only
feasible in some very specific applications, for example to study the effect of car crash
impacts on brain damage. As it is not possible to measure all soft tissue deformations,
some assumptions have to be made to be able to study human movement.

intention CNS activation muscles forces skeletal movement

system

sensory physiological
feedback sensors

Figure 1. Schematic block diagram of the human motor control system

The most important assumption in rigid body dynamics is that movement only
occurs in the joints. As a consequence of this assumption, the structures in-between the
joints, the so-called segments, are assumed to be rigid, so they have an infinite large
stiffness. All passive soft tissue structures are considered to be rigid as well. The
muscles are replaced by the forces they produce. So it is assumed that the human body
behaves like a segments model or a linked system of rigid bodies. Apart from the
enormous reduction of complexity, this viewpoint has other advantages as well: The
rigid body equations of motion, when applied to each segment, are sufficient to
describe the dynamics of the entire system.
Rigid body dynamics describes the kinematics (movement) and kinetics (forces) of
a given segments model. The equations of motion are the link between these two,
which is a set of two first-order nonlinear differential equations for each movement
component (or, alternatively, a single second order equation that includes a dependency
on the second time derivative, the acceleration). For a given segments model the forces
and movements provide interchangeable information: The forces can be calculated
when the movements are known (i.e. inverse dynamics approach) and the movements
can be calculated when the forces are known (i.e. direct or forward dynamics approach).

2. Kinematics

2.1. Medical Motion Description

Two bones can move with respect to each other by virtue of the joint in between. In the
human body, only the synovial joints allow for large movements. In principle, a bone
has six Degrees-of-Freedom (DOF) of motion with respect to the other bone: three
rotations and three translations. The motions of the joint are limited by passive
structures like the articular surfaces and the ligaments. These passive structures pose
 H.F.J.M. Koopman / Dynamics of Human Movement 29

restraints to the joint motions: Though motions in the direction of the restraint are still
possible, these motions will be very small.
For example, cartilage can be compressed a few millimeters. Most of the time
these small motions are neglected, the motion is said to be constrained. For each
constraint the number of DOF diminishes by one. As soon as constraints come into the
picture, one has already started modeling the joint. In Figure 2 all combinations of
(constrained) rotations and translations are shown. Many of them are merely
hypothetical, and will not be found in the human body.

Figure 2. Combinations of rotations and translations in joints. Many joints

shown are hypothetical and are not found in the human body.
30 H.F.J.M. Koopman / Dynamics of Human Movement

Figure 3. Schematic drawing of 5 joint types, ball and socket (1), condyloid (2),
saddle (3), hinge (4) and pivot joints (5).

Traditionally, joints have been studied (e.g. by Fick [4]) by comparing them with
standard revolute joints: Hinges, spherical joints, ellipsoidal or condyloidal joints,
saddle joints (see Figure 3). The shoulder joint and the hip joint behave approximately
as spherical joints only permitting three rotational DOF. The elbow joint and finger
joints can be regarded as hinge joints with one rotational DOF. The first metacarpo-
phalangeal (thumb) joint and the ankle joint resemble a saddle joint, having two non-
intersecting rotational axes. But often one cannot derive the potential motions from the
shape of the articular surfaces.
Since anatomists were the first to study joint motions, medical definitions still
dominate the way joint motions are described. The goal of such medical definitions is
to distinguish between pathological and normal motion, and to evaluate the outcome of
treatment: Is there improvement in the range of motion or not.
As a starting position the anatomical position (Figure 4) is used. From this position
the motion is defined for each single rotation. For spherical joints, the rotation axes are
defined along the axis of the global coordinate system: A vertical axis and two
horizontal axes, pointing backward-forward (antero-posteriorly) and side-side (medio-
laterally, from the centre of the body to the left or right side). Commonly used terms
are flexion-extension, abduction-adduction, and medial and lateral rotation. A problem
occurs if the motion is not about just one of the standardized axes, but is a combination
of rotations. The order of rotations is not defined, and it is not clear whether the
rotation axes move with the bone or not. This causes much confusion about rotation
angles, and makes comparison between studies often impossible.
 H.F.J.M. Koopman / Dynamics of Human Movement 31

Figure 4. The anatomical positions

2.2. The Segments Model

All models to analyze or simulate human movement are based on segments. These are
coupled systems of rigid bodies, with dimensions and inertial properties that are
deduced from the dimensions and inertial properties of the human body. Initially, these
segments models were used to dimension the dummies used in car crash experiments
[3] and were therefore based on averaged data of human bodies. With the introduction
of the segments models to computer simulations, a wider variety became possible. The
segments models became scalable, with inertial properties usually depending on local
segment dimensions and total body weight. In this way, the segments model could be
matched to fit each individual [2].
The choice of the number of segments should be large enough to simulate the
movement adequately. Too large a number of segments, however, would lead to
unnecessary complexity and larger computational efforts. To simulate walking,
segments models varying from 3 segments [10] up to 17 segments [6] have been
proposed. To further reduce complexity, symmetry between right and left leg is often
assumed [1] and the movement is often restricted to the sagittal plane only.
In a model for normal walking, there are segments for the thighs, shanks and feet.
The head, arms and trunk (HAT) and the pelvis are modeled as two separate segments.
The segments are connected to each other at the joints (Figure 5). Although it is
possible to have more than two joints in a segment (e.g. the pelvis), each joint is
32 H.F.J.M. Koopman / Dynamics of Human Movement

connecting just two segments. To define the position of the segments in space, an
absolute or reference frame is attached to the floor, with the x-axis pointing in the
walking direction, the y-axis pointing upward and the z-axis perpendicular to the xy-
plane in the lateral direction. Figure 5 shows an 8-segmental model with numbering of
the segments, the joints and definition of the reference frame. It should be noted that
the shape of the segments is of no importance as long as the positions of the joints and
the mass properties are well defined. In each segment, a local frame is defined,
standardized with the help of some bony landmarks on the segment.

Figure 5. The segments model for the body at rest. From left to right: Numbering of segments and joints,
definition of the reference frame, definition of local frames and positions of the centers of mass.

The segments model must include the following mechanical properties, which are
dependent on individual characteristics: The dimensions and positions of the segments
in space, the positions of the joints, and the mass properties of the segments (mass,
position of the center of mass and moment of inertia tensor). Other properties, such as
the ranges of motion of the joints, the maximal moments of force that can be exerted on
the joints, and muscle models, may also be included [8].

2.3. Mathematical Motion Description

The segments model is the basis for a sound technical description of human movement.
In general, a movement of a rigid body is defined as the displacement and orientation
of the local frame relative to the reference frame or another local frame. Therefore, any
description of a (body or joint) rotation implicitly assumes that some kind of segments
model is defined, otherwise the rotation has no unique meaning. The anatomical
position is usually taken as the offset position for the segments model where the
relative rotations are defined as zero.
 H.F.J.M. Koopman / Dynamics of Human Movement 33

Figure 6. Transformation of position vector Ap from local coordinate system

A to position vector Gp in the global coordinate system G.

Mathematically, the motion of a body with respect to another body is described

unambiguously by a [3x3] rotation matrix R and a translation vector t [5], see Figure 6.
This results in a linear equation for an arbitrary position vector p. The rotation matrix
can be viewed as some multi-dimensional function that, when it operates on the vector
p, it changes the orientation of p. Note that p has 3 components, and each component
may change with respect to each of the 3 dimensions, so the rotation matrix represents
9 (or [3x3]) dependencies.
G G A A
( p − t )= GA R ⋅( p − t )
(1)
G G A
p = t + GA R⋅ p

Bold capitals are used to denote matrices; underlined lowercase characters are used
to denote vectors. Gp is the position vector of point P in the global coordinate system G,
A
p is the position vector of point P in the local coordinate system A. Gt is the translation
vector from the origin of A expressed in coordinate system G. Obviously At, the origin
of A expressed in coordinates of A equals the zero vector 0. GAR is the rotation matrix
that describes the rotation from the local coordinate system A to the coordinate system
G (the global coordinate system).
Element-wise, equation (1) looks like

r13 ⎤ ⎡ A p x ⎤ ⎡ t x + r11 p x + r12 p y + r13 p z ⎤

⎡ G p x ⎤ ⎡ G t x ⎤ ⎡ r11 r12 G A A A
⎢G ⎥ ⎢G ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ p y ⎥ = ⎢ t y ⎥ + ⎢r21 r22 r23 ⎥ ⋅ ⎢ A p y ⎥ = ⎢ G t y + r21 A p x + r22 A p y + r23 A p z ⎥ (2)
⎢ G p ⎥ ⎢ G t ⎥ ⎢r r r33 ⎥⎦ ⎢ A p z ⎥ ⎢⎢ G t z + r31 A p x + r32 A p y + r33 A p z ⎥⎥
⎣ z⎦ ⎣ z ⎦ ⎣ 31 32 ⎣ ⎦ ⎣ ⎦
in which the elements of the rotation matrix are cosines of the angles between the
axes of the global and local coordinate system (see Figure 6):
34 H.F.J.M. Koopman / Dynamics of Human Movement

⎡ r11 r12 r13 ⎤ ⎡cos(θ11 ) cos(θ12 ) cos(θ13 )⎤

GA ⎢ ⎥ ⎢ ⎥
R = ⎢r21 r22 r23 ⎥ = ⎢cos(θ 21 ) cos(θ 22 ) cos(θ 23 )⎥
⎢⎣ r31 r32 r33 ⎥⎦ ⎢⎣cos(θ 31 ) cos(θ 32 ) cos(θ 33 )⎥⎦
A G
θ11 = ∠( x, x ), (3)
A G
θ 21 = ∠( x, y ), etc.

Although the rotation matrix contains 9 elements, there exist dependencies

between them because we assume rigidity of the segment it represents. Ultimately,
there are only three independent variables present. Together with the three elements of
the translation vector, these six variables describe the potential 6 DOF motion of the
segment or between two bones. To better visualize motion and comparison between
subjects, the rotation matrix is reduced to three successive rotations about well-defined
axes, so-called Euler angles. This can be viewed as three hinge joints in succession.
Rotation of the first hinge will change the orientation of the second and third, rotation
of the second hinge will change the orientation of the third hinge, and rotation about
the third hinge will result in the desired orientation of the local coordinate system.
In Biomechanics, usually Cardan angles are used, which are Euler angles with
successive rotations about the local x-, y- and z-axes. Instead of the order x-y-z, any
other combination of these rotations could have been chosen. Since matrix
multiplication is not a commutative operation (a different order of the rotations, e.g. y-
x-z instead of x-y-z, will give different outcomes), each combination of rotations will
result in other values of the angles. Other parameterizations for the orientation may be
used as well. For example, a well-known parameterization for rigid body movement is
the Finite Helical Axis (FHA) definition, which utilizes helical angles.

3. Dynamics

The next step after defining the movement of rigid bodies is to derive the equations of
motion that define the relations between the acting forces and torques and the
movement. A straightforward way to do this is to apply the Newton-Euler approach.
Although alternative formulations exist that have specific applications, such as the
Lagrange approach, the Newton-Euler approach is simplest because:
The method is identical for each segment.
It is based on the Free Body Diagram of the segment. With this Free Body
Diagram there is a simple recipe to obtain the equations of motion.
The equations of motion for the entire body is the sum of the equations for each
segment.
Newton formulated the equations of motion for systems of mass particles. Euler
recognized that a rigid body is a special case for such a system: The positions of the
particles are constrained with respect to each other. This leads to the notion that the
internal forces (the forces acting between the particles) do not perform work and do not
contribute to the equations of motion for the entire system, the rigid body. Since a rigid
body has six degrees of freedom (DOF), there must be six equations describing the
relation between forces and motion. This leads to the formulation of the Newton-Euler
equations of motion for each segment:
 H.F.J.M. Koopman / Dynamics of Human Movement 35

d 2p
CM
F CM = m
dt 2 (4)
d ( J CM ω )
M=
dt
Where FCM is the resulting external force acting on the center of mass (CM) of the
rigid body with mass m and pCM is the position of CM. The first derivative with respect
to time of pCM reflects the change of position in time, or the velocity of CM. Likewise,
the second derivative of pCM (as in equation 4) reflects the change of velocity in time,
or the acceleration of the center of mass. Any force, acting on the rigid body, can be
divided in a force, acting on CM, and a moment of force M. The second vector
equation in (4) resembles the first, as the resulting torque equals the product of rotation
inertia tensor JCM and angular acceleration (or the first time derivative of the angular
velocity vector ω (omega). However, unlike the mass m, JCM is in general not constant
in each coordinate system. This leads to considerable complications in 3-D; in 2-D on
the other hand JCM reduces to a single constant component.

MP FP

P rAP
A θ

rA
G
y

G mg
G
x

Figure 7. Free body diagram of a rigid body A with joint P.

Consider now the 2-D situation of a (foot) segment A with mass m and moment of
inertia JA during the swing phase of walking. A local frame is attached to the center of
mass with an angle θ to the global (inertial) frame G. The segment connects to another
segment in joint P, the ankle joint (Figure 7). Of course, the lower leg segment applies
some connection forces on the foot in the ankle joint, expressed by the joint force FP
and joint moment of force MP. Finally, a gravity force applies to the foot, thus
completing the free body diagram. When all vectors, which have both magnitude and
direction, are expressed in global coordinates, equation (4) reduces to 3 scalar
equations – scalars have magnitude only:
36 H.F.J.M. Koopman / Dynamics of Human Movement

G d 2 G rAx
FPx = m ⋅
dt 2
G
d 2 G rAy
FPy − m ⋅ g = m ⋅ (5)
dt 2
dω
M P +( G rAPx ⋅G FPy − G rAPy ⋅G FPx ) = J A ⋅
dt

Note that g is the acceleration of gravity constant. The expression between

brackets in the moment of force equation is the vector product of force and moment
arm. Note also that the vector ArAP (expressed in the local frame A) is a constant but
G
rAP (The same vector but now expressed in global frame G) depends on the angle θ
(and thus on time) as defined by equation (3). The vector GrA may be viewed as the
translation vector in equation (1). With equations (1), (2) and (3), equation (5) may be
expressed as three differential equations in the (time-dependent) degrees of freedom rAx,
rAy and θ.
When the movement of the segment is known the second derivatives of the
position and rotation coordinate may be calculated. Equation (5) may then be used to
calculate the (unknown) joint forces and moments of force. Likewise, when additional
known forces would be present, equations similar to (5) may be derived to compute the
unknown forces at the ankle. For example, in an inverse analysis of the walking
movement, the measured ground reaction forces are known. These would then
contribute to the resulting forces and torques acting on the foot segment, and the
equations of motion (4) are used to solve for the ankle force and torque. From these and
the equations of motion of the shank segment, the knee forces are calculated, and so on.

4. Application to Human Gait

Bipedal walking is a complex movement, a balancing performance against the constant

pull of gravity. Numerous definitions of walking are made, such as:
“In bipedal locomotion, man is continuously preventing a fall by placing one foot
in front of the other.”
“The two basic requisites of walking are the continuing ground reaction forces
that support the body and the periodic movement of each foot from one position of
support to the next in the direction of progression”
Apart from these formulations, walking can be quantified with a number of
parameters. These are shown in the next sections: The step-parameters to describe the
timing of the movement; kinematics for the movement itself (the joint rotations) and
dynamics to describe the forces and moments of force that cause this movement.
The two ways to apply the equations of motion are usually referred to as the
inverse dynamics and the direct (or forward) dynamics approach. In the inverse
dynamics approach, the movement is assumed to be known and the forces and
moments of force, needed to bring about that movement, are calculated (e.g. Koopman
[9]). Inverse dynamics is applied in gait analysis, the equations of motion are usually
derived with the Newton-Euler formulation. The estimated internal forces can be
 H.F.J.M. Koopman / Dynamics of Human Movement 37

further processed in muscle models to estimate the distribution of muscle forces, which
allows for a validation with measured EMG patterns.
In the direct dynamics approach, the movements of the segments are calculated by
integrating the equations of motion, usually based on a Lagrangian formulation. This is
only possible when the joint moments of force are known or assumed to be zero. The
latter is the case in ballistic walking [10]. The joint moments of force can be found by
trial and error such that a normal walking pattern results [12], from estimations of the
muscle forces [11], as the result of an inverse dynamics model or by optimization
techniques. This direct dynamics modeling will not be discussed further.

Figure 8. The walking cycle

4.1. Step-Parameters

The step-parameters are used for a general characterisation of the walking pattern.
They are determined by the points of heel contact (HC) and toe-off (TO) of the left and
right feet. For a walking cycle beginning with left heel contact (HCL), this is followed
by right toe-off (TOR), right heel contact (HCR) and left toe-off (TOL). A cycle is
completed with HCL again (Figure 8).
These points divide the walking cycle in four different phases. For the left leg, the
stance phase is from HCL to TOL and the swing phase (or the single stance phase of
the right leg) is from TOL to HCL. There are two double stance phases where both feet
are on the floor. The left double stance phase is from HCL to TOR. One stride consists
of two steps: The left step is usually defined from HCR to HCL.
For the step-parameters, a distinction is made between the time-parameters and the
distance-parameters. With the cycle beginning at HCL (tHCL=0), four other time-
parameters will suffice to define the points of HC and TO. These may be normalized
with the stride or cycle time T to make these time parameters dimensionless for
comparison purposes. Likewise, the distance parameters can be made dimensionless
with the stride length S. For symmetrical walking, the step-parameters are not
independent any more.
38 H.F.J.M. Koopman / Dynamics of Human Movement

Two step-parameters are derived from S and T, which are the average forward
velocity v and the step ratio r. These are calculated from:

v=S (6)
T
r= 1
4S ⋅T (7)

The step ratio is usually defined as the step length divided by the step frequency.
In contrast to this, definition (7) does not depend on differences between right and left
leg for asymmetrical walking. The reason for introducing the step ratio is that this
parameter is shown to be reasonably constant for a wide range of walking velocities
and for different subjects [13]. For normal walking, r ranges from 0.39 to 0.44 m·s for
men and from 0.34 to 0.40 m·s for women [7, 15]. For a constant r, S and T are
determined by the forward velocity only:

T =2 r (8)
v

S = 2 r ⋅v (9)

Figure 9. Joint rotations in the sagittal plane for a cycle beginning with heel contact (in degrees). Hip flexion,
knee flexion and ankle dorsiflexion; average values for 19 subjects with measuring deviations. From (Winter,
1991).

4.2. Kinematics

A more detailed description of the walking movement is obtained when the joint
rotations are measured at each point of time of the walking cycle. The hip, knee and
ankle flexion are well documented in the literature, so measured data can be compared
 H.F.J.M. Koopman / Dynamics of Human Movement 39

with "standard" data. As an example, the rotations that are shown in Figure 9 are
average values for normal walking, measured by Winter [16].
The other rotations that also contribute to the walking movement (e.g. hip
adduction, pelvic rotations) are less well documented; accepted average values have yet
to be established. The kinematics of the walking movement could also be described
with the displacements of the joints as time functions or with a combination of
displacements and rotations. The choice of one of these possibilities mostly depends on
the measuring system that is available.
Note that a joint angle (e.g. hip angle) describes the orientation of one bone
relative to another bone, or more precisely, the orientation of one local frame with
respect to the other local frame of two connecting rigid bodies. A segment angle, on the
other hand (e.g. the foot angle) describes the orientation of a local frame with respect to
the global frame.

Figure 10. Forward (- - -) and vertical (─) ground reaction forces, normalized with body weight, ensemble
average from 19 subjects. From Winter (1991).

4.3. Kinetics

An alternative description of the walking movement is provided with the internal and
external forces and moments of force, acting on the body segments. Of these forces
only the ground reaction forces can be measured directly. The average ground reaction
forces, as measured by Winter [16], are shown in Figure 10.
Note that the average vertical ground reaction force should equal body weight.
Variations from this average must result from accelerations of the total body centre of
mass, according to equation (4).
The joint moments of force are determined by applying the Newton-Euler
equations of motion (4). This always implies that some modeling assumptions,
concerning the mechanical properties of the human body, have to be made. The joint
moments of force are computed from the measured ground reaction forces and
40 H.F.J.M. Koopman / Dynamics of Human Movement

accelerations. Where the variation in the ground reaction forces between different
subjects is reasonably small, due to the assumptions there may be a larger variation in
the joint torques. Figure 11 shows some of the joint moments of force, as determined
for four different subjects [7]. The joint moments of force can be viewed as the net
result of all muscular, ligament and frictional forces acting on the joint. For example,
the push-off with the calf musculature is clearly visible in the ankle torque of Figure 11.
For some applications, the muscular forces can be related to electromyographic (EMG)
signals, which may validate the modeling assumptions.

Figure 11. Joint moments of force for four different subjects (in Nm).
From Inman et al. (1981).

4.4. Energy Expenditure

A distinction is usually made between mechanical energy and metabolic energy

expenditure. Mechanical energy can only be determined by modeling and is based on
computed kinetic and potential energies of segments or on joint powers.
Metabolic energy expenditure is measured by oxygen uptake during walking and is
usually expressed in energy per unit time (Ew) or energy per unit distance walked (Em).
Ralston [14] first showed that Ew is proportional to the square of the walking velocity:

E w = a + bv 2 (10)
This relation is confirmed by various investigators [7]. When Ew is expressed in
watt per kg body mass and v is in m/s, the experimental values for the constants a and b
are a = 2.24 W/kg and b = 1.26 Hz. Em is defined by
Ew a
Em = = + b⋅v (11)
v v
Ew and Em are shown in Figure 12.
The optimal velocity vopt is defined as the velocity where Em is minimal.
Differentiating equation (11) with respect to v and equating to zero yields vopt = 1.33
 H.F.J.M. Koopman / Dynamics of Human Movement 41

m/s. The comfortable walking velocity vcomf is the velocity a person tends to adopt in a
natural walk. vcomf is found to have an average value of 1.39 m/s [7], which differs from

Figure 12. Metabolic energy expenditures Ew and Em for normal

walking at different velocities.

vopt by only 4 %. Inman et al. called this an example of a fundamental feature of human
motor behavior: "In freely chosen rate of activity, a rate is chosen that represents
minimal energy expenditure per unit task". In the case of walking a speed is adopted,
with a specific stride length S and cycle period T, such that each meter is covered as
cheaply as possible. The energy expenditure Em can in this sense be interpreted as
inversely proportional to the efficiency of walking.
In Figure 12 is seen that Em varies only little for a wide range of velocities: The
sensitivity of the efficiency to the velocity is small in this range. When the velocity is
enforced, for example in a treadmill, the stride length, cycle period and joint rotations
can still be chosen freely. The choice is such that the efficiency of the movement is
maximized.

4.5. Work Balance

By applying the equations of motion to the moving segments model in an inverse

dynamics approach, the internal forces and moments of force are calculated. The
product of the joint moment of force and the angular velocity equals the power output
at the joint (Figure 13). A positive power reflects energy generation; a negative power
reflects energy dissipation at the joint. On a muscular level, this is comparable to
concentric (i.e. the muscle shortens while pulling) and eccentric (i.e. the muscle
lengthens while pulling, force opposite to movement) muscle contractions respectively.
These joint powers finally result in an increase (or decrease) of the mechanical energy
(kinetic or potential) of the segments. However, since walking is a cyclic movement,
the total power output of all the joints together in one cycle must equal zero for level
walking. If for example the power output of one cycle would be positive, than the total
mechanical energy of the segments model would be increased during one cycle. An
increase of kinetic energy or walking speed is contradictory to the assumption of a
cyclical movement; an increase of potential energy is only possible for walking uphill.
42 H.F.J.M. Koopman / Dynamics of Human Movement

Figure 13. Joint powers (in W/kg). Average values for 19 subjects.
Obtained from Winter (1991).

To be able to tell something about the work done at the joints, usually a distinction
is made between the positive and the negative work (Wj+ and Wj-) at joint j with power
Pj:
T
⎧ P j ( P j > 0)
W j+ = ∫ Pj+ dt; Pj+ = ⎨
0 ⎩ 0 ( P j ≤ 0)
(12)
T 0 ( ≥ 0 )
⎧ P j
W j− = ∫ Pj− dt; Pj− = ⎨
0 ⎩ Pj ( P j < 0)
This definition results in a work balance as shown in table 1. From the table it is
obvious which joints in general generate energy and which joints dissipate energy. In
most cases, the hip joints generate as much as they dissipate, the knee joints are mainly
dissipaters (in the beginning of the stance phase to accept body weight) and the ankle
joints are mainly generators (at the end of the stance phase to push off, also visible in
Figure 12). The total work for all joints during one cycle should equal zero. In table 1,
the total Wj++Wj- equals –2 J/cycle. This value reflects some inaccuracies of the
calculation, especially the fact that the properties of the segments model do not
perfectly match the properties of the human body.
The last column shows the absolute amount of mechanical work done. When it is
assumed that mono-articular muscles do all the work, no energy is stored in passive
structures, and the efficiency equals 1 for concentric contractions and –1 for eccentric
contractions, this number can be related to the metabolic energy consumption. The
metabolic energy consumption for a person of 80 kg with a stride length of 1.5 m/cycle
is about 3.4×80×1.5=408 J/cycle (see Figure 11).
 H.F.J.M. Koopman / Dynamics of Human Movement 43

Table 1. Typical work balance for normal level walking.

joint normal walking [J/cycle]

j Wj+ Wj- Wj++Wj- Wj+-Wj-

R hip 18 -14 -4 32

L hip 18 -14 -4 32

R knee 17 -32 -15 49

L knee 17 -32 -15 49

R ankle 13 -3 10 16

L ankle 13 -3 10 16

total 96 -98 -2 194

About half this amount is needed for basal metabolism, the energy consumption at
zero velocity. This leaves about 200 J/cycle spendet on walking alone. This value is
close to the estimated mechanical work (table 1).
However, if more realistic values for the efficiencies are assumed (i.e. 0.3 and -1.0
for concentric and eccentric contractions respectively, see McMahon [10], the total
mechanical work predicts a metabolic energy consumption of 418 J/cycle. This implies
that energy transfer between joints through biarticular muscles and energy storage in
passive structure are important mechanisms to reduce the over-all metabolic energy
consumption. The work balances are especially useful when analyzing situations that
deviate from normal.

References

[1] R.A. Brand, R.D. Crowninshield, C.E. Wittstock, D.R. Pedersen, C.R. Clark and F.M. van Krieken, A
model of lower extremity muscular anatomy. J. Biomech. Eng. 104 (1982), 304-310.
[2] R.F. Chandler, C.E. Clauser, J.T. McConville, H.M. Reynolds and J.W. Young, Investigation of the
inertial properties of the human body. Report DOT HS-801430, National Technical Information Service,
Springfield Virginia 22151, U.S.A, 1975.
[3] C.E. Clauser, J.T. McConville and J.W. Young , Weight, volume, and center of mass of segments of the
human body. Aerospace Medical Research Laboratory TR-69-70 (AD 710 622), Wright-Patterson Air
Force base, Ohio, 1969.
[4] R. Fick, Handbuch der Anatomie und Mechanik der Gelenke (Handbook of joint anatomy and
mechanics), Gustav Fischer, Jena. 1911
[5] H. Goldstein, Classical mechanics (second edition). Addison Wesley Publishing company, Reading,
Massachusetts. ISBN 0-201-02969-3, 1980.
[6] H. Hatze, Quantitative analysis, synthesis and optimization of human motion. Hum. Movem. Sc. 3
(1981), 5-25.
[7] V.T. Inman, Ralston, H.J., Todd, F., Human walking, Baltimore, Williams & Wilkins, 1981.
[8] B. Koopman, Grootenboer H.J., Jongh H.J. de, An inverse dynamic model for the analysis
reconstruction and prediction of bipedal walking. J. Biomech 28 (1995), 1369-1376.
[9] H.F.J.M. Koopman, The three-dimensional analysis and prediction of human walking, Ph.D.
Dissertation, University of Twente, Enschede, 1989.
[10] T.A. McMahon, Muscles, reflexes, and locomotion. Princeton University Press, Princeton, New Jersey,
1984.
44 H.F.J.M. Koopman / Dynamics of Human Movement

[11] S.J. Olney and D.A. Winter, Predictions of knee and ankle moments of force in walking from EMG and
kinematic data. J. Biomech. 18 (1985), 9-20.
[12] M.G. Pandy and N. Berme, A numerical method for simulating the dynamics of human walking. J
Biomech 21 (1988), 1043-51.
[13] R.H. Rozendal, P.A.J.B.M. Huijing, Y.F. Heerkens, R.D. Woittiez, Inleiding in de kinesiologie van de
mens. (Introduction in the human kinesiology) Culemborg: Educaboek, 1990.
[14] H.J. Ralston, Energy-speed relation and optimal speed during level walking. Int. Zeitschrift für
angewandte Pysiologie 17 (1958), 277.
[15] R.L. Waters, B.R. Lunsford, J. Perry, R. Byrd, Energy-speed relationship of walking: standard tables. J
Orthop Res 6 (1988), 215-22
[16] D.A. Winter, The Biomechanics and motor control of human gait, Waterloo (Canada), University of
Waterloo Press, 1991.
Basic Engineering for Medics and Biologists 45
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-45

I.4. Biofluid Mechanics & the Circulatory

System
Pascal VERDONCKa and Kris DUMONT b
ab
Institute Biomedical Technology, Ghent University, Belgium

Abstract. A fluid is a medium which deforms, or undergoes motion, continuously

under the action of a shearing stress and includes liquids and gases. Applying
biofluid mechanics to the cardiovascular system requires knowledge of anatomy
and geometry, pressure data and blood flow, volume and velocity measurements.
A good example is the assessment of the haemodynamics of biological and
mechanical heart valves.

Keywords. Pressure gradient, Bernoulli, Doppler measurement, Effective orifice

area and performance index, heart valve, regurgitation, laminar flow, turbulent
flow, Reynolds number

Introduction

The performance of a heart valve depends on the flow of blood passing through it.
Cardiac output is the amount of blood pumped by each ventricle in one minute. It is a
function of heart rate (HR), the number of heart beats per minute, and stroke volume
(SV), the amount of blood pumped out by each ventricle per minute. The stroke
volume (SV [ml]), times the number of beats per minute (heart rate, HR), equals the
cardiac output (CO [l/min]; equation 1). Cardiac output in humans varies between 4-6
l/min at rest to 20-35 l/min during exercise, with heart rate variations from 40-70 bpm
at rest to 170-200 bmp during exercise and SV from 60-125 ml during rest to 100-200
ml during exercise. A well trained person develops a larger SV and so CO can be
achieved with a lower HR. [13].

CO = SV ⋅ HR (1)

Changes in either stroke volume or heart rate will alter cardiac output. The SV can
be calculated from velocity measurements over the valve (equation 2):

SV = VTI ⋅A (2)

with VTI, the velocity time integral (VTI [cm]) and A the valve area in [cm2]. A
velocity-time curve is obtained using Doppler ultrasound and the VTI (= ∫ v ⋅ dt ) is the
area under the velocity curve.
46 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System

1. Pressure Gradient

The resistance of the valve against blood flow can be quantified by pressure/energy
losses over the valve.

1.1. Pressure gradient from catheterization

A dynamic systolic gradient is found between left ventricular and aortic pressures
(Figure 1a). Assuming proper alignment of the pressure tracings, a series of
instantaneous pressure differences can be measured, and the mean systolic gradient
ΔPmean can be calculated. The gradient ΔPpeak between the peak systolic left
ventricular pressure and the systolic aortic pressure can also be determined, even
though these two peaks are non-simultaneous. Note that the maximum instantaneous
gradient ΔPmax is always larger than the peak-to-peak gradient. Similar gradients can
be determined for the mitral valve during ventricular filling or diastole [16].

1.2. Pressure gradient from Doppler measurements

The pressure gradient can be calculated using the simplified Bernoulli equation. This
equation is derived from the Bernoulli equation which is based on the assumptions that
flow through the stenosis is laminar and inviscid (viscous forces negligible).
Bernoulli’s hydraulic formula states the conservation of mechanical energy between
two points along a streamline (Figure 1b) [22]; equation 3:

2
1 dv
P1 − P2 = ρ(v 22 − v12 ) + ρ g ( z 2 − z1 ) + ρ∫ ds (3, Bernoulli’s equation)
2 1
dt

where P1 [Pa], v1 [m/s], z1 [m] are the pressure, velocity, and height from a reference
level at the upstream location, and P2 [Pa], v2 [m/s], z2 [m] are the pressure, velocity,
and height at the downstream location. The integral term accounts for the flow
acceleration or deceleration between the two locations. If the equation is applied at
peak systolic velocity, this term becomes zero, because the rate of change of velocity,
the derivative, becomes zero. When further applied on two points along a stream line
located at the same height, one finds that [22]; equation 4:

1
P1 − P2 = ρ (v22 − v12 ) (4)
2

When the downstream velocity is much higher than the upstream velocity
(v2 >> v1), v1 can be neglected. Filling in the value for the density of blood at 37°C
(1060 kg/m3) ρ (rho), and converting pressure from Pa to mmHg (133 Pa = 1
mmHg), results in a factor of about 4, in the so-called simplified Bernoulli’s
equation [22]; equation 5:

ΔPDoppler [mmHg ] = 4v22 (5, simplified Bernoulli’s equation)

 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System 47

a. Example of pressure gradient measurements [5].

b c

Figure 1. Pressure gradient.

where ΔP is in [mmHg] and v2 is in [m/s] [22]. This equation has found wide
application clinically in determining pressure drops across severe stenoses from
noninvasive velocity measurements using Doppler ultrasound. Bernoulli’s equation
is not valid in cases in which viscous forces are significant, such as in long
constrictions with a small diameter, or in flow with separation regions. In addition,
it does not account for turbulent energy losses in cases of severe stenosis. Such
losses should be taken into account, because they reduce the energy content of the
48 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System

fluid [22]. In normal individuals, there is a very slight (1-2 mmHg) pressure
difference between the left ventricle and aorta during ejection that helps drive the
blood across the aortic valve. In contrast, very high pressure gradients are observed
in patients with a stenotic valve. Patients with a pressure gradient ΔP of 100
mmHg, in the case of severe aortic stenosis, and of 16 mmHg in the case of mitral
stenosis, are referred for clinical assessment..

2. Pressure Gradient in Experimental and Clinical Studies

Results from in vitro experiments by Feng et al [5] are displayed in Figure 1a. The
SJM HP (St. Jude Medical Hemodynamic Plus) and On-X valves produced the lowest
mean positive pressure drop, and the CM (CarboMedics) valve produced the highest
under every condition. The differences in these pressure drops were related to the
geometric orifice diameters and the degrees of valve opening. That the On-X valve
produced the smallest pressure drop is mainly due to its larger internal orifice diameter
and the parallel opening of its leaflets. The SJM HP valve also benefits from a larger
internal orifice diameter and a large opening angle. The ATS valve has a lower mean
positive pressure drop than the CM valve in pulsatile flow, despite the findings that its
maximal opening angle is less than that of the CM valve. Furthermore, it can be
observed from Figure 1a, that the pressure drop over the valves increases with the
cardiac output. In the clinical setting, this maximum pressure gradient cannot be
measured invasively because it is not possible to position a catheter across or between
mechanical prosthetic valve leaflets in patients [18]. Doppler gradients across the
central orifices are significantly higher than the transvalvular and net catheter pressure
gradients measured across the valve (Figure 1b). These differences are due to
downstream pressure recovery (Figure 1c) [6, 18], which gives the pressure at a certain
time along the axis of the valve. The pressure shows a drop, but recovers at a certain
distance downstream the valve, due to conversion of kinetic energy into potential
energy upon deceleration of the flow through the valve.

3. Effective Orifice Area and Performance Index

When blood is flowing through an orifice with cross section area A [m 2], convergence
of streamlines make that only part of the cross section is effectively used for flow
passage. This is particularly important in valve dynamics where the effective valve area
should be maximized to avoid high velocities, turbulence, shear, and associated high
pressure drops. The effective orifice area (EOA) is the standard parameter for the
clinical assessment of the severity of the valve stenosis. It is also a parameter that can
be used to compare performance of different mechanical and biological valve
prostheses.
As demonstrated in Figure 2, the flow through the valve is narrowing, using only
the EOA, i.e. a fraction of the total geometric orifice area (GOA). The effective orifice
area (EOA [cm2]) is calculated - based on the continuity equation - as the ratio of
forward stroke volume and VTI; equation 6:

EOAContinuity = SV / VTI (6)

 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System 49

Figure 2. Geometric and effective orifice area (GOA and EOA).

Besides the continuity equation, the Gorlin equation provides an alternative way to
calculate the effective orifice area (EOAGorlin). With Q the flow rate [m3/s] and v the
maximal velocity [m/s] in the vena contracta (narrowed vessel), the EOA is given by

EOA = Q / [ C1 v ] (7)

where C1 is a coefficient to adjust for the velocity profile in the vena contracta.
Through the orifice, there is a conversion of potential energy to kinetic energy, and this
conversion is described by [22]:

v = C2 2 ⋅ g ⋅ Δh (8)

with the pressure head across the orifice Δh = Δz + ΔP/(ρg) and where C2 is a
coefficient to adjust for loss in the conversion of energy from potential to kinetic
energy. With Δz = 0 and assumption of C1·C2 = 1, the formula can be written as:

Q Q Q
EOA = = = (9)
v ΔP ρ Hg ⋅ g ⋅ ΔhHg
2 2
ρ ρ

The commonly used SI units [m2], [m3/s], [Pa] are converted into clinical used
units [cm2], [ml/s], [mmHg] for the area, the flow and pressure respectively.

EOA [cm2] = 104 · EOA [m2],

Q [ml/s] = 106 · Q [m3/s], (10)
ΔP [mmHg] = 103 · ΔhHg [mHg]
50 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System

ρHg = 13600 [kg/m3]

g = 9.81 [m/s2] (11)

Combining equations 9 and 10 results in:

Q ⋅ 10 −6
EOA ⋅ 10 − 4 = (12)
ρ Hg ⋅ g ⋅ 10 −3 ⋅ ΔP
2
ρ

And thus finally the Gorlin equation can be written as:

Q
EOA Gorlin = (13, Gorlin equation)
51,6 ΔP

with EOA in [cm2], ΔP in [mmHg] and the mean forward flow Qfwd,mean in [ml/s],
calculated as the ratio of the “positive” area under the flow curve (forward stroke
volume SV) divided by the duration of forward flow. EOAGorlin is a fraction θ (theta), of
the available geometric orifice area GOA, so that EOAGorlin = θ ·GOA. The
performance index (PI [no dimension]) of the valve is calculated as the ratio of the
effective orifice area and the geometric orifice area (GOA):

PI = EOA / GOA (14)

3.1. Effective Orifice Area

Many clinical and experimental studies are done in order to estimate the EOA and
calculate the performance index PI which enables clinicians and engineers to compare
different types of heart valves. The higher the PI, the better the hydrodynamic
performance is of the valve. This means the contraction of flow through the valve is
minimized and the flow utilizes a high percentage of the available geometric orifice
area. Bech-Hanssen et al. showed similar effective orifice areas in vitro for St Jude
Medical and Omnicarbon valves (Figure 3) [1].

4. Regurgitation

Regurgitation is reversed flow through the ‘one way’ valve. The percentage of
regurgitation (% reg) in relation to the total SV is calculated as

% reg = Vreg / (Vreg + SV) (15)

Vreg [ml] is the total volume of regurgitation and can be split into V close and Vleak.
Vclose is the volume of regurgitation due to the closing of the valve. V leak is the volume
of regurgitation due to leakage of the closed valve (Figure 4).
 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System 51

Figure 3. Effective orifice area of two different mechanical valve types [1]. Results are shown for the in vivo
(black o) and in vitro (open o) studies. Bars indicate mean ± SD. Boxes show the geometric orifice area.

(a) Caged ball valve. (b) Tilting disc valve. (c) Bileaflet valve. (d) Bioprosthesis.

Figure 4. Velocity profiles in cm/s based on in vitro measurements on 27 mm aortic valve designs.
Measurements were done downstream the valve on a centerline at peak systole,
at a cardiac output of 6 l/min and a heart rate of 70 beats/min [21].

4.1. Regurgitation

Figure 5 shows regurgitation data from different types of bileaflet valves [5]. As
reported by Wu et al. [20], the closing volume, expressed as a percentage of the
forward flow volume, increased with decreasing cardiac output. The closing volumes
52 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System

of the valves, shown in Figure 5, are within the acceptable range (<8%) [5]. The SJM
HP valve under low cardiac output condition at 100 bpm showed a somewhat higher
closing volume of 11.4%. Closing volume is believed to be proportional to opening
angle of the mechanical heart valve [5].

Figure 5. Closing volumes of different bileaflet heart valves (ATS: Advancing The Standard, CM:
Carbomedics, SJM HP: St. Jude Medical Hemodynamic Plus, On-X) studied in mitral position in an
experimental setup at 70 beats/min and at 100 beats/min [5].

5. Flow Patterns and Shear Stresses

5.1. Laminar and Turbulent Flow

The Reynolds number (Re) is a quantity that engineers use to assess whether fluid flow
is laminar or turbulent. This is important, because increased mixing and shearing occur
in turbulent flow. This results in increased energy losses which affects the efficiency of
heart valves. A good example of laminar and turbulent flow is the rising smoke from a
cigarette. The smoke initially travels in smooth, straight lines (laminar flow) then starts
to ”wave” back and forth (transition flow) and finally seems to randomly mix
(turbulent flow). The dimensionless Reynolds number Re [-] is defined as:

U ⋅D
Re = ρ (16)
μ

with ρ (rho) the density of the fluid (kg/m3), U the velocity of the flow (m/s), D the
diameter of the vessel (m) and µ (mu) the dynamic viscosity, or internal friction, of the
fluid (mPa·s). Figure 6 demonstrates the different flow regimes in function of Re for
the flow in a straight tube. Generally, a fluid flow is laminar in a stiff tube from Re = 0
 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System 53

to some critical value (Re < 2000) at which transition flow begins. Transition flow is
fluctuating.

Figure 6. Diagram of flow regimes in pipe flow. between laminar and turbulent flow (2000 < Re < 4000).

Fluid flow becomes unstable for higher Reynolds numbers (Re > 4000). In
turbulent flow, there is increased mixing that results in viscous losses, which are
generally much higher than those in laminar flow [19]. The Reynolds number can reach
up to 4500 at peak flow in the normal aortic valve. The Strouhal number gives an
indication of a steady or transient regime of the flow. The Strouhal number is defined
as

D
Sr = (17)
Tp ⋅U

with Tp the period of time of the observed flow phenomenon. A very small Strouhal
number (Sr << 1) is considered quasi-steady. For a Strouhal number close to one, the
flow is considered transient.

Figure 7. Shear stress represents the frictional force exerted by the

flowing blood on the endothelial surface of the wall.
54 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System

5.2. Shear Stresses

As blood flows through a vessel or valve, it exerts a physical force on the vessel wall or
the valve leaflet. This force can be resolved into two principal vectors (Figure 7). Shear
stress, being tangential to the wall, represents the frictional force exerted by the flowing
blood on the endothelial surface of the wall. The shear stress on the vessel wall or on
the valve leaflet is normally called wall shear stress or WSS. Normal stress, or pressure,
is perpendicular to the wall. In the case of laminar flow shear stress is calculated with
equation 18.
r
r ∂U
τ lamin ar = μ = μ ⋅ γ& [Pa] or [N/m2] (18)
∂y

with τ (tau), the shear stress, μ (mu) the dynamic viscosity of the fluid, U the velocity
vector and γ& the shear rate. Turbulence in the blood system increases resistance to flow,
resulting in higher pressure gradients. Turbulent shear stresses are calculated in a
slightly different manner, as shown in equation 19 [3]:
r
r ∂U
τ turbulent = μ − ρu ′v ′ (19)
∂y

with u’ and v’ the turbulent fluctuations of the velocities u and v respectively. The
vascular and valvular endothelial cells are subjected at all times to shear forces that act
on the vessel surface as a result of the flow of viscous blood. Fluid shear stress
regulates endothelial phenotype by altering its gene expression profile, including
growth factors [11,12]. Fluid shear stress transforms polygonal, cobblestoneshaped
endothelial cells of random orientation (Figure 8a) into fusiform endothelial cells
aligned in the direction of flow (Figure 8b). High shear stress (> 40 N/m 2) can cause
direct endothelial injury [11]. Low and oscillating shear stress regions (< 0.4 N/m 2) can
lead to atherosclerotic lesions or plaques, mainly at arterial bifurcations and in the
coronary arteries [11].

a. Physiological arterial haemodynamic shear b. Low arterial haemodynamic shear stress (τ

stress (τ > 1.5 N/m2) [1]. ~ 0-0.4 N/m2) [1].

Figure 8. Transformation of endothelial cell morphology by fluid shear stress: aortic endothelial cells
exposed to low shear stress are randomly oriented (a), while those exposed to physiological shear stress (1.5
N/m2, right panel) for 24 hours align in the direction of blood flow [11,12]
 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System 55

High shear stresses in the blood may create platelet activation [10,15] leading to
thrombosis [23], and the subsequent risk of embolism. Blood platelet damage starts to
occur at shear stress values of 10 N/m2 [8]. Furthermore, the magnitude, exposure time,
and spatial distribution of the shear stresses coincide with damage to red blood cells.
Shear stresses above 200 N/m2 will cause hemolysis [2,4,14]. Hemolysis can also
occur at lower shear stress values if there is a long exposure time [3,7]. Therefore
recirculation zones in prosthetic heart valves are to be avoided, unless they are washed
out by the next heart beat [9], so that the residence time of particles remains limited. In
vitro studies of aortic prosthetic heart valve designs have revealed shear stress values
sufficiently high to cause lethal or sublethal damage to blood cells [17,24]. Shear stress
cannot be measured directly in vitro nor in vivo on a moving heart valve leaflet. For
that reason computational fluid dynamics is used to estimate the shear stress.

6. Flow Patterns and Shear Stresses

6.1. The Caged Ball Valve

Natural heart valves allow blood to flow straight through the center of the valve. This
property, known as central flow, minimises the work done by the heart. With non-
central flow, the heart works harder to compensate for momentum lost to change in
direction of the fluid. Caged ball valves completely block central flow (Figure 9a);
blood therefore requires more energy to flow around the occluder. In addition,
collisions with the occluder ball caused damage to blood cells. Caged-ball valves are
also notorious for stimulating thrombosis, requiring patients to take lifelong
prescriptions of anticoagulants.
High turbulent shear stresses are observed at the edges of the jet (Fig 9a). The
intensity of turbulence during peak systole does not decay very rapidly downstream of
the valve. Elevated shear stresses occurred during most of systole [21]. This type of
valve is no longer used clinically.

6.2. The Monoleaflet Tilting Disc Valve

This valve design allows more central flow while still preventing backflow. However,
flow is not completely central (Figure 9b), because blood has to make a significant
directional change to flow around the occluder disc. Newer models of monoleaflet
valves improve this shortcoming. In the major orifice region (Figure 9b), high turbulent
shear stresses are confined to narrow regions at the edges of the major orifice jet.
During the acceleration and deceleration phases, the turbulent shear stresses are
relatively low [21]. High turbulent shear stresses are more dispersed in the minor
orifice than those in the major orifice regions (Figure 9b)

6.3. The Bileaflet Valve

These valves provide the closest approximation to central flow yet achieved in a
mechanical heart valve (Figure 9c). The two leaflets block flow to some extent, leading
to a three-part flow pattern. High turbulent shear stresses occur at locations
immediately distal to the valve leaflets (Figure 9c).
56 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System

(a) Caged ball valve. (b) Tilting disc valve. (c) Bileaflet valve. (d) Bioprosthesis.

Figure 9. Turbulent shear stress profiles in 0.1·N/m2 based on in vitro measurements on 27 mm aortic valve
designs. Measurements were taken downstream of the valve on a centerline at peak systole, at a cardiac
output of 6 l/min and a heart rate of 70 beats/min [21].

6.4. Biological Valves

Biological valves, optimally mimic native heart valves and provide central flow (Figure
9d). Turbulent shear stress measurements during the deceleration phase show low shear
stresses spread out over a wide region [21]. During peak systole, the high turbulent
shear stresses were confined in a narrow region (Figure 9d).

7. Conclusions

This chapter illustrates the importance of biofluid mechanics in understanding

hemodynamics in general and the flow across heart valves in particular. Also the
impact of biomechanical loading, both normal and shear stress, on heart valve flow is
discussed.

References

[1] O. Bech-Hanssen, K. Caidahl, I. Wallentin, P. Ask and B. Wranne, Assessment of effective orifice area
of prosthetic aortic valves with doppler echocardiography: An in vivo and in vitro study, The Journal of
Thoracic and Cardiovascular Surgery 122(2), (2001) 287–295.
[2] P. Blackshear, F. Dorman and E. Steinbach, Shear, wall interaction and hemolysis, Transactions of the
American Society of Artificial Internal Organs 12, (1966) 113–120.
[3] D. Bluestein, Y. Li and I. Krukenkamp, Free emboli formation in the wake of bi-leaflet mechanical
heart valves and the effects of implantation techniques, Journal of Biomechanics 35(12), (2002), 1533–
1540.
[4] J.T. Ellis, T.M. Wick and A.P. Yoganathan, Prosthesis-induced hemolysis: mechanisms and
quantification of shear stress, Journal of Heart Valve Disease 7(4), (1998) 376–386.
 P. Verdonck and K. Dumont / Biofluid Mechanics & the Circulatory System 57

[5] Z. Feng, T. Nakamura, T. Fujimoto and M. Umezu, In vitro investigation of opening behavior and
hydrodynamics of bileaflet valves in the mitral position, Artificial Organs 26 (2002), 32–39.
[6] D. Garcia, J. Dumesnil, L.G. Durand, L. Kadem and P. Pibarot, Discrepancies between catheter and
doppler estimates of valve effective orifice area can be predicted from the pressure recovery
phenomenon: Practical implications with regard to quantification of aortic stenosis severity, Journal of
the American College of Cardiology (JACC) 41 (2003), 435–442.
[7] L. Goubergrits and K. Affeld, Numerical estimation of blood damage in artificial organs, Artificial
Organs 28(5), (2004) 449–507.
[8] J. Hellums and R. Hardwick, The Rheology of Blood, Blood Vessels and Associated Tissues, chapter
Response of Platelets to Shear Stress - a Review, 160–183, Sijthoff & Noordhoff, Alphen aan den Rijn,
1981.
[9] S. Kelly, P. Verdonck, J. Vierendeels, K. Riemslagh, E. Dick and G.V. Nooten, A threedimensional
analysis of flow in the pivot regions of an ATS bileaflet valve, International Journal of Artificial
Organs 22(11), (1999) 754–763.
[10] M. Kroll, J. Hellums, L. McIntire, A. Schafer and J. Moake, (1996) Platelets and shear stress, Blood
88(5), (1996) 1525–1541.
[11] A.M. Malek, S.L. Alper and S. Izumo, Hemodynamic shear stress and its role in atherosclerosis,
Journal of the American Medical Association (JAMA) 282(21), (1999) 2035–2042.
[12] A. Malek, and S. Izumo, Mechanism of endothelial cell shape change and cytoskeletal remodeling in
response to fluid shear stress, Journal of Cell Science 109, (1996) 713–726.
[13] K. Matthys, Assessment of Vascular Hemodynamics: Investigation of non-invasive and minimally
invasive methods for assessment of vascular function at rest and during cardiovascular challenge, Ph.D.
Dissertation, Ghent University, 2004.
[14] A. Nevaril, E. Lynch, C. Alfrey and J. Hellums, Erythrocyte damage and destruction induced by
shearing stress, Journal of Laboratory and Clinical Medicine 71(5), (1968) 781–790.
[15] Z. Ruggeri, Mechanisms of shear-induced platelet adhesion and aggregation., Thrombosis and
Haemostasis 70 (1), (1993) 119-123.
[16] K. Schmailz and O. Ormerod, ed., Ultrasound in Cardiology, Blackwell Science, 1998.
[17] W. Tillmann, H. Reul, M. Herold, K. Bruss and J. van Gilse, In vitro wall shear measurements in aortic
valve prostheses, Journal of Biomechanics 17(4), (1984) 263–279.
[18] P.M. Vandervoort, N.L. Greenberg, M. Pu, K.A. Powell, D.M. Cosgrove and J.D. Thomas, Pressure
recovery in bileaflet heart valve prostheses: localized high velocities and gradients in central and side
orifices with implications for doppler-catheter gradient relation in aortic and mitral position,
Circulation 92 (1995), 3464–3472.
[19] J. Welty, C. Wicks, R. Wilson and G. Rorrer, Fundamentals of momentum, heat, and mass transfer,
Wiley Text Books; 4 edition, ISBN 0471381497, 2000.
[20] Z. Wu, B. Gao, J. Slonin. and N. Huang, N., Bileaflet mechanical heart valves at low cardiac output,
ASAIO Journal 42(5), (1996) 747–749.
[21] A.P. Yoganathan, The Biomedical Engineering Handbook, Volume I: Second Edition., chapter Cardiac
Valve Prostheses, CRC Press LLC, ISBN 0-849-38594-6, 2000.
[22] A.P. Yoganathan and G. Chatzimavroudis, PanVascular Medicine: Integrated Clinical Management,
Chapter 7: Hemodynamics, Springer, 2002.
[23] A.P. Yoganathan, T. Wick and H. Reul, H., Current issues in heart valve disease, chapter Thrombosis,
Embolism and Bleeding, London: ICR, 1992.
[24] A.P. Yoganathan, Y.R. Woo and H.W. Sung, Turbulent shear stress measurements in aortic valve
prostheses, Journal of Biomechanics 19(6), (1986) 433–442.
58 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-58

I.5. Biomechanics of Implants

Jan G. HAZENBERG a, Johannes SCHMID b, T. Clive LEE c and Gijsbertus J.
VERKERKE d,e
a
IVAX Pharmaceuticals, Waterford Industrial Estate, Waterford, Ireland
b
University of Applied Sciences, dept of Mechanical Engineering, Regensburg,
Germany
c
Royal College of Surgeons in Ireland, Department of Anatomy, Dublin, Ireland
d
University Medical Center Groningen, University of Groningen, Dept of Biomedical
Engineering, Groningen, the Netherlands
e
University of Twente, Dept of Biomechanical Engineering, Enschede, the Netherlands

Abstract. For simple constructions a mechanical analysis to determine internal

stresses and deformation is possible using theoretical formulas. However, for
complex constructions, like joint prostheses, this is not possible. Numerical
simulation of internal stresses and deformations offers a solution for these
constructions. The so-called Finite Element Analysis divides the complex structure
in simple ones (elements), applies the mechanical formulas and adds the effect on
each element to predict the behaviour of the complex contruction.

Keywords. Finite Element Analysis, mechanics, differential equation, node,

stiffness matrix

Introduction

Human life expectancy is increasing and so too, therefore, is the need for joint
replacements to maintain mobility. However, the functional life expectancy of the
implants is limited to 10-20 years, thus in many patients the original implant must be
removed and replaced. One of the reasons for the limited lifetime of these devices is
the fact that bone adapts its shape to a change in load distribution as the geometry of an
artificial joint differs from the original, particularly in its stem.. As the bone adapst its
shape, the fixation of implants is affected resulting in movement of the artificial joint,
malfunction and pain, making revision surgery necessary. Such revision surgery is
difficult and dangerous, so an increase in the lifetime of joint prostheses is preferable.
To improve those implants, an analysis of the loading situation and response of the
implant is required However, the geometry of such an implant and the surrounding
bone is complicated, and bone is a complex material, with non-homogenous properties.
The Finite Element Analysis (FEA), which originates from the aerospace and nuclear
industries, offers a solution, Finite element packages avail of advanced imaging
 J.G. Hazenberg et al. / Biomechanics of Implants 59

techniques and are capable of solving the most sophisticated problems, not just in
structural analysis, but for a wide range of phenomena such as static and dynamic
loading of orthopaedic implants and fluid flow problems in cardiovascular applications.
Its popularity is due to a number of factors, which include:
• A variety of loading conditions and environmental parameters can be analysed
• A variety of materials can be analysed and tested under in vivo circumstances
• Geometrical modifications are easily made
• It reduces the number of animal and pre-clinical tests required
• It is cost and time efficient

1. How does FEA work in practice?

1.1. Geometry

It typically starts with the generation of a geometrical model. Currently a variety of

techniques is available which include μCT and MRI scanning. These machines are able
to provide data files, which contain three-dimensional data points of the geometry,
which can be a femur, mandible or artery. These data point are than linked together,
resulting in a wire frame which represents the geometry. Using these lines, areas can be
created from which a solid 3D model is generated. In Figure 1A, a wire frame model is
shown of half a mandible, in which a plane fracture (blue arrow) and a mini plate for
reconstruction (red arrow) are modelled. The angle of the fracture can be modified in
the solid model (Figure 1B).

-20° 0° 20°

A B C

Figure 1. From the MRI or μCT scan a model is created of a mandible (A). The data points are
connected through lines and a solid model is created (B). Finally the solid model is meshed (C).

To carry out a finite element analysis, the model must be divided into a number of
small pieces known as finite elements. In simple terms, a mathematical net or "mesh" is
required to carry out a finite element analysis. If the system is one dimensional, it can
be represented by line elements, if it is two dimensional, then a 2D mesh is required.
60 J.G. Hazenberg et al. / Biomechanics of Implants

Correspondingly, if the problem is complex and a 3D representation of the continuum

is required, then we use a 3D mesh. Figure 1C shows the 3D-meshed model of a
mandible.

1.2. Material Properties

The next step is to give the finite elements mechanical properties such as Young’s
modulus, which determines its elasticity. The final step is to place a load (or different
loads) at one or more nodes, or intersection points of the mesh, and to fix other nodes
to the environment.
Now the load on all elements will be computed and can be compared to the
maximal allowable load before failure. If this load is exceeded in some element, we
know that the structure will collapse.
If the maximal load is not exceeded, the resulting deformation on every element
can be calculated and, by adding all deformations, the total deformation of the
complete structure is found..
In more detail, the finite element is a mathematical method for solving ordinary
and partial differential equations. As these types of equations occur naturally in
virtually all fields of the physical sciences, the applications of the finite element
method are limitless as regards the solution of practical design problems. Within each
element, the variation of displacement is assumed to be determined by simple
olynomial shape functions and nodal displacements. Equations for the strains and
stresses are developed in terms of the unknown nodal displacements. From this, the
equations of equilibrium are assembled in a matrix form, which can easily be
programmed and solved on a computer. After applying the appropriate boundary
conditions, the nodal displacements are found by solving the matrix stiffness equations.
Once the nodal displacements are known, element stresses and strains can be calculated.
From the elements and their nodes the 'stiffness' matrix is formulated. This square
matrix contains details of the material properties, the model geometry and any
assumptions of the stress-strain field.

Figure 2. Single line element.

A mesh is essentially an assembly of springs. For a spring in general we can say

that the force F is proportional to the spring constant k times the displacement u,
or F = k ⋅ u . Let us consider a single line element, consisting of two nodes, ‘i’ and ‘j’,
 J.G. Hazenberg et al. / Biomechanics of Implants 61

Figure 2. Two these nodes we can apply forces, fi and fj respectively, resulting in a
displacement ui and uj. The displacement of these two nodes will be dependant on the
stiffness of the spring ‘k’. In any case where linear elastic material properties can be
assumed, the displacement u increases proportionally with the force f (if k is constant).
In order for this spring to be in a state of equilibrium, the forces fi and fj have to be
of equal magnitude in opposite directions. Mathematically we can say that:

f i = − F = −k (u j − u i ) = ku i − ku j , for node i, (1)

f j = − F = −k (u j − u i ) = −ku i + ku j at node j. (2)

This can also be expressed in matrix form, as is usual in FEA, as follows:

⎧ fi ⎫ ⎡ k − k ⎤ ⎧ ui ⎫
⎨ ⎬=⎢ ⎥•⎨ ⎬
⎩ f j ⎭ ⎣− k k ⎦ ⎩u j ⎭ (3)

This is similar to F = K ⋅ u in this case. Here F is the matrix of the load vectors,
derived from the loads of every cases, u is the matrix of the displacement of nodes and
K is the coefficient matrix of the system, the so called stiffness matrix. Since FEA does
not use one single element but several, we can do this in the same way for a spring
assembly.

1 k1 2 k2 3

F1,u1 F2,u2 F3,u3

Figure 3. Two line elements.

In Figure 3 we have two elements; one is connected by nodes 1 and 2, while the
second element is connected through the elements 2 and 3. This can be expressed in
matrix form as follows:

⎧⎪ f 11 ⎫⎪ ⎡ k 1 − k 1 ⎤ ⎧ u1 ⎫
⎨ 1⎬ = ⎢ ⎥ • ⎨ ⎬ for element 1 and (4)
⎪⎩ f 2 ⎪⎭ ⎣− k1 k1 ⎦ ⎩u 2 ⎭

⎧ f 12 ⎫ ⎡ k 2 − k 2 ⎤ ⎧u 2 ⎫
⎨ 2⎬ = ⎢ • ⎨ ⎬ for element 2. (5)
⎩ f 2 ⎭ ⎣− k 2 k 2 ⎥⎦ ⎩u 3 ⎭
62 J.G. Hazenberg et al. / Biomechanics of Implants

The equilibrium forces at node 1 are therefore F1 = f 11 , at node 2, F2 = f 21 + f 12 and at

node 3, F3 = f 22 . These matrices can now be combined to form the stiffness matrix
for these to elements, which look like:

⎧ F1 ⎫ ⎡ k1 − k1 0 ⎤ ⎧ u1 ⎫
⎪ ⎪ ⎢ ⎥ ⎪ ⎪
⎨F2 ⎬ = ⎢− k1 k1 + k 2 − k 2 ⎥ • ⎨u 2 ⎬ (6)
⎪F ⎪ ⎢ 0 − k2 k 2 ⎥⎦ ⎪⎩u 3 ⎪⎭
⎩ 3⎭ ⎣

In this two-dimensional example we have relatively little to calculate. Due to

increased computational power, more sophisticated three-dimensional models can be
produced, which can consist up to 100,000 elements and require several hours to solve
a particular problem.
Once the stiffness matrix is created, external loads are applied to evaluate the
displacements of the structure (hence the term displacement analysis). These loads can
be static, dynamic or represent impact loading, depending on the location and function.
Initially, the construction of a ‘free-body-diagram’ in which all muscle forces, joint
load and reaction forces are drawn is useful. In cases where dynamic loads are to be
investigated a ‘kinetic diagram’, which is similar to the ‘free-body-diagram’, can be
made in which the expected accelerations can be mapped. This information can than be
inserted into our meshed model (see Figure 4).

Figure 4. On the left-hand side (A) an image of a free-body-diagram in which P is the posterior part of
temporalis, M1 is the superficial part of masseter, M2 is the deep part of masseter, Pm is the medial
pterygoid and J is the lateral pterygoid muscle. On the right-hand (B) side is an image of the forces applied to
the meshed model of the mandible.
 J.G. Hazenberg et al. / Biomechanics of Implants 63

On evaluation of the displacements, they are differentiated to give six strain

distributions, 3 mutually perpendicular direct strains and 3 corresponding shear strains.
Finally, six stress distributions are determined via the stress/strain relationships of the
material. Figure 5 shows the stress and shear distributions in a 2D case. A point to note
is that at least one of the displacements must be known before the rest can be
determined (before the system of equations can be solved). These known displacements
are referred to as boundary conditions and are oftentimes a zero value. An alternative
solution may be obtained via the force matrix method. In the previous description, the
displacements were the unknown, and solution is obtained via the stiffness method. In
the force method, the forces are the nodal unknowns, while the displacements are
known. The solution is obtained for the unknown forces via the flexibility matrix and
the known displacements. The stiffness method is more powerful and applicable than
the flexibility approach.
Typically in implant design, account has to be taken of the non-linear behaviour of
biological tissues. In order to explain non-linearity in stress analyses, let us focus on
the nature of linear solutions. The primary assumption in linear stress analysis is that
the stress/strain relationship is directly related to the deformation. In general, there are
four causes for nonlinear behaviour. It could be that the material deforms in a non-
linear way. This is where the material stress-strain relationship is actively nonlinear. In
this case, material behaviour depends on the current deformation state and possibly on
the past history of the deformation.

Figure 5. Stress and shear distributions in 2D

Material nonlinearity can be observed in structures undergoing nonlinear elasticity,

plasticity, viscoelasticity, creep or other inelastic effects. There could be geometrical
causes, which may be due to large strains (in impact trauma analysis) or small strains
64 J.G. Hazenberg et al. / Biomechanics of Implants

but with large displacements (in the analysis of muscles). Muscles contractions are also
a good example where nonlinear forces are applicable. This is where the magnitude or
direction of the applied forces change with application to the structure. Displacement
boundary condition nonlinearities, such as occurs in modelling of friction in knee joints.
This is where the displacement boundary conditions depend on the deformation of the
structure. The most important and obvious application is in contact problems, the
displacement is highly dependent on the relationship between two contact surfaces (e.g.
a fracture site in bones). It is important to note that the bodies in contact could be in a
state of linear elastic stress; the nonlinearities all come from the contact definition in
this case.
All non-linearities are solved by applying the load slowly (dividing it into a
number of small load increments). The model is assumed to behave linearly for each
load increment, and the change in model shape is calculated at each increment. Stresses
are updated from increment to increment, until the full applied load is reached. In a
nonlinear analysis, an initial condition at the start of each increment is the state of the
model at the end of the previous one. This dependency provides a convenient method
for following complex loading histories. At each increment, the solver iterates for
equilibrium. Due to the iterative nature of the calculations, non-linear FEA is
computationally expensive, but reflects the real life conditions more accurately than
linear analyses.

2. The Power of FEA

The previous section has provided some of the background regarding FEA. But why is
it commonly used in the analysis of implant biomechanics and implant design? These
days it is very easy to obtain accurate models of various body parts of interest using
MRI and CT-scanning. This means that irregular shaped objects are easy to analyse.
Anatomical studies can be used as a great source of information regarding muscle
attachments, insertions and directions. Analysis can be done for static as well as for
dynamic loading conditions. All these features contribute to more sophisticated models
that mimic in vivo situations in a more realistic way. But the true power of FEA lies in
the ability to change parameters relatively easily once the model has been created.
In the case of the analysis of a fractured mandible, we can choose a specific
material for the plate that stabilises the fracture site. The analysis will indicate what the
local stress or strains are in the tissue. Perhaps one wants to investigate what happens if
a softer and more flexible material is used. This would only require altering the
material properties and rerunning the analysis. The same applies if the location and size
of the plate or the fracture angle is to be investigated. This can provide valuable
information, even at tissue level. The phenomenon of stress shielding is a common
problem in implant design, which can result in tissue loss or altered structure. Typically
a tissue is stimulated by the daily loads to which it is subjected. However, following the
insertion of an implant, this load is partly reduced since it is shared by the implant.
Factors that are of importance here are the geometry and the fit of the implant with the
natural structure. Interface conditions between the implant and its surrounding tissue
might be designed to promote bounding between them and therefore realise a better
stress distribution from the implant to the surrounding tissue and so increase the
lifetime of the implant.
 J.G. Hazenberg et al. / Biomechanics of Implants 65

References

[1] T.M. Wright and S.B. Goodman, Implant Wear in Total Joint Replacement: Clinical and Biological
Issues, Material and Design Considerations, American Academy of Orthopaedic Surgeons, 2002
[2] P.J. Prendergast, Finite element models in tissue mechanics and orthopaedic implant design,. Clin
Biomech (Bristol, Avon) 12 (1997),.343-366.
This page intentionally left blank
Basic Engineering for Medics and Biologists 67
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-67

Introduction to
Chapter II. Bioelectronics
Richard B. REILLY and T. Clive LEE (eds.)

Bioelectronics can be defined as the interface area between electronics, photonics,

computer science, and mathematics, on the one hand, and the nervous system at the
molecular, cellular, organ, and systemic levels, on the other. Therefore it is the aim of
this chapter to explore this definition and introduce bioelectronics from the
fundamental properties of electricity, sensors for measuring biological activity and the
analysis of electrical signals acquired from the body.
There have been numerous successes in bioengineering due to bioelectronics
including the ability to measure brain activity using the electroencephalogram to
provide information on the functioning of the brain, implantable bladder stimulators
which can dramatically improve quality of life for paraplegics, cochlear implants which
can restore partial hearing to the profoundly deaf, cardiac pacemakers, cardiac
defibrillators and deep brain stimulation which improves the quality of life for those
suffering from Parkinson’s Disease.
Despite these successes which have had a profound effect on quality of care and
quality of life for patients, bioelectronics is at a relatively primitive stage of
development, but is recognised as being a strategically important area at the beginning
of the 21st century.
As our understanding of human physiology advances, along with new
developments in biomaterials and mathematical analysis, the impact of bioelectronics
will increase. Therefore, an appreciation of the fundamental properties of bioelectricity
is essential to apply bioelectronic approaches and devices in the clinic, while
appreciating the trends in this area into the future.
This page intentionally left blank
Basic Engineering for Medics and Biologists 69
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-69

II.1. Elementary Electrodynamics

Jacques JOSSINET
Inserm, U556, Lyon, F-69003
University of Lyon, Lyon, F-69003, France

Abstract. The paper describes the physical phenomena involved in the conduction
of electricity, with particular reference to living tissue. The conduction of
electricity depends on the flow of charge carriers in the material, while the
dielectric properties are due to the rotation of dipoles that can align along an
applied electric field. The relation between the electric variables in a conducting
medium and their physical meanings are explained. The phenomena responsible
for the electric and dielectric properties of living tissues are described. The
presence of cells limits the flow of charge carriers, in particular at low frequency,
and the membranes are responsible for dielectric relaxation. The passive response
of cell membranes to weak applied signals enables bioelectrical tissue
characterisation. Practical tools for electrical impedance spectroscopy are given
with an overview of the most recent applications.

Keywords. Electric conduction, dielectric permittivity, bioimpedance, tissue

characterisation

1. Definitions

1.1. Electric Charge (q)

Electric charge (coulomb, C) is a property of matter and is transported at the atomic

level by ions: elementary particles and charged molecules. There are two kinds of
electric charge, positive and negative. The term "charge" may also represent a
dimensionless point bearing a given electric charge. Electricity is quantized, meaning
that charge always comes in integer multiples of the elementary charge (the electron
charge, e = 1.6×10-19 C). Due to the great number of atoms in any piece of matter,
electricity is handled at the macroscopic level as a continuous variable. The net charge
of an object is the sum of the positive and negative charges its carries. The normal state
of matter is electric equilibrium, with equal number of positive and negative charges.
The conservation of electricity is verified in any process or reaction.

1.2. Electric Field E

The electric field is a property of space in the presence of an electric charge. A point
charge q0 produces an electric field which is a vector, denoted E , passing by the point
charge and verifying the following equation
70 J. Jossinet / Elementary Electrodynamics

(
E = q 0 / 4π ε 0 d 2 ) (1)

The quantity ε 0 is the dielectric permittivity of free space and d is the distance
to the source. The major property of the electric field is the force, f = q E , it exerts on
a charge. Two charges exert opposite forces on each other. The electric force is
attractive when the charges have opposite signs and repulsive when the charges have
identical signs.

1.3. Electric Potential or Voltage (v)

Voltage or electric potential, u, at a given point is the electric potential energy per unit
charge (in joules per coulomb = volts). The definition of potential is generally
presented under the form: E = − grad ( v ) , where grad is the vector operator gradient.
This vector represents the magnitude and direction of the greatest rate of change of the
potential. This definition indicates that the existence of electric field in a medium
implies a non uniform potential distribution. Conversely, if the potential is uniform, the
field is equal to zero at any point (e.g. in a conductor at electric equilibrium).

1.4. Electric Current (i)

Electric current is the ensemble movement of charged particles, termed charge

carriers. The electrons in a metal or the ions in an electrolyte are examples of charge
carriers. The intensity of the current (ampere, A) is the quantity of electricity that is
transported per unit time: current is the time derivative of electric charge:

i = ∂q / ∂t (2)

1.5. Current density j

It is often useful to consider the flow of electric charge per unit surface. This quantity is
a vector termed current density, j (A/m2). The current density through a surface
element of area dS, is given by:

j = n.di / dS (3)

where n is the unity vector perpendicular to the surface. The dot product shows that
only the orthogonal component of the flow gives rise to a current density through the
surface.
The current density determines the effects of electricity in a medium. A practical
example is the heating of a wire by the Joule effect. For a given current, the heating is
greater in a conductor of small cross section than in a conductor of larger cross section.
In body tissues, the desired or adverse effects of electrical current depend on how the
current concentrates in certain pathways or spreads in a larger section, resulting in
different current densities.
 J. Jossinet / Elementary Electrodynamics 71

1.6. Electric Conductivity

Electric conductivity (siemens, S) characterises the ability of a material to conduct

electricity. The electric conductivity of an insulating material is equal to zero. The
conduction of electricity through a medium implies the presence of charge carriers able
to move within the medium. Charges than cannot move (bound charges) do not
contribute to the conduction of electricity. If the charge carriers can move freely (e.g.
electrons in metals) the medium is highly conductive. The ability of a charged particle
to move within a given medium is characterised by its "mobility". Mobility is the limit
speed reached by this particle in a given medium in the presence of a unit uniform and
constant electric field. This speed is reached when the viscous force (increasing with
the speed of the particle) becomes equal to the electric force q E .

1.7. Static Conductivity of Solutions

In solution of dissolved electrolytes, the quantity of electricity that can be carried by an

ionic species is proportional to its number of charge per ion, the concentration of
dissolved ions and their mobility. The overall conduction σ 0 , of a solution containing
several species is the sum of the conduction of each species α i. This is written as

σ 0 = F ∑ γ i α i C i ni μ i (4)
i

The activity coefficient, γ i , ( γ ≤ 1) accounts for the deviation from linear behaviour
for increasing ionic concentrations. For a given species, Ci is the concentration of this
species in the solution, ni the number of charges of the ion, and μ i is the mobility and
Σ denotes summation. The above equation deals with the static conductivity for direct
current. However, the viscous forces delay the movement of charges carriers and the
establishment of the steady state is not instantaneous. This effect however becomes
appreciable for rapidly varying fields. Hence, in biomedical applications the
conductivity of ionic solutions can be considered independent of the applied field
frequency.

1.8. Ohm's Law

Ohm's law relates to the proportionality between the DC current, denoted I, and the DC
voltage difference, denoted V, across an ideal conductor, circuit or current pathway.
The general expression of Ohm's law is

V = R× I (5)

The constant of proportionality, R, is termed the "resistance", (ohm, Ω ) of the circuit.

If the resistance of a conductor is constant over a wide (ideally infinite) range of
voltages, the conductor is said to be "ohmic".
The reciprocal of the resistance is termed the "conductance", G (siemens, S).
Ohm's law also applies at the macroscopic scale and then is written differently.
72 J. Jossinet / Elementary Electrodynamics

Considering the flow of charge in an elementary volume element, it can be

demonstrated that the "local" expression of Ohm's law is:

E = J / σ = ρJ (6)

The quantity ρ , reciprocal of the conductivity σ , is the resistivity of the medium

(ohm.metre). This equation shows that for the current density, J (A/m2), i.e. the
quantity of electricity passing though a unit surface per unit time is proportional to the
electric field and the electric conductivity of the medium.
The heat production per unit time (energy) due to the current in a circuit is the
electrical "power" W, measured in watts. In a circuit of resistance R passed by a current
I the dissipated energy is RI2. Using Ohm's law this also writes V2/R, with V and R
defined above. Instead of a circuit's resistance, one may consider the energy dissipated
within a medium. The energy dissipated in a volume element of volume dt, within a
medium. In such an element, the energy dissipated per unit volume is given by

dW = J2 / σ  (7) 

(where σ is the electric conductivity and J the current density). This equation shows
that in a heterogeneous medium, the local Joule effect depends on the conductivity and
current density at each point.

1.9. Impedance, Admittance,

In a circuit passed by a sinusoidal signal, the voltage and the current vary in synchrony,
but are shifted in time, according to the properties of the circuit. The voltage-to-current
ratio is termed impedance.
A sinusoidal signal (voltage or current) is characterised by its magnitude and phase
angle and is represented by a complex number formed by a real part and an imaginary
part.
It is expressed as a complex number denoted Z. The real part and the imaginary
part of Z are denoted R and X respectively.
The magnitude |Z| of this complex number is the ratio of the magnitudes (V/I, V in
volt and I in ampere).The argument, θ , is the phase difference (in radians) between
voltage and current. Simple mathematical considerations lead to the following classic
equations.

( )
Z = R 2 + X 2 , R = R Z cos θ and X = Z sin θ (8a)

The real part, R, of the impedance is termed resistance and its imaginary part, X, is
termed its reactance. Impedance is the generalisation to sinusoidal signals for the
notion of resistance for DC current. The reciprocal of impedance is also a complex
number denoted Y ( Y , θ' ), termed admittance, and verifying:

Y = 1/Z, Y = 1 / Z and θ' = −θ (8b)

 J. Jossinet / Elementary Electrodynamics 73

The impedance of a capacitor is given by X = 1/(j2 π fC), where C is the capacity

(farad, F) and f the applied signal frequency (hertz, Hz).

1.10. Dielectric Permittivity

Dielectric properties are observed in a medium containing bounded electric dipoles. An

electric dipole is the rigid system of two equal and opposite point charges, located at
points Q-, Q+, and bearing opposite charges of equal magnitude, -q and +q,
respectively. The dipole is characterised by its moment which is the vector defined by

M = q × Q − Q+ (9)

The dipoles in a dielectric medium cannot move but can rotate in the presence of an
electric field. The two charges experience opposite forces. It can be shown that these
forces create a torque tending to align the dipole along the direction of the electric
field.
The rotation of dipoles creates the dielectric polarisation of the material. Near the
boundary surface, the orientation of dipoles makes charges of a given polarity come
closer to the surface and the charges of the opposite polarity move away from it. This
results in a distribution of charges at the outer surface of the material. This distribution
depends on the concentration of dipoles and their ability to rotate. A dielectric material
is characterised either by its dielectric permittivity ε (farad/metre) or by its "dielectric
constant" ε r = ε ε 0 (dimensionless).

1.11. Capacitive Coupling C

The polarisation of a dielectric slab in the presence of an orthogonal field results in

surface charges of opposite signs on each side of the slab. These surface charges can
attract or repel charges from the outer circuit connected to the slab. Similarly, the
external circuit can bring or remove charges at the surface and produce polarisation of
the dielectric slab. In both cases, the corresponding flows of charges is termed
"displacement current". It can be shown that the charges on both sides are of opposite
signs and equal magnitudes and vary in synchrony. This effect is termed the "capacitive
coupling", as the charges seem to pass from one side to the other one. In practice,
capacitive coupling appears when a membrane separates two conductors, forming a
capacitor. The "capacity" (farad, F) is the ratio of the magnitude of the charge
accumulated on either side to the voltage across the dielectric. The capacity of a
capacitor increases with the area of the plates and decreases with the separation
between the plates according to the relation

S
C = ε0ε (10)
d

S is the area of charged surfaces, d is the distance between them and ε is the dielectric
constant of the material. The main feature of a capacitor is to block DC signals and be
permeable to varying signals.
74 J. Jossinet / Elementary Electrodynamics

1.12 Dielectric Polarisation

The rotation of dipoles in a dielectric is hindered by viscous forces exerted by the

medium. Dielectric permittivity has therefore a dependence on frequency. This
dependence is called the dielectric dispersion. The associated time constant is called the
"relaxation time" of the medium. The term dielectric relaxation denotes the delayed
polarisation in response to the application of an electric field.
The time lag between electrical field and dipole rotation implies irreversible
energy losses (dielectric losses). For ideal dielectrics, containing dipoles with a unique
time constant, τ , the relaxation is described in terms of permittivity as a function of
frequency, which can be described by Debye's equation:

Δε
ε(ω) = ε ∞ + (11)
1 + jωτ

Δε is the dielectric increment, ε ∞ the high frequency limit permittivity (both

characterising the material), ω is the angular frequency ( ω = 2π f ) and j the base of
complex numbers (j2 = -1).

2. Electric Properties of Cells and Tissues

The conduction of electricity in the human body presents a particular set of features
that is not met is other materials, except in animal tissues and, to some extent in plants.
The main feature is the simultaneous existence of electric conduction and dielectric
behaviour. The presence of cells is responsible for the remarkable electric and
dielectric properties of tissue.

2.1. The Cell

The cell is the basic building block of the human body. The living cell is delimited by
the plasma membrane providing rigidity and protection. The cell membrane also binds
different cells together. The membrane consists of an asymmetric molecular lipid bi-
layer which is basically impermeable to ions. However, protein molecules inserted in
the bi-layer ensure the transport of materials, including ions, across the membrane. The
interior of the cell contains the nucleus and the cytoplasm which is a jelly-like material
consisting of the cytosol and the organelles. The cytosol is made up of water, salts, and
organic molecules and is therefore conductive. The content and structure of the interior
of a cell depend of each cell type.

2.2. Membrane potential

Under normal conditions of activity, there is a difference in ion concentrations on both

sides of the cell membrane. This charge separation results in an electrostatic field and a
voltage difference across the membrane. The equilibrium voltage (or resting potential)
corresponds to the balance between the flow of ions due to the electrostatic field and
 J. Jossinet / Elementary Electrodynamics 75

the flow produced by the ionic concentration gradient. This potential difference is
called the Nernst potential. The resting value for nerve cells is about 70 mV.

2.3. Biological Tissue

A tissue is formed by cells performing a similar function. An organ is a structure

containing several types tissues working together to carry out specific functions. The
archetypal representation of a tissue is a population of cells separated by the interstitial
space. However, its structure may be complex, particularly in mammalian tissues. The
extracellular matrix is a structural entity supporting cells; it is composed of 3 major
classes of biomolecules including structural proteins (collagen and elastin), specialized
proteins and proteoglycans (composed of a protein core attached to
glycosaminoglycans, long chains of repeating disaccharide units). There is a great
variability in cell shape, concentration and arrangement, depending on each tissue's
structure, function and physiological state. The interstitial space between cells contains
a conductive medium the conductivity of which depends on the particular tissue. For
these reasons, the dielectric properties of body tissue are highly variable from one
tissue to another.

2.4. Cell Membrane Passive Properties

The overall conduction of electric signals through a tissue depends on the conductivity
of the interstitial space, that of the intracellular medium and the capacitive behaviour of
the cell membranes. The membrane is practically impermeable to DC current, except
for the active transport of materials through it by the trans-membrane proteins. For
applied signals, the membrane acts like a passive insulator separating two conducting
media and causes capacitive coupling between the interstitial space and the interior of
the cell. The capacitance of the cell membrane is about 1 µF/cm2 [7, 12]. Finally, DC
and low frequency signals do not penetrate into cells. For increasing applied signal
frequency, the interior of cells is progressively involved in the overall conduction due
to capacitive coupling through the membrane.

2.5. Dielectric Relaxation in Tissue

Observation shows that body tissues are dispersive media. In a dispersive dielectric
medium, the permittivity and dielectric constant vary with the applied field frequency.
In general, the dielectric relaxation in body tissue does not follow in the Debye
equation. Experiments show that, in the frequency range from a few hertz to many
gigahertz (1 GHz = 109 Hz), the dielectric permittivity of tissues is a decreasing
function of frequency and shows several dispersion domains. Several types of
phenomena, preponderant in different frequency domains, are responsible for the
observed dispersion [6, 14, 15].
2.5.1. α -Relaxation
This relaxation occurs at low frequency roughly below 1 Hz. This effect is due to the
counter-ion atmosphere surrounding the membrane surface. In the presence of an
electric field, the ions in the cell vicinity accumulate on diametrically opposed
locations and then form a cell-size dipole. Such a dipole cannot follow rapidly varying
76 J. Jossinet / Elementary Electrodynamics

signals. This low frequency relaxation does not give information on either cell
membrane or cellular content.
2.5.2. β -Relaxation

This structural relaxation is also termed Maxwell-Wagner relaxation. It may occur up

to the radio-frequency range, depending on the properties of each particular tissue. The
basic mechanism is the accumulation of charge at interfaces such as cell membranes. In
body tissue, the time constant of this relaxation depends on the conduction in the media
and the capacitive coupling through the interfaces.
2.5.3. γ -Relaxation

This relaxation is due to the orientation of dipoles. Several types of dipoles contribute
to the polarisation of the medium. It takes place typically in the microwave frequency
domain (f < 1 GHz). The permanent dipoles include dipolar molecules (mainly H2O),
hydrogen ionic or covalent bonding. Induced dipoles result from the temporary
separation from the barycentre (the point between two objects where they balance each
other) of positive and negative charges in non-polar molecules by either neighbouring
molecules or the applied field. Transient dipoles (also called London forces or van der
Waals forces) result from the random mutual perturbation of electron clouds of two
atoms or molecules. When an uneven distribution occurs, a temporary dipole is created.
In polar molecules, transient dipoles are usually smaller than the permanent dipolar
moment.
The above relaxation phenomena give tissues high dielectric constants compared
to usual materials (Table 1).

2.6. Equivalent circuit model

An equivalent circuit model (ECM) is a circuit yielding the same impedance as a tissue
in a given frequency domain. There are lumped circuits and discrete element circuit
models. The most commonly used discrete element ECM are composed of two resistors
and one capacitive element. The response of such circuits is characterised by a single
central frequency; denoted f0 according to the general equation [10, 11]:

R0 − R∞
Z = R∞ + (12)
1 + j ( f / f 0 )β

Z denotes the overall impedance of the circuit, R0 the low frequency limit resistance, R∞
the high frequency limit resistance and j the base of the complex numbers (j2 = -1).
Exponent β is termed the "fractional power" ( β ≤ 1). At frequency f0 the magnitude
of the imaginary part passes by a maximum. The plot in the complex plane of the
imaginary part against the real part (termed an Argand diagram) is an arc of circle the
centre of which is shifted from the horizontal axis (Fig. 1).
 J. Jossinet / Elementary Electrodynamics 77

Table 1. Example values of electric conductivity (S.m-1) and dielectric constant in tissues and some usual
materials at 50 Hz or "below 100 Hz" [2]. The above figures are supplied for illustration and different values
can be found according to measurement frequencies and experimental conditions [12].

Conductivity Dielectric constant

Siemens/metre
Vacuum 0 1
-15
Dry air 3×10 1
6
Copper 56×10 1
−6
PolyVinyl Chloride 10 3.4
-6
Water at 25 °C. 5.5×10 78
NaCl saline 0.9 g/l 1.68 77
7
Kidney 0.089 1.0×10
6
Fat 0.02 1.5×10
6
Nerve 0.027 1.6×10
6
Liver 0.037 1.8×10
7
Grey matter 0.075 1.2×10
7
Muscle 0.23 1.8×10
7
Tendon 0.27 1.7×10
3
Whole blood 0.7 5.3×10
Cerebro-spinal fluid 2 109

Figure 1. Sketch of Eq. (13) in the complex plane: Horizontal axis is the resistance and vertical axis is the
reactance. The reactance of body tissues is negative. The white dots represent particular measurement
frequency points. The abscissa of the intercepts with the horizontal axis are the so-called “limit resistances”
at low frequency ( R0 ) and at high frequency ( R∞ ). The bottom of the arc corresponds to f 0 in Eq. (13).
The shift of the centre from the horizontal axis is a function of the fractional power β.
78 J. Jossinet / Elementary Electrodynamics

There are two types of 3-elements circuits giving rise to the above response in Fig.
2. In these circuits, Zcpa is a "pseudo-capacitance" (subscript "cpa" stands for "constant
phase angle"). Its frequency response is given by Zcpa = K ( j ω )−β . The exponent β is
the fractional power in (12). A constant phase element is a hypothetical element that
does not exist in practice, but is useful to describe the frequency response of body
tissues. If the particular case where β is equal to unity, Zcpa is equivalent to a capacitor
of capacity C=1/K. The relation between ECMs' elements and the variables in (12) are
given in Table 2.

Figure 2. Equivalent circuit models comprising 3-element and yielding identical frequency responses. Type
(a) circuit is generally used for the deep tissue, with Re representing the resistance of the interstitial space
and Ri the intracellular resistance. Type (b) circuit is generally used to represent the impedance of layered
media such as skin.

In practice, fitting an arc of circle to the experimental data points enables the
calculation of R0, R∞, and β (Table 2). The calculation of these parameters makes it
possible to compare data sets obtained at different measurement frequencies. Table 3
gives the equations enabling the calculation of the elements of ECM using the four
parameters resulting from the fit of a circular arc to the experimental data. The table
confirms that, with appropriate values of their elements, both circuits can model a
given response (same R0, R∞, K and β ).

Table 2: Equations for R0, R∞ in equation (12) for the equivalent circuit models of Fig. 2.

ECM R0 R∞ FT
1β
R0 = Re R∞ = Re Ri (Re+ Ri ) 1 ⎛⎜ R0 − R∞ ⎞
⎟
(a) FT= K
2π ⎜⎝ R02 ⎟
⎠
1β
R0 = Rp + Rs R∞ = Rs 1 ⎛ K ⎞
(b) FT= ⎜ ⎟⎟
2π ⎜⎝ R0 − R∞ ⎠
 J. Jossinet / Elementary Electrodynamics 79

Table 3: Equations to calculate the elements of the equivalent circuit models of Fig. 2a and Fig. 2b from the
variable in Equ. (12) Quantity Xc denotes the ordinate of the centre of the circular arc. Centre's abscissa is
equal to (R0+R∞)/2.

ECM β K

Ri = 2 ⎛ R − R∞ ⎞
(a) Re = R0 atan⎜⎜ 0 ⎟ (R0 − R∞ )(2πFT )β
R0 R∞ (R0 − R∞ ) π ⎝ 2X C
⎟
⎠

Rp = 2 ⎛ R − R∞ ⎞
(b) Rs = R∞ atan⎜⎜ 0 ⎟
⎟ (R0 − R∞ )(2πFT )β
R0 − R∞ π ⎝ 2X C ⎠

3. Applications

The measurement of the electric and dielectric properties of tissues has given rise to
range of biomedical applications [1, 8]. The most widespread are EIS, EIT and BIA.
Electrical Impedance spectroscopy (EIS) is the study of the frequency response of a
tissue or an organ to characterise its physiological and/or pathological state (e.g.
detection of oedema, ischaemia, inflammation, and tumour growth) [7]. Electrical
Impedance Tomography is an imaging method that appeared in the 1980s. The purpose
is the reconstruction of the distribution of conductivity inside the body from multiple
surface measurements. The main features of this method are the relatively small
number of measurements, limited spatial resolution, high time resolution, real-time
imaging capability and true3-D image reconstruction. This method can be associated
with EIS [3, 4, 5, 9].
Impedance measurements at the body level or across limb segments are termed
Bio-Impedance Analysis (BIA) [13]. The purpose is to monitor the changes in fat free
mass and/or total body water. This method has given rise to devices for various
applications including fitness, sports and nutrition. One popular implementation is the
incorporation of BIA system in personal weighing scales to monitor the effects of
exercise and/or nutritional regimes.
The above applications involve measurements using low intensity currents (of the
order of milliamps or less) avoiding any significant interaction with the explored tissue
at the used measurement frequencies. Therapeutic devices are based on the interaction
of relatively strong signals with the body. Typical applications include electrosurgery,
transcutaneous electric stimulation of muscles or nerve (TENS), cardiac pacing and
defibrillation.
80 J. Jossinet / Elementary Electrodynamics

References

[1] J. Ackman and M. Seitz, Methods of complex impedance measurements in biologic tissue, CRC Crit.
Rev. in Biomed. Eng. 11(1984), 281-311.
[2] D. Andreuccetti, R. Fossi and C. Petrucci, Internet resource for the calculation of Dielectric Properties
of Body Tissues, Italian National Research Council, Institute for Applied Physics "Nello Carrara",
Florence.
[3] K. Boone, D. Barber and B.H. Brown, Imaging with electricity: report of the European concerted
action on impedance tomography, J. Med .Eng. Technol. 21(1997), 201-232.
[4] S.G. Dawids, Evaluation of applied potential tomography: a clinician's view, Clin. Phys. Physiol.
Meas., 8A(1987), 175-180.
[5] A.M. Dijkstra, B.H. Brown et al., Clinical applications of electrical impedance tomography, J. Med.
Eng. Technol. 17(1993), 89-98.
[6] K.R. Foster and H.P. Schwan1, Dielectric properties of tissues and biological materials: a critical
review, CRC Critical Reviews in Biomedical Engineering, 17(1989), 25-104.
[7] J.P. Grant and N.M. Spyrou, Complex permittivity differences between normal and pathological
tissues: mechanisms and medical significance, J. Bioelectricity, 4(1985), 419-458.
[8] P. Heroux and M. Bourdages Monitoring living tissues by electrical impedance spectroscopy, Ann.
Biomed. Eng., 22(1994), 328-337.
[9] H.C.N. Jongschapp, R. Wytch, J.M.S. Hutchinson andV. Kulkarni, Electrical impedance tomography:
A review of current literature, Europ. J. of Radiology, 18(1994):165-174.
[10] J.R. MacDonald, Impedance spectroscopy, Ann. Biomed. Eng., 20(1992), 289-305.
[11] J.P Morucci, M.E. Valentinuzzi, B. Rigaud, C.J. Felice, N. Chauveau and P.M. Marsili, Bioelectrical
Impedance techniques in Medicine, Critical Reviews in Biomedical Engineering, J.R. Bourne Ed.,
24(1996), ISSN 0278-940X.
[12] R. Pethig, Dielectric properties of biological materials, Clin. Phys. Physiol. Meas, 8A(1984), 5-12.
[13] P.J. Riu, J. Rosell, R. Bragosand and O. Casas, Electrical Bioimpedance Methods, Ann. NY Acd. Sci.,
873(1999), ISBN 1-57331-191-X.
[14] H.P. Schwan, Electrical properties of tissue and cell suspensions, Advan. Biol. Med. Phys., 5(1957),
147-208.
[15] R.D. Stoy, K.R. Foster and H.P. Schwan, Dielectric properties of mammalian tissues from 0.1 to 100
MHz:A summary of recent data, Phys. Med. Biol., 27(1982), 501-513.
Basic Engineering for Medics and Biologists 81
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-81

II.2. Electrical Safety

Jacques JOSSINET
Inserm, U556, Lyon, F-69003,
University of Lyon, Lyon, F-69003, France

Abstract. Correct use of medical equipment within the clinical environment is of

prime importance. This includes awareness of the safety issues regarding
equipment, particular when it is an electrically powered device. Incidents can
occur in the clinic in which a medical device is suspected of contributing to patient
or staff injury. It is important that one can identify in advance any potential
hazards which may arise with electrical equipment due to technical or
environmental factors. This paper gives an overview of electrical safety.

Keywords. Electrical Safety, electrical injury, shock, prevention, first aid

Introduction

The contact with either a man-made or natural source of electrical energy can create
injury. Such a contact is termed "electrification". The term "electrocution" should be
reserved for cases resulting in a victim's death. The detrimental effects of electrification
depend on several parameters including the source of electrical energy and current
conduction through the body.
For electrification to occur, the victim must complete the circuit: at least two
contact points are necessary with sufficient voltage difference to drive the current
through the circuit. Electrification can have direct and indirect effects. For instance, the
explosive force produced by lightning can throw the victim a distance and the sudden
muscular spasm caused by the current can cause the person to fall from a height.
However, most lesions and trauma due to electrification are the direct effects of
current. It must be pointed out that it is the current that produces lesions. This is
confirmed, for instance, by the harmlessness of the static discharge, where the voltage
is high, but the current is limited in intensity and duration. This feature is reminded by
the familiar slogan:
"The volts jolt, the mills kill", mills signifying milliamps, the unit of current
For a given applied voltage, the current is controlled by the series combination of
the impedance of the sources and the impedance of the body through the contact points.
In lightning and electrostatic discharge (ESD), the resistance of the body is small
compared to that of the total circuit and has very little influence on current intensity. In
contrast, body impedance is the most important variable when considering
electrocution by power lines.
Note: In the literature on electrical safety, tissues are often characterised by their
resistances. This is incorrect use of the terminology since for AC current a tissue is
characterised by its impedance. However, at the frequencies used for power distribution
82 J. Jossinet / Electrical Safety

(e.g.: 50 Hz in Europe or 60 Hz in the U.S.A), the imaginary part (the reactance) of

tissue impedance is small and is generally ignored.
There is a wide variation in body resistance between people; therefore the same
voltage level may result in different effects. Skin impedance is normally high and acts
as an insulating barrier. This feature, however, disappears if skin is wet or damaged.
For given tissue resistances, the nature and importance of lesions created in the body by
electric current depends on the electrical parameters of the source of electrification
[11].

1. Sources of Electrical Hazard

1.1. Electrostatic Discharge

Electrostatic electricity is that which accumulates at the surface of persons or objects

(e.g.: carpets, pieces of furniture and clothing). It is mainly produced by friction
triboelectricity. Contact with charged objects results in electric shocks due to the
discharge current through the body. The voltage can reach several kilovolt or more, but
the duration of the current is so short that there is no dangerous effect on a person.
Electrostatic discharge, however, can damage electronic components and produce
sparks powerful enough to cause a fire or explosion in the presence of an explosive
atmosphere. Specific equipment for ESD protective workstations includes a dissipative
floor or table mat, dissipative work surface, ground connected wrist strap and a
common point ground system.

1.2. Lightning

Lightning is the natural and violent discharge between an electrically loaded cloud
(cumulonimbus) and the ground. Once the "stepper" has established the path, the
current flows along the "arc", i.e. a tube of plasma (high temperature, highly
conductive ionised matter). Lightning voltages are typically greater than 20 kV and can
reach millions of volts. The average current during the discharge is about 20-30 kA.
The lightning generally consists of several strokes and lasts from about 0.2 to 1 sec.
Thunder is due to the shock wave resulting from the rapid temperature change of the air
in the close vicinity of the arc.
Direct stroke is not the only danger of lightning. The lightning current spreads
through the ground and gives rise to gradients in ground potential. The "step voltage" is
defined as the voltage difference in electrical potential between a person’s feet. As a
body is a far better conductor of electricity than the earth, a current can flow through
the legs and cause lesions, trauma and even death, especially if excessive current passes
through the heart [8].

1.3. Power Lines

By construction, power lines (and the mains) maintain a constant voltage even if large
current is current is drawn either in the circuit connected to it or through the body of a
person. The contact with power lines can therefore result in severe electrical injury, as
a large current can continue to pass through the body as long as the contact is
 J. Jossinet / Electrical Safety 83

maintained. This is particularly critical when the victim is not able to release the
contact, above the so-called "let go threshold".
High-voltage power lines can also give rise to arc-flash between the line and
conducting objects connected to the ground. In air, this may occur for distances shorter
than 1 cm for voltage levels of approximately 4000 volts. The arc is made of plasma
and gases at high temperature produce second and third degree burns by direct contact
or by ignition clothing. Arcs can also produce pressure wave caused by the rapid
thermal expansion of gases, giving rise to the explosion of circuit components.

2. Electrical injury

The detrimental effects of electricity through the body can range from a barely
perceptible tingle to immediate cardiac arrest [2, 12]. The undesirable effects of electric
current consist of burns and electric shocks. Burns are due to the excessive electrical
energy dissipated in the tissue. An electrical current flowing through the body can be
hazardous or fatal if it causes local current densities in vital organs that are sufficient to
interfere with their functions.

2.1. Burns

Burns are produced by the heating of body tissues (Joule effect). The appreciable
conductivity differences between organs result in an inhomogeneous distribution of
current density. Consequently, the thermal effects on electrification (burns) depend on
the pathway of the current through the body. Electrical burns are usually most severe at
the contact points, where current concentration results in high current density. This
effect is exploited in electrocautery devices for the cauterization of tissue, where the
tissue is burned by an electrical current. Electrosurgery uses alternating current to
directly heat the tissue itself. This is known as diathermy.

2.2. Electric shock

Electric shock is the physiological and pathological effect of the passage of an

electrical current through the body. The electric current may result in action potentials
and produce nerve stimulation and muscle contraction (also termed tetany). According
to the intensity and path, electrification can produce spasm of the muscles, lung
paralysis, loss of consciousness, irregular heartbeat, heart fibrillation and/or cardiac
arrest [4, 5, 10].

3. Factors Influencing Electrical Injury

The severity of electrical injury depends on the type of source, the intensity of the
current, the pathway through the body and the duration of the contact. Other factors are
the applied current frequency, the phase of the heart cycle when the shock occurs and
the general health status of the person.
The effect of electric shock decreases with applied signal frequency [7]. High
frequencies currents do not excite muscles and do not cause cardiac arrhythmias.
84 J. Jossinet / Electrical Safety

Unless otherwise noted, the values quoted here are implicitly for the frequency of
power distribution.
The current and pathway of the current through the body depend on the location
and impedance of contact points and the resistance of the body between these points.
This latter issue depends on each subject's morphology and body composition. Finally,
it must be kept in mind that there are no absolute thresholds or known values
quantifying the exact injury from a given current or voltage.

3.1. Voltage

Sources producing electrical injury are generally classified into "low voltage" (less than
500 volts) and "high voltage" (greater than 1000 volts). Both high and low voltage can
be fatal. Low voltage injuries occur at home or in a residential environment.
Electrocutions in bathtubs and by electric dryers are the most common causes of low-
voltage deaths.
Low voltage does not imply low hazard. In most cases, electrification by the mains
does not cause skin damage unless the contact point is small or the victim has delicate
skin. A low-voltage source can, however, produce major cardiopulmonary
complications and death if a sufficient current passes across the chest (depending on
the resistance) to induce ventricular fibrillation (depending on the pathway).
High-voltage injuries generally occur in an outdoor environment near power
sources and power lines, either by direct contact or arc flashing. High-voltage has a
greater potential for tissue destruction and can be responsible for severe injuries
leading to amputations and tissue loss. However, remembering that the adverse effect
of electrification is due to current a shock of 100 volts is not less dangerous than a
shock of 10000 volts. Individuals have been electrocuted by appliances connected to
the mains.

3.2. Amperage

The term "amperage" refers to the intensity of the current passing through the body.
The "let-go threshold" is the limit at which a person becomes unable to let go off the
current source because of muscular tetany. Typical values of let-go currents for men,
women and children are about 9, 7 and 4 mA, respectively. Electricians familiar with
this effect often refer to an immobilized victim of electric shock as being "frozen on the
circuit." Typical effects of electrification of the human body by 50 or 60 Hz AC
currents are summarized in Table 1.

3.3. Source Type

The type of circuit involved, either direct current (DC) or alternating current (AC) is
one of the factors affecting the nature and severity of electrical injury. Direct current
(ESD, lightning, batteries) tends to throw the victim from the source after. Alternating
current (power lines) is more dangerous for it causes muscle contraction that can
maintain the contact with the power source. High-voltage AC injuries are more severe
than DC ones mainly because of increased risk of muscular tetany and prolonged
contact. AC exposures are said to be three times more dangerous than DC exposure at
the same voltage.
 J. Jossinet / Electrical Safety 85

Table 1. Archetypal effects of electric current on the human body against the applied currents for frequency
of 50/60 Hz

Current (mA) Effect

0.1 - 1 Perception threshold

1-4 Faint tingling sensation

1 - 10 Muscular contraction, slight shock, not painful but disturbing, “no let go”
danger

6 - 30 Painful shock, muscular contraction, muscular control lost

20 - 75 Breathing becomes difficult due to thoracic muscle tetany

Extreme pain, severe muscular contraction, respiratory arrest, ventricular

50 - 150
fibrillation above 100 mA, death is possible

100 - 200 Currents between 100 mA and 200 mA are lethal

Severe burns, unconsciousness, strong muscular contraction, heart

200 - 1000 clamped during the shock (no ventricular fibrillation), good chances for
survival if the victim is given immediate attention

Muscular contraction, nerve damage, ventricular fibrillation, death is most

1000 - 4300
likely

10000 Cardiac arrest, severe burns and probable death

3.4. Duration of Exposure

The severity of injury is increases with the time of current flow through the body. The
effects of short duration current are more shock than burns. For instance, the electric
ray produce discharges of several hundred volts. This fish, however, does not produce
lethal shocks since the discharge duration, of the order of tens of microseconds, is too
short.
However, even brief exposures to very high amperage can produce important tissue
damage. Lightning current is seldom long enough to cause severe burns, but it produces
electric shock than can be fatal. Thoracic tetany can occur at levels just above the let-
go current and result in respiratory arrest.

3.5. Current Pathway

The pathway of the current between the contact points determines the tissues at risk,
the type of injury regardless of whether high, low, or lightning voltages are being
86 J. Jossinet / Electrical Safety

considered.
The hand is the most common site of source contact point. Bone, tendon, and fat
have a very high resistance and tend to heat up and coagulate rather than transmit
current. Similarly high resistance skin resistance dissipates much of the energy,
producing surface burns and limiting internal damage. If the current were delivered
through low resistance contact points, it can produce cardiac arrest but no surface
burns, such as in a bathtub injury.

3.6. Contact Impedance

The impedance at the contact point depends on the piece of material in contact with the
skin and the properties of the skin. In particular, the contact impedance depends on the
contact surface area: the wider the area, the lower the contact resistance. Hence, the
contact impedance can be different if the victim is grasping a handle or coming
accidentally into contact with an electrified object. In practice, the contribution of skin
resistance to the contact impedance is preponderant.
For usual contacts, practical values of skin resistance vary from 1000 Ω for wet
skin to over 500 KΩ for dry skin. When a victim comes into contact with a high-
voltage power source, the epidermis may be destroyed by heat within milliseconds.
Moisture, sweating and immersion in water can dramatically reduce skin resistance.
Pure water is a poor conductor, but tap water is almost as conductive as body tissues. In
practice, water always increases the risk of electrical injury by decreasing the resistance
of contact points.

3.7. Tissue Resistance

Body tissues other than skin are relatively good conductors; large currents can then
pass through them and produce tissue damage [1,6]. The resistances of the organs
determine the distribution of the current through the body for the current tends to pass
through low resistance regions. Nerves, muscles, and blood are better electrical
conductors than bone, tendon, and fat. Other body tissues have intermediate
resistances.
The actual resistance of the body varies depending upon the points of contact and
the skin condition (moist or dry). For instance, skin resistance excluded, the internal
resistance between the ears is about 100 ohms, while the resistance from hand to foot is
about 500 ohms.

4. Symptoms of Electrical Injury and First Aid to Victims

The outcome of an electric shock depends on the speed and adequacy of the treatment
[3,9]. The current's path through the body determines for prognosis and therapy. It is
impossible to know the exact amperage because of the variability of the resistance.
However, an estimate can be calculated knowing the voltage of the source. The
pathway can be estimated knowing the contact points.
Pain is the least significant result of electric shock. The outside of the victim’s
body may appear to have only minor lesions, but the internal injuries may still be
significant. More generally, the symptoms include:
 J. Jossinet / Electrical Safety 87

 Skin burns,
 tingling, weakness, unconsciousness,
 Muscle contraction, muscular pain,
 Bone fractures
 Headache, hearing impairment, seizures
 Respiratory failure, arrhythmias, cardiac arrest.

4.1. Cares to Victims

The classical dos and don'ts can be summarized as follows:

1. Switch off the electrical current at the control box, if safely possible.
2. Call for emergency medical help.
3. If the current can't be switched off, use an insulator (dry rope, cloth, broom
handle…) to drag the victim away from the contact with the source.
4. Once the victim is free from the source of electricity, check the victim's breathing
and pulse. If either has stopped or seems dangerously slow or shallow, initiate first
aid..
5. Give first aid for burns
6. If the victim is faint, pale, or shows other signs of shock, lay the victim down, and
cover the person with a blanket or a coat.
7. Stay with the victim until medical help arrives.
8. Avoid moving the victim's head or neck if a spinal injury is suspected. Give first
aid needed for other wounds or fractures.
DO NOT get close to a victim in contact with a high-voltage source
DO NOT touch the victim in contact with the source of electricity with bare hands
DO NOT attempt to rescue a victim near active high-voltage lines
DO NOT move a victim of electrical injury unless there is immediate danger.
DO NOT remove dead skin or break blisters if the victim has acquired burns.
DO NOT apply ice, butter, medications, cotton dressings or adhesive bandages to a
burn.

5. Protection and prevention measures

Simple rules improve electrical safety at home and at the laboratory. The use of
electricity at the working place must be in agreement with the general safety
regulations and the specific ones applying to every particular type of activity (e.g. use
of chemical, flammable gases, protection against ESD). However, the application of
simple basic rules minimizes electrical hazards at home and in the laboratory. Some of
them are given below.

5.1. General Electric Safety Tips

 Unplug electrical appliances before cleaning

 Always unplug an appliance that overheats and have it checked by a qualified
repair person before using it again,
88 J. Jossinet / Electrical Safety

 Check appliances and extension cords for fraying or cracking,

 Always pull the plug and not the cord; never carry an appliance by the cord

5,2, Tips for the laboratory

 Maintain equipment properly

 Learn the location of electrical panels and shut-off switches for a quick power
disconnection in the event of an emergency
 Minimal clearance of a 1 metre around electrical panels for ready access
 Don't overload circuits by using power strips or multiple outlets on regular
sockets.
 Carefully place power cords so they don’t come in contact with water or
chemicals
 Do not allow cords to dangle from counters or hoods
 Do not allow cords to contact hot surfaces, water or chemicals
 Do not place electrical cables where can be walk on or tripped over
 Never attach an exposed connector such as an alligator clip to a power supply
 No jewellery or other metal objects around electricity
 Don't work with electricity with wet hands or feet, or if the floor is wet.

5.3. Problems Needing Inspection orRepair

The device must be inspected by a qualified repair person in case of:

 Recurring problems with blowing fuses or circuit breakers,
 Discoloration of wall outlets,
 Burning smell or unusual odour coming form an appliance or wiring,
 Feeling a tingle or a shock when you touch an electrical appliance,
 Sizzling sound at wall switches or outlets,

Bibliography

[1] C. Biegelmeier, New knowledge of the impedance of the human body, in Electric shock safety criteria,
J. Bridges, L. Ford, L. Sherman, M. Vainberg, eds, Pergamon Press, 1985.
[2] C.F. Dalziel, The threshold of perception currents, in: IEEE Trans. Power Apparatus and Systems .
73(1954), 990-996.
[3] A.R. Dimick, Electrical Injuries, in Harrison's Principles of Internal Medicine, Anthony S. Fauci et al.
Eds, McGraw-Hill, New York, 1997.
[4] R. Fish, Electric shock- Part I: Physics and pathophysiology, J Emerg Med., 11(1993), 309-312.
[5] R. Fish, Electric shock, Part II: Nature and mechanisms of injury, J Emerg Med., 11(1993), 457-462.
[6] IEC Technical Specification, Electrical impedance of the human body, effects of sinusoidal alternating
current in the range of 15 Hz to 100 Hz, effects of direct current, IEC publication IEC/TS 60479, Part
1, (1994-09).
[7] IEC Technical Specification, Effects of alternating current with frequencies above 100 Hz, Effects of
special waveforms of current, Effects of unidirectional single impulse currents of short duration", IEC
publication IEC/TR 60479, Part 2 (1987-03).
[8] IEC Technical Specification, Effects of lightning strokes on human beings and livestock, IEC
publication IEC/TR 60479, Part 4 (2004-07).
[9] A.C. Koumbourlis, Electrical injuries, Crit Care Med., 30(2002), 424-430.
 J. Jossinet / Electrical Safety 89

[10] R.C. Lee, D. Zhang and J. Hannig, Biophysical injury mechanisms in electrical shock trauma, Ann. Rev
Biomed Eng., 2(2000), 477-509.
[11] D. Leibovici, J. Shemer, S.C. Shapira, Electrical injuries: current concepts, Injury, 26(1995), 623-627.
[12] J.P.Reilly, Scales of reaction to electric shock, in Electrical Injury, Annals of N.Y. Acad. Sci.,
720(1994), 21-37.
90 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-90

II.3. Electrograms (ECG, EEG, EMG,

EOG)
Richard B. REILLY and T. Clive LEE
Trinity Centre for BioEngineering, Trinity College, Dublin 2, Ireland
Department of Anatomy, Royal College of Surgeons in Ireland, Dublin 2, Ireland

Abstract. There is a constant need in medicine to obtain objective measurements

of physical and cognitive function as the basis for diagnosis and monitoring of
health. The body can be considered as a chemical and electrical system supported
by a mechanical structure. Measuring and quantifying such electrical activity
provides a means for objective examination of heath status. The term electrogram,
from the Greek electro meaning electricity and gram meaning write or record, is
the broad definition given to the recording of electrical signal from the body. In
order that comparisons of electrical activity can be made against normative data,
certain methods and procedures have been defined for different electrograms.
This paper reviews these methods and procedures for the more typical
electrograms associated with some of the major organs in the body, providing a
first point of reference for the reader.

Keywords. Biopotential, membrane potentials, electrograms, cardiogram,

encephalogram, myogram, oculogram

1. Biopotentials

The term biopotential refers to the electric potential that is measured between points in
living cells, tissues, and organisms and which accompanies all biochemical processes.
Biopotentials from the organs of the body provide rich physiological and clinical
information, playing often significant role in diagnosis.
Bioelectric potentials are produced as a result of electrochemical activity in a
certain class of excitable cells. These cells are components of nervous, muscular, and
glandular tissues. Electrically, these cells exhibit a resting potential and, when
appropriately stimulated, an action potential. Their activity is necessary for information
transfer (e.g. sensory information in the nervous system or coordination of blood
pumping in the heart).

1.1. Mechanism Behind Biopotentials

Neurons are designed to respond to stimulation, which they do by generating electrical

impulses. These impulses are expressed as changes in the electrical potentials
conducted along the plasma membranes of the dendrites, cell body, and axon of each
neuron. The difference in potential across the plasma membrane of the neuron results
from differences in the concentration of certain ions on either side of the membrane.
Cell potential is a function of membrane permeability and concentration gradient to
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 91

various molecules (i.e. K+, Na+, Cl-, and Ca2+). The cell membrane is a very thin (7-15
nm) lipoprotein complex that is essentially impermeable to intracellular protein and
other organic anions (A-). The membrane in the resting state is only slightly permeable
to sodium Na+ and rather freely permeable to potassium K+ and chlorine Cl-. The
permeability of the resting membrane to potassium ions is approximately 50~100 times
larger than its permeability to sodium. The diffusion and electrical forces acting across
the membrane are opposed to each other, and a steady state is ultimately achieved. This
steady membrane potential is called the equilibrium potential for potassium on account
of the main ion potassium K+ involved in the resting state. As a result an approximation
of steady membrane potential can be calculated using the Nernst equation , Eq. (1). The
Nernst equation provides a quantitative value for potential generated, EK given the
intra- and extracellular concentrations of potassium.

RT [K ] o [K ] o (1)
EK = ln = 0.0615 log10
nF [K ] i [K ] i

where n is the valence of potassium, [K]i and [K]o are the intra- and extracellular
concentrations, R is the universal gas constant, T is the absolute temperature in Kelvin,
F is the Faraday constant [1].
A more accurate expression for the membrane equilibrium potential can be
calculated from the Goldman- Hodgkin-Katz equation, Equations 2 and 3, which
provide a more accurate estimate of the potential E, by taking into consideration the
intra- and extracellular concentrations of sodium and chlorine ions as well as
potassium.

⎡ ⎤ (2)
RT ⎢ PK [K ]o + PNA [Na ]o + PCl [Cl ]i ⎥
E= ln
F ⎢ PK [K ]i + PNA [Na ]i + PCl [Cl ] ⎥
⎣⎢ o ⎦⎥

⎡ ⎤ (3)
P [K ]o + PNA [Na ]o + PCl [Cl ]i ⎥
E = 0.0581 log10 ⎢ K
⎢ PK [K ]i + PNA [Na ]i + PCl [Cl ] ⎥
⎢⎣ o ⎥⎦

Here P is the permeability coefficient of the given ion.

When membrane stimulation exceeds a threshold level of about 20 mV, an action
potential occurs. Sodium and potassium ionic permeabilities of the membrane change.
Sodium ion permeability increases very rapidly at first, allowing sodium ions to flow
from outside to inside, making the inside more positive. The more slowly increasing
potassium ion permeability allows potassium ions to flow from inside to outside, thus
returning membrane potential to its resting value. At rest, the sodium and potassium
(Na-K) pump restores the ion concentrations to their original values. The number of
ions flowing through an open channel is greater than 106/sec.
The change in the electrical potential difference across a plasma membrane is the
key factor in the initiation and subsequent conduction of a nerve impulse. A stimulus
that is strong enough to initiate an impulse in a neuron is called a threshold stimulus.
When such a stimulus is applied to a polarized resting membrane of an axon, sodium
92 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

ion channels into the cell open, and sodium ions rush in, reversing the electrical charge
at the point of stimulus (In the l millisecond that a channel is open, about 20,000
sodium ions flow through). Thus, at the point of stimulus, the inside of the membrane
becomes positively charged relative to the outside, a condition known as
depolarization. When a stimulus is strong enough to cause depolarization, the neuron is
said to fire.
Once a small area on the neuron is depolarized, it stimulates the adjacent area, and
an action potential, or nerve impulse, is initiated and conducted along the plasma
membrane.
Shortly after depolarization, tbe original balance of sodium and potassium ions is
restored by the action of the membrane pumps; the relative electrical charges inside and
outside the membrane are also restored. The membrane is then said to be repolarized.
The transmission of a nerve impulse along the plasma membrane may be visualized as
a wave of depolarization and repolarization.
After each firing, there is an interval of from 0.5 to 1 millisecond before it is
possible for an adequate stimulus to generate another action potential. This period is
called the refractory period. Most nerve fibers are capable of generating about 300
impulses per second
In the resting state there exists a steady electrical potential difference between
internal and external environments—typically between -70 to -90mV, relative to the
external medium.
In the active state an electrical response is instigated by adequate stimulation. It
consists of an “all-or-none” action potential after the cell threshold potential has been
reached.
The body is an inhomogeneous volume conductor and these ion fluxes create
measurable potentials on the body surface. By recording these biopotentials, a process
known as electrography, an analysis of the underlying physiological can be carried out
providing information for clinical diagnosis.
Biopotentials have extremely low voltages values, typically in the order of micro
or millivolts. Amplification of these biopotentials is required prior to their acquisition
by computer and subsequent analysis.

2. Bioamplifers

A bioamplifier is an electronic instrument used to acquire and increase the signal

amplitude of biopotential electrical activity for output to various sources. The design of
biopotential amplifiers requires great technical expertise, as besides a linear
amplification of biopotential amplitude to improve the signal-to-noise ratio, the device
must maintain signal fidelity in terms of time resolution. The design parameters of
biopotential amplifier are specific to the biopotential under investigation. A more
detailed review of bioamplifers in provided in the accompanying related paper on
Biosensors.
There are many electrograms, however the most commonly used in the clinic are
the electrocardiogram (ECG), electroencephalogram (EEG), electrooculogram (E0G),
electromyogram (EMG), and electroretinogram (ERG).
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 93

3. Electrocardiogram (ECG)

An electrocardiogram is the most familiar electrogram and is the graphical output of an

electrocardiograph, which is the temporal variation of electrical activity of the heart.
The term cardio derives from the Greek for heart and the word electrocardiogram is
often abbreviated to ECG or EKG. The ECG possesses great information content, due
to the ease at which can it can measured in different circumstances and it is pivotal in
providing clinical diagnoses. By analyzing the perturbations in normal electrical
activity, the ECG can provide critical information on an evolving myocardial
infarction, different cardiac arrhythmias, the effects of hypertension, as well as
important information for cardiac rehabilitation and exercise.
The first practical electrocardiogram was recorded by the Dutch doctor and
physiologist Willem Einthoven in 1903. He was awarded the Nobel Prize in Physiology
or Medicine in 1924 “for his discovery of the mechanism of the electrocardiogram”.

3.1. Basis of the ECG

The heart consists of four chambers: the left and right atria and also the left and right
ventricles. Blood enters the right atrium and passes through the right ventricle, which in
turn pumps the blood to the lungs where it becomes oxygenated. The oxygenated blood
is brought back to the heart by the pulmonary veins which enter the left atrium, flowing
then into the left ventricle. The left ventricle pumps the blood to the aorta which will
distribute the oxygenated blood to all parts of the body.
There are three groups of cells in the heart. Pacemaker cells, Electrical Conducting
Cells and Myocardial Cells [2]. In the resting state these cells are electrically polarised,
but it is their depolarisation that is the fundamental electrical event of the heart.
Depolarisation is propagated from cell to cell, producing a wave of depolarisation that
is transmitted across the heart. This wave sets up a biopotenital across the heart,
resulting in the flow of electrical current [3]. This depolarisation can be detected by a
number of electrodes attached at specific points on the body. Once the depolarisation
has occured, repolarisation commences to restore the cardiac cells to their resting state.
This repolarisation phase can also be recorded with the same electrodes. The ECG is a
recording of the depolarisation and repolarisation of the myocardial cells of heart,
which makeup the majority of heart cells and represent the heart’s contractile
machinery. The ECG indicates the overall rhythm of the heart and weaknesses in
different parts of the heart muscle [4].
Since the human body can be regarded as a volume conductor, changes in potential
are transmitted throughout the body. Therefore to be able to record myocardial activity,
one needs to be able to detect small changes in potential on the body surface. These
waves can be measured at electrodes attached to the skin. An ECG displays the voltage
between pairs of these electrodes and the muscle activity in different locations.
In order to record the ECG, each lead pair is connected to a bioamplifier. Bipolar
leads measure the difference in electrical potential between any two points on the body,
unipolar leads measure the potential at a single point on the body and a virtual
reference point, with zero electrical potential, located in the centre of the heart.
A lead records the electrical signals of the heart from a particular combination of
recording electrodes that are placed at specific points on the patient's body. In various
cases, the detected signals have the following polarities (see Figure 1):
94 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

• When a depolarization front propagates toward a positive electrode, it

creates a positive deflection on the ECG in the corresponding lead.
• When a depolarization wavefront moves away from a positive
electrode, it creates a negative deflection on the ECG in the corresponding lead.
• When a depolarization wavefront moves perpendicular to a positive
electrode, it creates an biphasic complex on the ECG. This will be positive as the
depolarization wavefront approaches (A), and then become negative as it passes by
(B).

Figure 1. Graphic showing the relationship between positive electrodes, depolarization wavefronts (or mean
electrical vectors), and complexes displayed on the ECG.

3.2. Measuring ECG

3.2.1. ECG Leads
In 1908 Willem Einthoven published a description of the first clinically important ECG
measuring system. The Einthoven triangle is an imaginary equilateral triangle having
the heart at the centre and formed by lines that represent the three standard limb leads
of the electrocardiogram (Figure 2).
3.2.2. Standard Limb Leads
The most basic arrangement of ECG leads is the three limb lead configuration, which
are bipolar leads measuring the potential difference between the right arm, left arm, and
left leg. They are positioned as follows:
Lead I – Difference between the left arm and right arm, the left arm being positive
Lead II – Difference between the left leg and right arm, the left leg being positive.
Lead III – Difference between the left leg and left arm, the left leg again being positive.
A fourth electrode is placed on the right leg to use as reference or “electric ground”.
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 95

Figure 2. Einthoven lead system

The Einthoven limb or Standard leads are defined in the following way:
Lead I: VI = Φ L - Φ R
Lead II: VII = Φ F - Φ R
Lead III: VIII = Φ F - Φ L
where
VI = the voltage of Lead I, VII = the voltage of Lead II,
VIII = the voltage of Lead III
Φ L = potential at the left arm (LA)
Φ R = potential at the right arm (RA)
Φ F = potential at the left leg (LL)
According to Kirchhoff's law, these lead voltages have the following relationship:

V I + V III = V II (4)

Hence only two of these three leads are independent.

In clinical practice, ECG is recorded using extra leads in addition to the standard
three. This is necessary due to the three dimenstional nature of the heart. The addtional
leads provide important information on the heart’s electrical activity in three
orthogonal directions.
3.2.3. Augmented Limb Leads
The same three leads that form the standard leads also form the three unipolar leads
known as augmented leads. These are referred to as “augmented” because they are
unipolar. These three leads are referred to as aVR (right arm), aVL (left arm) and aVF
(left leg) and also record a change in electric potential in the frontal surface of the
chest.
96 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

3.2.4. Precordial Leads

For measuring the potentials close to the heart, Wilson introduced the precordial leads
(chest leads) in 1944. These leads (V1 to V6) are located over the left chest as shown in
Figure 3.

Figure 3. Precordial leads.

Placement of Precordial Leads
th
V1 – 4 intercostal space, just to the right of the sternum
V2 – 4th intercostal space, just to the left of the sternum
V3 – Halfway between V2 and V4
V4 – 5th intercostal space in the mid-clavicular line
V5 – Halfway between V4 and V6
V6 – 5th intercostal space in the mid-axillary line

3.2.5. The 12-Lead System

The most commonly used clinical ECG-system, the 12-lead ECG system, consists of
the following 12 leads:
• I, II, III
• aVR, aVL, aVF
• V1, V2, V3, V4, V5, V6
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 97

Of these 12 leads, the first six are derived from the same three measurement points.
Therefore, any two of these six leads include exactly the same information as the other
four.

3.3. Formation of the ECG Waveform

In order to interpret the ECG, one needs to understand the electrical events that
underpin the cardiac cycle of contraction (systole) and relaxation (diastole) [2]. The
dominant pacemaker cells in the heart, are known as the Sinoatrial Node. These are
located at the top right hand side of the heart, in the right atrium. These cells fire
spontaneously at a rate of 60 to 100 times per minute, and a wave of depolarization
begins to spread outward into the atrial myocardium cells. It is the depolarization of
atrial myocardium cells that results in atrial contraction. Electrodes on the surface of
the body will record this contraction as a burst of electrical activity lasting a fraction of
a second. This is called the P wave. The atria begin contracting around 100 ms after
the start of the P wave. Once the atrial depolarisation is complete the ECG returns to
baseline.
Following this the ventriclar myocardial cells depolarise, causing a ventriclar
contraction. Electrodes on the surface of the body will record this contraction as a large
increase in amplitude. This is known as the QRS complex. The amplitude of the QRS
complex is larger than the P wave as the mass of the ventricular muscle is much larger
than that of the atria. The ventricles begin contracting shortly after the R wave. The
smaller T wave indicates ventricular repolarization. The atria repolarize during the
QRS complex and therefore this repolarization cannot be observed separately.

Figure 4. Heart excitation related to electrocardiogram (ECG)

98 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

Therefore a typical ECG tracing of a normal heart cycle consists of a P wave, a

QRS complex and a T wave. A small U wave, caused by repolaristaion of Purkinje
fibers the nerve cells of the heart, is normally visible in 50 to 75% of recorded ECGs.
The baseline voltage of the electrocardiogram is known as the isoelectric line.
Typically the isoelectric line is measured as the portion of the tracing following the T
wave and preceding the next P wave. Figure 4 shows the important features of a sample
ECG.

3.3.1. Electrical heart axis

When the heart depolarizes and repolarizes, it is convenient to represent the electrical
activity as an electric dipole or mathematically as a vector. A vector has both
magnitude and an angle of orientation. The magnitude of the heart vector represents a
maximum amplitude of the recorded electrical activity, while the angle of orientation
represents the average direction of the current flow in the heart. The magnitude and
direction of this heart vector can be beneficial in diagnosis.
The augmented and standard leads allow analysis of the heart on what is known as
the frontal plane. Therefore, the heart vector refers to the direction of the electrical
depolarization obtained from the sum of all different vectors in the frontal plane. It is
usually oriented in a right shoulder to left leg direction. The normal angle of
orientation of the QRS vector generally lies between 0 and 90 degrees. A change in the
angle of orientation provides diagnostic information. For example, a left deviation (-30 o
to -90o) may indicate left anterior fascicular block or Q waves from inferior myocardial
infarction (MI). Right axis deviation (+90o to +180o) may indicate left posterior
fascicular block, Q waves from high lateral myocardial infarction, or a right ventricular
strain pattern.

3.3.2. ECG – Clinical significance

ECG is the standard method for the diagnosis of cardiac arrhythmias (Figure 5).
Pathology that can be identified and studied using ECGs includes rhythm disturbances,
ischaemia and infarction, chamber enlargements, electrolyte abnormalities, and drug
toxicities.

Some abnormalities that may indicate illness:

• An extended P-R interval may be diagnosed as AV node block
• Widening of the QRS complex conduction problems in the bundle of His
• Elevated ST segment may indicate occurrence of MI
• Negative polarity T wave may be due to coronary insufficiency. QRS
amplitude, polarity, time domain, PR interval (indicator of heat beat per min.
& T-wave amplitude are some very important distinctive features.
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 99

Figure 5. Clinical significance of ECG ( Source: www.bioscience.org)

100 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

4. Electroencephalography

Hans Berger was the first to standardize recordings of the electrical activity in the
brain. In a series of experiments in 1920s, he pioneered electroencephalography (EEG),
by placing electrodes on the scalp and measuring electrical activity during different
tasks. The name is derived from the Greek encephalo meaning brain.
The conventional EEG is recorded from scalp electrodes, and shows cortical
electrical activity. This includes cortical manifestations of the sub cortical regions
(projection pathways, thalamus, reticular formation, mesencephalon or midbrain) [5].
As the EEG reflects the functional status of the brain, it can be used to monitor its
functional integrity and thus assist clinicians in the diagnosis of a variety of
neurological problems. It is very important to relate scalp potentials to the underlying
neurophysiology. The pathological states most commonly diagnosed using EEG
include, common headaches and dizziness, seizure disorders, stroke, brain tumours,
epilepsy, multiple sclerosis, sleep disorders and movement disorders [5].
A method similar to the EEG is intracranial EEG (icEEG), also referred to as
subdural EEG (sdEEG) and electrocorticography (ECoG). The electrodes are placed
directly on the exposed surface of the brain to record electrical activity from the
cerebral cortex. ECoG is an invasive procedure in which electrodes are placed over a
specific brain area. In a very small selected group of patients being considered for
epilepsy surgery, ECoG may be recorded over several days to map the distribution and
spread of seizure activity.
EEG records the spontaneous activity in the brain. The amplitude the signal record
on the scalp is typically 100 µV, while that of ECoG from the surface of the cortex it is
in the order of 1-2 mV.

4.1. Electrode Montages

There is a standard system for electrode placement, known as the 10-20 international
system which includes 64 electrodes. In this system the electrodes are located on the
surface of the scalp, as shown in Figure 6. The electrode positions are determined as
follows: Reference points are nasion, which is the delve at the top of the nose, level
with the eyes; and inion, which is the bony lump at the base of the skull on the midline
at the back of the head, the external occipital protrusion. The first mark is placed 10%
of the distance along this line and others are arranged at 20% intervals. With this
electrode placement system, activity at specific brain locations can be recorded with
specific electrodes and thus can easily be compared across individuals [5].
When carrying out EEG experimentation, it is typical to record specific sensory
processing areas of the cortex: visual, auditory or somatosensory. For example, visual
processing is carried out in the occipital regions at the back of the brain. Activity at
electrodes O1 and O2 would be more active during visual processing and these
electrodes would be termed active electrodes while the brain is engaged in processing
visual information.
64-channel EEG systems are typical, but some high-density systems exist, with
128 and 256 electrodes. Placement is based again on the 10-20 system, with the new
electrodes assigned to locations in between the 32 electrodes of the 10-20 system.
In order to obtain good signal quality, impedance between the scalp and each
electrode should ideally be kept below 5kOhms. If the electrode impedance increases,
background noise and movement artifacts may obscure the EEG signal. As a
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 101

consequence, the signal-to-noise ratio decreases and the ability to extract diagnostic
information may be limited.

Figure 6. The international 10-20 system seen from (A) left and (B) above the head. A = Ear lobe, C =
central, Pg = nasopharyngeal, P = parietal, F = frontal, Fp = frontal polar, O = occipital.

Electrode placements and the different patterns of combining electrode pairs to

measure potential differences on the scalp constitute the electrode montage. There are
two basic types of EEG montage: Referential and Bipolar. In the referential montage
the potential difference is measured between an active electrode and an inactive
reference electrode. While, with the bipolar montage, the potential difference is
measured between two active electrodes.

4.2. Basis of the EEG

The EEG signal recorded on the scalp or on the cortex is generated by the summation
of the synchronous activity of thousands of neurons that have similar spatial orientation
and placement radial to the scalp. An action potential in a pre-synaptic axon causes the
release of neurotransmitter into the synapse. The neurotransmitter diffuses across the
synaptic cleft and binds to receptors in a post-synaptic dendrite. The activity of many
types of receptors results in a flow of ions into or out of the dendrite. This results in
currents in the extracellular space. It is these extracellular currents which are
responsible for the generation of EEG potentials.

4.3. Processing of EEG

The character of the EEG signal is highly dependent on the degree of the activity of the
cerebral cortex. EEG observed during states of wakefulness and sleep are remarkably
102 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

different. EEG is typically described in terms of patterns of rhythmic activity and

transients. The rhythmic activity is divided into bands by frequency. Most of the
cerebral signal observed in the scalp EEG falls in the range of 1-40 Hz. EEG frequency
ranges are categorized as Delta (σ), Theta (τ), Alpha (α) and Beta (β) and Gamma (γ)
[6].
Alpha rhythm is usually observed between 8 and 12Hz. Alpha is usually best
observed in the posterior regions of the head on each side. It tends to dominate during
eyes closed and by relaxation, and tend to diminish with eye opening or cognition
processing (thinking, calculating). It is the major rhythm seen in normal relaxed adults.
It is present during most of life especially beyond the age of thirteen.
Beta activity is a higher frequency rhythm than alpha, with a frequency between
14-20 Hz and is classed a fast activity. It is usually most evident frontally. It is
accentuated by sedative-hypnotic drugs especially the benzodiazepines and the
barbiturates. It may be absent or reduced in areas of cortical damage. It is generally
regarded as a normal rhythm. It is the dominant rhythm in patients who are alert or
anxious or who have their eyes open.
Theta activity has a frequency between 3.5 to 7.5 Hz and is classed as slow
activity. It is abnormal in awake adults but is perfectly normal in children up to 13
years of age and also in sleep.
Delta activity is 3 Hz or below. It tends to be the greatest in amplitude of all EEG
rhythms. It is the dominant rhythm in infants up to one year and in stages 3 and 4 of
sleep. It is usually most prominent frontally in adults and posteriorly in children.
Gamma activity is associated with 40 Hz and up to 70Hz. Activity in this band can
often be from 24Hz. Gamma activity is thought to be associated with perception and
consciousness. Many encephalographers do not distinguish gamma waves as a distinct
class but include them in beta brain waves.
Since the interpretation of EEG is time consuming and requires much experience,
different processing methods have been developed to facilitate its interpretation.
Typically, Fourier Analysis is applied to recorded EEG. Fourier analysis provides an
estimate of the frequency content of a signal and results in the power spectrum of that
signal. In the case of EEG, the spectrum of EEG can be divided into the specific
spectral bands: delta, theta, alpha, beta and gamma. In terms of diagnostic information,
of great interest is the ratio of power of one frequency band to another during a specific
brain state, for example the ratio of alpha power to beta power. Another quantity of
diagnostic interest is the change over time in power of a frequency band during a
specific brain state.

4.4. Evoked Potential

The EEG reflects the functional status of the brain. The strict definition is continuous
EEG, as it represents ongoing brain activity. Another useful electrical signal that can be
recorded is the evoked or event related potential. Evoked potentials arise in response to
a stimulus (auditory, visual or somatosensory, etc.). They provide great diagnostic
information on sensory processing. They are much smaller in amplitude than
continuous EEG and are not readily distinguished. Therefore, a train of stimuli is often
applied to the individual and signal averaging method used to improve the signal-to-
noise ratio of the recorded signals to allow interpretation [5],[6].
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 103

4.4.1. Visual Evoked Potentials

These are tests of the pathway between the eye and the back of the brain. The
individual being examined fixates on the centre of a computer monitor that has a
moving black and white checkerboard pattern. Each eye is tested separately, with a
patch covering the eye not being investigated. The recorded visual evoked potential
(VEP) is achieved by averaging over a number of responses to checkerboard pattern
and are of the form shown in Figure 7.

Figure 7. A normal pattern reversal VEP

Measurements are made of the latencies and amplitudes of particular peaks. In the
normal VEP, a very distinguishable peak is found approximately 100ms after
stimulation, known as the P100. There are also two clear troughs called the N75 and
N135, around 75ms and 135ms after stimulation. The latencies and amplitudes of these
peaks provide objective diagnostic information on the brain’s ability to process visual
information. Individual data can be with compared to normative data.
4.4.2. Auditory Evoked Potentials
Audio evoked potentials (AEP) reflect the function of the auditory pathway. AEP can
be evoked by repeated clicks of short duration (100-500 ms) given into an ear piece.
Trigger-synchronized averaging of a defined number of EEG-segments is used to
extract the AEP by reduction of the underlying EEG signal (background noise). The
reduction in background noise is proportional to the square root of the number of
averaged segments; the more averaged segments, the better the quality of the AEP. The
extracted AEP signal consists of a number of peaks and troughs. As with VEP analysis
measurements are made of latencies and amplitudes of particular peaks. Three main
groups of peaks can be distinguished and they can be correlated to the anatomical
structures:
104 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

• Brainstem AEP (BAEP) with latencies shorter than 10 milliseconds.

Anatomical structures: cochlea, acoustic nerve (BAEP wave I, II), brainstem
(BAEP wave III-V)
• Middle latency AEP (MLAEP) with latencies of 10-50 milliseconds.
Anatomical structures: medial geniculate body and primary auditory cortex
(temporal lobe).
• Late cortical waves with latencies over 50 milliseconds. Anatomical
structures: frontal cortex, association fields.

4.5. Interpretation of EEG

Changes in EEG are not necessarily specific to underlying mechanisms [5]. For
example, slowing of EEG may reflect either changes in anesthetic concentrations or
cerebral ischemia. EEG has poor spatial resolution, meaning that it not possible to
know exactly the source of electrical activity from the recordings made on the scalp.
Mathematically one can consider the electrical activity being generated by an electric
dipole. Considerable research attention is focused on providing estimates of the
location of electric dipoles that would generate the potential similar to those recorded.
The magnitude and orientation of these dipoles can be used as a basis for diagnosis in
neurological conditions.
Despite these disadvantages, EEG has considerable advantages over neuroimaging.
Its poor spatial resolution is offset by excellent temporal resolution, which is in
milliseconds. The ease of recording EEG makes it of great clinical importance.

5. Electromyography (EMG)

Electromyography (EMG) is the study of the mechanical properties of muscles at rest

and in contraction. The electrical activity associated with muscle is recorded using two
groups of electrodes over the muscle site. These can be indwelling (intramuscular)
electrodes or can be surface electrodes for non-invasive recordings [7].
EMG is used to diagnose two general categories of disease: neuropathies and
myopathies. EMG may aid with the diagnosis of nerve compression or injury (such as
carpal tunnel syndrome), nerve root injury (such as sciatica), and with other problems
of the muscles or nerves. Less common medical conditions include amyotrophic lateral
sclerosis, myasthenia gravis, and muscular dystrophy.

5.1. Basis of the EMG

Muscle consists of motor units, which is areas consisting of a motor neuron and the
muscle fibres it innervates. Each motor unit is controlled by the motor neurons. The
motor end plate is the synaptic junction between the motor neuron and the controlled
motor unit. Depolarization of the post synaptic membrane arises in case of activation of
the motor unit. End plate potential (EPP) is the potential that is recorded. The
depolarization wave moves along the direction of the muscle fibers. The signal between
the EMG electrodes corresponds to the depolarization wave front and to the subsequent
repolarization wave.
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 105

The electrical source is the muscle membrane potential of about -70mV. Due to
movement of the muscle, the measured potentials range between less than 50 μV and
20 to 30 mV. EMG signals are made up of superimposed motor unit action potentials
(MUAPs) from several motor units. For a thorough analysis, the measured EMG
signals can be decomposed into their constituent MUAPs.

5.2. Location and orientation of the electrode

The electrode should be placed between a motor point and the tendon insertion or
between two motor points, and along the longitudinal midline of the muscle. The
longitudinal axis of the electrode should be aligned parallel to the length of the muscle
fibers. The reference electrode also called the ground electrode should be placed as far
away as possible and on electrically neutral tissue [7].
The voltage waveform that is recorded is the difference in potential between the
two electrodes. The electrical signal generated in the muscle fibers is called a muscle
action potential (MAP). As mentioned above, the surface electrodes or indwelling
electrodes record the algebraic sum of all MAPs that are being transmitted along the
muscle fibers.

6. Electrooculography (EOG)

Electrooculography (EOG) is the study of retina function by recording changes in

steady, resting electric potentials of the eye. The eye can be considered to be a fixed
electric dipole with a positive pole at the cornea and a negative pole at the retina. The
electric potential is not generated by excitable tissue but is attributed to the higher
metabolic rate in the retina. The magnitude of the corneoretinal potential is in the range
of 0.4-1.0 mV.
The electric potential is independent to light stimulus. The field may be detected
with the eye in total darkness and/or with the eyes closed. This potential difference and
the rotation of the eye form the basis for a signal measured at a pair of periorbital
surface electrodes. Electrodes are placed near the canthi, the corner of the eye where
the upper and lower eyelids meet The resulting signal is called the electrooculogram. If
the eye travels from the center position in the direction of one electrode, this electrode
perceives the positive side of the retina while the opposite electrode views the negative
side of the retina. As a result, a potential difference occurs between the electrodes. If
the resting potential is constant, the recorded potential is a measure for the eye position
(Figure 8).
The EOG has clinical importance in the study of eye movement. The small
movements of the eye, termed nystagmus, are measured with the help of EOG. The
resulting signal is called an electronystagmogram.
The retina is the site of cells that are sensitive to the incident light energy; as with
other peripheral nerve cells, they generate receptor potentials. The collective behavior
of the entire retina is a bioelectric generator, which sets up a field in the surrounding
volume conductor. This potential field is normally measured between an electrode on
the cornea (contact-lens type) and a reference electrode on the forehead. The recorded
signal is known as the electroretinogram (ERG).
106 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

Figure 8. An illustration of the electrooculogram (EOG) signal generated by horizontal movement of the
eyes. Two electrodes are employed and each electrode connected to the input of a bioamplider. The polarity
of the signal is positive at the electrode to which the eye is moving

EOG is a sensitive electrical test for detection of retinal pigment epithelium

dysfunction. The main applications are in ophthalmological diagnosis and in recording
eye movements.

7. Conclusion

Electrical activity associated with the brain, heart, muscle and eye can be recorded and
analysed. Comparisons against normative data provide important clinical information.
It is important to relate the electrical activity to the underlying physiology so that this
information can be used for diagnosis.

References

[1] Deutsch S. and Deutsch A. (1993), Understanding the Nervous System-An Engineering Perspective,
IEEE Press, New York.
[2] Katz A.M. (1986), Physiology of the Heart, Raven Press, New York
[3] DiFrancesco D. and Noble D. (1985) “A model of cardiac electrical activity incorporating ionic
pumps and concentration changes”, Phil Trans R. Soc, London [B] 307, 307-353
[4] Geselowitz D.B. (1989) “On the theory of the electrocardiogram”, Proc IEEE 77, 857.
[5] Nunez P.L., Srinivasan R. (2005) Electric Fields of the Brain: The Neurophysics of EEG, Oxford
University Press, USA.
[6] Buzsaki G. (2006), Rhythms of the Brain, Oxford University Press, USA.
[7] Katirji B. (2007) Electromyography in Clinical Practice: A Case Study Approach, 2nd Edition,
Mosby.
 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG) 107

Glossary of Terms

Action potential: An action potential is initiated by a stimulus above a certain intensity

or threshold.
Atrium: There are four chambers in the human heart, two atria and two ventricles.
The right atrium receives de-oxygenated blood from the superior vena cava, inferior
vena cava and coronary sinus. The left atrium receives oxygenated blood from the left
and right pulmonary veins.
Artery: A vessel that carries blood away from the heart to the farthest reaches of the
body
Atrioventricular node: An electrical relay station between the atria (the upper) and the
ventricles (the lower chambers of the heart).
Aorta: The large artery that receives blood from the left ventricle of the heart and
distributes it to the body.
Biopotentials: electric potential that is measured between points in living cells, tissues,
and organisms as a result of electrochemical activity of excitable cells.
Bioamplifier: Device used to gather and increase the signal integrity of human
neurophysiologic electrical activity for output to various sources.
Bipolar lead: Registration of the potential difference between two electrodes.
Diastole: Phase of cardiac relaxation
Einthoven triangle: An imaginary equilateral triangle having the heart at its center and
formed by lines that represent the three standard limb leads of the electrocardiogram
Electrical heart axis: The direction of the electrical depolarization obtained from the
sum of all different vectors in the frontal plane.
Electrocardigraphy (ECG): Electrical recording of the heart and is used in the
investigation of heart disease.
Electrocorticography (ECoG): Graphical recording of electrical activity in the brain
by placing electrodes in direct contact with the cerebral cortex
Electroencephalograph (EEG): Graphical recording of electrical activity in the brain
recorded by scalp electrodes.
Electromyography: Graphical recording of the electrical activity of muscle
Electrooculography: Graphical recording technique for measuring the resting
potential of the retina.
Electroretinography: Graphical recording electrical responses of various cell types in
the retina.
Elevation: The rise of the ST-segment above the iso-electrical line.
Electrode: A metallic part which is used for transmitting the electrical activity from
the body to the input circuit of a bioamplifier.
Electrolyte: It is a conducting solution (gel or paste) or may be fluid of living tissue as
when electrode inserted below skin.
Excitable cells: Those that can be stimulated to create a tiny electric current.
Frequency: The number of beats per minute
Frontal plane leads: The frontal plane leads are the leads I, II, III, aVR, aVL and aVF.
His, bundle of: Fast conducting bundle which runs from the AV-node to the cells of
the ventricles.
Infarction: The formation of an infarct, an area of tissue death due to a local lack of
oxygen.
Interval: A specific distance on the ECG.
108 R.B. Reilly and T.C. Lee / Electrograms (ECG, EEG, EMG, EOG)

Ischemia: Shortage of oxygen in a tissue, caused by insufficient blood flow towards

that tissue.
Isoelectric line: The baseline voltage of the electrocardiogram
Lead: Electrode for registering the electrical potentials.
Membrane potential: Potential difference across the cell membrane
Myocardial infarction (MI): Medical condition that occurs when the blood supply to
a part of the heart is interrupted.
Precordial leads: Leads placed on the chest to record the ECG.
P-wave: The depolarization of both atria.
Physiology: The study of how living organisms function including such processes as
nutrition, movement, and reproduction.
Repolarization: Recovery of the resting potential.
Rhythm strip or an ECG strip: The printed record of the electrical activity of the
heart is called a rhythm strip or an ECG strip
Sinus node: Primary pacemaker of the heart located in the right atrium - sinoatrial
(SA) node.
Systolic: Phase of cardiac contraction
Transmembrane potential: Resting potential across the cell membrane.
T-wave: The repolarization of both ventricles.
Unipolar lead: Lead in which the potential differences are registered in one point
compared to the central terminal.
U-wave: Repolarization of the Purkinje fibbers
Vector: Electrical force in a specific direction and with a specific magnitude.
Basic Engineering for Medics and Biologists 109
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-109

II.4. Biosensors
Richard B. REILLY and T. Clive LEE*
Trinity Centre for BioEngineering, Trinity College, Dublin 2, Ireland
*Department of Anatomy, Royal College of Surgeons in Ireland, Dublin 2, Ireland

Introduction

The ability to accurately measure physiological and chemical changes in the body is
fundamental to the development of new diagnostic and therapeutic methods. In order to
carry out such measurements, specific tools are necessary to detect physiological and
chemical changes and to transform these changes into a form that can easily be
analyzed and processed. A biosensor is one such tool and can be defined as an
analytical device which converts a biological response into an electrical signal. The
name signifies that the device is a combination of two components, a biological
element and a sensor or transducer element.
The biological element may be tissue, living cells, an enzyme, antibody or antigen.
The sensor/transducer element includes electric current, electric potential, intensity and
phase of electromagnetic radiation, mass, conductance, impedance, temperature and
viscosity.
This paper will outline the components and characteristics of a biosensor, describe
some of the typical transducers found in biomedical engineering and biomedical
instrumentation amplifiers. A focus is placed on sensing within electrophysiology.
How transducers are used in medicine to measure key biophysical characteristics in the
clinic will form the concluding section.

1. Components of a Biosensor

The basic components of a biosensor include a biological element and the

physiochemical transducer, Figure 1. The output of the transducer is passed to the
detector for further processing and analysis.
The biological element is the target system under investigation, such as tissue,
microorganisms, cell receptors, enzymes, antibodies, nucleic acids or biologically
derived material. The transducer transforms the signal from the sensor into a more
easily measurable and quantifiable signal. Typically, a computer or microprocessor
device acquires (more formally termed samples), amplifies and processes the signal
from the transducer. Following signal processing, the data are often converted to other
units and transferred to a display or/and data storage device.
The electrical signal from the transducer is often low and may be superimposed
upon a relatively high and noisy baseline. One of the more fundamental signal
110 R.B. Reilly and T.C. Lee / Biosensors

processing steps involves subtracting a 'reference' baseline signal, derived from a

similar transducer without being coupled to the sensor, from the sampled signal.

Figure1. Components of a biosensor

1.1. General Biosensor Characteristics

An ideal biosensor must fulfill a number of criteria to measure a physical quantity [1].
The biosensor must be able to provide a signal that is directly proportional to the
quantity under investigation, providing all the useful information about the physical
event. The biosensor must also meet requirements connected with measurement such as
repeatability, sensitivity, reproducibility, selectivity, specificity, good response times
and linearity of response. These are defined as follow:
Specificity/selectivity: Ability to recognize a single event/condition among other
events/conditions in the same sample/signal. The selectivity of the biosensor is
determined by both the sensor and transduction method.
Sensitivity: The sensitivity of a sensor is measured by the change in its response as
a function of the analogous change in the quantity examined. The sensitivity s of a
biosensor is defined by the ratio

Δoutput
s=
Δinput
(1)

where
Δoutput = change in the amplitude of output and
Δinput = change in the amplitude of input.
The sensitivity determines the aptness of the sensor for a particular application.
Repeatability: The results are deemed to have high repeatability if two sets of
results, obtained using the same sensor and the same sample, are numerically close to
each other.
Reproducibility: The method is reproducible if similar results are obtained by other
scientists at different laboratories.
Linear response: The linear region of a biosensor is obtained from a calibration
curve of its response to different inputs. A good calibration curve is also indicative of
stability of the biosensor’s response, which should neither drift nor oscillate with time.
 R.B. Reilly and T.C. Lee / Biosensors 111

A biosensor having a linear response can be more easily described mathematically,

than one which has a nonlinear response.
Response time: Response time is defined as the time taken to reach a steady state
output from the instant of variation in the input value to the sensor.

1.2. Transducer Principles

The transducer is responsible for converting a biological signal to a measurable

electrical signal. The transducer is a device, usually electrical, electronic, electro-
mechanical, electromagnetic, photonic, or photovoltaic that converts one form of
energy to another. The transducer can take many forms depending upon the parameters
being measured - electrochemical, optical, mass and thermal changes are the most
common. The signal produced by a transducer after a change is measured, may or may
not need to be converted to an electrical signal, depending on the type and application
of the biosensor.

1.3. Types of Transducers

There are several types of transducers and they can be classified based on their
mechanism of action. Some of the types are listed below with examples.
Displacement Transducers: A displacement transducer measures distance or angle
traversed by the sensor. A device that produces movement in a straight line is a linear
displacement transducer and if it measures movement through an angle it is angular
displacement transducer. Displacement transducers are widely used in kinesiology, the
study of human movement.
Resistive Transducers: A resistive transducer contains a translational or angular
displacement sensing shaft. This shaft drives a wiper in the transduction element that
slides along the resistive layer. The resistance measured between the wiper and one of
the ends of the layer is a function of the position, motion, or displacement. Resistive
transducers are often used in kinesiological studies.
Inductive Transducers: Inductive transducers are used to sense small
displacements of movements and are based on the inductance L of a coil. This type of
transducer measures movement by changing the value of L, in proportion to the
movement of a ferrite or iron core in the bore of the coil assembly. L is given by the
following equation L = n2 x G μ, where n is the number of turns in the coil, x the
distance of the ferrite core within the coil assembly, G is the Geometric form constant,
and μ is the permeability of magnetically susceptible medium inside the coil. Inductive
transducers are often used to measure respiration, where the sensor is connected to a
flexible woven belt worn around the chest.
Capacitive Transducers: The sensing shaft in a capacitive transducer changes the
position of the dielectric between the capacitor's plates in the transduction element, or it
changes the distance and area between the plates. A change in these three parameters
leads to a change in capacitance, which is then measured.
Piezoelectric Transducers: Piezoelectric crystals (e.g. quartz) vibrate under the
influence of an electric field. The frequency of this oscillation depends on their
thickness and cut, each crystal having a characteristic resonance frequency.
Piezoelectric crystals can generate an electric field when subjected to physical pressure.
As a result, piezoelectric crystals can sense light pressures, such as touch. They are the
112 R.B. Reilly and T.C. Lee / Biosensors

basis of tactile sensors. They can also be used to provide vibrotactile stimulation in the
study of movement disorders and in force plates to measure ground reaction forces.
Transducers in temperature Measurement: The most popular method of
temperature measurement is by using a mercury thermometer, but this is slow, difficult
to read, not reliable and prone to contamination. In many cases, continuous monitoring
of temperature is required, as in an operating theatre or intensive care unit. Electronic
thermometers are convenient, reliable and hence used for continuous monitoring. They
use probes incorporating a thermistor or thermocouple sensor, which is an electronic
component that has rapid response characteristics of voltage with temperature.
Thermocouples measure the temperature difference between two points and not
absolute temperature.
Electrochemical Transducers: The underlying principle for this class of biosensors
is that many chemical reactions produce or consume ions or electrons which in turn
cause some change in the electrical properties of the solution. These electrical
properties can be measured.

2. Sensing and Instrumentation in Electrophysiology

Sensing electrical activity within the body is an important task in clinical diagnosis.
This section focuses on the physics underlying the acquisition of this electrical activity.

2.1. Electrodes

Sensing is typically achieved with a metal plate electrode, consisting of a metallic

conductor, often silver, in contact with the skin with an electrolyte gel in between.
These electrodes are often made using a metal foil for flexibility and sometimes in the
form of a suction electrode. Silver-silver-chloride (Ag-AgCl) electrodes have been
shown to have electrically superior characteristics to silver electrodes, especially when
recording low level AC and DC potentials. Chlorided silver electrodes present less low
frequency "noise" than silver electrodes. Such electrodes are employed in
electrocardiography. A metal disk electrode with a gold surface of conical shape is used
in electroencephalography. The gel is used to provide a good electrical contact between
the skin and the metal.
A hydrogel electrode is a film saturated with electrolytic solution and made up of
sticky materials placed on the electrode surface. The opposite side of the gel layer can
be connected to the skin. In this case, the electrolyte directly sticks to the body and
reduces motion artifact. This type of electrode is more typically used when the subject
is required to move or perform exercise. Hydrogel electrodes are typical in
electrocardiography and electromyography.

2.2. Basic Biosensor Amplifier:

The amplifier that is often used in conjunction with transducers is the operational
amplifier or Op-Amp [2]. An op-amp is a differential amplifier that amplifies the
difference between two inputs. One input has a positive effect on the output signal; the
other input has a negative effect on the output. The op-amp is powered by a dual
polarity power supply in the range of +/- 5 volts to +/- 15 volts. The theoretically
 R.B. Reilly and T.C. Lee / Biosensors 113

perfect op-Amp has an infinite voltage gain, theoretically being able to amplify a signal
to any level. It also theoretically has infinite bandwidth, being able to amplify any
frequency without distortion. It also theoretically has infinite input impedance, being
able to sense an input voltage level without distorting that voltage in any way. The
perfect Op-Amp also has zero-Ohm output impedance, allowing it to be connected to
any other electronic device without distortion. In reality such characteristics are not
met, but real op-amps do achieve great performance when correctly configured.

2.3. Inverting Amplifier

The inverting amplifier is one such configuration [2]. The op-amp is connected using
two resistors RA and RB such that the input signal is applied in series with RA and the
output is connected back to the inverting input through RB. The non-inverting input is
connected to a reference voltage, typically electrical ground or zero volts. This can be
achieved by connecting to the centre tap of the dual polarity power supply (Fig. 2a).
During operation, as the input signal becomes positive, the output will become
negative. The opposite is true as the input signal becomes negative. The amount of
voltage change at the output relative to the input depends on the ratio of the two
resistors RA and RB. As the input changes either positively or negatively, the output
will change in the opposite direction, so that the voltage at the inverting input remains
constant or zero volts in this case. If RA is 1K and RB is 10K and the input is +1 volt
then there will be 1 mA of current flowing through RA and the output will have to
move to -10 volts to supply the same current through RB and keep the voltage at the
inverting input at zero. The voltage gain in this case would be R B/RA or 10K/1K = 10.
Since the voltage at the inverting input is always zero, the input signal will see input
impedance equal to RA, or 1K in this case. For higher input impedances, both resistor
values can be increased.

2.4. Non-inverting Amplifier

Another configuration is the non-inverting amplifier, where the input signal is

connected to the non-inverting input and the input resistor RA is at electrical ground
(Fig. 2b). As the input signal changes either positively or negatively, the output will
follow in phase to maintain the inverting input at the same voltage as the input [2]. The
voltage gain is always greater than 1 and can be calculated as Vgain = 1+ RB/RA.

2.5. Voltage Follower

The voltage follower, also called an electrical buffer, provides high input impedance, a
low output impedance, and unity gain. As the input voltage changes, the output and
inverting input will change by an equal amount (Fig. 2c). The voltage follower is often
used to electrical isolate the transducer from a computer, to allow subsequent analysis
and signal processing to take place.
114 R.B. Reilly and T.C. Lee / Biosensors

Figure 2. (a) Inverting Amplifier

Figure 2. (b) Noninverting Amplifier

Figure 2. (c) Voltage Follower

2.6. Input Impedance

Impedance (symbol Z) is a measure of the overall opposition of a circuit to current. It is

like resistance, but it also takes into account the effects of capacitance and inductance.
Impedance is measured in Ohms. Impedance is more complex than resistance because
the effects of capacitance and inductance vary with the frequency of the current passing
through the circuit and thus implies that impedance varies with frequency. Input
impedance (ZIN) is the impedance resulting from the connection of any input to a
circuit or device (such as an amplifier). It is the combined effect of all the resistance,
capacitance and inductance connected to the input inside the circuit or device. The
effects of capacitance and inductance are generally most significant at high frequencies.
For good signal amplification, the input impedances should be high; at least ten times
the output impedance of the transducer supplying a signal to the input. This ensures
that the amplifier input will not overload the transducer producing the signal and thus
reduce the voltage of the signal by a substantial amount.
Human skin presents large impedance if connected via an electrode to an amplifier.
The impedance is greatly reduced if the surface layer of the skin is removed. This layer
is known as the stratum corneum. In practice, for electrocardiography or
electromyography, the stratum corneum is removed by lightly abrading the skin. The
skin contact with the electrolyte gel and silver-silver chloride electrode now presents
significantly lower impedance. For electroencephalography, input impedance at each
electrode site is kept below 5kOhms.
 R.B. Reilly and T.C. Lee / Biosensors 115

2.7. Instrumentation Amplifiers

An instrumentation amplifier is a type of differential amplifier that has been

specifically designed to have characteristics suitable for use in measurement and test
equipment. Its characteristics include very low DC offset, low electrical drift, low
noise, very high gain, very high common-mode rejection ratio, and very high input
impedances [1][2]. Instrumentation amplifiers are used where great accuracy and
stability of the circuit, both short- and long-term are required. The ideal common-mode
gain of an instrumentation amplifier is zero. Instrumentation amplifiers can be built
with individual op-amps and precision resistors, but are also available in integrated
circuit form.

2.8. Filtering

Filtering is a necessary processing step with any biosensor, in order to reduce noise to a
minimum and accentuate frequencies of interest. Filtering may be carried out in
software using digital signal processing, but it can also be accomplished in hardware
[3].

When designing a filter, the range of frequencies we wish to pass through the filter
without attenuation is known as the passband. The range of frequencies we wish to
attenuate is known as the stopband. The steepness of the filter’s transition from the
passband to the stopband, is related to the filter order. The higher the filter order, the
sharper the transition and more precise is the filter in pass and attenuating frequencies.
Three 2nd order filters are described here. A low pass, high pass, and bandpass
filter. Each of these filters will attenuate frequencies outside their passband at a rate of
12dB per octave equivalent to 25% of the voltage amplitude for each octave of
frequency increase or decrease outside the passband (Fig. 3(a) and Fig. 3(b) ).
First order low or high pass cutoff frequency is given by the choice of a resistor
and capacitance value.

1
FrequencyCutoff =
2πRC
(2)

2nd order low or high pass cutoff frequency (-3dB point) is also chosen by resistor
and capacitance values.

1
FrequencyCutoff =
2π R1R2C1C 2
(3)

A bandpass filter passes a range of frequencies while rejecting frequencies outside

the upper and lower limits of the passband Fig 3(c). The range of frequencies to be
passed is called the passband and extends from a point below the centre frequency to a
point above the centre frequency where the output voltage falls to approximately 70%
of the output voltage at the centre frequency. These two points are not equally spaced
above and below the centre frequency but will look equally spaced if plotted on a log
graph. The percentage change from the lower point to the centre will be the same as
116 R.B. Reilly and T.C. Lee / Biosensors

from the centre to the upper, but not the absolute amount. The filter bandwidth (BW) is
the difference between the upper and lower passband frequencies. A formula relating
the upper, lower, and center frequencies of the passband is

Centre Frequncy = LowerFrequency ∗ Upper Frequency

R1 C1

R1 R2

C1 C2
C2
R2

Figure 3. (a) High Pass Filter Figure 3. (b) Low Pass Filter

R1
C1

R1 R2
C1 C2

R2 C2

Figure 3. (c) Band Pass Filter

2.9. Electrical Safety with respect to electrophysiology:

Electrical safety is very important in hospitals as patients may be undergoing a

diagnostic or treatment procedure where the protective effect of dry skin is reduced.
Burn injuries can occur when using skin-electrode interfaces. This is specifically
relevant to electromyography or functional electrical stimulation (FES). In FES,
electrical current is injected into a muscle to initiate a muscle contraction. Burn
injuries can occur when there is small contact between the skin and the electrode. The
injury received from electric current depends on the magnitude of current, the pathway
that it takes through the body and the time for which it flows. As a result, physically
large electrodes, which provide a wide area for the current to pass are necessary in FES
to form a large area of the current to pass and reduce the risk of injury.
 R.B. Reilly and T.C. Lee / Biosensors 117

3. Examples of Biosensors in Use

This section provides an example of the use biosensors in clinical engineering, a branch
of biomedical engineering which is concerned with equipment routinely used in
hospitals and clinics [4].

3.1. Direct BP Measurements:

Direct blood pressure measurements are made by introducing a cannula (needle) into an
artery and then coupling this to a pressure transducer. The advantage of this approach is
that pressure is constantly monitored beat-by-beat and a clinically important waveform
(a graph of pressure against time) can be displayed. Patients with invasive arterial
monitoring require very close supervision, as there is a danger of severe bleeding if the
cannula line becomes disconnected. It is generally reserved for critically ill patients
where rapid variations in blood pressure are anticipated.

3.2. Indirect BP Measurements:

Indirect measurement involves use of an inflatable cuff coupled to a pressure gauge

(sphygmomanometer). In the standard sphygmomanometric approach, the cuff is
wrapped round the arm (placed at about heart level) and a pressure, higher than the
expected blood pressure, applied to the artery. A stethoscope placed over the brachial
artery is used to listen to the changes in sounds as the cuff is slowly deflated. The first
Korotkoff sounds occur when the systolic pressure, the highest pressure reached when
the ventricles contract and restrict blood flow, first exceeds the pressure in the cuff so
that blood once again flows through the artery beneath the stethoscope. At systolic
pressure a clear tapping sound is heard in time with the heart beat. The Korotkoff
sounds become muffled and disappear when the pressure in the cuff drops below the
diastolic pressure, the minimum pressure that occurs at the end of ventricular
relaxation.
Another indirect approach of measuring blood pressure is the oscillometric
method. Here a microprocessor periodically inflates and deflates the cuff. When the
blood breaks through the occlusion caused by the cuff, the walls of the artery begin to
vibrate slightly due to the turbulent nature of the blood flow. The onset of these
oscillations in pressure correlates with systolic pressure. Diastolic pressure corresponds
to the pressure when the amplitude of the oscillations drops to zero. It is the
oscillometric method that is employed in Holter monitors, which record blood pressure
over a 24-hour period while the patient-under-observation goes about their daily
activity.

3.3. Respiratory System Measurements (Spirometry)

Spirometry is the measurement of lung function and is the most common pulmonary
function test. It is an important tool used in assessing conditions such as asthma,
pulmonary fibrosis, cystic fibrosis, and chronic obstructive pulmonary disease (COPD).
There are many designs of spirometers, but most modern devices are electronic based.
They all involve the measurement of the volume and flow of breath during inhalation
and exhalation. The most common electronic based devices consist of a small handheld
118 R.B. Reilly and T.C. Lee / Biosensors

tube into which the patient exhales as hard as possible and for as long as possible,
having been just requested to take the deepest breath possible. A flow sensor in the
tube measures the direction and volume of the breath flow. Following exhalation, the
patient rapidly inhales (inspiration). Inspiration is important when assessing possible
upper airway obstruction and is sensed in the tube by the directional sensor.
Sometimes, the test will be preceded by a period of quiet breathing in and out from the
tube, so that sensor can measure tidal volume. During the test soft nose clips may be
used to prevent air escaping through the nose.

4. Conclusions

As can be appreciated, the majority of biosensors are essentially the same as sensors
used in other commercial or industrial applications. The distinctive aspect about
biosensors is their application. However, there are problems that are encountered by
biosensors that are unique to them. These problems are associated with the interface
between the biosensor and the biological system being measured. The presence of
foreign materials, especially implanted materials, can affect the biological environment
surrounding the sensor. Therefore one must be aware of potential physical or chemical
reactions as a result of the sensing process. The presence of the sensor may also hinder
the subject’s physical and cognitive ability. Therefore the placement of the sensor on
the body must be given due care and attention.
Despite these issues, biosensors have an enormous clinical impact both in
diagnosis and therapy, as well for being critical for data collection for biomedical
research.

References

[1] Geddes L.A. and Baker L.E., (1989), Principles of Applied Biomedical Instrumentation, Wiley-
Interscience, New York.
[2] Horowitz P. and Hill W., (1989), The Art of Electronics, Cambridge University Press.
[3] Wise D.L. (1991), Bioinstrumentation and Biosensor, Marcel Dekker, New York
[4] Harsanyi G. (2000), Sensors in Biomedical Applications: Fundamental Technology and Applications,
CRC, Florida.
Basic Engineering for Medics and Biologists 119
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-119

Introduction to
Chapter III: Medical
Informatics for Biomedical
Engineering
Paul McCULLAGH and T. Clive LEE (eds.)

Medical Informatics comprises the theoretical and practical aspects of information

processing and communication, based on knowledge and experience derived from
processes in medicine and healthcare.
Chapter III.1 will introduce the student to the range of computing devices, their
operating systems and programming environment. Most devices no longer work in
isolation so the option for networking will be explored. When data is moved from one
system to another across a network security issues such as privacy, authentication, and
data integrity are important. A Case study will illustrate the role of ICT and the
Internet in the rehabilitation of stroke patients.
Data should be structured for storage and retrieval, as indicated in Chapter III.2.
The database is the fundamental storage repository. Data may be stored as free text, and
in image format, but coding is important to aid analysis and audit. The chapter will
discuss the most important formats for coding to assist classification. Unambiguous
description and standards are important. To date interoperability has been most
advanced in ECG equipment. A Case study on a markup language to describe the
electrocardiogram (ecgML) will be presented.
As the amount of data becomes ever larger, mining techniques are required to
extract relevant information from the database. A goal of medical informatics is that
‘knowledge’ can be deduced which will aid in decision support. Chapter III.3 explores
these topics and discusses a Case Study which uses a machine learning approach to
discover the optimal placing of electrodes for the recording of the ECG.
Advances in Information and Communication technology (ICT) have resulted in
the use of the Internet to support medical applications (eHealth) along with the use of
high speed communication links to support Telemedicine and Telecare. This is
particularly important when large distances separate the healthcare professional and the
patient along with the provision of home based assistive services. Chapter III.4
illustrates these issues by using a Case Study to support the home based management
and dispensing of medication.
This page intentionally left blank
Basic Engineering for Medics and Biologists 121
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-121

III.1. Medical Informatics and eHealth

Paul J. McCULLAGHa, Huiru ZHENGa, Norman D. BLACKa, Richard DAVIESa,
Sue MAWSONb and Kieran McGLADEc
a
Faculty of Engineering, University of Ulster at Jordanstown,Co. Antrim, N Ireland
b
Centre for Health and Social Care Research, Sheffield Hallam University, UK
c
Department of General Practice, Queens University of Belfast, N Ireland

Introduction

Medical Informatics combines the disciplines of medicine and computing, and

knowledge of this area is a fundamental requirement for a student of biomedical
engineering, providing support for software development, information gathering and
critical assessment, and evidence based medicine [1]. This review assumes the
following definition:
Medical Informatics comprises the theoretical and practical aspects of information
processing and communication, based on knowledge and experience derived from
processes in medicine and healthcare [2].
One often hears the term “eHealth”. This is a related area, usually understood to
apply to direct applications of medical informatics to health care. It often involves use
of the Internet and includes the activities of patients and clients as much as health care
professionals. We will use the following definition:
“eHealth” is concerned with the application of electronic information systems in
the organization of and access to health care including such examples as booking and
referral systems; it covers integrated information tools which allow secure access to
personal health data for all those who need it to deliver optimal healthcare; and it
includes complex clinical applications which can support the clinician in diagnosis and
treatments and ultimately supports the citizen in their own environment [3].
Essentially eHealth places the citizen at the centre of their own well being and
care. In this review we will introduce computing devices, their architecture, operating
systems and programming environments. As most devices no longer work in isolation,
the options for connectivity and networking will be explored. When data are moved
from one system to another across a network, security issues such as confidentiality,
incorporating preservation of privacy, authentication and data integrity are important.
An eHealth case study will illustrate the role of Information and Communication
Technology (ICT) and the Internet, utilizing a biomedical engineering solution to assist
the rehabilitation of stroke patients.

1. Role of the Computer in Medical Informatics and eHealth

The computer is ubiquitous in everyday-life and is an essential tool for the biomedical
engineer. It is fundamental to the acquisition, storage, processing and analysis of data,
122 P.J. McCullagh et al. / Medical Informatics and eHealth

and can be used to summarize and display information via the human computer
interface (HCI) so that it is comprehensible to users. This interface is normally
graphical nowadays and can be very sophisticated, e.g. 3-dimensional image
construction of the fetus in the womb, by processing reflections of externally applied
ultrasound, providing the means for clinical assessment. Indeed processing capacity is
such that images can be constructed and displayed in real-time to provide a non-
invasive ‘movie’ of the fetus [4]. Alternatively, a more straightforward example is the
use of automated speech recognition to allow a specialist to record clinical observations
directly into the computer. An example of this is the reporting of diagnostic imaging
results in radiology [5].
The ability to be connected to private networks and the Internet is essential for
information sharing and for information retrieval from large authoritative information
repositories. The Electronic Patient Record (EPR), which is now commonplace in
Primary Care, is an example of the former, with many additional, often incompatible
systems existing in Secondary Care departments of large hospitals. In the United
Kingdom, a key aim of the government's Connecting for Health program [6] was to
establish a central database of 50 million patients' medical records, accessible over
NHSnet1. This is the world’s largest Information Technology development program,
with an estimated cost from six to twenty billion pounds. The United States National
Library of Medicine’s publicly accessible database PUBMED [7, 8] with over 16
million citations from MEDLINE is an example of an Internet based resource. Others
include large data banks such as the human genome project.
Van Bemmel has provided an important conceptual model for structuring
computer applications in health [9]. It comprises six levels, with the lowest level being
largely automated and the higher levels requiring progressively more human
interaction. This model provides a useful context in which to discuss the interaction of
computers, networks and eHealth. We will use the recording, storage and interpretation
of the electrocardiogram (ECG), an electrical recording used to assess the physiology
of the heart, as an example.
Level 1: Communication and telematics deals with data acquisition and transfer,
e.g. the recording of the ECG potential waveform with electrodes and physiological
amplifier and its subsequent transfer, in a standardized format, to a computer screen for
human interpretation. The computer can be thousands of miles distant, a capability
which has promoted the emergence of telemedicine as a discipline.
Level 2: Storage and retrieval deals with the storage of information in a data
repository, e.g. multimedia EPR which capitalizes upon advances in memory storage
capacity and database technology to form digital libraries comprising labeled text,
physiological recordings and images.
Level 3: Processing requires a computer program to extract information from the
captured data. For an ECG, this could be important diagnostic parameters such as heart
rate, or pattern abnormalities such as ST elevation2, which may have some diagnostic
significance. Similarly for an image (e.g. an echocardiogram), the computer can be
used to segment boundaries and highlight potential abnormalities, for human
verification.
Level 4: Diagnosis combines features extracted from the processing of acquired
data with existing medical knowledge to assist with decision making, e.g. the ‘ECG is

1
NHSnet is a private network connecting hospitals and general practitioner sites in the UK
2
The well know ECG complex is designated by QRST markers
 P.J. McCullagh et al. / Medical Informatics and eHealth 123

abnormal’ or ‘ECG indicates an arrhythmia’ and some therapy is required. A Decision

Support System (DSS) includes an knowledge-base (KB) containing information such
as statistical norms for sex and age. It can provide alerts, critiques and advice. The DSS
is a useful tool to support a diagnosis. However, the treating physician may be aware of
many additional case-specific factors, such as previous illness and medication, which
may have significant influence far beyond the current knowledge in the KB. The
physician always makes the final diagnosis. It is possible to conclude that a physician’s
diagnosis may be as much an art as a science.
Level 5: Treatment utilizes the information derived from the diagnosis to formulate
an intervention, e.g. fit an on-demand pacemaker, which will stimulate the patient’s
heart when the cardiac beat becomes irregular. A computer algorithm will form part of
the embedded control system.
Level 6: Research allows users to interact with data to formulate hypotheses and
build computer models, e.g. to simulate the electric depolarization of the heart, which
can then be used to promote understanding of normal and abnormal function. The
possibilities for research are only limited by prevailing knowledge, tools and the
ingenuity of the human mind.

1.1. Architecture of the Computer

Logically, the computer comprises three elements: (i) a processor (control unit and
arithmetic and logic unit (ALU)), (ii) random access memory, some form of persistent
storage, currently a hard disk (memory) and (iii) input/output devices. This is the
classical von Neumann architecture [10] or "stored-program computer", which has
been in use for decades and forms the design of most of today’s computers. Parallel
architectures are also in use, particularly for high performance applications. At the
heart of a computer is an algorithm, which is a set of programmed instructions to carry
out some task. The design uses a single store to hold both instructions and data, and the
control unit instructs the ALU to execute instructions sequentially.
Any hardware specification quickly becomes outdated due to the pace of
technological progress. This can be summed up by Moore's Law (proposed by Gordon
E. Moore a co-founder of Intel, the major silicon chip manufacturer) which observes
that the transistor density of integrated circuits, with respect to minimum component
cost, doubles every 24 months [11]. A similar law has held for hard disk storage cost
per unit of information, and indeed random access memory (RAM) storage capacity
has increased at the same rate as processing power. Kurzweil’s expansion of Moore's
Law [12] shows that the underlying trend has held true from integrated circuits to
earlier transistors, vacuum tubes, relays and electromechanical computers. He projects
continued progress until 2019 with transistors just a few atoms in width, and predicts
that it is possible that Moore's Law may be extended by a paradigm shift into new
technology, e.g. quantum computers.
The computer has many different incarnations: supercomputer, grid computer,
mainframe, server, desktop, laptop, personal data assistant, smart phone and embedded
device. A summary of characteristics is provided in Table 1.
124 P.J. McCullagh et al. / Medical Informatics and eHealth

Table 1: Classifying computers by scale and application

Computer Description Example Application Possible

type Operating
System (OS)
Supercomputer High power number Modelling (computationally Proprietary,
cruncher using an array intensive), simulation, research e.g. Unix, or Linux
of powerful processors identifying genetic disposition for
various disease from the human
genome
Grid Virtual cluster of Modelling (computationally Heterogeneous
desktop computers for intensive) or access to information or operating
sharing resources computational resources systems, linked
by ‘open’
middleware using
standards
Mainframe Large central computer, Administration and storage of data, UNIX, VMS or
accessed by ‘terminals’ Hospital Information System, EPR, other proprietary
Patient Administration System, systems
Picture archive and retrieval
Server High performance Web, Database, email, etc UNIX, Window
computer NT
Desktop, General computer for Multi-purpose, office, web access, Windows, Linux,
Workstation personal use and storage email, program development. Apple
of information; Workstation suited to image
workstation may have processing applications e.g.
enhanced graphics Radiology department
Laptop, Tablet Similar specification to Multi-purpose, as above, normally Windows, Linux,
PC desktop but with with built in connectivity Apple
portability the main
requirement
Personal data Computer in your hand Cut down versions of office Windows CE,
assistant programs, specialized ‘calculator’ Palm OS
programs, normally with built in
connectivity.
Smart phone Telephony, with storage Organizer, small programs, with Symbian OS
and limited processing connectivity to the cellular phone
(e.g. games) systems (GSM, GPRS, 3G)
Embedded Dedicated control of ECG monitor / defibrillator. Bespoke Windows CE or
device hardware software development Linux

The supercomputer and grid perform specialized tasks, typically in the research
world, where enormous processing is required. For example, supercomputers have
been used to model avian flu progression by scientists from Los Alamos National
Laboratory [13], and to model HIV's vulnerability to a class of drugs known as protease
inhibitors [14] by researchers at Stony Brook University. The former application was
implemented on a platform developed for the nuclear weapons programme comprising
2048 processors, the largest cluster in the world. The latter simulation took 20,000
hours of processing (about three months) on the USA National Center for
 P.J. McCullagh et al. / Medical Informatics and eHealth 125

Supercomputing Applications. However, it would have taken more than a year to

complete the work on a conventional computing system.
Grid computing [15] involves sharing heterogeneous resources (based on different
platforms, hardware/software architectures, and computer languages), located in
different places belonging to different administrative domains over a network using
open standards. This offers a model for solving huge computational problems by
making use of the resources of large numbers of computers, treated as a virtual cluster
embedded in a distributed telecommunications infrastructure. It has the goal of solving
problems too big for any single supercomputer, whilst retaining the flexibility to work
on multiple smaller problems. Secure authorization techniques allow remote users to
control computing resources. The availability of large amounts of data (clinical,
genomic) in heterogeneous sources and formats, and the progress in fields such as
computer based drug design, medical imaging and medical simulations have lead to a
growing demand for large computational power and easy accessibility to data sources.
Functional imaging research in schizophrenia [16] is an example application,
integrating of large numbers of medical images, with patient’s medical and biological
data, to assess genetic pre-disposition to the condition.
In order to help combat malaria, which kills more than 1 million people annually,
CERN (Conseil Européen pour la Recherche Nucléaire, the European Organization for
Nuclear Research) has launched a grid computing effort to run a simulation program
called MalariaControl.net [17], developed by researchers at the Swiss Tropical
Institute. The program simulates how malaria spreads through Africa, enabling
researchers to better understand the impact of introducing new treatments. On a test
phase of a few months with 500 volunteers, the grid was able to run simulations
equivalent to 150 years of processing time on a single computer.
Mainframes are enterprise computers with high specification and associated high
cost. In the context of health they are used for large Hospital Information Systems to
store patient administration systems and more recently the EPR, with technical support
from an Information Systems department.
By comparison to the previous systems, a server is a smaller computer system with
a dedicated function, such as a database, email or web servicing. The hardware and
software specification tends to be more demanding than the conventional desktop, as it
may need to serve multiple simultaneous connections, and secure data backup is often a
requirement. The workstation can be a dedicated high specification desktop with for
example high specification graphics, as required for imaging applications. The desktop
provides cost effective processing power with significant storage and easy connectivity
for stationary users. The system unit typically interfaces to a keyboard, mouse and
screen display. Printers and scanners are common peripherals and high volume backup
may be achieved by compact disk or digital versatile disk rewriter. The floppy disk
drive has been replaced by a memory stick (or card) with capacity in Giga (10 9) bytes
for file transfer, ad-hoc storage and backup3. The interface options to external devices
are typically Universal Serial Bus (USB2.0), with firewire (IEEE 1394 High Speed
Serial Bus) and memory card readers as alternatives. Local area network (LAN)
connectivity to a fixed wired Ethernet typically provides a data rate of between 10-

3
Terms for storage: Terabytes (1012 bytes ), Gigabytes (109 bytes), Megabytes (106 bytes), Kilobytes (103
bytes),
126 P.J. McCullagh et al. / Medical Informatics and eHealth

1000 Mbps (106 or million bits per second). While this provides adequate quality of
service for medium sized networks (up to say 30 users), the growing demand for
multimedia applications (including the transfer of sound, images and movies, all of
which are required in the health domain) continues to challenge the performance of the
network.
Mini desktop PCs are beginning to emerge. These have smaller footprints, and
significantly lower energy requirements than the conventional tower or desktop.
Average Power Consumption of only 29Watts, is approximately 30% less energy than
a standard PC. This is important in the reduction of the carbon footprint of computer
technology. Virtualisation lets one larger computer do the job of multiple computers,
by sharing the resources of a single computer across a network. A small energy
efficient ‘thin client’ (no hard disk and minimal peripherals) is required on the desktop.
In addition to energy savings and lower cost due to more efficient use of hardware,
virtualisation provides better desktop management, increased mobility and security,
and improved disaster recovery.
The home computer sacrifices little functionality by comparison to the desktop.
The network connection was initially via a telephone dial-up (Public Switched
Telephone Network) to an Internet Service Provider (ISP) at 56 kbps (10 3 or thousand
bits per second) but increasing is via broadband (Asymmetric Digital Subscriber Line
or cable television) at speeds typically of 1-10 Mbps for download, and 512 kbps for
upload. The speeds have increased as sophisticated signal processing yields more bits
per second from the available bandwidth.
For mobility, a number of options are available. The laptop comprises practically
all the functionality of the desktop, but with a slightly inferior specification-price ratio.
Wireless Bluetooth using radio waves or infrared light (The Infrared Data Association,
IrDA) provide unfettered desktop connectivity. Both wired Ethernet and wireless
Ethernet (known as wireless fidelity or WiFi) are normally supported. Wireless offers
nominal speeds of the order of 11-54 Mbps, depending on the characteristics of the
network access point, and the wireless interface built into the laptop. This is inferior to
the wired network speed of typically 100 Mbps. Reliability of connection is probably a
more important issue for the user, with the wired network providing better service. The
tablet PC is a variant of the laptop, with automatic handwriting, a standard input
feature. A major issue with portable computing is the capacity of the battery, typically
supporting a couple of hours of unfettered computing and hence a major design issue is
the conservation of power. This has spawned research into harnessing power from the
environment, e.g. from kinetic energy or solar power.
The personal data assistant (PDA) provides computing power which can be stored
in the pocket. This is very appealing in the area of medicine, e.g. a ward round in a
hospital where even a lightweight laptop can be obtrusive. The PDA comprises similar
connectivity interfaces to the laptop and supports similar applications to desktop such
as word processing, spreadsheet, database, calendar, Internet browsing, and email,
normally with limited functionality. However, due to the small screen size and small
keypad, usability is the major difficulty. Some models have virtual keypads, others
have small keyboards or incorporate speech recognition. Many medical applications
have been developed for the PDA, including dictionaries, calculators, guidelines and
databases, see [18].
In many respects, from technology and user perspectives, the PDA and the smart
phone have converged. The smart phone primarily provides telephony within a Global
System for Mobile communication (GSM) cell (IS-95 in USA). However the use of
 P.J. McCullagh et al. / Medical Informatics and eHealth 127

data services such as Short Message Service (SMS), and Internet access using Wireless
Access Protocol (WAP) or I-Mode4 using General Packet Radio Service (GPRS) are in
widespread use. The telephone companies have introduced third generation (3G)
cellular systems based on Universal Mobile Telecommunications System (UMTS),
with data rates of 384 kbps and beyond, which can support video conferencing and
hence basic telemedicine. Further technological enhancements offering performance
approaching LAN speeds, e.g. High Speed Packet Download Access (HSPDA) are
under development (download speed of 3.6 Mbps with 14.4 Mbps planned for the
future). The smart phone also contains a processor and can be programmed for bespoke
applications. By enhancing a PDA with GPRS connectivity, the PDA essentially
becomes a telephone. Similar convergence from ICT means that a computer, laptop or
PDA connected to a ‘hot-spot’ can utilize voice over internet protocol (VOIP) to
provide telephony or ‘chat’ services.
Of course, embedded devices with microprocessor controllers are also ubiquitous
in biomedical engineering. These can utilize advances in miniaturization to perform
dedicated functions, under the control of a program. An example is an ECG monitor,
which uses purposely designed hardware and software to acquire and display the
electrocardiogram. If a new clinical measure is required, then an updated program can
be downloaded from a web site to the device firmware. Semi-automated ECG
defibrillation systems rely more heavily on embedded software for decision support.

1.2. Operating Systems

An operating system (OS), is a computer program that manages the hardware and
software resources of a computer. The OS performs basic tasks such as controlling and
allocating memory, prioritizing system requests, controlling input and output devices,
facilitating networking, and managing files. It may provide a graphical user interface.
Microsoft Windows is the dominant OS on the PC and Apple Macintosh platforms,
with 90% market share. The main competition comes from Open Source5 protagonists
who support Unix-variants such as Linux. On larger mainframe computers Unix and
proprietary systems such as OpenVMS dominate. The latest Windows OS, Windows 7,
includes added functionality in security and network administration.
The most common operating systems used by smart phones are Symbian, Linux,
Windows Mobile, and Palm OS. Windows Compact Edition (CE) is an OS for minimal
computers, such as PDAs. It is optimized for devices that have minimal storage and
may run in under a megabyte of memory. Embedded systems normally use dedicated
operating systems and versions of Linux.
Programming environments for the smart phone and PDA are Microsoft.NET
Compact Framework and Java6. For the former, applications may be authored in
languages such as C# or Visual Basic.NET. For the latter, Java programs (know as
MIDlets) may be authored for embedded devices.

4
NTT DoCoMo's I-Mode is a wireless Internet service made popular in Japan, providing access to a number
of consumer sites
5
Open principles and practices are applied to the development of source code that is made available for
public collaboration, and released as open-source software
6
Java is an open source programming language from Sun Microsystems
128 P.J. McCullagh et al. / Medical Informatics and eHealth

1.3. Client-Server Computing

Client/server is a network-based architecture, see Figure 1, which separates the client

computer (often an application that uses a browser) from the server computer. It uses
the HyperText Transfer Protocol (HTTP), using methods to publish and retrieve HTML
pages. The destination server, which stores or creates resources such as HTML files
and images, is called the origin server. An HTTP client initiates a request by
establishing a Transmission Control Protocol (TCP) connection to a particular address
(called a port) on a remote host (the address is port 80 by default). An HTTP server
listening on that port waits for the client to send a request message. Upon receiving the
request, the server sends back a confirmation, and a message of its own, the body of
which is perhaps the requested file, an error message, or some other information.
Resources to be accessed by HTTP are identified using Uniform Resource Identifiers
(URIs).
Each instance of the client software can send requests to a server, e.g. a file server,
terminal server, or mail server. Examples of servers programs are Apache and
Windows Internet Information Services (IIS). The Apache Server Project [19]
maintains an open-source HTTP server for modern OSs, including UNIX and Windows
NT. IIS provides similar services, but restricted to the Microsoft platform.
The three-tier architecture separates the user interface, functional process logic,
and data storage/access. The programs are developed and maintained as independent
modules (possibly on separate platforms), so that any of the tiers may be upgraded or
replaced independently as requirements or technology change. Typically, the user
interface runs on a desktop PC or workstation and uses a standard graphical user
interface. Functional logic may consist of one or more separate modules running on an
application server, and a Relational Database Management System (RDBMS) on a
database server contains the data storage logic. Structured Query Language (SQL) is
used to create, modify, retrieve and manipulate data from relational database
management systems.
Interoperability is important when interacting with heterogeneous platforms. A
Web service is a software system designed to support interoperable machine-to-
machine interaction over a network, using a distributed function (or method) call
interface. Simple Object Access Protocol (SOAP) is a protocol for exchanging
(extensible markup language) XML-based messages over the computer network,
normally using the HTTP protocol.
Whilst client-server is the dominant architecture, peer to peer systems have
become more popular for file sharing and ‘chat’ applications on the internet. In this
case, files may be transferred without the involvement of a server. Often a server-based
database holds the location of information, e.g. music files, but not the actual files,
permitting a subsequent peer to peer connection for transfer.

1.4. Computer Networks

Networks may be categorized by scale as: personal area networks, local area networks
and wide area networks.
A personal area network (PAN) is a computer network used for communication
among computer devices (including telephones and personal digital assistants) a few
meters (normally 10m but up to 100m is possible) apart. PANs can be used for
communication among the personal devices themselves, or for connecting to a higher
 P.J. McCullagh et al. / Medical Informatics and eHealth 129

level network and the Internet. Personal area networks may be wired (USB and
FireWire) or wireless (WPAN) using technologies such as IrDA and Bluetooth. ZigBee
is the name of a specification for a suite of high level communication protocols using
small, low-power digital radios based on the IEEE 802.15.4 standard for wireless
personal area networks. ZigBee operates in the Industrial, Scientific and Medical (ISM)
radio bands; 868 MHz in Europe, 915 MHz in the USA and 2.4 GHz worldwide. The
technology is simpler and cheaper than Bluetooth.

Request for:
http://www.esem.org/index.html
Client Server
PC PC
Document index.html
returned to client

Figure 1: Example of Client Server Interaction

A local area network (LAN) is a computer network normally covering an office,

but it is increasingly used in the home. Current LANs are most likely to be based on
switched IEEE 802.3 Ethernet running at 100 or 1,000 Mbit/s or on WiFi technology.
The defining characteristics of LANs in contrast to WANs (wide area networks) are:
their much higher data rates; smaller geographic range; and that they do not require
leased telecommunication lines.
A wide area network (WAN) covers a broad geographical area, e.g a country.
WANs are used to connect local area networks (LANs) together, so that users and
computers in one location can communicate with users and computers in other
locations. Many WANs are built for one particular organization and are private. Others,
built by Internet service providers, provide connections from an organization's LAN to
the Internet. WANs are most often built using telephone lines. At each end of the
130 P.J. McCullagh et al. / Medical Informatics and eHealth

leased line, a router connects to the LAN on one side and a hub within the WAN on the
other. Network protocols including TCP/IP deliver transport and addressing functions.
The Internet is the worldwide, publicly accessible network of interconnected
computer networks that transmit data by packet switching using the standard Internet
Protocol (IP). It is a "network of networks" that consists of home, academic, business,
and government networks, which together carry various information and services, such
as electronic mail, online chat, file transfer, and the interlinked Web pages.

1.5. Security

In eHealth, sensitive confidential information about a person’s health status, may be

stored in computer systems, or transferred across computer networks. Security deals
with the unauthorized access to computers, networks or data. Basic security requires
that access to a computer system is controlled by username and password, preventing
unauthorized access to programs and data. Enhanced access control measures include
the use of a smart card and card reader and the use of biometric data such as fingerprint
scanning.
Encryption is the process of obscuring information to make it unreadable without
special knowledge of the algorithm used and the key to unlock it. Encryption is now
employed in protecting widely-used systems, such as Hospital records, Internet e-
commerce, mobile telephone networks and bank automatic teller machines. Encryption
can use either secret key algorithms such as Data Encryption Standard (DES) and
Advanced Encryption Standard (AES) or public/private key algorithms such as RSA,
named after its inventors Rivest, Sahmir and Adelman. Encryption can be used to
ensure secrecy, but other techniques are still needed to make communications secure,
particularly to verify the integrity and authenticity of a message; for example, a
message authentication code (MAC) or digital signature. In computer security,
authentication is the process of attempting to verify the digital identity of the sender of
a communication such as a request to log in. The sender being authenticated may be a
person using a computer, a computer itself or a computer program. Authentication
verifies a person's identity, while authorization is the process of verifying that a known
person has the authority to perform a certain operation. Authentication, therefore, must
precede authorization.
Security measures include the use of firewall technology to restrict access to
computer systems, except via an ‘electronic drawbridge’. This permits control of
accepted programs, ports and IP addresses. Anti-virus software is also important to
prevent malicious infiltration of and attack on computer systems. Good security also
dictates that users adhere to strict guidelines. For example wireless networks are
normally considered to be less secure than wired equivalents and may permit the
opportunity to by-pass the secure firewall. Hence Wired-Equivalent Privacy (WEP)
should be utilized on wireless devices. This security discussion is probably equally
relevant to the home and office environments.

1.6. eHealth

eHealth is health care practice which is supported by electronic processes and

communication. eHealth can use advances in ICT to empower citizens and patients,
enabling them to play an active role in decisions relating to their health. The European
Union has launched a portal [20] to provide citizens with access to comprehensive
 P.J. McCullagh et al. / Medical Informatics and eHealth 131

information on Public Health initiatives, to positively influence behaviour and promote

the steady improvement of public health in Europe. Possible services are summarized
in Table 2. eHealth may be particularly appropriate to the management of chronic
disease in an era when the population is ageing and the costs of providing support are
becoming prohibitive.

Table 2: eHealth examples

eHealth Component Explanation

Electronic Medical Records Communication of data between patient and healthcare
professionals e.g. GP, specialists, care team, pharmacist
Telemedicine Physical and psychological measurements that do not require
a patient to travel to a specialist
Evidence Based Medicine Healthcare professional can look up whether his/her diagnosis
and treatment are in line with current scientific research.
Citizen-oriented Information Provides both healthy individuals and patients with
information on medical topics
Specialist-oriented Information Overview of latest medical journals, best practice guidelines
Provision or epidemiological tracking
Virtual healthcare Healthcare professionals who collaborate and share
information on patients through digital equipment

2. CASE Study: SMART Rehabilitation: Tele-Rehabilitation and eHealth

We will use the following case study to emphasize the importance of computers and
networking to the eHealth paradigm and relate this to the van Bemmel conceptual layer
model [2], introduced in section 2.

2.1. Stroke and Rehabilitation

In the United Kingdom, stroke is the most significant cause of severe disability
affecting a quarter of a million people [21]. At six months post-stroke, around half of
patients need help with bathing, a third need help with dressing and a third need help
with feeding [22]. Rehabilitation is primarily concerned with maximising the functional
and cognitive ability of the patient combating social exclusion and increasing patient
independence [22]. The National Framework for Older People recommends that
rehabilitation should continue until maximum recovery has been achieved [24].
Research suggests that intensive, task specific and repetitive training may be necessary
to modify neural organisation [25]. However, due to cost factors, in-patient length of
stay is decreasing and outpatient rehabilitation facilities are limited. Furthermore,
current rehabilitation techniques used in hospitals and rehabilitation centres require
access to specialised equipment and a dedicated laboratory set-up [26]. Therefore there
is a strong argument for the need to develop a low-cost, accessible system that can
132 P.J. McCullagh et al. / Medical Informatics and eHealth

augment existing rehabilitation services for post-stroke patients. Currently, two main
approaches have been explored, namely tele-rehabilitation and home-based
rehabilitation.
Tele-rehabilitation, was firstly proposed in 1997 by the National Institute on
Disability and Rehabilitation Research (United States Department of Education), it is
defined as:
the use of communication and information technologies to deliver rehabilitation
services and exchange information over geographical distances [27].

The advantages of rehabilitation in a home environment are [28]:

• Enhanced evaluation of daily living activities, which improves outcome;
• More frequent and timely treatment that would otherwise burden health care
systems;
• Patients can stay at home for extended periods, reducing costs;
• Feedback can enhance user motivation;
• Empowerment of the user by enabling them to make decisions about the time
and frequency of their intervention;
• Chronic patients can benefit from regular and frequent intervention.

SMART rehabilitation: technological applications for use in the home with stroke
patients, examined the scope, effectiveness and appropriateness of eHealth to support
home-based rehabilitation for older people and their carers [29]. It utilized Information
and Communication Technology and motion sensors to provide tele-rehabilitation
alongside home-based rehabilitation.
The design and development phases were guided by an expert panel of national
and international academics. In addition an expert group of stroke patients and their
carers contributed to developments by way of focus groups and consultations [30].
Throughout the development, neurological physiotherapists engaged with decisions
about the interface and the feedback given to the user. A biomedical engineering design
team was involved in the development of the sensor attachment to clothing ensuring
that the clothing was acceptable and usable by the stroke patients.

2.2. SMART Rehabilitation System Overview

The SMART rehabilitation system employs an activity monitoring system linked to a

decision support platform that provides therapeutic instruction, supports the
rehabilitation process and monitors the effectiveness of rehabilitation interventions on
patient function (Level 5). Information relating to this process is fed back to patients,
their carers and/or health care professionals.
The system consists of three primary components; a motion tracking unit, a
computer base station and a web-server shown in Figure 2. The motion tracking unit
provides the ability to monitor a patient’s upper limb during exercise such as reaching
and drinking. The motion information, which includes both position and rotation, is
transmitted via Bluetooth to a base station (Level 1) for visualisation and analysis.
 P.J. McCullagh et al. / Medical Informatics and eHealth 133

PC

Figure 2: SMART System Architecture

(Level 3). The decision support platform (Level 4) operating at the base station provides
various types of feedback to user, such as movement display, exercise history and
variable measurement results. A personal computer and touch screen monitor make up
the base station which is connected to a central server providing web services to the
user. Patients and their carers can view their activities in a three dimensional (3D)
environment on the ICT platform, and can communicate with their therapists using the
web services (Level 1).

2.3. Motion Tracking Wnit

Various technologies have been applied in the area of motion tracking and
rehabilitation:
• Switches: pedometer and functional electrical stimulation (FES);
• Video systems;
• Gyroscopes (angle / rate of turn);
• Accelerometers (velocity);
• Robot arms.
In the SMART project, inertial sensors (Xsens Dynamics Technologies,
Netherlands, known as MT9) were used to track patients’ upper limb movements. The
MT9 sensor consists of a three axis accelerometer, a three axis gyroscope, and a three
axis magnetic field sensor, which can be used to measure rotation rates and linear
134 P.J. McCullagh et al. / Medical Informatics and eHealth

accelerations. For a more comprehensive review of accelerometers and general motion

sensing techniques see [31].
As illustrated in Figure 3, two MT9 sensors may be attached to a patient’s upper
limb via wearable harnesses to record movement during daily activities. The MT9s
record movement information, such as positional and angular data relating to upper
limb joints - the wrist, elbow and shoulder. The data are then transmitted wirelessly to
the base station via a digital box called the XBus which is worn on a belt around the
patient’s waist. The tracking unit utilises sensor fusion and optimisation techniques and
is implemented using Visual C++, based on a PC with a 1.2GHz CPU (Level 2).
Motion data are generated continuously with a sampling rate of 25Hz.

Figure 3: Accelerometer based motion tracking unit in Smart rehabilitation system with embedded
processing

2.4. ICT Decision Support Platform

The ICT platform consists primarily of five modules: database, interface, decision
support, communication and a feedback module.
The database module (Level 2) maintains information such as system information,
personal details, rehabilitation movement parameters; comments and instructions from
healthcare professionals and movement analysis. The system information stores a series
of questions designed to ascertain the patient’s ability to safely complete the daily
exercises. A calendar service stores a patient’s entire history of rehabilitation. The
database module also provides access to relevant personal information on the patient.
The design of the interface module was largely informed by the fact that users
would interact with the software via a touch screen display. Its aim is to provide easily
accessible and understandable menus. One such menu is illustrated in Figure 4
allowing the user to start an exercise, view their calendar and diary and obtain some
feedback on their rehabilitation progress.
The decision support module (Level 4) analyses the movement data and provides
key outcome variables relating to physical performance such as length of reach, elbow
angle. The communication module manages the transfer of information between the
base station and the central server. The feedback module provides different types of
 P.J. McCullagh et al. / Medical Informatics and eHealth 135

information to patients, namely 3D movement information, comments and instructions,

and analysis results presented as a 3D visualisation.

Figure 4: Welcome screen after user login. Menus are designed using large icons to accommodate for touch
screen technology, and promote usability.

The visualisation replays the movement of rehabilitation exercises to users in a 3D

environment using a virtual body and arm. Stored movement templates can be overlaid
or mirrored on the screen as a reference to help the patient improve the exercise. The
target movement template is personalised and adaptive, and is selected by the therapist
according to the patient’s rehabilitation progress (Level 5). Figure 5 displays an upper-
limb arm movement during a reach exercise.

2.5. Web-Based System

All the SMART home-based systems are linked to a central server via the public
Internet, using a virtual private network. Patients’ movement data and comments or
queries can be sent to the central server. A therapist can log into the central server to
view his/her patients’ movement data and their comments, provide instructions after
analysing their exercises and provide feedback on their queries.
There are three processes that run on the server to manage the system; MySQL
database server to control the storage of data, Apache TomCat web server (see section
2.3) to deal with Internet traffic and Matlab to provide a tool for analysing the
movement data (Level 3).
136 P.J. McCullagh et al. / Medical Informatics and eHealth

Figure 5: Software screen shot of 3D rendering providing personalized feedback to promote rehabilitation

2.6. User centered Design Strategy

The user groups involved in the rehabilitation system are post-stroke patients, their
carers, and therapists. The system must be as simple as possible to use, and adaptable
to individual needs. Special care must be taken with people who have had a stroke as
they have complex impairments often incorporating cognitive difficulties such as
problems with perception, attention, information processing, language and memory. In
order to provide a user friendly rehabilitation system, the user-centered design strategy
was applied throughout the projects development.
In the early stages of the project, focus groups were held with patients and
healthcare professionals to ensure that proposed technical solutions, methodology and
outputs were acceptable. In the later stages, a group of expert users provided feedback
on key aspects of the system such as user interface, type of feedback and computer
interface. These data were collected by qualitative researchers, summarised and fed
back to the engineering teams.

3. Summary

In this review, we have examined computers and communication which are drivers for
medical informatics and the eHealth paradigm. A conceptual model for medical
informatics, proposed by van Bemmel, is applicable to eHealth. Having provided the
 P.J. McCullagh et al. / Medical Informatics and eHealth 137

background on the technology, we use a case study of home-based tele-rehabilitation to

illustrate the potential of eHealth in an era when conventional interventions for chronic
conditions will become increasingly more difficult to support. The case study provides
examples of hardware, software, secure communication over a network and a 3-tiered
client server Internet based approach. It comprises many of the model sub components
and indeed its purpose was to research the area of tele-rehabilitation and stroke (Layer
7), in order to provide new knowledge that can be shared with the multi-disciplinary
team comprising biomedical engineers, computer scientists and healthcare
professionals.

Acknowledgement

The SMART consortium consists of Royal National Hospital for Rheumatic Diseases,
University of Bath, Sheffield Hallam University, University of Essex, University of
Ulster and The Stroke Association. We would like to thank the patients and carers, who
have given their time and support to the project. We would also like to thank Design
Futures from Sheffield, Xsens, Charnwood Dynamics Ltd. and the Stroke Association
for their support. SMART was funded under the EQUAL 4 (extend quality of life)
initiative from the UK Engineering and Physical Sciences Research Council (EPSRC).

Glossary

3G Third generation cellular telephony

ADSL Asymmetric Digital Subscriber LIne
ALU Arithmetic and Logic Unit
Bluetooth Personal area network standard
CERN Conseil Européen pour la Recherche Nucléaire
CPU Central Processing Unit
DSS Decision Support System
ECG Electro-CardioGram
eHealth Electronic Health
EPR Electronic Patient Record
GPRS General Packet Radio Service
GSM Global System for Mobile communication
HCI Human Computer Interface
HSPDA High Speed Packet Download Access
HTML Hyper Text Markup Language
HTTP Hyper Text Transfer Protocol
ICT Information and Communication Technology
IEEE Institution of Electrical and Electronic Engineers
IIS Internet Information Services
I-Mode Internet service designed for mobile phones
IrDa The Infrared Data Association
IS-95 Cellular telephony system in USA
ISM Industrial Scientific and Medicine
KB Knowledge Base
LAN Local area network
138 P.J. McCullagh et al. / Medical Informatics and eHealth

MAC Message Authentication Code

MySQL An Internet database system
NHSnet UK’s National Health Service private network
OFDM Orthogonal Frequency Division Modulation
OS Operating system
PDA Personal Data Assistant
PSTN Public Switched Telephone System
PUBMED Publicly accessible database from US National Library of Medicine
RDBMS Relational Database Management System
SMS Short Message Service
SOAP Simple Object Access Protocol
SQL Structured Query Language
TCP/IP Transport Control Protocol/ Internet Protocol
Tomcat A web based server from Apache Corporation
UMTS Universal Mobile Telecommunications System
URI Uniform Resource Identifier
USB Universal Serial Bus (connectivity interface for PC)
VOIP Voice over internet protocol
WAP Wireless Access Protocol
WiFi Wireless Fidelity, a wireless local area network
WWW World Wide Web
XBus Open Source routing and transformation system for connected
devices
XML eXtensible Markup Language
ZigBee Personal area network technology

References

[1] D.L. Sackett, W.M.C. Rosenberg, J.A. Muir Gray, R.B. Haynes, W. Scott Richardson, Evidence based
medicine: what it is and what it isn't. BMJ 312, (1996), 71-72.
[2] J.H. van Bemmel, The structure of medical informatics. Med Informics 9, (1984), 175-180.
[3] P. Wilson, C. Leitner, A. Moussalli, Mapping the Potential of eHealth: Empowering the citizen through
eHealth tools and services. European Institute of Public Administration 2004.
[4] 4D Ultrasound, http://www.4dfetalimaging.com/, accessed Oct 2007.
[5] R. Morin, S. Langer. Speech Recognition System Evaluation. Journal of the American College of
Radiology 2(5), 449-451, May 2005.
[6] Connecting for health, http://www.connectingforhealth.nhs.uk/ accessed Dec 2007.
[7] S. Kotzin, Medline and PubMed will be able to synthesise clinical data BMJ. 324 (2002), 791.
[8] PUBMED, National Institute of Health, http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?DB=pubmed,
accessed Oct 2007.
[9] J.H. van Bemmel and M.A. Musen (eds.), Handbook of Medical Informatics, Springer 1997.
[10] W. Aspray, John Von Neumann and the Origins of Modern Computing. Cambridge, Massachusetts:
The MIT Press, 1990.
[11] G.E. Moore, Cramming more components onto integrated circuits. Electronics Magazine 19 April
1965.
[12] R. Kurzweil, The Law of Accelerating Returns,
http://www.kurzweilai.net/articles/art0134.html?printable=1, March 2001.
[13] Los Alamos National Laboratory, http://www.nigms.nih.gov/News/Results/FluModel040306, accessed
Oct 2007.
[14] M.L. Baker Supercomputer Serves as Weapon in AIDS Fight,
http://www.eweek.com/article2/0,1759,1972257,00.asp, accessed Oct 2007.
[15] Health Grids, http://www.healthgrid.org/, accessed Oct 2007.
[16] Biomedical Informatics Research Network, http://www.nbirn.net/, accessed Oct 2007.
 P.J. McCullagh et al. / Medical Informatics and eHealth 139

[17] http://www.malariacontrol.net/, accessed Dec 2007.

[18] PDA software available from www.tucows.com, accessed Dec 2007.
[19] Apache web server project, http://httpd.apache.org/, accessed Dec 2007.
[20] EU Health portal, http://ec.europa.eu/health-eu/index_en.htm, accessed Dec 2007.
[21] http://www.stroke.org.uk/noticeboard/obesity.htm, accessed Oct 2007.
[22] M. Walker, Stroke rehabilitation. The British Journal of Cardiology, 9(1) (2002) 23-30.
[23] DoH (2000), National Service Framework for older people. London.
[24] J.M.Winters, Telerehabilitation research: emerging opportunities, Annu. Rev. Biomed. Eng. 4, (2002),
287-320.
[25] H.R.Miltner, H. Bauder, Effects of constraint-induced movement therapy on patients with chronic
motor deficits after stroke, Stroke 30(1999), 586-92.
[26] P.M. Rossini, C. Calautti, Post-stroke plastic reorganisation in the adult brain. The Lancet Neurology
3(2003), 493-502.
[27] Natl. Inst. Disbil. Rehabil. Res. Request for applications for Rehabilitation Engineering Research
Center on Telerehabilitation. 1998. Fed. Regist. June 12, 32526-39.
[28] H. Zheng, N.D. Black, N. Harris, Position-sensing technologies for movement analysis in stroke
rehabilitation, Medical & Biological Engineering & Computing 43(4) (2005), 413-420.
[29] SMART PROJECT website, (http://www.shu.ac.uk/research/hsc/smart/), accessed December 2006.
[30] G.A.Mountain, P.M.Ware, J. Hammerton, S.J.Mawson, H. Zheng, R.Davies, N.D. Black, H. Zhou, H.
Hu, N. Harris and C. Eccleston, The SMART Project: A user led approach to developing and testing
technological applications for domiciliary stroke rehabilitation, 3rd Cambridge Workshop on Universal
Access and Assistive technology (CWUAAT 2006).
[31] H. Zhou and H. Hu, Inertial motion tracking of human arm movements in home-based rehabilitation,
Proceedings of IEEE Int. Conf. on Mechatronics and Automation, Ontario, Canada, 29 July – 1 August,
2005, 1306-11.
140 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-140

III.2. Data Structures, Coding and

Classification

Huiru ZHENGa, Haiying WANGa, Norman D. BLACKa and John WINDERb

a
Faculty of Engineering, University of Ulster at Jordanstown, Co. Antrim, N Ireland
b
Faculty of Life and Health Sciences, University of Ulster at Jordanstown, Co. Antrim,
N Ireland

Introduction

Medical data are at the centre of many healthcare-related activities, ranging from
providing medical care for each individual patient to supporting scientific research and
education. This review offers an introduction to medical data structures and
management. Examples of coding and classification systems are also presented.

1. Medical Data

Medical data include a broad range of data types, including text-based alphanumeric
data, recorded physiological signals and medical picture/images. In addition to
traditional clinical data, recent progress in medical science and technology has led to
the accumulation of a tremendous amount of genomic data in support of individualised
healthcare and personalised therapies. In this section, the characteristics of typical
clinical data are introduced, followed by a brief introduction to genomic data.

1.1. Clinical Data

In general, there are three types of clinical data in the practice of medical and health
science 0, [2].
• Alphanumeric data are typically recorded as unstructured, free text in the
medical record. They account for a large portion of the information that is
gathered verbally in the care of patients, such as patient demographics, a
description of patients’ illnesses and responses to physician’s questions.
Results of physical examination such as vital signs and laboratory tests, and
other narrative descriptions, such as treatment plans and surgical procedures,
also belong to this category.
• Physiological signals, captured from the human body by using appropriate
biomedical sensors, are usually represented by a series of values along two
dimensions with the x-axis representing time and the y-axis displaying voltage
values associated with the underlying physiological activity. Advances in
physiological signal acquisition techniques have generated a vast quantity of
 H. Zheng et al. / Data Structures, Coding and Classification 141

physiological recordings, including electrocardiogram (ECG),

electroencephalogram (EEG), gait, respiration, blood pressure tracings, and
oxygen saturation. As illustrated in Figure 1, an ECG records the electrical
activity of the heart as a function of time. Each ECG beat represents one
cardiac cycle. By convention it includes the P wave, the QRS complex. T
wave, and sometimes a small U wave may be seen following the T wave.
• Medical images, as illustrated in Figure 2. With the ability to reveal the areas
of interest in great detail, medical images are a valuable resource in clinical
diagnosis and medical research. Radiological images based on X rays,
computed tomography and magnetic resonance imaging techniques are the
most well known examples. Using the electromagnetic spectrum, radiological
images provide a means to visualize the condition of internal organs or an area
that is not externally visible. Light images may also be produced by skin
photography and endoscopy. Instead of using electromagnetic radiation,
ultrasound imaging is based on mechanical energy in the form of high-
frequency sound waves.

Figure 1. Illustration of an ECG signal: (a) a typical ECG wave, (b) ECG intervals. RR, QT and PR
represent the intervals between two respective wave peaks.

Figure 2. An example of ultrasound images and Magnetic Resonance (MR) images.

142 H. Zheng et al. / Data Structures, Coding and Classification

1.2. Genomic Data

Advances in biological science are fostering a new clinical practice, where clinical
diagnosis and treatment will be supported by the information encoded in relevant
genomic data, such as DNA (deoxyribonucleic acid) sequences, protein structures and
microarray gene expression data. Examples of genomic applications in medical
research can be found in the Programs for Genomic Applications (PGAs) funded by the
National Heart, Lung and Blood Institute (NHLBI) of the NIH (National Institutes of
Health) (http://www.nhlbi.nih.gov/resources/pga/).
Each type of genomic data has a unique data structure. For instance, DNA may be
represented by a linear sequence from the alphabet {A, G, T, C}. Each of these
characters refers to a nucleotide. A typical microarray experiment dataset includes
expression levels of thousands of genes in a number of experimental samples
(conditions). These samples may correspond to different conditions or serial time
points taken during a biological process. An expression dataset can be summarised by a
matrix, in which the horizontal rows represent genes, one gene in each row, and the
vertical columns contain the various samples corresponding either to the time points or
to various conditions, with one sample in each column. Thus, various models of
electronic medical records have been developed to integrate traditional clinical data and
diverse genomic data 0.
The strong interest in integrating genomic data into clinical research is driven by
the hope that it will result in the prevention and diagnosis of complex genetic disease
and the design of highly targeted therapies [4]. For example, with DNA microarray
technology, scientists are able to measure gene expression of thousands of genes under
selected physiological conditions at a given time. Thus, by comparing expression
profiles of an individual gene under various conditions or by comparing expression
profiles of samples to understand their similarities and differences and to find the
relationships among genes and samples, many important physiological and medical
phenomena can be studied at the molecular level.

1.3. Characteristics of Medical Data

Apart from privacy and confidentiality requirements, medical data have important and
unique features. For example, a patient record is a mixture of opinion and factual data
generated from a wide range of clinical professionals. Some of these data, such as
observations and interpretations, are difficult to record in a structured form. Recorded
physiological signals and medical images are contaminated by noise generated from the
machine recording the data or from the patient. For instance, an ECG record could be
contaminated by physiological variability, baseline wander noise, 50 or 60 Hz power
line interference, and muscle noise 0. Due to the quantum noise inherent in photo
detection, images generated by computed tomography-based techniques such as
positron emission tomography (PET) and single photo emission tomography (SPECT)
have a high noise content 0.

2. Data Structures and Management

The characteristics of medical data pose a great challenge to efficient data storage and
management. In this section, electronic patient records are described. The examples of
 H. Zheng et al. / Data Structures, Coding and Classification 143

management and storage of medical images and physiological signals are introduced
respectively.

2.1. Electronic Patient Records

A patient record contains all the clinical data relevant to the care of one patient. Given
the increasing volume and scope of medical data available nowadays, traditional paper-
based patient record systems are not sufficient to address many issues related to
efficient medical data management. The Electronic Patient Record (EPR) is
revolutionising storage, management and communication of medical information. With
EPR comes all related medical data, including patient history, physical examination,
drug prescriptions, and recorded biological signals, linked into one record that can be
easily accessed for different purposes. The ability to store all patient-related medical
information in one record, which may be accessed at any time or from any place,
ultimately contributes to improvements in healthcare delivery.
It has been shown that an EPR-based system offers several significant advantages
in the process of data management and analysis 0]. One of the most obvious benefits is
that the speed of data retrieval is significantly improved. A physician should be able to
search through thousands of records for specific attributes such as medication or
diagnosis with little effort. With structured data entry, medical data collected in an EPR
are well structured and organised, allowing the electronic sharing of patient recording
among different clinical professionals. Moreover, an EPR-based record system has the
inherent flexibility to provide task-specific views on all the available patient data to
meet the needs of different clinical tasks, as illustrated in Figure 3.

Figure 3. An EPR provides an integrated view of patient data and has the flexibility to generate task-specific
views on all the available data to meet the needs of different tasks.

One of the key components in an EPR is a database management system (DBMS)

0. DBMS is a software system designed to manage the storage, retrieval, and
organization of data in a database. Different data models have evolved in recent years
to structure the logical view of the DBMS, including relational, hierarchical, text-
144 H. Zheng et al. / Data Structures, Coding and Classification

oriented, and object-oriented data models. Among these, the relational data model is
most widely used.
A relational data model, developed by E.F. Codd [8], is one in which the data and
relations between them are organised into a series of tables. Each table is made up of
columns and rows. All values in a given columns have the same data type, which could
be numeric or textual. All values in a row conceptually belong together. Typically, each
row, known as a record, is uniquely identified by a primary key which may contain a
single column or multiple columns in combination. Figure 4 shows an example of a
relation database with two tables: PatientDemographics and PhysicalExamination
tables.

Figure 4. Example of a relational database with two tables.

2.2. Medical Images - DICOM

Digital Imaging and Communications in Medicine (DICOM) is the healthcare industry

standard for medical image data file storage and transfer. It was first conceived in 1983
by the American College of Radiology (ACR) and the National Electrical
Manufacturers Association (NEMA) and first known as the ACR-NEMA standard.
This standard was revised in 1993 and named DICOM. The standard defines principles
and methods for the storage and transfer of medical images (mainly radiological) in a
multi vendor environment (http://medical.nema.org). Virtually all radiological imaging
modalities are now digital and a large teaching hospital may generate around 2-3 Tera
(1012) bytes of image data per annum. The use of medical imaging is increasing, as is
 H. Zheng et al. / Data Structures, Coding and Classification 145

the resolution of imaging systems, therefore we can expect to see the computer network
and storage requirements also increase.
The main DICOM functions are to transmit and store medical images, to enable
searching and retrieval, to facilitate printing of hard copy, to support workflow
management and to ensure quality and consistency of image presentation 0. The
DICOM file format is well defined and contains two specific sections. The first part of
the file, called the header, holds patient demographics, hospital information and details
of the referring physician. Also, the header contains detailed data on the actual specific
imaging modality parameters for each patient scan. For example, in a magnetic
resonance (MR) image data file, the slice thickness, field of view, pulse sequence
details, table position, and scan date are stored. The header also contains hidden fields
used by the equipment vendor to store proprietary information on image reconstruction
algorithms. Another feature of the DICOM header is that it contains a unique
identification number that identifies the scanner manufacturer, the individual scanner,
the modality types and the patient study number. This is called the unique identifier
(UID). DICOM data files written to storage media like DVD or to another computer
will be named using the UID as part or the entire file name.
Figure 5 shows a screen shot of part of a DICOM image header from OSIRIS
developed by The University Hospital of Geneva, Switzerland. Osiris is a free DICOM
viewer available at http://www.sim.hcuge.ch/osiris/01_Osiris_Presentation_EN.htm.
The second part of the DICOM data file is the image pixel data where pixel values are
stored using between 4 and 32 bits, depending on the imaging modality. A typical file
size for MR is 0.5 MBytes (one patient examination may contain up to 1000 images),
whilst a Computed Radiography chest X-ray may be 32 MB. The DICOM standard
includes the functionality to compress medical images using lossless or “lossy”
techniques 0. JPEG 2000 is the common standard achieving a lossless compression of
approximately 3:1. Further compression is possible, however, there is some reluctance
to reduce file size further as diagnostic integrity of the original image may be lost after
significant compression.
The benefit of DICOM is the ability to connect a number of different vendor
medical imaging modalities to a Picture Archive and Communication System (PACS).
This is the main image database and transmission system in the health environment.
This system enables multiple copies of images to be available at any time and at any
place that has the appropriate network connections and permissions. PACS also enables
the creation of large digital libraries for teaching and research.

2.3. SCP-ECG

The SCP-ECG (the Standard Communications Protocol for Computer-Assisted

Electrocardiography) was developed by the Comité Européen de Normalisation
Technical Committee 251 (CEN/TC251) project team in 1993 exclusively for the
purpose of the exchanging, encoding and storage of 12-lead ECGs 0. The data level in
the SCP-ECG standard includes the ECG signal data, patient demographics, ECG
measurements, ECG interpretation results, as well as schemes for ECG waveform
compression. In the SCP-ECG standard, various parts of data related to an ECG are
specified in different data sections with different encoding forms as illustrated in
Figure 6.
146 H. Zheng et al. / Data Structures, Coding and Classification

Figure 5. Typical information stored in the DICOM file header.

Figure 6. Structure of a SCP_ECG record

 H. Zheng et al. / Data Structures, Coding and Classification 147

2.4. XML based systems

Due to its ability and flexibility to define markups for specific type of data, XML
(eXtensible Markup Language, http://www.w3.org/xml/) and its related techniques
have gained great attention in medical data representation and management. Currently,
various organizations within healthcare such as HL7 (health level 7) and ASTM
(American Society for Testing and Materials) are working on recommendations of
XML-based e-healthcare technologies within the medical domain. For example, HL7
has published the Clinical Document Architecture (CDA), which uses XML as the
representation format for medical and administrative data
(http://xml.coverpages.org/ni2004-08-20-a.html). Along with the CDA, HL7 version 3
messaging standard has also been developed using XML as the standard exchange
format. The use of XML syntax for the exchange of electronic patient records was
illustrated in the EU project Synapses (https://www.cs.tcd.ie/synapses/public/) and its
implementations [12]. The application of XML in DICOM can be found in several
related projects [13]. The U.S Food and Drug Administration (FDA) Centre for Drug
Evaluation and Research has proposed recommendations for the XML-based
techniques for the exchange of time-series data. In December 2003, the FDA XML
format was finalized and the Annotated ECG (aECG) format is now part of the HL7
family of standards
(http://www.hl7.org/v3annecg/foundationdocuments/welcome/index.htm).

3. Coding and Classification

Clinical concepts can be represented and described in many ways. Consequently,

standards for codes/terminology have become an essential element in the development
of electronic patient records. In this section, several widely-used coding and
classification systems are introduced.

3.1. The International Classification of Diseases

The International Classification of Disease (ICD) is a coding system published by the

World Health Organization (WHO, http://www.who.int/en/ ). Its main purpose is to
allow morbidity and mortality data to be collected from around the world in a standard
format. Since its first edition was published in 1900, ICD has become the most widely
used pathology classification system and de facto reference point for many healthcare
terminologies 0. The code has been updated approximately every 10 years to reflect the
medical advances in the past years and the current version is ICD-10.
The classification of ICD-10 is structured into 21 chapters, identified by 21 Roman
numerals: I, II, III, IV, V, VI, VII, VIII, IX, X, XI, XII, XIII, XIV, XV, XVI, XVII,
XVIII, XIX, XX, and XXI. Among these, Chapters I-V, XV-XVII and XIX cover
special diseases, such as neoplasms, pregnancy, childbirth and the puerperium, and
injuries, poisoning and certain other consequences of external agents. Diseases
associated with body systems are found in Chapters VI to XIV. For example, Chapter
VI covers diseases of the nervous system, and diseases of the genitourinary system are
assigned to Chapter XIV. Symptoms, signs and abnormal clinical and laboratory
findings are given in Chapter XVIII. Other external factors affecting morbidity and
mortality are included in Chapters XX and XXI.
148 H. Zheng et al. / Data Structures, Coding and Classification

Each chapter is divided into blocks and categories. Unlike ICD-9, in which
diseases are coded with three digits, category codes in ICD-10 start with an
alphabetical character followed by two numeric characters, ranging from A00 to Z99.
For instance, code I01 represents rheumatic fever with heart involvement. This greatly
increases the number of available codes, which can be used to code common diseases
in more detail.
Most 3-character categories in ICD-10 are further subdivided into up to 10
subcategories by adding a fourth digit after a decimal point. This can be used, for
instance, to classify varieties of a disease, e.g. as illustrated in Figure 7, acute
rheumatic pericarditis is coded with I01.0, while acute reheumatic endocarditis is
represented by I01.1.

Figure 7. Example of ICD-10 code.

3.2. The Clinical Terms (Read Codes)

Read Codes were initially introduced into the UK in the middle 1980s as a 4-byte set
terminology to summarise patient treatment, including clinical process and
administrative data, in primary care. This version only consists of about 40000 codes.
Its subsequent Version 2 released in 1990 was based on the content of ICD-9 and
Classification of Surgical Operations and Procedures, Fourth Revision (OPCS-4). Read
Codes in this version were extended to 5-digit alphanumeric characters, each
representing one level of hierarchy. While being easy to understand and implement,
such a rigid hierarchical classification structure poses some difficulties in the
 H. Zheng et al. / Data Structures, Coding and Classification 149

representation of clinical terms. For example, without giving two Read Codes, a
clinical term can only be located within one hierarchy.
Due to increasing demands from clinical professionals, Read Codes were further
expanded by the Clinical Terms Project, a joint project established by the UK’s
Conference of Medical Royal Colleges and the government’s National Health Service
(NHS). As a result, the Clinical Term Version 3 (CTV3) was developed and first
released in 1994.
In CTV3, a concept refers to any clinical disorder, procedure or observation and
can be described by more than one clinical term. Like its previous version, CTV3
adopts a 5-digit alphanumeric code, giving more than 650 million possible codes to
represent clinical concepts. However, CTV3 implements a new, more flexible structure
to address the problems caused by its early version. For example, the coded-based
hierarchical structure has been abandoned. A Read Code is only used to uniquely
identify a concept and no longer represents its hierarchy position. Instead, hierarchical
relationships between concepts are defined by a set of parent-child links described in a
Hierarchy file.
Core release files included in CTV3 are Concept file, Term file, Descriptions file,
Hierarchy file, key file, Specialty files, Cross-mapping files, Template file, Redundant
codes mapping file and Description Change file. The structures and examples of
Concept file (describing concepts), Term file (describing terms), Descriptions file
(linking terms to Read Codes) and Hierarchy file (describing a hierarchy) are illustrated
in Figure 8. The reader is referred to [14] for a more detailed description of CTV3.

3.3. Systematized Nomenclature of Medicine

The Systematized Nomenclature of Medicine (SNOMED), developed and maintained

by the College of American Pathologists, is generally regarded as one of the most
comprehensive standardized medical coding systems available today
(http://www.snomed.org/). It uses a hierarchical, multi-axial method to encode clinical
data, allowing for the representation of various aspects of a disease. For example,
SNOMED International is organized around 11 independent axes: T (Topography), M
(Morphology), L (Living organisms), C (Chemical), F (Function), J (Occupation), D
(Diagnosis), P (Procedure), A (Physical agents, forces, activities), S (Social context), G
(General syntactic). Each axis represents a unique hierarchical classification system.
Thus, a disease in SNOMED can be described using a morphologic code, a topographic
code, an etiological code, and a function code. Moreover, these codes across different
axes can be cross-referenced, leading to a better understanding of each code. For
example, the disease pneumonia could be coded as T-28000 (topology code for lung),
M-4000 (morphology code for inflammation) and L-25116 (etiological code for
Streptococcus pneumoniae).
To reflect the advances in medical informatics, SNOMED has evolved further with
the release of SNOMED RT (Reference of Terminology) in 2000. To provide a unified
international language that supports the electronic patient record and decision support
system, the College of American Pathologists and the UK NHS announced their
intention to unite SNOMED RT and CTV3. As a result, SNOMED CT (Clinical Term),
which incorporates the content and structure of CTV3, was released in 2002.
150 H. Zheng et al. / Data Structures, Coding and Classification

Figure 8. The structures and examples of (a) Concept file; (b) Term file; (c) Descriptions file; and (d)
Hierarchy file.

One of the important features of SNOMED CT is that it utilizes description logic

to describe the scope of a concept. Its core content includes concepts, descriptions and
the relationships between them 0. As illustrated in Figure 9, a concept in SNOMED CT
can be described by terms in one or more descriptions. SNOMED CT contains more
than 300000 concepts with unique meanings and formal logic-based definitions, which
are organized into top-level hierarchies such as Clinical findings and Body Structure.
The relationships between concepts are described in relationship tables. There are two
types of relationships in SNOMED CT. While IS-A relationships connect concepts
within the same hierarchy, Attribute relationships describe the relationships between
concepts located in different hierarchies. For example, the relationships between the
 H. Zheng et al. / Data Structures, Coding and Classification 151

disease concepts: Arthritis of knees, Arthritis and Arthropathy belong to IS-A

relationships. Examples of Attribute relationships in SNOMED CT include Finding
site, Procedure site, Associated morphology, and Method. The relationship between the
concepts Appendicitis that is a disease concept and Inflammation that belongs to the
Body structure hierarchy can be described using Associated morphology.

Figure 9. The outline of SNOMED CT core structure.

The current version of SNOMED CT includes more than 300,000 clinical concepts,
770,000 descriptions and 900,000 semantic relationships. SNOMED CT can also map
to other medical classification systems such as ICD-10.

4. Case Study: An XML-based representation of ECG data

The ecgML model, a markup language for supporting representation and coding of
ECG data, was first published in 2003 [15]. Based on XML, ecgML offers several
advantages over existing ECG coding systems. For example, it includes all the
components required for ECG representation. The terms and structure used can be
further expanded or reviewed at any stage. This protocol can be used as a common
coding platform between different ECG acquisition devices and visualization programs.
152 H. Zheng et al. / Data Structures, Coding and Classification

4.1. Hierarchical Presentation of ecgML

The main components included in ecgML are one PatientDemographic, optional

element MedicalHistory, one or more Record components, and an optional Diagnosis
element, as illustrated in Figure 10. To facilitate the inclusion of multiple time-related
patient’s ECG data, each Record element, which consists of zero-or-one
RecordingDevice, zero-or-one ClinicalProtocol, and one-or-more RecordData, is
uniquely identified by AcquisitionDate and AcquisitionTime. As a key component in
ecgML, each RecordData includes three main subcomponents (Waveforms,
Annotations and Measurements) to represent original ECG waveform data, the
annotations and measurements respectively.

Figure 10. The main structure of the ecgML model.

ecgML incorporates several international standards to describe relevant

information. For example, it applies the DICOM lead labeling format to define channel
names associated with RecordData element. The Unified Code for Units of Measure
(UCUM) scheme is used to define measurement units when appropriate. Moreover,
different coding and classification schemes can be incorporated to define medical terms
at different levels. One example is to utilize terminologies included in SNOMED CT to
describe the diagnostic interpretation of the ECG (Diagnosis element) and medical
history of patient’s clinical problems (MedicalHistory element).
A portion of an ecgML-based ECG recording generated from the MIT-BIH
Arrhythmia Database is illustrated in Figure 11.
 H. Zheng et al. / Data Structures, Coding and Classification 153

Figure 11. An example of an ecgML-based ECG representation.

4.2. Encoding ECG Waveform Data in ecgML

ECG waveform data are a key component of an ECG record. Given that a wide range
of ECG recoding devices are available today, the size of waveform data can vary
dramatically. Moreover, they are usually expressed in binary format. In order to
accommodate the full spectrum of ECG data, ecgML provides the following three
options to handle ECG waveform summarized in Figure 12.

• For large data files, typical of Holter recording, ECG data are maintained in an
external file, which can be referenced in ecgML using a FileLink element;
• Using RealValue elements to directly include actual waveform values in
ecgML as the content of the element;
154 H. Zheng et al. / Data Structures, Coding and Classification

• To encode binary waveform data using a BinaryData element. ecgML is based

on XML techniques to encode and represent ECG records. Due to the valid-
character restriction posed by XML specification, simply embedding binary
ECG waveform data within ecgML may cause the parser to encounter invalid
sequences and fail to achieve its intended goal. Thus, each BinaryData
element is associated with a specified encoding scheme, which may be Base64
or hexadecimal.

Figure 12. Framework for handling ECG waveform data in ecgML.

4.3. Accompanying Tools

A series of Java-based, user-friendly tools are being developed to assist users in

exploiting ecgML-based application [16],[17]. These include ecgMLgenerator, ecgML
parser and ecgMLBrowser.
The ecgMLgenerator produces ecgML-based ECG record from existing ECG
databases. While the ecgMLparser allows the user to read the ECG record and access
their contents and structure, ecgMLbrowser provides onscreen display of the collected
waveform data as shown in Figure 13. The hierarchical structure of the ecgML-based
ECG record is displayed on the left hand side. It can be expanded and shrunk at any
level. The waveform data are shown on the right hand pane. The ecgMLeditor allows
an authorized user to modify the contents of an ecgML-based ECG record. Some
components such as the raw waveform data are not allowed to be changed.
 H. Zheng et al. / Data Structures, Coding and Classification 155

Figure 13. A screenshot of ecgMLbrowser.

5. Summary

The above coding systems are developed for a specific purpose - they are goal-oriented.
For example, ICD takes a pathophysiological and aetiological view of medicine and
classifies diseases primarily according to the organ system involved with some
important exceptions such as infectious diseases [18] It is unable to fulfill the needs of
all users.
Another problem is that the terms are changing over time, thus the meaning of a
same term can be different over time, new terms are created and old terms may be
replaced. When the meaning of a standard code is changed, the data will be interpreted
incorrectly. When a standard code is removed from the coding system, the data are no
longer interpretable. Additionally there is a lack of comprehensive, standardized
medical terminology, and highly skilled experts are required to interpret the systems.
Despite these limitations, coding and classification systems have been successfully
applied in health care. A widely accepted coding system is essential for storing and
exchanging patient records efficiently.

Glossary

ACR American College of Radiology

ASTM American Society for Testing and Materials
CDA Clinical Document Architecture
CEN/TC251 Comité Européen de Normalisation Technical Committee 251
CTV3 Clinical Term Version 3
156 H. Zheng et al. / Data Structures, Coding and Classification

DBMS Database Management System

DICOM Digital Imaging and Communications in Medicine
DNA DeoxyriboNucleic Acid
ECG ElectroCardioGram
EEG ElectroEncephaloGram
EPR Electronic Patient Record
FDA U.S Food and Drug Administration
HL7 Health Level 7
ICD International Classification of Disease
MR Magnetic Resonance
NEMA National Electrical Manufacturers Association
NHLBI National Heart, Lung and Blood Institute
NHS National Health Service
NIH National Institutes of Health
OPCS-4 Classification of Surgical Operations and Procedures, Fourth
Revision
PACS Picture Archive and Communication System
PET Positron Emission Tomography
PGA Programs for Genomic Application
SCP-ECG Standard Communications Protocol for Computer-Assisted
Electrocardiography
SNOMED Systematized Nomenclature of Medicine
SNOMED CT Systematized Nomenclature of Medicine Clinical Term
SNOMED RT Systematized Nomenclature of Medicine Reference of Terminology
SPECT Single Photo Emission Tomography
UCUM Unified Code for Units of Measure
UID Unique IDentifier
WHO World Health Organization
XML eXtensible Markup Language

References

[1] E. Shortliffe, L. Perreault, G. Wiederhold, L. Fagan, eds., Medical Informatics: Computer Applications
in Health Care and Biomedicine, New York: Springer-Verlag;Second Edition 2001.
[2] W. Tompkins, Biomedical Digital Signal Processing: C lauguage Examples and Laboratory
Experiments for the IBM PC. Englewood Cliffs, N.J.; London : Prentice Hall, 1993.
[3] B. Robson, R. Mushlin, Genomic Messaging System and DNA Mark-Up Language for Information-
Based Personalized Medicine with Clinical and Proteome Research Applications. J. Proteome Res. 3(5),
(2004) 930 – 948.
[4] E. Coiera, Guide to Health Informatics. Second Edition. London: Hodder Arnold, 2003.
[5] Y. Pu, R. Patterson, Comparison of R-wave detection errors of four wireless heart rate belts in the
presence of noise. Physiol. Meas. 24, 913-924, 2003.
[6] P. Razifar, M. Lubberink, H. Schneider, B. Langstrom, E. Bengtsson, and M. Bergstrom. Non-isotropic
noise correlation in PET data reconstructed by FBP but not by OSEM demonstrated using auto-
correlation function. BMC Medical Imaging, 5:3, (2005).
[7] J. H. van Bemmel, M.A. Musen, Handbook of medical informatics. Heidelberg, Germany ; AW
Houten, Netherlands : Springer Verlag, 1997.
[8] E.F. Codd, A relational model of data for large shared data banks. Communications of the ACM, 13(6),
(1970), 377–387.
[9] A.E. Flanders and J.A. Carrino, Understanding DICOM and IHE. Seminars In Roentgenology, 38(3),
(2003) 270-281.
 H. Zheng et al. / Data Structures, Coding and Classification 157

[10] Graham RNJ, Perriss RW, Scarsbrook AF. DICOM demystified: A review of digital file formats and
their use in radiological practice. Clinical Radiology 60, (2005), 1133 – 40.
[11] ENV 1064 standard communications protocol for computer-assisted electrocardiography. European
Committee for Standardisation(CEN), Brussels, Belgium, 1996.
[12] B. Jung, E.P. Andersen, J. Grimson, Using XML for Seamless Integration of Distributed Electronic
Patient Records. In Proceedings of XML Scandinavia 2000 conference, Gothenburg, Sweden, May
2000.
[13] A. Tirado-Ramos, J. Hu, K.P. Lee. Information object definition-based unified modeling language
representation of DICOM structured reporting: a case study of transcoding DICOM to XML. J Am Med
Inform Assoc, 9(1), (2002), 63-71.
[14] http://www.connectingforhealth.nhs.uk/terminology/readcodes/, accessed Mar 2008.
[15] H.Y. Wang, F. Azuaje, B. Jung, N.D. Black, A markup language for electrocardiogram data acquisition
and analysis (ecgML). BMC Medical Informatics and Decision Support, 3 (4), 2003.
[16] H.Y. Wang, B. Jung, F. Azuaje, N.D. Black. ecgML: Tools and technologies for multimedia ECG
presentation. In the Proc. of XML Europe Conference, London, United Kingdom, May 2003.
[17] H.Y. Wang, F. Azuaje, G. Clifford, B. Jung, N.D. Black, Methods and tools for generating and
managing ecgML-based information. In the Proc. of 2004 Computers in Cardiology, IEEE Press,
Chicago, September 2004.
[18] http://www.opengalen.org/, accessed Mar 2008.
158 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-158

III.3. Mining, Knowledge and Decision

Support
Dewar D. FINLAY, Chris D. NUGENT, Haiying WANG, Mark P. DONNELLY and
Paul J. McCULLAGH
School of Computing and Mathematics, University of Ulster, Shore Road, Belfast,
Co. Antrim, N Ireland, BT37 OQB

1. Introduction to Decision Support

A decision support system (DSS) usually consists of a computer program which

provides assistance in the decision making process. DSSs have found application in
many domains and have seen particular prominence in medicine. This is an obvious
application due to the large amounts of data (e.g. laboratory measurements such as
blood pressure, heart rate, body-mass index) and information (e.g. patient history,
population statistics based on age and sex) that must be considered before diagnosing
any disease. As well as assisting in primary diagnosis, a DSS can reduce medical error,
assist compliance with clinical guidelines, improve efficiency of care delivery and
improve quality of care. Like their human counterparts, clinical DSSs can also be
designed with medical specialities in mind.
In this review we look at the various traits of decision support in clinical practice.
We examine two broad types of support, classification and prediction, and summarize
various automated techniques that form the building blocks for these processes. We
conclude with a discussion of some of the current trends in decision support along with
a case study looking at how techniques commonly used in decision support can be
employed in data mining and knowledge discovery.

1.1. The Modern Clinical DSS

Modern commercial clinical DSSs have evolved from the work of early investigators
who set about developing systems that could flag the presence and genre of disease
using the computational tools available at that time. An example of such a system was
that developed by DeDombal in 1972 [1]. This system, designed to diagnose acute
abdominal pain, used statistical techniques to suggest to which of 13 diagnostic
categories a patient was most likely to belong. Although most early systems utilised
statistical techniques, investigators soon began applying other methods. An example of
this was in the implementation of MYCIN [2], a rule-based system designed to
diagnose and recommend treatment for certain blood infections and other infectious
diseases. In this system clinical knowledge was modelled as a set of IF-THEN rules
with certainty factors attached to diagnoses. More recent systems have also been
developed with the ability to provide support over a comprehensive range of
conditions. The INTERNIST system [3] was capable of providing hundreds of
 D.D. Finlay et al. / Mining, Knowledge and Decision Support 159

diagnoses in internal medicine, based on thousands of symptoms. This has since been
superseded by Quick Medical Reference (QMR) which contains information on almost
700 diseases and more than 5,000 symptoms. Iliad, DXplain and DiagnosisPro are
similar interactive consultation systems, which use clinical findings to produce a
ranked list of diagnoses which might explain the clinical manifestations.

1.2. Categories of Decision Support System

In Table 1, adapted from [4], likely examples of DSS utility are presented. From this
table it is evident that DSSs can be designed to provide a wide range of functionality
and hence clinical support. We want, however, to give further consideration to two of
the most common and hence important genres of functionality, namely “classification”
and “prediction”. These forms of support are analogous to the processes of diagnosis
and prognosis respectively.
1.2.1. Classification
Classification is the term used to describe the process of analysing data to form
decisions regarding the class membership of individual instances. An instance can be
thought of as the individual case or record that is observed while a class relates to the
possible categories to which the instance can belong. For example, consider a patient
who has arrived at the emergency department with chest pain. This chest pain could
result from some form of cardiac pathology or perhaps be attributed to less serious
causes such as indigestion or a panic attack. The information collected through the
medical assessment of the patient is used to describe the instance. Using this
information, a diagnosis is made which places the instance into some predefined class,
such as those outcomes outlined above. The actual diagnosis is formed by comparing
and locating similarities from the details of the medical assessment with previously
documented assessments for which there exists a known diagnosis.
1.2.2. Prediction
Similar to classification, prediction involves the analysis of measured information in
order to form a decision. While classification is synonymous with diagnosis, prediction
can be compared to prognosis. In this sense the analysis of information recorded to date
is used to produce a forecast of the expected outcomes in the future or over time.
When dealing with chronic diseases, for example, medical staff often refer to the
records (instances) of patients who have, in the past, suffered from the same disease.
These cases can then be used to predict, for example, the life expectancy or treatment
outcomes for current patients.

2. Automated Decision Making

Regardless of functionality rendered, all decision support systems are based upon the
same underlying processes. These processes are summarised in the block diagram
presented in Figure 1, which consists of three stages. The first stage involves the
measured data and its treatment and should result in useful ‘features’. The second stage
involves the interpretation of these features to provide some outcome. The last stage
considers the relevance of the outcomes with some quantification of performance.
160 D.D. Finlay et al. / Mining, Knowledge and Decision Support

Table 1: Examples categories of clinical decision support systems, after Coeira [1]

Type of support Clinical example

Alerts Changes in a patient’s vital signs can provide an alert, requiring immediate
intervention, e.g. an ECG monitor on a ward can alert nursing staff to cardiac
arrhythmia
Reminders Digital storage of patient data can be used as an aide-memoir for doctors, e.g.
during a surgical procedure the Electronic Patient Record (EPR) may generate a list
of immunizations that each patient on the schedule requires based on medical
guidelines
Diagnosis The DSS can use clinical findings (signs, symptoms, laboratory data) to produce a
assistance ranked list of likely diagnoses which might be associated with the clinical
manifestations. Rare presentations of common illnesses, or illnesses which require
specialised expertise can be flagged. Indeed further tests which would confirm or
eliminate a diagnosis can be suggested.
Critiquing The DSS can be used to critique the professional, e.g. if a patient's haemoglobin
level is above the normal transfusion threshold, a clinician requesting a transfusion
should receive a critique and must justify the procedure, by providing additional
information such as ‘patient has active bleeding’ [2]
Therapy planning Planning systems can propose a treatment based on a patient’s condition and
accepted treatment guidelines
Prescribing Prescribing systems can assist with complex drug interactions, dosage errors, and
contraindications such as allergies
Information Intelligent information retrieval systems can assist with the generation of clinical
retrieval questions and act as information filters to reduce the number of documents found in
response to a query
Image recognition Image processing algorithms can detect potentially abnormal images for more
detailed human attention and interpretation, e.g. high density irregularities in
mammograms can prompt further investigation

In the following paragraphs, we describe the various stages beginning with a

description of features and their manipulation. We then introduce some of the
fundamentals of automated interpretation algorithms. This is a short description, as we
devote the entire following section to a more comprehensive description of widely used
algorithms. We conclude this section by introducing common performance metrics.

2. 1. Features

Automated classification relies on a useful set of measurements or variables which

accurately describe the prediction problem at hand. We refer to such measurements or
variables as ‘features’. In some, usually very simple, prediction problems we can easily
identify the variables that are most useful and those which are not. Using the detection
of diabetes as an example, we know that blood glucose level is a fairly sound indicator.
It is also fairly obvious, even to those with the most basic clinical knowledge, that
looking at the colour of a patient’s hair is unlikely to inform us of the presence of this
particular disease. In more complex problems, however, there may be many
 D.D. Finlay et al. / Mining, Knowledge and Decision Support 161

measurements or variables and therefore many possible features. Additionally, there

may also be little understanding of how these measurements relate to the likely disease.
In such cases, where there are many poorly understood measurements, care must be
taken that the actual interpretation algorithm is not overloaded with redundant
information. The terms ‘feature extraction’ and ‘feature selection’ both refer to the
process of reducing the number of variables which are to be considered in an
interpretation problem. As illustrated in Figure 1, we refer to these processes
collectively as ‘feature processing’.

Measured/
recorded Decision Outcome
data

Feature Interpretation Performance

processing algorithm metrics

Figure 1. Steps involved in decision support

2.1.1. Feature Extraction

In feature extraction, a new set of features is obtained from the original features
through some functional mapping [5], [6]. In this technique, the original features are
not retained, but are transformed into a smaller feature space. Although feature
extraction techniques are suited to many classification problems, there is a
disadvantage in that when original features are transformed to a smaller feature space,
all appreciation for the original features is lost. This is a particular disadvantage if the
significance or contribution of the original features to the classification outcome is of
interest. This will be exemplified in the case study at the end of this section.
2.1.2. Feature Selection
Feature selection works in a similar way to feature extraction by reducing the number
of features considered in classifier development. In feature selection, however, the aim
is not to transform the original features into a smaller feature space, but is to eliminate
redundant features, whilst retaining those which provide the most discriminatory
information. This process is said to result in the ‘best subset’ of original features [7].
This is particularly useful because as well as giving a useful insight into the nature of
the prediction problem at hand [8], determining the most relevant features can provide
an indication of which features should initially be measured.
162 D.D. Finlay et al. / Mining, Knowledge and Decision Support

2.2. Interpretation Algorithms

The interpretation algorithm can be considered the core component of a DSS as it must
take the presented features and make a decision based upon their values. Thinking
again of the simple example of analysing blood glucose levels for the presence of
diabetes, a very simple algorithm could be implemented that would flag the presence of
diabetes if a certain threshold has been reached. The threshold which dictates the
presence of this disease can easily be specified by anyone with a prior knowledge of
the domain. In more complex diseases, where there are perhaps many features and a
less comprehensive knowledge of the disease process, development of the decision
algorithm is more difficult. Specifically, in cases where the meaning of the features and
the relationships between them are not obvious, it is necessary to develop an algorithm
that will learn how to make decisions based on a pool of exemplar data. In this case,
we say that the knowledge that is contained within the decision algorithm is ‘learned’
from the existing data. The process of ‘learning’ knowledge involves exposing an
algorithm to a set of ‘example cases’ that represents the classification task at hand.
Subsequently, when unseen data in the form of ‘test cases’ are presented, the algorithm
will suggest which group or category each case belongs to, based on the previous
learning. There are many ways in which interpretation algorithms can learn, these are
broadly categories as either ‘supervised’ or ‘unsupervised’.
2.2.1. Supervised Learning
In the supervised approach, the algorithm is exposed to a number of example cases.
Each will consist of a set of features along with the actual outcome that was observed
when those features were measured. After application of a suitable number of training
examples, the algorithm is considered ‘trained’. Subsequent to this, when the algorithm
is exposed to a new example case which will consist of an unseen set of features, it
should be able to indicate what the output would be. The number of training examples
required to adequately train the algorithm relies on many factors. These include the
type of algorithm used, the number of features and the overall complexity of the
problem. The final performance of the algorithm is also dependent on many such
factors. A diagram illustrating supervised learning is displayed in Figure 2.

Actual Desired
Respons Respons
Input Algorithm e e Teacher

Input and
Error desired
output

Figure 2. Diagrammatic representation of the supervised learning approach (Σ represents summation)

 D.D. Finlay et al. / Mining, Knowledge and Decision Support 163

2.2.2. Unsupervised Learning

In unsupervised learning, the algorithm is also presented with a number of example
cases, each consisting of the relevant features. As well as being more difficult to
conceptualise and visualise, a big difference with this approach is that the algorithm is
not given the corresponding outcome for each training example. Instead, based on the
presented features, examples are clustered. When new unseen cases are subsequently
presented, these are added to an already existing cluster or group. The algorithm itself
determines possible segregating properties.Even though unsupervised learning does not
require details of the desired output, some methods require details or guidelines on how
the final groupings are to be organised, to avoid the situation where the end result may
not be successful.

2.3. Performance Metrics

2.3.1. Accuracy
When trying to understand how well a given interpretation algorithm performs it is
desirable to quantify how well it predicts the presence or absence of disease. This
quantification is based on a comparison with the actual or true outcomes. The easiest
way to achieve this is to calculate the percentage of cases that have been correctly
classified. Refereed to as ‘accuracy’ (ACC), this can be defined as [9]:

x
ACC(%) = 100 × (1)
N

where x is the number of subjects that have been correctly classified and N is the total
number of subjects in the dataset. Although this measure gives an indication of overall
interpretation performance, no information can be attained from this metric regarding
the breakdown of the performance of a particular group within the dataset, e.g. how
many ‘normals’ (i.e. people without the disease) have been correctly assigned, or how
many ‘abnormals’ (i.e. people with the disease) have been correctly assigned. To
achieve this further measures of ‘sensitivity’ and ‘specificity’ are required.
2.3.2 Sensitivity and Specificity
Sensitivity (SEN) is a measure of how well a given interpretation algorithm can predict
the presence of disease. Specificity (SPE), on the other hand, is how well a given
interpretation algorithm can predict the absence of disease (SPE).

These are defined as [10]:

a
SEN (%) = 100 × (2)
a+c
d
SPE(%) = 100 × (3)
b+d
where a is the number of true positives, indicated as positive by the given algorithm, b
is the number of false positives (i.e. true negatives that are indicated as positive by the
given algorithm), c is the number of false negatives (i.e. true positives that are indicated
as negative by the given algorithm), and d is the number of true negatives indicated as
164 D.D. Finlay et al. / Mining, Knowledge and Decision Support

negative by the given algorithm. In general a high level of SEN will be exhibited by an
algorithm that performs well in the detection of presence of disease, a high level of
SPE shall be exhibited by a classifier that performs well in the detection of absence of
disease.

3. Interpretation Algorithms

In this section, we look at actual interpretation algorithms. We begin by discussing

common algorithms that avail of supervised learning.

3.1. Examples of Supervised Methods

3.1.1. Statistical
The main objective of the statistical approach is the allocation of a test case to one of a
number of possible diagnostic categories, with minimum probability of miscalculation
[11]. When using such a classification technique, the background knowledge which is
known is used to manipulate the data into a number of groups within the data set as
opposed to using the knowledge to design the ‘rules’ for the algorithm itself. Once the
data has been arranged, it is then possible to calculate a number of statistical measures
relating to how likely a specific feature is in belonging to a specific category. These
types of ‘probability’ measurements form the underlying basis of the statistical
interpretation process.
3.1.2. Induction Algorithm
An induction algorithm builds a set of rules from a set of training examples which can
then be used to assign unseen examples. After development, these rules are often
represented as a ‘decision tree’ structure. In this type of system, a hierarchy of decision
nodes connected by branches forms a decision tree. Each node represents a binary test,
the result of each test determines the path through the tree network. After the rule set
has been established, this type of algorithm is easy to implement, although the
induction algorithms themselves are complex. The main disadvantage of the resulting
rule set is their rigid nature i.e. only a very strict set of conditions can be tested for with
sharp boundaries between normal and abnormal.
3.1.3. Neural Networks
Neural Networks (NNs) are vast interconnections of elements called neurons. The
biological neuron itself is a building block of the brain and in NNs an attempt is made
to replicate the behavior of these neurons. An artificial neuron consists of a number of
inputs which are weighted before reaching a main body or processing unit, the signal is
processed and passed to an output. The operation of the NN depends greatly on how
the neurons are interconnected. After design, a NN must be trained. During this process
the NN adjusts its own internal parameters causing the network to converge to the
desired response. Although in this section we are focusing on supervised techniques,
NNs can also be configured to work in an unsupervised manner.
 D.D. Finlay et al. / Mining, Knowledge and Decision Support 165

3.1.4. Combining Classifiers/Mixture of Experts

In this approach, several interpretation algorithms are employed simultaneously with
the aim of providing superior performance in solving complex problems. The first of
two such approaches is known as ‘Ensemble Combination’ [12]. Here, each algorithm
in the ensemble grouping attempts to solve the same problems and derives its own
solution. A voting system is then employed to determine which outcome was the most
popular among the majority of techniques. This outcome is then taken forward as the
final accuracy.
In another approach, known as modular learning [12], each algorithm is presented
with a different subsection of a particular problem. In this approach, each algorithm
usually only focuses on a relatively minor part of the signal or image being classified
and this often provides superior accuracy because the overall problem is divided into
several simpler problems.

3.2. Examples of Unsupervised Methods

Traditional unsupervised methods include hierarchical clustering, k-means and the

Kohonen Self-Organizing Feature Maps (SOM) [13]. These common techniques are
described below.
3.2.1. Hierarchical Clustering
The hierarchical clustering approach constructs clusters in either a bottom-up
(agglomerative) or a top-down (divisive) manner. The agglomerative method starts
with every data sample in a single cluster. Then, it repeatedly merges the closest pair of
clusters into a single cluster until all data samples are in one cluster. Depending on the
way of defining similarity between clusters, an agglomerative clustering approach can
be implemented in several ways, such as single linkage, average linkage and complete
linkage. As a top-down technique, a divisive method starts with all the data samples in
one single cluster, and then successively splits them into smaller clusters until each
cluster contains only one sample. This method is outlined in Table 2. Hierarchical
clustering is conceptually simple and relatively easy to implement. By visualising a
dendrogram, which diagrammatically summarize the whole hierarchal clustering
process, basic relationships between all the data samples can be obtained. However, the
deterministic nature of this technique makes it impossible to make any adjustments and
corrections once a data sample is assigned to a node.
3.2.2. K-Means Clustering
A k-means method is an algorithm that divides a data set into k disjoint partitions such
that a certain metric relative to the centroids of k clusters is minimized. The algorithm
is composed of several steps. In the first step, k initial cluster centroids are randomly
determined. Then each sample is assigned to the closest centroid based on a distance
function. Once this is done, the new centroid is recalculated for each cluster by taking
the mean of all the cluster members. The process is iterative until no change in the
centroids is observed. For a large data set, k-means technique is computationally faster
than hierarchical clustering. Nevertheless, the performance of k-means clustering
greatly depends on selection of initial seeds. It could get stuck at a local minimum with
poor quality. Moreover, the number of clusters needs to de defined at the onset. In
166 D.D. Finlay et al. / Mining, Knowledge and Decision Support

some cases, it could be difficult to predict what the number of k should be. Fixed
number of clusters can make it difficult to predict what k should be.

Table 2: Basic Divisive Hierarchical Clustering Algorithm

1: Initialise: Start with one all-inclusive cluster

2: Repeat
3: Calculate pair-wise distances within each cluster
4: Find two data samples that are in the same cluster but have a largest distance
5: Use these two samples as seed points to create two new clusters
6: Assign all samples in the original cluster to two new clusters that have the closest seed
7: Until each cluster contains only one sample

3.2.3. Self-Organising Map

Based on the combination of a competitive learning principle with a topological
structuring of neurons, the SOM is perhaps one of the most widely used unsupervised
neural networks in the literature. It implements an ordered dimensionality-reducing
mapping of high-dimensional data sets. A SOM generally involves an architecture
consisting of a 2-dimensional Kohonen layer, in which each neuron receives the same
input from an input layer, as illustrated in Figure 3. Upon completing a learning
process, the SOM is able to provide a graphical representation of clustering distribution
with neighbouring neurons tending to define related clusters. A detailed description of
this algorithm, along with the selection of its learning parameters can be found in [13].
Similar to the k-means algorithm, the SOM inherently requires users to specify the
network topology and the number of neurons in advance. Such a fixed, predetermined
grid structure has negatively influenced its applications in the context of knowledge
discovery.
3.2.4. Utility of Unsupervised Techniques
The potential for the use of unsupervised algorithms is often much less obvious than
that of their supervised counterparts and in the following paragraphs we provide some
examples of their utilisation as reported in the literature. The ability to reveal the
patterns hidden in a given data set certainly lends unsupervised learning techniques to
many applications. From a knowledge discovery perspective, various clustering
methods have received great attention. Bortolan and Pedrycz [14] presented a SOM-
based interactive framework for ECG signal analysis, which allows users to gain an
insight into the structure of the data under investigation. Douglas et al. [15] used the k-
means algorithm to perform automatic segmentation of magnetic resonance images
(MRI) of the residual limb of trans-femoral amputees into limb and background. Other
studies include the utilization of the SOM for spatiotemporal analysis and classification
of body surface potential map (BSPM) data investigated by Simelius et al. [16]. This
study indicates that, unlike traditional QRS integral maps which miss time-dependent
 D.D. Finlay et al. / Mining, Knowledge and Decision Support 167

features of activation process and have to compress the spatiotemporal information of

the whole QRS complex into one spatial distribution, the SOM-based approach can
incorporate the temporal dynamics and thus take advantage of both the whole spatial
distribution and the temporal development during the QRS.

Figure 3. A typical SOM architecture consisting of two layers: an input layer with i input samples and a 2D
layer which contains a grid of neurons. The rectangle in dotted lines represents a topological neighborhood,
Nc(t), of winning neuron, Wj, which is monotonically decreased as training progresses.

In bioinformatics, pioneering studies include investigations by Eisen et al. [17] and

Golub et al. [18]. Eisen et al. [17] used an agglomerative hierarchical clustering
method to group genes according to their gene expression similarity. By visualizing the
output dendrogram, the inherent data structure encoded in microarray data can be
conveyed in a form that is intuitive for biologists. Golub et al. [18] successfully applied
SOMs to distinguish types and subtypes of leukemia based on gene expression patterns.

4. Future Trends and Possible Applications of DSS Techniques

In the introductory paragraphs of this article we have discussed DSSs mainly in the
context of clinical care. There is also scope for utilizing DSSs in everyday life to
support independent living [19]. We now live in an age where we can take more control
of our own health by using readily available devices which support personal vital sign
monitoring. The availability of such devices provides the opportunity for the
deployment of decision support systems outside the boundaries of primary or
secondary healthcare establishments. If we consider very basic home based measuring
168 D.D. Finlay et al. / Mining, Knowledge and Decision Support

devices such as the blood pressure monitor or blood glucose monitor, it is possible to
add a simple algorithm which would sound an alarm if an excessive measurement were
recorded. Obviously this intermediate form of assessment is merely a pre-cursor to
assessment by a healthcare professional; however, it is a step forward in terms of the
automated analysis of vital signs outside of the conventional clinical setting. In the
following paragraphs we discuss three, albeit closely related, areas which we believe
define new trends and challenges in decision support.

4.1. Wearable Computing

Recent technological advancements in the domain of ‘wearable computing’ have the

potential to streamline the process of vital signs monitoring and subsequently decision
support. In particular, developments in ‘smart clothing’ allow sensors and electrodes to
be seamlessly woven into garments. This provides the ability to measure a wide range
of vital signs from a subject when they wear a specially designed shirt. Obviously this
streamlines the process of vital signs monitoring as the subject need not worry about
the attachment of a number of discrete sensors and recording apparatus.This potentially
provides much more freedom and scope for vital signs monitoring as much more
information can be recorded in a less obtrusive manner whilst the patient goes about
their everyday tasks. Although adding additional complexity to the design of the
decision support system, there is now the potential to provide a fuller assessment of a
patient’s status.

4.2. Smart Homes

Whether it is through ‘smart clothing’ or conventional recording apparatus, vital signs

monitoring has traditionally been an integral part of home based monitoring. However,
it is now also possible to record more general activities of the person within their home
and their interaction with domotic devices. It has emerged that much more can be
gained by complementing and correlating traditional vital signs with this behavioral
information. For example, a scenario of an increased heart rate should not be
considered as alarming if it is also known that the person has just run up a flight of
stairs. On the other hand, there may be reason for concern if a person’s blood sugar
levels are classified as being abnormal and the person has not been in the kitchen to
cook food all day. Although this approach has the potential to define a new paradigm in
healthcare delivery, a big challenge will be processing the endless information and the
complex scenarios that are possible. It is believed that manipulation and further
development of existing decision support systems will make this realizable.

4.3. Internet Based Healthcare

With all of the aforementioned scenarios, the decision support systems will be required
to analyse information from a number of different sources. Given complementary
advances within the telecommunications domain it is possible for all of the information
to be analysed on the device or system within the home environment or relayed to some
central healthcare facility. The advantage of the latter is the remote support which may
be offered to a person within the comfort of their own home. For example, a decision
support system based in a hospital may receive information regarding a possible alarm
for a person’s high blood pressure. In addition, the system may receive information
 D.D. Finlay et al. / Mining, Knowledge and Decision Support 169

suggesting the person has not taken their anti-hypertensive medication. Based on this
information the decision support system could raise an alarm resulting in a healthcare
professional calling the person at home to check on their general status. Although this
is a simplistic scenario, it conveys the end-to-end element of healthcare monitoring and
delivery which is directing the future development and deployment of medical decision
support systems.

5. Case study: Mining for Electrocardiographic Information

DSSs have the ability to assist the decision making process in many applications. These
range from classical clinical decision making to more contemporary applications
associated with independent living. The developments in DSSs have also spurred the
development of elaborate computational techniques and algorithms which in turn have
created new opportunities evident in the emergence of new domains such as knowledge
discovery and data mining. These domains and the tools used have also opened up new
possibilities for gaining an understanding of and streamlining the process of healthcare
delivery. In this case study we look at how some of the techniques that form part of the
DSS process can be utilized for solving less conventional problems. We approach this
discussion not by focusing on the technical rigors of this application, but on the
application and need itself.
A medical discipline that has gained a lot from these developments is cardiology,
particularly electrocardiology. This particular area saw one of the first applications of
computers in medicine [20], and computerized electro cardiology is now a domain in
its own right [21]. The application of computers in this area is widespread with
computerized enhancement of the acquisition process right through to computerized
interpretation and diagnosis being commonplace. In more recent years computers and
particular data mining techniques have been applied in an attempt to gain more of an
understanding of the principles of electrocardiology and of the underlying
electrophysiology. To provide a typical example of how these techniques can be
applied we look at the application of data mining techniques in the selection of
electrocardiographic recording sites. The most effective way to capture
electrocardiographic information is to record information from recording sites which
are positioned all over the torso. Such an approach is referred to as body surface
potential mapping and can involve recording information from in excess of 200
recording sites. This technique is deemed to provide the most diagnostic information,
as effectively all information as projected on to the body surface is recorded. Although
comprehensive, this technique is seldom used in clinical practice due to the excessive
number of recording sites required and the associated complexity of the hardware. A
more favored technique is the 12-lead ECG which as the name suggests renders just 12
channels of information for interpretation. Although widely used, the 12-lead ECG is
also met with some skepticism as it is well appreciated that the locations of the
recording sites are known to be sub optimal. There therefore is a very valid research
question of where should electrocardiographic information be recorded from a patient’s
torso to increase diagnostic yield. Assuming that something around the number of
recording sites that are required to record the 12-lead ECG is suitable, many
investigators have set about analyzing body surface potential maps to find which of the
200+ recording sites available yield the most information. Historically this analysis has
170 D.D. Finlay et al. / Mining, Knowledge and Decision Support

been conducted using statistical methods [22]. However, recently more contemporary
techniques have been endorsed.
The problem of recording site selection can be described as trying to locate the
leads in a body surface potential map that yield the most diagnostic information. If we
take this a step further and think of this problem in the context of automated
classification, we can think of the information from each electrocardiographic lead as a
feature and we want to reduce the number of these features prior to classification. We
have therefore defined a ‘feature selection’ problem. As described earlier, feature
selection deals with the removal of redundant variables, and the retention of those
which are deemed more useful. This is in contrast to feature extraction, which
transforms the features and hence will not tell us which electrocardiographic leads we
should measure.
Regardless of the problem domain, feature selection would involve exhaustively
evaluating all possible combinations of input features and choosing the best subset. In
reality, the computational cost will be prohibitive if there are many features to be
considered. This is likely to be the case in the ECG lead selection problem where, for
this example, we have in excess of 200 features and hence leads. For this reason, there
has been an interest in developing algorithms and strategies that locate optimal features
at low computational cost [23].

6. Summary

In this review, we have examined the role of decision support in medicine.

Development of DSS is often undertaken by the biomedical engineer, medical physicist
or medical information, in collaboration with clinicians. The transfer of domain
knowledge is a key element to this process. Selection of important features for the
decision making process is crucial to the success of this. Performance can be measured
by statistical metrics: accuracy, sensitivity and specificity. Interpretation algorithms
may use supervised or unsupervised learning. A case study to support decision making
in the domain of electro cardiology has been presented.

Glossary

ACC Accuracy
BSPM Body surface potential map
DSS Decision support system
EPR Electronic patient records
MRI Magnetic resonance imaging
NN Neural Network
QMR Quick medical reference
SEN Sensitivity
SOM Self organizing map
SPE Specificity
 D.D. Finlay et al. / Mining, Knowledge and Decision Support 171

References

[1] F.T. de Dombal, D.J. Leaper, J.R. Staniland, A.P. McCann and J.C. Horrocks, Computer aided
diagnosis of acute abdominal pain, British Medical Journal 2 (1972), 9-13.
[2] E.H. Shortliffe, Computer Based Medical Consultations: MYCIN, Elsevier, North Holland, New York,
1976.
[3] J.D. Myers, The Background of INTERNIST-I and QMR, in A History of Medical Informatics, eds.
Bruce I. Blum and Karen Duncan (New York: ACM Press, 1990), 427-433.
[4] E. Coiera, Guide to Medical Informatics, The Internet and Telemedicine, Chapman & Hall Medical,
1992.
[5] N. Wyse, R. Dubes, and A. K. Jain, A critical evaluation of intrinsic dimensionality algorithms, in
Pattern Recognition in Practice, E. S. Gelsema, and A. K. Jain, Eds. Morgan Kaufmann Publishers,
Inc, 1980, 415-425.
[6] L. Huan, and H. Motoda, Feature transformation and subset selection. IEEE Intelligent Systems, 13
(1998), 26-28.
[7] A. Jain, and D. Zongker, Feature selection: Evaluation, application, and small sample performance,
IEEE Transactions on Pattern Analysis and Machine Intelligence 19 (1998), 153-158.
[8] J. Reunanen, Overfitting in making comparison between variable selection methods, Journal of
Machine Learning Research, 3 (2003), 1371-1382.
[9] C. D. Nugent, PhD. Dissertation, University of Ulster, 1998.
[10] P. M. Rautaharju, H. W. Blackburn, and J. W. Warren, Review: The concepts of sensitivity, specificity
and accuracy in evaluation of electrocardiographic, vectorcardiographic and polarcardiographic criteria,
Journal of Electrocardiology, 9 (1976), 275-281.
[11] J. A. Kors, J. H. van Bemmel, Classification Methods for Computerised Interpretation of the
Electrocardiogram, Methods of Information in Medicine, 29 (1990), 330-336.
[12] J.C. Sharkley, Combining Artificial Neural Nets: Ensemble and Modular Multi-Net Systems. London:
Springer, 1999.
[13] T. Kohonen, Self-Organising Maps. Heidelberg, Germany: Springer-Verlag, 1995.
[14] G. Bortolan and W. Pedrycz, An interactive framework for an analysis of ECG signals, Artificial
Intelligence in Medicine, 24 (2002), 109-132.
[15] T. S. Douglas, S. E. Solomonidis, V. S. P. Lee, W. D. Spence, W. A. Sandham and D. M. Hadley,
Automatic segmentation of magnetic resonance images of the trans-femoral residual limb, Medical
Engineering & Physics 20 (1999), 756-763.
[16] K. Simelius, M. Stenroos, L. Reinhardt, J. Nenonen, I. Tierala, M. Mäkijärvi, L. Toivonen, and T.
Katila, Spatiotemporal characterization of paced cardiac activation with body surface potential mapping
and self-organizing maps, Physiological Measurement 24 (2003), 805-816.
[17] M. B. Eisen, P. T. Spellman, P. O. Brown, and D. Botstein, Cluster analysis and display of genome-
wide expression patterns, Proc Natl Acad Sci USA, 95 (1998) 14863-14868.
[18] T. R. Golub, D. K. Slonim, P. Tamayo, C. Huard, M. Gassenbeck, J. P. Mesirov, H. Coller, M. L Loh,
J. R. Downing, M. A. Caligiuri, C. D. Bloomfield and E. S. Lander, Molecular classification of cancer:
class discovery and class prediction by gene expression monitoring, Science 286 (1999), 531-537.
[19] C. D. Nugnet, J Augusto, Smart Homes and Beyond. Amsterdam, Netherlands: IOS Press, 2006.
[20] R.J.A. Schijvenaars, Intraindividual variability of electrocardiogram. Assessment and exploiuting in
computerized ECG analysis, PhD Dissertation, University Rotterdam, 2000.
[21] www.isce.org, accessed Mar 2008.
[22] F. Kornreich, P. M. Rautaharju, J. Warren, T. J. Montague, and B. M. Horacek, Identification of best
electrocardiographic leads for diagnosing myocardial infarction by statistical analysis of body surface
potential maps, American Journal of Cardiology, 56 (1985), 852-856.
[23] D. D. Finlay, C. D. Nugent, P. J. McCullagh, and N. D. Black, Mining for diagnostic information in
body surface potential maps: A comparison of feature selection techniques, Biomedical Engineering
Online, 4 (2005) 51.
172 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-172

III.4. Remote Healthcare Monitoring and

Assessment
Chris D. NUGENTa, Dewar FINLAYa, Richard DAVIESa, Mark DONNELLYa,
Josef HALLBERGb, Norman D. BLACKa and David CRAIGc
a
School of Computing and Mathematics, Faculty of Engineering, University of
Ulster, Shore Road, Northern Ireland, BT37 0QB
b
Luleå University of Technology, Sweden
c
Belfast City Hospital/Queens University of Belfast, Northern Ireland

1. Introduction to Remote Healthcare Monitoring

Remote healthcare monitoring is the process of assessing the well-being of a patient

when the patient themselves and their healthcare professional are not physically
together in the same room. Conventionally, patient assessment in Primary and
Secondary healthcare provision involves a face-to-face consultation between the patient
and the healthcare professional (General Practitioner, Consultant etc.). Advances in
technology, specifically medical devices, sensors and high speed fixed and wireless
communication networks have now made it possible to bring the assessment process to
the patient, as opposed to limiting the assessment to the constraints of hospitals and
doctors’ surgeries. As a result, it is now possible for a patient to be assessed whilst in
the comfort of their home or to have their vital signs monitored whilst, for example,
shopping or at work. In addition, it is possible for patients to visit their local Doctor’s
surgery and benefit from expert consultants and receive their advice, without having to
have a face-to-face meeting with them.
It is the aim of this review to introduce the concepts of remote healthcare
monitoring, discuss the drivers which have lead to the realisation and uptake of this
approach and to look at the benefits that evolving technologies can offer in this domain.

1.1. Changes in Population Demographics

We are at present witnessing a change in the way our population is represented, in

terms of age profile. In the 1950s the largest percentage of the population was made up
by those in the category of 0-9 years of age. At the turn of the century this category was
amongst one of the largest, with a similar level of numbers being found in other age
categories. Predictions for the next thirty years have shown that this trend will continue
and infact result in a profile of age categories with almost similar sizes across all age
groups (Figure 1). People are living longer and the total population size is increasing.
In addition, the percentage of our population represented by those aged 65 and over is
steadily increasing [1]. From a healthcare perspective this offers a number of
challenges. In the first instance, we are faced by the simple fact that the population is
 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment 173

now larger and hence places a larger demand on healthcare services. Secondly, there is
an increased prevalence of healthcare problems and instances of disease within the
elderly and thirdly there are now fewer younger people to care for the elderly [2].
Taking all of this into account has resulted in healthcare providers, governments
and members of society, searching for new paradigms of healthcare delivery.

Population Catergories
85+ 2030

70-80
2000

40-50

1950's

0-10
Population (Males) Population (Females)

Figure 1 Estimated profile of age categories in 1950, 2000 and predictions for 2030.

1.2. The Impact of Technology from Social and Economic Perspectives

We have previously introduced the concept that the increase in the size of the
population has a direct impact on the prevalence of instances of healthcare problems
and of long term chronic diseases. In addition to placing a resource burden on primary
and secondary healthcare organisations, this increase in population size has additional
societal and economic implications. There are benefits from both perspectives which
would support the introduction of remote healthcare monitoring.
From a societal perspective, people are wishing to play a more active role in their
own healthcare. This can be facilitated, at least notionally, via the realisation of remote
healthcare monitoring where the patient is given a higher degree of responsibility to
become more proactive with the management of their own healthcare and its
assessment. Involvement in such a manner can also lead to the patient having a greater
understanding of not only their condition, but also its means of assessment. In addition,
there is the general appreciation that people recover more quickly in their own homes
and would prefer to remain in their own homes, even if they are at greater risk [3].
Remaining at home for an extended period of time can also be linked to improvements
in the patient’s perceived quality of life. Taking this into account, the use of remote
healthcare monitoring has the potential to avoid instances of institutionalisation when
the patient requires high levels of healthcare assessment. It can also avoid lengthy and
174 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment

inconvenient trips being made for the purposes of routine clinical assessments. Patients
can also benefit from remote expert clinical diagnosis. For example, it is possible to
send a patient’s clinical details from one institution to another and receive expert
clinical assessment via remote means.
In addition to the societal benefits, deployment of technology to support healthcare
assessment via remote means has a number of economic benefits. If a person can have
their healthcare assessed within their own homes, removing the requirement of
temporary or permanent institutionalisation, then significant healthcare savings can be
gained. Additionally, those patients receiving clinical care within a hospital setting can
be discharged in a much shorter space of time and continue to have routine assessments
conducted via remote means. Finally, consultants can offer remote consultations to
patients all over the world without having to leave their own office. This offers benefits
to the patient in terms of them receiving treatment from 'world class' specialists who,
because of geographical distances, would otherwise not be able to provide such
treatment.

1.3. Models of Patient-Healthcare Professional Interaction

Traditionally, the means by which those involved in healthcare provision interact with
their patients is based on a bi-lateral communication model (Figure 2 (a)). In this
respect it is normal for the patient to have direct contact with a number of healthcare
providers, however, commonly there is a potential lack of a unified infrastructure to
support communication between healthcare providers. Deployment of technology and
the establishment of care models facilitates communication channels to be established
between both patients and healthcare providers as well as intercommunication between
the healthcare providers (Figure 2 (b)). This offers an improved means of healthcare
delivery from the patient’s perspective as all those involved in the delivery of their care
can interact.

Physician
Physician
Professional
Healthcare
Carer

Patient
Healthcare
Patient
Professional

Pharmacist

(b) Enhanced stakeholder

(a) Interaction between interaction facilitated via
stakeholders: the classic technological communications
situation

Figure 2 Examples of Patient-Healthcare Professional interactions.

 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment 175

Although the aforementioned care models can be used to facilitate the deployment
of technology, their conceptualisation can at times be faced with resistance. The
deployment of technology within the realms of healthcare provision can offer benefits
to the patients, however, the changes of practice which may be required from the
healthcare providers’ perspective, in addition to the problem of new service integration,
creates major challenges concerning the widespread deployment of the technology. On
the one hand, drivers such as the benefits of improved treatment results and alleviation
of the problems associated with the shortage of physicians and nurses encourage such
developments. Nevertheless, such innovations are faced with the constraints such as
technology costs, reimbursement issues and legal issues, to name but a few. This
inevitably results in a ‘push-pull’ scenario (Figure 3) from a technological perspective.
The development of new technology results in new paradigms being pushed into the
healthcare market. These can be considered as high risk solutions. If, on the other hand,
a user driven approach is adopted, a pull effect can be witnessed. This may be viewed
by some as the desired effect, where the end solution will be based on mature
technology, which, from a technological perspective may not offer high levels of
innovation, however, from the healthcare perspective can be seen as a low risk solution
meeting the needs of the end users. In order to provide the patient with the best solution
to their needs, a user centered design process offers many benefits. Adopting this
approach allows the patient to be involved during the early stages of the design process.
s
iont
ca

Push Effect
pli
Ap

High risk enabling

w
Ne

new solutions

Pull Effect
gy

Driven by user requirements and

ol o

mature solutions
n
ch
Te
w
Ne

Figure 3 ‘Push-Pull’ effect of technology within healthcare market place.

176 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment

1.4. The impact of the Internet on Healthcare Delivery

The Internet is one of the most widely used and accepted forms of modern day
communication, as well as a rich source of information. Patients can now seek
healthcare advice, research their symptoms and determine possible diagnosis, in
addition to researching available courses of treatments. Additionally, the Internet can
be used as a forum to support communities offering discussion groups dealing with
health and healthcare topics. Furthermore, the Internet offers great potential for
distributed healthcare, which is important especially for sparsely populated rural areas,
and for people with limited mobility.
The Internet makes it possible for user-centric healthcare, where the user’s needs
are in focus. More demands are also placed on being able to locate the necessary
information and help in user understanding. This has created more demands on
information and communication resources, as well as on available healthcare
professionals. Nevertheless, this also means that for many health-related issues, the
user can receive help from centralized healthcare resources. For example, a parent
concerned about a child's skin rash could be taught how to acquire high-resolution
pictures of the affected area and subsequently forward such pictures to a centralized
healthcare service. This approach would offer a fast and efficient means of receiving
direct feedback regarding the condition of the skin, without the necessity of the parent
and child having to travel to the hospital for a physical examination.

2. Telemedicine

This Section aims to introduce the notion of Telemedicine and how technology can be
deployed to link healthcare professionals together in an effort to improve the quality of
the healthcare delivery process.

2.1. What is Telemedicine?

Telemedicine has received many definitions over the years. Most notably it can be
described as the remote use of medical expertise at the point of need [4]. This results in
the use of telecommunications to facilitate the sharing of medical knowledge and the
delivery of healthcare over a distance [5]. With this approach medical expertise can be
delivered to remote sites, rural areas or understaffed regions. lthough Telecare may also
be considered as a form of Telemedicine, for the purposes of this article we will treat
these two topics individually and focus within this Section on the element of
Telemedicine and the associated technology which fosters a collaborative working
environment between medical experts.
Telephony: A rudimentary means of providing telemedicine is via telephone
communications. This can be in the form of two specialists discussing a patient's
diagnosis or prognosis over large distances or similarly can relate to a consultation
between a patient and a specialist. Other examples include automated services which
patients access, via the telephone, to provide details regarding their current state of
health. An example of this type of system is an automated diabetes service which
permits patients to verbally enter information regarding their condition via an
automated telephone system [6]. Such a system facilitates the routine assessment of a
 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment 177

patient’s condition without the need for direct consultation thus alleviating pressure on
healthcare services.
Video Conferencing Systems: Video Conferencing Systems have become of one the
most popular platforms used to facilitate telemedicine and subsequently improve
communications between healthcare professionals [7]. Such systems typically include
the ability to send and receive video, audio, and text between one or more users.
Additionally, some systems provide tools, such as shared whiteboards, which facilitate
the communication of ideas and thoughts via informal means. The main purpose of
using video conferencing systems in telemedicine is to provide a sense of presence
through the provision of visual instructions, or demonstrations to patients, and to
visually present or describe problems and symptoms to healthcare professionals.
Video conferencing provides a wide range of services that healthcare providers can
avail of for example nurses can use video conferencing to make house-calls while
keeping contact with the doctor, whilst patients can use the system to set up meetings
with healthcare professionals from their home, instead of having to travel to the
hospital.
2.1.1. Point-to-Point Ssystems
Similar to video conferencing, point-to-point systems offer commutations between
stakeholders, however, they only provide communications between two points. At
present, several such systems are freely available. While these systems are often
sufficient in providing simple communication between a patient and the doctor, they do
not offer the support group functionality that video conferencing systems do.
2.1.2. Remote Surgery
At the other end of the spectrum from telephony systems exists the concept of remote
surgery. This technology provides the means to support the delivery of remote care
through robotics [8]. An obvious advantage of this approach is that it permits world
renowned surgeons from across the globe to provide patient care. The first remote
surgery was conducted in 2001 and was referred to as "Operation Lindbergh". Through
the use of three robotic arms setup in an operating theatre in France, surgeons located
in New York where able to remotely perform a gallbladder operation over a fibre optic
connection. Since this date, several other remote surgical procedures have taken place.
Store and Forward Services: In addition to real-time telemedicine systems such as
those already described, offline telemedicine is also possible. Such care delivery does
not require both parties to be present at the same time of communication. For example,
technology within one consultant’s office can be used to record some patient
information which can be subsequently stored and forwarded at a set time to be
reviewed offline by another consultant.

3. Telecare

This Section aims to introduce the notion of Telecare and how technology can be
deployed to link remote patient information with healthcare providers in an effort to
improve the quality and the healthcare delivery process.
178 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment

3.1. What is Telecare?

Telecare combines the usage of sensing devices and telecommunications to support the
remote monitoring of a patient within their home environment with the goal of
providing a means of care support. Therefore, via remote patient monitoring it is
possible for a Telecare service to react to abnormal situations which are a cause for
concern. In response, a Telecare service can issue an alert to a caregiver or a family
member. It is also possible for the person themselves to raise an alarm in instances
when they may require support, for example, following a fall.
3.1.1. Technology Platforms Available for Home Based Monitoring
To provide a home based monitoring system to patients who require additional health
care services requires exploiting current technology platforms that are already in place.
These should be widely available and practical in terms of cost, performance and
speed.
The Public Switched Telephone Network (PSTN) is a commonly used means of
supporting the transmission of information from the patient’s home to a healthcare
organisation. The PSTN is available throughout the world and although originally
based on analogue lines, is now almost entirely digital at its core. The primary use of
the PSTN is the ability to connect people together to provide a phone based voice
service. However, in more recent times, the PSTN has been used as a digital
communication channel allowing information to be transferred to and from patients’
homes. The PSTN itself has a number of drawbacks in comparison with more recent
technologies based on the same infrastructure. It can be more difficult to set up and is
expensive to run if the application is resource intensive and operates at relatively slow
speeds making it suitable for only a small percentage of applications.
Broadband technology is another possibility. As its name suggests, broadband
technology offers a wider band of information. One example of broadband is a Digital
Subscriber Line (DSL) which encodes digital information at a higher frequency on one
channel and sends voice on a lower frequency forming another voice based channel.
DSL technology offers high availability, as it is based upon the already existing PSTN
system which has worldwide coverage. There are a number of advantages with
broadband technology. Firstly, it is easy to set up and easy to use and can even support
wireless connectivity. Although dependent on line conditions such as the distance from
the exchange, the speed and reliability are greatly improved. Broadband technology,
and DSL in particular, are essential to providing the underlying communication
infrastructure to allow home based services to become successful.
3.1.2. Telecare Services
One of the most common forms of Telecare services exist in the form of alarm based
pendants. Alarm based pendants are essentially a device worn by a patient which can
be used in either a passive or active manner. Their primary use is within the home
environment and can be used to trigger various alarms. A passive pendant is one which
involves no interaction between the patient and the device; this could take the form of a
fall detector which passively monitors the orientation of the patient during home based
activities. Falls are common place among elderly patients living alone and are the
leading cause for such people having to enter into permanent institutional care. Such a
device could alert a relative or health care personnel to a serious situation. A more
obvious and active approach of an alarm pendant is one that is activated by the patient
 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment 179

whenever a security situation arises. Once the pendant has been activated, an
emergency message is relayed to a healthcare professional or family member so that
the necessary intervention can take place.
In addition to pendant alarms, Telecare services can be based upon a number of
devices which may exist within the home environment, for example medication
management devices, cognitive prosthetics, fire/smoke/water alarms and various door
and device interaction sensors. The exchange of physiological data recorded from the
patient in their home and transferred to a remote site can be referred to as Telehealth.
Although the patient and the communication infrastructure for both Telecare and
Telehealth are the same, it has been usual in the past to keep these two terms separate.
Typical parameters which can be monitored within the auspices of Telehealth include
blood pressure, electrocardiogram, weight and blood glucose levels.

4. Ambient and Pervasive Computing

The way in which healthcare can be delivered is changing rapidly. The convergence of
mobile communications, decision support systems and Internet computing are offering
a wealth of new healthcare paradigms. In this Section, we focus on a number of new
and emerging technologies and demonstrate how they can be deployed within the
realms of home based healthcare.

4.1. Intelligent Homes

Perhaps the most notable advance in care delivery has been through the introduction of
the Intelligent Home environment [9]. This environment aims to promote the use of
embedded technology to support a person within their own living environment. The
desired goal is to offer a means of independent living and extend the period of time a
person can remain in their own home prior to institutionalisation {2}. Within such an
environment, it is common to find two types of devices; sensors and actuators. Sensors
are the devices which can record information about the person in the environment.
These may be motion sensors detecting which room the person is in, or they may be
pressure sensors detecting whether the person is in, for example, a chair or their bed.
Other types of sensors include temperature sensors, door sensors, water sensors,
appliance sensors etc. Such sensors can provide sufficient information so that the
current status of the person can be inferred and subsequently be reused to recommend
how the environment should be modified or controlled. The environment itself is
managed through the use of actuators. These can, for example, control the ambient
temperature if the environment becomes too warm, or raise an alarm if a hazardous
situation arises. For example, a person turns on the cooker and then turns on the taps in
the bath. The challenge at present is to find the correct balance between the information
gathered by the sensors and the ensuing processing to support the dynamic change of
the environment through the use of the actuators to support the changing needs of the
person.
180 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment

4.2. Wearable Systems and Smart Clothes

The ability to monitor a number of vital signs from a patient in a pervasive manner has
been realised through the introduction of Wearable Computing or Smart Clothes [10].
Smart clothing is a result of the consolidation of sensor technologies, advanced textiles
and information and communication technologies [11]. Sensing devices can be
embedded into clothing to offer a means of recording information such as heart rate,
perspiration levels, body motion / movement and breathing rates. Smart clothing offers
the ability to record from a larger surface area on the body, for example the torso,
hence the limitations of taking recordings from the finger or the wrist can be avoided
[12]. It is possible, during the manufacture of the garment, to fabricate the sensors by
either coating a textile with a form of sensing material or to form a sensing material
which can then be knitted or woven into the garment. Although many challenges now
exist regarding which are the optimal places to locate the recording electrodes [13] and
what information should be recorded, smart clothing provides a realistic solution to
continuously monitor the person in an unconstrained manner and either provide a
means of local processing or relay the information to a central location for assessment
by healthcare staff. Smart clothing has huge potential in conjunction with the
aforementioned intelligent environments as an augmented means, not only to assess the
activities of the person, but also, their health status.

4.3. Data Processing and Context Awareness

We have described a number of paradigms whereby information relating to the patient

can be shared between medical professionals or can be viewed remotely by a medical
professional to offer a means of support. Given the ability to store large amounts of
data about the person, their activities and their general health conditions through a
number of vital sign markers, it has now become possible to consider deploying data
processing systems with the ability to automatically understand the current status of the
person. One of the biggest areas of interest at present is the ability to monitor changes
in lifestyle and adapt the technology within the environment to support the changing
requirements of the patient. As such, much effort is being directed towards the
development of data processing systems which aim not only to detect trends within a
person’s behaviour but to try to understand the cause of such behaviour and attempt to
correlate this with other social or environmental behaviours.
In addition to data processing, the introduction of Context Aware computing
within healthcare and intelligent environments has also become more prevalent [14].
The term ‘context’ refers to any information which can be used to characterise the
situation of a person or computational entity. Context aware computing is used to
extract, interpret and then use contextual information in such a way that the system or
environment can adapt its current state of operation to match the current context of its
use.

5. Case Study: Home Based Medication Management

We will now show how some of the aforementioned topics have been addressed in a
real project scenario – Home Based Medication Management. Within this context we
present an exemplar of how the problem domain has been identified, how a technical
 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment 181

solution has been deployed and how emerging technologies have been integrated into
the second generation of the system.

5.1. Problem Domain

It is a well recognised problem that patients do not fully adhere to their medication
regimen having received a prescribed course of medication. This results in a
significant burden being placed on the healthcare service due to extended patient
welfare and healthcare costs. According to the World Health Organisation (WHO)
adherence is defined as “the extent to which a person’s behaviour – taking medication,
following a diet, and/or executing lifestyle changes, corresponds with agreed
recommendations from a healthcare provider” [15]. There have been numerous
attempts to quantify the impact of non-adherence to medication, including:
- one third of people take all their medication, one third take some and one third do
not take any at all.
- as many as 125,000 deaths may occur each year in the US as a direct result of
non-adherence to medication.
- up to 23% of admissions to nursing homes can be attributed to non-adherence.
Adherence to medication is a complex issue and stems from problems such as the
patient not understanding their medication regimen, perceived side effects and financial
issues [16]. These can also be further complicated by any confusion which may arise
through the context in which the medication is prescribed, delivered and taken by the
patient.

5.2. Potential Deployment of Technology

Given the huge impact associated with non-adherence, many efforts have been made
to deploy technology as a potential means to alleviate the effects of the problem [17].
Technology has been deployed in three different manners. In the first instance, pill
holders can be offered to the patient which store the medication in labeled
compartments for various times throughout the day. The second type of device is an
extension of the first, however, electronic modules support the inclusion of pre-
programmable alarms. The third is a monitoring device which provides a remote means
to assess if the patient has taken their medication. This last type of device will be the
focus of this Case Study.

5.3. The MEDICATE system

The aim of the MEDICATE system was to develop an Internet based care model and
associated peripherals to support the needs of all stakeholders within the prescribe to
intake chain of medication. The anticipated benefits of the system were the ability to
improve and support the patient with the management of their medication in addition to
providing a means to support communications between all stakeholders. To achieve
this vision it was necessary to develop a suite of interfaces which could be used by
each of the stakeholders. The interfaces took the form of both custom electronic
devices and software interfaces all connected via an Internet based care model as
shown in Figure 4.
182 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment

Figure 4 Overview of MEDICATE care model to support home based medication management.

This care model supported communication between the patient within their home
and their remote formal or informal carer, in addition to a communication channel
between healthcare professionals. Figure 5 below shows the various stakeholder
interfaces. Briefly these include:
Patient: a mobile medication management system (Figure 5 (a)) and a base station
acting as a reservoir of medication and a means to connect the patient to the Internet
portal. These devices had the ability to store the medication and would remind the
patient at the appropriate time to take their medication. Any instances of non-adherence
would be recorded by these devices
Doctor: a web based interface (Figure 5 (b)) to support the prescribing of
medication and assessment of patient adherence. This interface provides a means for
the medication regime of the patient to be entered onto the system.
Pharmacist: a software interface (Figure 5 (c)) to support the filling of medication
containers to be used by the patient’s medication device according to their prescribed
medication regimen. This system has the ability to retrieve the information entered onto
the system by the Doctor.
Caregiver: this software interface (Figure 5 (d)) provides a means to allow the
patient’s adherence to their medication regimen to be monitored in real time and, in
instances of non-adherence, will raise the attention of the care giving staff.
Within this scenario we have demonstrated how the basic concepts of utilising
modern day communications have supported the integration of a number of
stakeholders in the supply-to-intake chain of medication. Evaluation of this system has
demonstrated its usefulness in terms of supporting a patient’s adherence to medication.
 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment 183

Nevertheless, it has also highlighted the importance which should be given to the
integration of such systems into existing practice.
The natural evolution of this system has involved the deployment of the concepts
of medication management onto mobile phone based platforms [17]. These facilitate a
similar means of providing management and of monitoring of a patient’s intake of
medication, in addition to maintaining the link with healthcare professionals. An
additional extension to the system would involve integration with context aware
services, where the means by which the reminder was offered to the patient would be
dependent upon their current activity.

a: Mobile medication device b: Doctor‘s interface

c: Pharmacists‘s interface d: Caregiver‘s interface

Figure 5 Stakeholder interfaces for the MEDICATE system.

5.4. Conclusions

Overall, within this review we have shown how technology can be used to establish a
number of care paradigms, all of which aim to improve the delivery of care provided to
the patient and to overcome geographical boundaries. In some cases existing
technological infrastructures may be used and in others new forms of technology are
required. Prior to widespread deployment, a number of challenges still remain. These
challenges are only partly related to the technology. Other challenges are more related
to organizational issues and issues relating to who is actually going to pay for the
184 C.D. Nugent et al. / Remote Healthcare Monitoring and Assessment

technology and its associated services. Nevertheless, it is widely accepted that there are
benefits to be accrued, hence it is anticipated that further effort will be directed towards
these areas in future years.

Glossary

PSTN Public Switched Telephone Network

DSL Digital Subscriber Line
WHO World Health Organisation

References

[1] Ageing and life course, World Health Organisation, http://www.who.int/ageing/en/ (accessed May
2008).
[2] M.E. Pollack, Intelligent technology for an aging population: The use of ai to assist elders with
cognitive impairment, AI Magazine, (2005), 9-24.
[3] D.J. Cook and S.K. Das, How smart are our environments? An updated look at the state of the art,
Journal of Pervasive and Mobile Computing, 3 (2007), 53-73.
[4] L. Belochi, Telemedicine Glossary: Glossary of Concepts, Technologies, Standards and Users, 5th
ed., European Commission, Brussels, 2003.
[5] H.R. Fischer, S. Reichlin, J.P. Gutzwiller, A. Dyson, C. Beglinger, Telemedicine as a new
possibility to improve health care delivery, in M-Health: Emerging Mobile Health Systems, eds.
RSH Istepanian, S Laxminarayan, CS Pattichis, (Springer, 2006), 203-218.
[6] L.A. Black, C. McMeel, M. McTear, N. Black, R. Harper, M. Lemon, Journal of Telemedicine and
Telecare, 11 (2005), 6-8.
[7] M. Drugge, J. Hallberg, P. Parnes, K. Synnes, Wearable systems in nursing home care: prototyping
experience, IEEE Pervasive computing, (2006), 86-91.
[8] J. Rosen and B. Hannaford, Doc at a distance, IEEE Spectrum, 43 (2006), 34-39.
[9] J.C. Augusto and C.D. Nugent, Designing Smart Homes: The Role of Artificial Intelligence, LNAI
4008, Springer, 2006.
[10] F. Axisa, P.M. Schmitt, C. Gehin, G. Delhomme, E. McAdams and A. Dittmar, Flexible
technologies and smart clothing for citizen medicine, home healthcare, and disease prevention, IEEE
Transactions on Information Technology in Biomedicine, 9 (2005), 325-336.
[11] C.D. Nugent, P.J. McCullagh, E.T. McAdams and A. Lymberis, Personalised Health Management
Systems: The Integration of Innovative Sesning, Textile, Information and Communication
Technologies, (Amsterdam: IOS Press, 2005).
[12] S. Brady, L. Dunne, A. Lynch, B. Smyth and D. Diamond, Wearable Sensors? What is there to
sense?, in Personalised Health Management Systems: The Integration of Innovative Sensing, Textile,
Information and Communication Technologies, (Amsterdam: IOS Press, 2005), 80-88.
[13] M. Donnelly, C. Nugent, D. Finlay, P. McCullagh and N. Black, Making smart shirts smarter:
Optimal electrode placement for cardiac assessment, International Journal of Assistive Robotics and
Mechatronics, 8 (2007), 53-60.
[14] S.S. Intille, Designing a home of the future, IEEE Pervasive Computing, 1 (2002), 76-82.
[15] World Health Organisation (WHO), Adherence to Long-Term Therapies, Evidence for Action, 2003.
[16] L. Blonde, Removing polytherapy as a barrier to adherence, Managed Care, Compliance and
Persistence with Medication Therapy, 9 (2000), 1.
[17] C.D. Nugent, D. Finlay, R. Davies, M. Mulvenna, J. Wallace, C. Paggetti, E. Tamburini, N. Black,
The next generation of mobile medication management solutions, International Journal of Electronic
Healthcare, 3 (2007), 7-31.
Basic Engineering for Medics and Biologists 185
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-185

Introduction to
Chapter IV: Biomaterials and
Tissue Engineering
Fergal J. O’BRIEN and Brian O’CONNELL (eds.)

The goal of tissue engineering is to develop cell, construct, and living system
technologies to restore the structure and functional mechanical properties of damaged
or degenerated tissue. The term “tissue engineering”, which is interchangeably used
with the more recent term of regenerative medicine, was officially coined at a National
Science Foundation workshop in 1988 to mean “the application of principles and
methods of engineering and life sciences toward fundamental understanding of
structure-function relationships in normal and pathological mammalian tissues and the
development of biological substitutes to restore, maintain or improve tissue function.”
While the field of tissue engineering may be relatively new, the idea of replacing tissue
with another goes as far back as the 16th century when an Italian, Gasparo Tagliacozzi
(1546-99), Professor of Surgery and Anatomy at the Bologna University described a
nose replacement that he had constructed from a forearm flap in his work ‘De
Custorum Chirurigia per Insitionem” (The Surgery of Defects by Implantation) which
was published in 1597. In modern times, the techniques of transplanting tissue from
one site to another in the same patient (an autograft) or from one individual to another
(transplant or allograft) have been revolutionary and lifesaving. However major
problems exist with both techniques. Harvesting autografts is expensive, painful,
constrained by anatomical limitations and associated with donor-site morbidity due to
infection and haemorrhage. Transplants have serious constraints. The major problem is
accessing enough tissue and organs for all of the patients who require them.
Transplants are strongly associated with rejection by the patient’s immune system and
they are also limited by the potential risks of introducing infection or disease.
The field of tissue engineering is highly multidisciplinary and draws on experts
from mechanical engineering, materials science, surgery, genetics, and related
disciplines from engineering and the life sciences. Tissue engineering technologies are
based on a biological triad and involve the successful interaction between three
components: (1) the scaffold that holds the cells together to create the tissue’s physical
form i.e. acts as a template for new tissue formation, (2) the cells that synthesise the
tissue and (3) signalling mechanisms (i.e. mechanical and/or chemical signals) that
direct the cells to express the desired tissue phenotype. The specific properties of each
of these three components is critically important in determining the quality of the
186 F.J. O’Brien and B. O’Connell / Biomaterials and Tissue Engineering

engineered tissue and therefore the potential for clinical success when the engineered
tissue has been implanted into an injured site in vivo. For example, the choice of
biomaterial and the fabrication process used to produce the porous scaffold which will
act as the template for in vitro tissue formation should be selected based on the specific
tissue type and anatomical site in which it will be implanted. Similarly, consideration
must be given to the choice of cell type used, for example, whether the cells should be
autologous or allogeneic and whether stem cells or primary mature cells should be used.
Finally, the process of cellular differentiation into a specific phenotype followed by the
production of extracellular matrix is stimulated by either the provision of specific
growth factors and cytokines and/or by subjecting a cell-seeded construct to
biophysical stimulation in the form of a bioreactor. The types of growth factors,
choice of bioreactor and duration of exposure all play a key role in determining the
success of the engineered tissue. The mechanical properties of the scaffold and
engineered tissue will govern the ability of the construct to function in vivo and to
allow infiltration of host cells and in most tissues (cartilage being an exception),
vascularisation. It is therefore important to develop techniques to test biomaterial
scaffold and tissue properties prior to implantation and to be able to predict how the
constructs will behave when implanted in vivo.
This chapter focuses on all three areas of the tissue engineering triad: Scaffolds &
Surfaces, Cellular & Molecular Biomechanics and Bioreactors in Tissue Engineering.
In addition, the fourth part of the chapter is devoted to Characterisation and Testing of
Biomaterials. It is pitched at a level that should allow either an engineer or a clinician
to understand the basic principles of tissue engineering and the fundamental interaction
between cells & scaffolds and their response to biophysical stimuli and the different
types of bioreactor that exist. The chapter finishes with a section which discusses the
fundamentals of materials testing which will be of interest to clinicians and biologists
involved not only in tissue engineering, but in any area of biomechanical analysis. The
contributors are all experts in their fields and are involved in research in tissue
engineering and a number of cognate disciplines and come from all across Europe.
They include an anatomist/bioengineer (O’Brien), dentist (O’Connell), materials
scientist (Partap), orthopaedic surgeon (Lyons), physiologist (Campbell), engineers
(Plunkett, Dendorfer and Hammer) and trauma surgeon (Lenich).
Basic Engineering for Medics and Biologists 187
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-187

IV.1. Scaffolds & Surfaces

Sonia PARTAP a,b, Frank LYONS a,b and Fergal J. O’BRIEN a,b
a
Department of Anatomy, Royal College of Surgeons in Ireland, 123 St. Stephen’s
Green, Dublin 2, Ireland
b
Trinity Centre for Bioengineering, Department of Mechanical Engineering,
Trinity College Dublin, Dublin 2, Ireland

Introduction

Every day thousands of clinical procedures are performed to replace or repair tissues in
the human body that have been damaged through disease or trauma. Current therapies
are focused on the replacement of the damaged tissue by using donor graft tissues
(autografts, allografts or xenografts). Problems associated with this approach include
shortage of donors or donor sites, the volume of donor tissue that can be safely
harvested, donor site pain and morbidity, the possibility of harmful immune responses,
transmission of disease and rejection of grafts [1]. Alternatively, the field of tissue
engineering (a phrase that is interchangeably used with regenerative medicine) aims to
regenerate damaged tissues instead of replacing them (with grafts) by developing
biological substitutes that restore, maintain or improve tissue function [2, 3].
In native tissues, cells are held within an extracellular matrix (ECM) which guides
development and directs regeneration of the tissue, serves to organise cells in space and
provides them with environmental signals to direct cellular behaviour. The goal of
tissue engineering is to synthesise substitutes that mimic the natural ECM to help guide
the growth of new functional tissue in vitro or in vivo. At a simplistic level, biological
tissues consist of cells, signalling mechanisms and extracellular matrix. Tissue
engineering technologies are based on this biological triad and involve the successful
interaction between three components: (1) the scaffold that holds the cells together to
create the tissues physical form, (2) the cells that create the tissue and, (3) the
biological signalling molecules (such as growth factors) that direct the cells to express
the desired tissue phenotype (Figure 1). Tissue engineering is a multidisciplinary field
that harnesses expertise and knowledge from a variety of fields, including those of the
medical profession, materials scientists, engineers, chemists and biologists.

Why are Scaffolds Required ? 2D vs. 3D Culture

There are differences in cell behavior in three dimensional (3-D) vs. two dimensional
(2-D) environments. In vitro 3-D cell culture conditions more accurately model in vivo
biological responses, as the conditions more closely resemble the natural structure and
function of tissues in vivo [4]. These conditions can be created by using a 3-D scaffold
that acts as a template, allowing cells to produce and deposit extracellular matrix
(ECM) that would not be possible in 2-D environments. 2-D cell culture does not allow
cells to move or assemble with the freedom they have in vivo, and thus cannot replicate
188 S. Partap et al. / Scaffolds & Surfaces

the effects of nutrient gradients, signal propagation or the development of bulk

mechanical properties. Studying these cells in 3-D models allows us to better
understand their biochemical and biophysical signaling responses as they would
normally occur in vivo, particularly the external signals occurring in the ECM, as well
as the mechanical and chemical signals arising from both adjacent and even distant
cells [5]. This approach can lead to the generation of more accurate cell- based assays
for engineering of suitable biomaterials that can be used to determine the cell-material
interaction.

Figure 1. The tissue engineering triad; factors that need to be considered when designing a suitable structure
for tissue engineering applications.

1. Properties of Scaffolds for Tissue Engineering

All scaffolds for tissue engineering applications are designed to perform the following
functions: (1) to encourage cell-material interactions i.e. cell attachment, differentiation
and proliferation, eventually leading to the deposition of extracellular matrix, (2) to
permit the transport of nutrients, wastes and biological signalling factors to allow for
cell survival, (3) to biodegrade at a controllable rate which approximates the rate of
natural tissue regeneration, and (4) to provoke a minimal immune and/or inflammatory
response in vivo. The following parameters must be considered when designing a
scaffold for tissue engineering.
 S. Partap et al. / Scaffolds & Surfaces 189

1.1. Biocompatibility

The implantation of a scaffold may elicit different tissue responses depending on the
composition of the scaffold. If the scaffold is non-toxic and degradable, new tissue will
eventually replace it; if it is non-toxic and biologically active then the scaffold will
integrate with the surrounding tissue. However, if the scaffold is biologically inactive,
it may be encapsulated by a fibrous capsule, and in the worst case scenario if the
scaffold is toxic, rejection of the scaffold and localised death of the surrounding tissue
can occur [6]. Biocompatibility is the ability of the scaffold to perform in a specific
application without eliciting a harmful immune or inflammatory reaction. For a
scaffold to positively interact with cells and with minimal disruption to the surrounding
tissue, it should have an appropriate surface chemistry to allow for cellular attachment,
differentiation and proliferation. Cells primarily interact with scaffolds via chemical
groups on the material surface or topographical features. Topographical features
include surface roughness and pores where cell attachment is favoured. Alternatively,
cells may recognise and subsequently bind to the arginine-glycine-aspartic acid (RGD)
cell adhesion ligand. Scaffolds synthesised from natural extracellular materials (e.g.
collagen) already possess this specific ligands, whereas scaffolds made from synthetic
materials may be designed to deliberately incorporate them.

1.2. Biodegradability

The severity of an immune or inflammatory reaction is not only determined by the

actual scaffold itself, but is also dependent on the scaffold’s degradation products.
Ideally, scaffolds are designed to be completely replaced by the regenerated
extracellular matrix by integrating with the surrounding tissue, eliminating the need for
further surgery to remove it [7]. Scaffolds should degrade with a controllable
degradation rate, (approximating the rate of natural tissue regeneration), as well as with
controllable degradation products. As it degrades, the breakdown products should be
non-toxic and easily excreted from the body via metabolic pathways or the renal
filtration system [8].

1.3. Mechanical Properties

The scaffold provides structural integrity to the engineered tissue in the short term.
Furthermore, it provides a framework for the three dimensional (3-D) organisation of
the developing tissue as well as providing mechanical stability to support the growing
tissue during in vitro and/or in vivo growth phases [9]. The mechanical properties of the
scaffold should be designed to meet the specific requirements of the tissue to be
regenerated at the defect site. Furthermore, at the time of implantation, the scaffold
should have sufficient mechanical integrity to allow for handling by the clinician, be
able to withstand the mechanical forces imposed on it during the implantation
procedure and survive under physiological conditions. Immediately after implantation,
the scaffold should provide a minimal level of biomechanical function that should
progressively improve until normal tissue function has been restored, at which point the
construct should have fully integrated with the surrounding host tissue.
190 S. Partap et al. / Scaffolds & Surfaces

1.4. Scaffold Architecture

Porous structures allow for optimal interaction of the scaffold with cells. The pore
architecture is characterised by pore size and shape, pore interconnectivity/tortuosity,
degree of porosity and surface area. The microstructure determines cell interactions
with the scaffold, as well as molecular transport (movement of nutrients, wastes and
biological chemicals e.g. growth factors) within the scaffold. Specifically, pore size
determines the cell seeding efficiency into the scaffold [10]; very small pores prevent
the cells from penetrating the scaffold, whilst very large pores prevent cell attachment
due to a reduced area and therefore, available ligand density. Subsequently, cell
migration within a scaffold is determined by degree of porosity and pore
interconnectivity/tortuosity. A scaffold with an open and interconnected pore network,
and a high degree of porosity (>90 %) is ideal for the scaffold to interact and integrate
with the host tissue [11].

1.5. Manufacturing Technology

In order for a scaffold or engineered construct to become commercially available in a

clinical setting, the cost effectiveness of it should be considered; particularly when it is
to be scaled up from making one at a time in a research laboratory to a production
process allowing small batch quantities of 100 to 1000 constructs to be made. In
addition, as clinicians ideally would prefer “off the shelf” products that may be used
routinely, it is important to take into consideration how the constructs will be
transported and stored in clinical environments. The cost effectiveness will be
determined by the choice of biomaterial, which will in turn affect the selection of
fabrication method. Many different techniques have been used to fabricate scaffolds
for tissue engineering (Figure 2). The following summarises the most commonly used
methods.
1.5.1. Particulate Leaching Methods
Particulate leaching is a technique that uses solid particles of a particular size to act as a
template for the pores; water soluble particles are frequently used as they can easily be
leached out of the final product by simply washing the final product with water. In
solvent casting-particulate leaching, a polymer dissolved in a solvent is mixed with salt
particles in a mould; the solvent is then evaporated to give a polymer monolith
embedded with the salt particles, these are then removed by washing the scaffold with
water, resulting in the formation of a porous scaffold [12]. Another variation of this
technique is melt moulding-particulate leaching: in this particular technique the
polymer is cast into a mould with the embedded solid porogen. The polymer is set by
applying heat and pressure, and again the porogen is leached away by washing the
resulting product with water to yield a porous polymer scaffold [13].
1.5.2. Phase Separation
Various forms of phase separation techniques enable the creation of porous structures.
A two phase polymer system that is homogenous can become thermodynamically
unstable by altering the temperature leading to (1) liquid/liquid or (2) liquid/solid phase
separations. In the first, a polymer is dissolved in a molten solvent, a liquid/liquid
phase separation (where one phase is concentrated in polymer whilst the other is not) is
achieved by lowering the temperature. The two phase liquid is quenched to yield a two
 S. Partap et al. / Scaffolds & Surfaces 191

phase solid, and the solvent is then removed yielding a porous polymer, this is known
as thermally induced phase separation (TIPS) [14]. In the second, a polymer is
dispersed in a solvent which is then frozen to induce crystallisation of the solvent to
form solvent crystals that act as templates for the pores. These crystals are then
removed by freeze drying to yield a porous foam. Manipulation of the processing
conditions enables the creation of different pore sizes and distributions [15].
1.5.3. Foaming
Foaming techniques use gaseous porogens that are produced by chemical reactions
during polymerisation, or are generated by the escape of gases during a temperature
increase or drop in pressure. Nam et al. 2000 [16] synthesised poly (lactic acid) [PLA]
scaffolds using ammonium bicarbonate which acted as both a gas foaming agent and as
a solid salt porogen, an increase in temperature caused the formation of carbon dioxide
and ammonia to create a highly porous foam. Also, high pressure carbon dioxide can
be used to foam polymers by saturating a prefabricated polymer monolith. A
subsequent reduction in pressure causes a decrease in solubility of the carbon dioxide
within the polymer, and as the carbon dioxide gas tries to escape it causes the
nucleation and growth of bubbles resulting in a porous microstructure [17].
1.5.4. Emulsion Templating
Porous structures can also be obtained by using emulsion templating techniques. The
internal phase of the emulsion acts as a template for the pores whilst polymerisation
occurs in the continuous phase (in which the monomer is dissolved). After
polymerisation, the internal phase is removed to give a templated porous material. The
resulting porous microstructures are replicas of the internal phase droplets around
which polymerisations were performed. The size of the emulsion droplets is preserved,
producing polymer foams with approximately the same size and size distributions as
that of the emulsion at the point of polymerisation [18].
1.5.5. Solid Free Form (SFF) Fabrication
Solid free form (SFF) fabrication or rapid prototyping (RP) technologies uses layer
manufacturing techniques to create three dimensional scaffolds directly from computer
generated files. There are a few techniques that come under this group including
stereolithography, selective laser sintering, fused depositional modeling and three
dimensional printing. However, all the techniques share the same principle where
powders or liquids are solidified one layer at a time to gradually build a three-
dimensional scaffold. The layering is controlled by computer assisted design (CAD)
programs where the scaffold architecture is designed and modelled. Data collected
from computed tomography (CT) or magnetic resonance imaging (MRI) scans may
also be used to create CAD models that are specific to the tissue to be regenerated [19].
1.5.6. Combination of Techniques
The techniques discussed above can also be combined with each other depending on
the exact requirements of the scaffold, e.g. phase separation (freeze drying) techniques
can be combined with emulsion templating processes. Whang et al. 1995 created an
emulsion that was quenched using liquid nitrogen, which was then freeze dried to
produce porous PLGA polymeric monoliths [20].
192 S. Partap et al. / Scaffolds & Surfaces

Figure 2 Scanning electron microscopy images of porous (a) collagen-GAG scaffolds made by freeze
drying15, (b) poly-L-lactide (PLLA) foams made by solvent casting-particulate leaching12, (c) alginate
scaffolds made by emulsion templating18 and (d) polycaprolactone–calcium phosphate composites made by
solid free form fabrication methods51.

2. Biomaterials in Tissue Engineering

A number of different categories of biomaterials are commonly used as scaffolds for

tissue engineering.

2.1. Ceramics

Ceramics (inorganic, non metallic materials) used within the biomedical field are
classified as being either bioinert or bioactive. The bioinert ceramics include materials
such as alumina and zirconia that are typically used as implants for musculoskeletal,
oral and maxillofacial applications whilst the bioactive group include the calcium
phosphates, the bioglasses and glass-ceramics [6]. All bioceramics are also further
defined as being osteoconductive (supporting bone growth) or osteoinductive
(stimulating bone growth); all types of bioceramics are osteoconductive as all support
the formation of bone, but not all are osteoinductive. The calcium phosphate based
bioceramics, bioglasses and glass-ceramics are commonly used as scaffolds for bone
tissue engineering as they have a compositional similarity to the mineral phase of bone
 S. Partap et al. / Scaffolds & Surfaces 193

[21]. Hydroxyapatite (HA) and tri-calcium phosphate (TCP) are two of the most
commonly used calcium phosphate bioceramics in tissue engineering applications. TCP
is used as a degradable scaffold, whilst HA, which is non-resorbable and has the added
advantage of being osteoinductive, is typically used for coating biomedical implants to
induce bone regeneration, allowing the implant to integrate with the surrounding tissue.
For this reason, HA has shown much popularity for use as a scaffold for tissue
engineering.

2.2. Synthetic Polymers

The mechanical, physical and biological properties of synthetic polymers can be

tailored to give a wide range of controllable properties that are more predictable than
materials obtained from natural sources. The advantage of using synthetic materials is
that the resulting properties can be customised by adjusting the ratios of the monomer
units (basic building blocks of the final polymer) and by the incorporation of specific
groups (e.g. RGD peptide that cells can recognise). Also, the degradation rate and
products can be controlled by the appropriate selection of the segments to form
breakdown products that can either be metabolised into harmless products or can be
excreted via the renal filtration system [8]. Among the many biodegradable synthetic
polymers used for tissue engineering applications, there are numerous reports on the
use of polylactic acid (PLA), polyglycolic acid (PGA) and their copolymers poly (DL-
lactic-co-glycolic acid) (PLGA), which are approved by the US Food and Drug
Administration (FDA). These polymers degrade by hydrolytic mechanisms and are
commonly used because their degradation products can be removed from the body as
carbon dioxide and water. However, a disadvantage is that there is a lowering of the pH
in the localised region resulting in inflammatory responses when they do degrade.
Polycaprolactone (PCL) has a very similar structure to PLA and PGA and is also
degraded via hydrolytic mechanisms under physiologic conditions (Figure 3). In
addition, it is degraded enzymatically and the resulting low molecular weight
fragments are reportedly taken up by macrophages and degraded intracellularly. It is
predominantly used for drug delivery devices because it has a slower degradation rate
than PGA and PLA. However, more recently, it is increasingly finding applications in
tissue engineering [22]. Traditionally, polyurethanes were used in the biomedical field
as blood contacting materials for cardiovascular devices, and were intended to be used
as non-degradable coatings. More recently they have been designed to be
biodegradable by being combined with degradable polymers such as PLA for soft
tissue engineering applications [14]. Poly(ethyleneglycol) [PEG] is a biocompatible,
non-toxic, water soluble polymer that is a liquid at cold temperatures and elastic gel at
37 oC [23]. PEG based copolymers have been used as injectable scaffolds for bone as
well as for drug delivery applications [24]. Also, copolymers of PEG and PLA have
been created where the degradation rate and hydrophilicity could be controlled by
adjusting the ratio of the hydrophilic (PEG) to hydrophobic (PLA) blocks.
194 S. Partap et al. / Scaffolds & Surfaces

Figure 3 Chemical structures of some biodegradable synthetic polymers used as scaffolds in tissue
engineering applications

2.3. Natural Polymers

Natural polymers offer an alternative to synthetic polymer systems (which intrinsically

lack cell recognition signals) as they can more closely mimic the natural extracellular
matrix of tissues. Alginate and chitosan are two natural polysaccharides that do not
exist within the human body but have been investigated for tissue engineering
applications because they are structurally similar to the glycosaminoglycans (GAGs)
found in the natural extracellular matrix of tissues i.e. skin, bone, and blood vessels.
Alginate originates from seaweed and is attractive because of its low toxicity, water
solubility and its simple gelation chemistry with calcium ions. Alginate hydrogels have
been investigated for use as scaffolds for cartilage [25] and liver regeneration [26], as
well as for wound dressings [27]. Chitosan is a derivative of naturally occurring chitin
which is found in the exoskeletons of crustaceans. It has a low toxicity and is
biocompatible. Chitosan scaffolds have been investigated for skin and bone tissue
engineering [28].
Given the importance of GAGs in stimulating normal tissue growth, the use of
GAGs as components of a scaffold for tissue engineering appears to be a logical
approach for scaffold development. Hyaluronic acid (sometimes referred to as
hyaluronan) is one of the largest GAG components found in the natural extracellular
matrix of all soft tissues and synovial fluid of joints [29]. The applications of pure
hyaluronic acid in tissue engineering applications are limited because of its easy
dissolution in water and fast biodegradation in biological environments. However, it
can be chemically modified to produce a more hydrophobic molecule, thus reducing its
solubility in water. Hyaluronic acid scaffolds are known to be biocompatible, and cells
easily adhere to and proliferate on this material. Hyaluronic acid also plays a significant
role in wound healing and can be modified for drug delivery applications.
 S. Partap et al. / Scaffolds & Surfaces 195

Structural proteins such as fibrin are also utilised in tissue engineering applications.
Fibrin can be used as a natural wound healing material, and has found applications as a
sealant and adhesive in surgery. It can be produced from the patient’s own blood, to be
used as an autologous scaffold. However, the stability of the material is limited as it
can be easily degraded unless apronitin, a protein inhibitor, is used to control the rate of
degradation. Fibrin hydrogels have been used to engineer tissues with smooth muscle
cells [30] and chondrocytes [31]. Alternatively, gelatin (a derivative of collagen) that is
produced by altering the helical structure of the collagen molecule by breaking it into
single strand molecules) has been investigated for cartilage tissue regeneration. [32].
However, as one of the main disadvantages of gelatin is its poor mechanical strength, it
has also been cross-linked with hyaluronic acid for skin tissue engineering, and with
alginate for wound healing applications [33]. Instead, collagen, the main component
found in the extracellular matrix of mammalian connective tissues has found use in
tissue engineering applications including skin substitutes [34], scaffolds for bone and
cartilage, vascular applications and as drug delivery systems. As is typical of all natural
polymers, collagen gels also display poor mechanical properties. However, these can be
improved by employing both chemical and physical crosslinking methods. Physical
crosslinking methods include UV radiation and dehydrothermal treatments, whilst
cross-linking agents such as glutaraldehyde and carbodiimides (EDAC) can be used to
produce chemically cross-linked collagen hydrogels with improved physical properties
(Figure 4).

Figure 4 Chemical structures of some natural polymers used as scaffolds in tissue engineering applications
196 S. Partap et al. / Scaffolds & Surfaces

2.4. Composites

Due to some of the problems associated with using scaffolds synthesised from a single
phase biomaterial (eg. poor mechanical properties and biocompatibility of natural and
synthetic polymers respectively, and poor degradability of bioceramics), a number of
researchers have developed composite scaffolds comprising of two or more phases to
combine the advantageous properties of each phase. For example, polymer/ceramic
composites of poly(lactic-co-glycolic acid (PLGA) and hydroxyapatite have been
investigated for tissue engineering applications [35], whilst Cui et al. [36] have
produced tri-phasic scaffolds by depositing nano-hydroxyapatite particles onto cross-
linked collagen-chitosan matrices. However, even though composite scaffolds such as
these have shown some promise as grafts for bone and cartilage, each one consists of at
least one phase which is not found naturally in the body and therefore has problems
with either biocompatibility or biodegradability or both. Table 1 summarises the
different types of biomaterials described above and lists the advantages and
disadvantages of each type for use as scaffolds in tissue engineering applications.

2.5. Case study: Collagen Scaffolds for Bone Tissue Engineering

From an engineering viewpoint, bone is a composite material made up of both organic

and inorganic phases embedded with bone cells and blood vessels. The main
components of the organic and inorganic phases are collagen and hydroxyapatite,
respectively. The collagen fibres impart tensile strength to the bone whilst the HA
crystals contribute to its stiffness. Based on this, collagen scaffolds are currently being
investigated for bone tissue engineering applications. In our laboratory, we are
currently using porous collagen-glycosaminoglycan (CG) composite scaffolds which
are produced using a lyophilisation (freeze drying) process. The final pore
microstructure of the scaffolds can be varied by controlling the rate and temperature of
freezing during fabrication and the volume fraction of the precipitate [15]. We have
shown that by varying the final freezing temperature during the lyophilisation process a
homologous series of scaffolds with a constant composition and solid volume fraction
with distinctly different pore sizes can be produced [10]. Additionally, experiments
performed in our laboratory using osteoblasts demonstrated that the fraction of cells
attaching to the scaffold decreased with increasing mean pore diameter, indicating that
scaffold ligand density is affected by pore size where an increase in ligand density
causes increased cell attachment. In another study, we have shown that collagen-based
scaffolds seeded with rat mesenchymal stem cells promoted differentiation along
osteogenic and chondrogenic lineages demonstrating their potential for orthopaedic
applications [37]. There is also evidence to suggest that non-seeded collagen scaffolds
with incorporated growth factors implanted into defects induce bone formation [38]. A
problem with collagen-based scaffolds, as with most natural polymer scaffolds, is their
poor mechanical properties. However, these can be improved through physical and
chemical crosslinking methods [39], and allowing bone cells to produce osteoid on the
scaffolds, enabling them to subsequently mineralise the scaffold in vitro prior to
implantation, also leads to improved mechanical properties. Alternatively, as
bioceramics are mechanically stronger and are known to enhance osteoblast
differentiation and proliferation, they have been combined with collagen scaffolds to
form mineralised collagen scaffolds that support cell growth [40, 41].
 S. Partap et al. / Scaffolds & Surfaces 197

Table 1 Properties, advantages and disadvantages of biomaterials used as scaffolds in tissue engineering
applications
198 S. Partap et al. / Scaffolds & Surfaces

There are also reports of triphasic scaffolds made from collagen, a bioceramic and
a synthetic polymer. Scaffolds made from nano-HA, collagen and PLA were placed in
defects of rabbit radius and they integrated with the defect site within 12 weeks [42].
These studies indicate that by finding an adequate balance between pore structure,
mechanical properties and biocompatibility, a collagen-based construct can potentially
support bone growth and may have real potential for bone tissue engineering.

3. Scaffolds: State of the Art and Future Directions

Economic activity within the tissue engineering sector has grown five-fold in the past 5
years. In 2007, approximately 50 companies offered commercially available tissue-
regenerative products or services, with annual sales recorded in excess of $1.3 billion,
whilst 110 development-stage companies with over 55 products in FDA-level clinical
trials and other preclinical stages spent $850 million on development [43]. The tissue
engineering approach was originally conceived to address the gap between patients
waiting for donors and the amount of donors actually available. To date the highest
rates of success have been achieved in the areas of skin regeneration where tissue-
engineered substitutes have been successfully used in patients [44].
However, much research still remains to be performed in all aspects of tissue
engineering [45]. Cellular behaviour is strongly influenced by signals (biochemical and
biomechanical) from the extracellular matrix, the cells are constantly receiving cues
from the extracellular matrix about their environment and are constantly remodelling it
accordingly. Therefore, an appropriate three dimensional structure that is
predominantly thought of as playing a mechanical role is not enough to promote the
growth of new tissue. It is important that the scaffold provides adequate signals (e.g.
through the use of adhesion peptides and growth factors) to the cells, to induce and
maintain them in their desired differentiation stage, for their survival and growth [46].
Thus, equal effort should be made in developing strategies on how to incorporate the
adhesion peptides and growth factors into the scaffolds, as well as in identifying the
chemical identity of adhesion peptides and growth factors that influence cell behaviour,
along with the distributions and concentrations required for successful outcomes. An
example would be to incorporate angiogenic growth factors in scaffolds for different
types of tissue in an attempt to generate vascularised tissues. Tissue vascularisation can
be used to establish blood flow through the engineered tissues and strategies involving
the incorporation of vasculature, as well as innervation will be of great importance [47].
Additionally, the incorporation of drugs (i.e. inflammatory inhibitors and/or antibiotics)
into scaffolds may be used to prevent any possibility of an infection after surgery [48].
The field of biomaterials has played a crucial role in the development of tissue
engineered products. An alternative to using prefabricated scaffolds is to use a polymer
system that is injected directly into the defect site which is polymerised in situ using
either heat [49] (thermoresponsive polymers) or light [50] (photoresponsive polymers).
The advantages for the patient with this approach over current therapies are that
injectable delivery systems fill both regularly and irregularly shaped defects (“get a
custom fit”), they represent a minimally invasive procedure therefore avoiding surgery
and the potential risks associated with it, eliminate the need for donor tissue or a donor
site, and waiting time for treatment is reduced, as it can be used whenever treatment is
required.
 S. Partap et al. / Scaffolds & Surfaces 199

At present, there is a vast amount of research being performed on all aspects of

tissue engineering/regenerative medicine worldwide. Thus, as the field progresses, one
of the key challenges is to try to mimic the sophistication of the natural ECM more
accurately in synthetic substitutes. As more advanced biomaterials and bioreactors are
developed, and as research leads to more knowledge on the cell signaling mechanisms
required to trigger the chain of tissue development, we will undoubtedly get closer
towards our goal of reducing the number of patients waiting for donor tissues.

References

[1] R. Langer, Biomaterials in drug delivery and tissue engineering: One laboratory’s experience, Acc
Chem Res 33 (2000), 94-101.
[2] A. Atala, Tissue engineering and regenerative medicine: Concepts for clinical application,
Rejuvenation Res 7 (2004), 15-31.
[3] L. J. Bonassar and C. A. Vacanti, Tissue engineering: The first decade and beyond, J Cell Biochem
Suppl 30-31 (1998), 297-303.
[4] M. P. Lutolf and J. A. Hubbell, Synthetic biomaterials as instructive extracellular microenvironments
for morphogenesis in tissue engineering, Nat Biotechnol 23 (2005), 47-55.
[5] L. G. Griffith and M. A. Swartz, Capturing complex 3d tissue physiology in vitro, Nat Rev Mol Cell
Biol 7 (2006), 211-24.
[6] L. L. Hench, Bioceramics, J. Am. Ceram. Soc 81 (1998), 1705-28.
[7] J. E. Babensee, A. G. Mikos, J. M. Anderson and L. V. McIntire, Host response to tissue engineered
devices, Adv. Drug Del. Rev. 33 (1998), 111-139.
[8] W. E. Hennink and C. F. van Nostrum, Novel crosslinking methods to design hydrogels, Adv Drug
Deliv Rev 54 (2002), 13-36.
[9] D. W. Hutmacher, Scaffolds in tissue engineering bone and cartilage, Biomaterials 21 (2000), 2529-
2543.
[10] F. J. O’Brien, B. A. Harley, I. V. Yannas and L. J. Gibson, The effect of pore size on cell adhesion in
collagen-gag scaffolds, Biomaterials 26 (2005), 433-41.
[11] T. M. Freyman, I. V. Yannas and L. J. Gibson, Cellular materials as porous scaffolds for tissue
engineering, Prog. Mater Sci. 46 (2001), 273-282.
[12] L. Lu, S. J. Peter, M. D. Lyman, H. L. Lai, S. M. Leite, J. A. Tamada, S. Uyama, J. P. Vacanti, R.
Langer and A. G. Mikos, in vitro and in vivo degradation of porous poly(dl-lactic-co-glycolic acid)
foams, Biomaterials 21 (2000), 1837-45.
[13] S. H. Oh, S. G. Kang, E. S. Kim, S. H. Cho and J. H. Lee, Fabrication and characterization of
hydrophilic poly(lactic-co-glycolic acid)/poly(vinyl alcohol) blend cell scaffolds by melt-molding
particulate-leaching method, Biomaterials 24 (2003), 4011-21.
[14] A. S. Rowlands, S. A. Lim, D. Martin and J. J. Cooper-White, Polyurethane/poly(lactic-co-glycolic)
acid composite scaffolds fabricated by thermally induced phase separation, Biomaterials 28 (2007),
2109-21.
[15] F. J. O’Brien, B. A. Harley, I. V. Yannas and L. Gibson, Influence of freezing rate on pore structure in
freeze-dried collagen-gag scaffolds, Biomaterials 25 (2004), 1077-86.
[16] Y. S. Nam, J. J. Yoon and T. G. Park, A novel fabrication method of macroporous biodegradable
polymer scaffolds using gas foaming salt as a porogen additive, J Biomed Mater Res 53 (2000), 1-7.
[17] D. J. Mooney, D. F. Baldwin, N. P. Suh, J. P. Vacanti and R. Langer, Novel approach to fabricate
porous sponges of poly(d,l-lactic-co-glycolic acid) without the use of organic solvents, Biomaterials 17
(1996), 1417-22.
[18] S. Partap, J. A. Darr, I. U. Rehman and J. R. Jones, "Supercritical carbon dioxide in water" Emulsion
templated synthesis of porous calcium alginate hydrogels, Adv. Mater. 18 (2006), 501-504.
[19] E. Sachlos and J. T. Czernuszka, Making tissue engineering scaffolds work. Review: The application
of solid freeform fabrication technology to the production of tissue engineering scaffolds, Eur Cell
Mater 5 (2003), 39-40.
[20] K. Whang, C. H. Thomas, K. E. Healy and G. Nuber, A novel method to fabricate bioabsorbable
scaffolds, Polymer 36 (1995), 837-842.
[21] K. A. Hing, Bioceramic bone graft substitutes: Influence of porosity and chemistry, Int. J. Appl.
Ceram. Technol. 2 (2005), 184-199.
200 S. Partap et al. / Scaffolds & Surfaces

[22] L. Savarino, N. Baldini, M. Greco, O. Capitani, S. Pinna, S. Valentini, B. Lombardo, M. T. Esposito,

L. Pastore, L. Ambrosio, S. Battista, F. Causa, S. Zeppetelli, V. Guarino and P. A. Netti, The
performance of poly-epsilon-caprolactone scaffolds in a rabbit femur model with and without
autologous stromal cells and bmp4, Biomaterials 28 (2007), 3101-9.
[23] P. J. Martens, S. J. Bryant and K. S. Anseth, Tailoring the degradation of hydrogels formed from
multivinyl poly(ethylene glycol) and poly(vinyl alcohol) macromers for cartilage tissue engineering,
Biomacromolecules 4 (2003), 283-92.
[24] F. Chen, T. Mao, K. Tao, S. Chen, G. Ding and X. Gu, Injectable bone, Br J Oral Maxillofac Surg 41
(2003), 240-3.
[25] W. J. Marijnissen, G. J. van Osch, J. Aigner, S. W. van der Veen, A. P. Hollander, H. L. Verwoerd-
Verhoef and J. A. Verhaar, Alginate as a chondrocyte-delivery substance in combination with a
nonwoven scaffold for cartilage tissue engineering, Biomaterials 23 (2002), 1511-7.
[26] J. Yang, M. Goto, H. Ise, C. S. Cho and T. Akaike, Galactosylated alginate as a scaffold for
hepatocytes entrapment, Biomaterials 23 (2002), 471-9.
[27] D. Bettinger, D. Gore and Y. Humphries, Evaluation of calcium alginate for skin graft donor sites, J
Burn Care Rehabil 16 (1995), 59-61.
[28] C. Mao, J. J. Zhu, Y. F. Hu, Q. Q. Ma, Y. Z. Qiu, A. P. Zhu,W. B. Zhao and J. Shen, Surface
modification using photocrosslinkable chitosan for improving hemocompatibility, Colloids Surf B
Biointerfaces 38 (2004), 47-53.
[29] J. L. Drury and D. J. Mooney, Hydrogels for tissue engineering: Scaffold design variables and
applications, Biomaterials 24 (2003), 4337-4351.
[30] C. L. Cummings, D. Gawlitta, R. M. Nerem and J. P. Stegemann, Properties of engineered vascular
constructs made from collagen, fibrin, and collagen-fibrin mixtures, Biomaterials 25 (2004), 3699-706.
[31] C. J. Hunter, J. K. Mouw and M. E. Levenston, Dynamic compression of chondrocyte-seeded fibrin
gels: Effects on matrix accumulation and mechanical stiffness, Osteoarthritis Cartilage 12 (2004),
117-30.
[32] M. S. Ponticiello, R. M. Schinagl, S. Kadiyala and F. P. Barry, Gelatin-based resorbable sponge as a
carrier matrix for human mesenchymal stem cells in cartilage regeneration therapy, J Biomed Mater
Res 52 (2000), 246-55.
[33] Y. S. Choi, S. R. Hong, Y. M. Lee, K.W. Song, M. H. Park and Y. S. Nam, Studies on gelatin-
containing artificial skin: Ii. Preparation and characterization of cross-linked gelatin-hyaluronate
sponge, J Biomed Mater Res 48 (1999), 631-9.
[34] I. V. Yannas and J. F. Burke, Design of an artificial skin. I. Basic design principles, J Biomed Mater
Res 14 (1980), 65-81.
[35] S. S. Kim, M. Sun Park, O. Jeon, C. Yong Choi and B. S. Kim, Poly(lactide-co-
glycolide)/hydroxyapatite composite scaffolds for bone tissue engineering, Biomaterials 27 (2006),
1399-409.
[36] K. Cui, Y. Zhu, X. H. Wang, Q. L. Feng and F. Z. Cui, A porous scaffold from bone-like powder
loaded in a collagen-chitosan matrix, J. Bioactive and Compatible Polymers 19 (2004), 17-31.
[37] E. Farrell, F. J. O’Brien, P. Doyle, J. Fischer, I. Yannas, B. A. Harley, B. O’Connell, P. J. Prendergast
and V. A. Campbell, A collagen-glycosaminoglycan scaffold supports adult rat mesenchymal stem cell
differentiation along osteogenic and chondrogenic routes, Tissue Eng 12 (2006), 459-68.
[38] M. Murata, B. Z. Huang, T. Shibata, S. Imai, N. Nagai and M. Arisue, Bone augmentation by
recombinant human bmp-2 and collagen on adult rat parietal bone, Int J Oral Maxillofac Surg 28
(1999), 232-7.
[39] M. G. Haugh, Jaasma, M.J. and O’Brien, F.J., Effects of dehydrothermal crosslinking on mechanical
and structural properties of collagen-gag scaffolds, J. Biomed. Mater. Res.: Part A 89 (2009), 363-369
[40] C. V. Rodrigues, P. Serricella, A. B. Linhares, R. M. Guerdes, R. Borojevic, M. A. Rossi, M. E. Duarte
and M. Farina, Characterization of a bovine collagen-hydroxyapatite composite scaffold for bone tissue
engineering, Biomaterials 24 (2003), 4987-97.
[41] A.A. Al-Munajjed, J.P. Gleeson and F. J. O’Brien, Development of a collagen calcium- phosphate
scaffold as a novel bone graft substitute, Stud Health Technol Inform 133 (2008) 11-20.
[42] S. S. Liao, F. Z. Cui, W. Zhang and Q. L. Feng, Hierarchically biomimetic bone scaffold materials:
Nano-ha/collagen/pla composite, J Biomed Mater Res B Appl Biomater 69 (2004), 158-65.
[43] Lysaght M.J., Jaklenec A. and D. E., Great expectations: Private sector activity in tissue engineering,
regenerative medicine, and stem cell therapeutics, Tissue Eng 14 (2008), 305-315.
[44] I. V. Yannas, E. Lee, D. P. Orgill, E. M. Skrabut and G. F. Murphy, Synthesis and characterization of a
model extracellular matrix that induces partial regeneration of adult mammalian skin, Proc Natl Acad
Sci U S A 86 (1989), 933-7.
[45] J. P. Vacanti, Editorial: Tissue engineering: A 20-year personal perspective, Tissue Eng 13 (2007),
231-2.
 S. Partap et al. / Scaffolds & Surfaces 201

[46] C. A. Pangborn and K. A. Athanasiou, Growth factors and fibrochondrocytes in scaffolds, J Orthop
Res 23 (2005), 1184-90.
[47] R. Langer, Tissue engineering: Perspectives, challenges, and future directions, Tissue Eng 13 (2007),
1-2.
[48] M. V. Risbud and M. Sittinger, Tissue engineering: Advances in in vitro cartilage generation, Trends
Biotechnol 20 (2002), 351-6.
[49] L. Klouda and A. G. Mikos, Thermoresponsive hydrogels in biomedical applications, Eur J Pharm
Biopharm 68 (2008), 34-45.
[50] K. T. Nguyen and J. L.West, Photopolymerizable hydrogels for tissue engineering applications,
Biomaterials 23 (2002), 4307-14.
[51] M. J. Mondrinos, R. Dembzynski, L. Lu, V. K. Byrapogu, D. M. Wootton, P. I. Lelkes and J. Zhou,
Porogen-based solid freeform fabrication of polycaprolactone-calcium phosphate scaffolds for tissue
engineering, Biomaterials 27 (2006), 4399-408
202 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-202

IV.2. Cellular & Molecular Biomechanics

Veronica A. CAMPBELLa,b and Brian O’CONNELLa,c
a
Trinity Centre for Bioengineering, Trinity College, Dublin 2, Ireland
b
Department of Physiology, School of Medicine, Trinity College, Dublin 2, Ireland
c
Dublin Dental School, Trinity College, Dublin 2, Ireland

Introduction

Cells exist in a mechanical environment and must be able to elicit an appropriate

response to strains evoked by fluid flow, compression, pressure and stretch. The
ability to detect a range of forces (from 10-4 through to 104 N m-2, reflecting a faint
sound and aortic pressure, respectively) is vital in order to elicit the appropriate cascade
of molecular events that facilitate the physiological processing of sensory information.
The detection of mechanical strain is mediated by mechano-sensitive components of
cells and is a property of a wide range of tissue types. Mechanosensors in the cells of
the skin and ear are critical for processing information about touch and hearing, while
other mechanosensors respond to blood pressure (baroreceptors), muscle stretch
(spindle receptors) and limb positions (proprioceptors). Bone tissue detects mechanical
stress, and this informs the bone remodeling process, while endothelial cells lining
blood vessels respond to shear stress evoked by fluid flow to initiate vascular
remodeling. The ability to respond to mechanical stimuli also regulates fundamental
cellular events such as proliferation, cell spreading, differentiation, motility and the
maintenance of cell shape. The precise cellular mechanisms underlying
mechanosensation remain to be fully resolved, but in recent years a number of surface
proteins have been identified that serve a role as mechanoreceptors. Biological systems
have evolved a complex array of mechanoreceptors to transduce force into an electrical
and/or intracellular biochemical cascade. Such signaling pathways are intricately linked
with cellular responsiveness and adaptation to mechanical stimulation.

1. Biomechanical Signal Transduction

1.1. Mechanosensitive (MS) Ion Channels

Mechanosensitive (MS) ion channels form pores that span the plasma membrane and
are therefore ideally situated to respond to external forces applied to the cell. These
channels serve as mechano-electrical switches converting a mechanical force into a
change in the cellular electrochemical gradient (see Figure 1), and they play a role in
coordinating diverse physiological events such as touch and hearing in mammals, to
adaptation to turgor in bacteria [1, 2]. The patch clamp technique has facilitated the
electrophysiological analysis of single MS channel currents and has demonstrated that
these channels open in response to a hypo-osmotic environment which causes the
 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics 203

membrane to swell, or by the direct application of suction [3]. It has been suggested
that MS channels may be gated directly by changes in the lipid bilayer tension, or in the
‘tethered model’ of gating, by displacement of the MS channel with respect to
components of the cytoskeleton or extracellular matrix [1]. Amongst the
mechanosensitive channels are the transient receptor potential channels (TRPs) [4] and
stretch-activated ion channels (SACs) [5]. However, mechanosensitive channels also
include other ion channels such as voltage-dependent Ca2+ channels [6], the NMDA
receptor [7] and Ca2+-dependent BK channels [8]. Many MS channels are permeable to
cations in general, with highest permeability to Ca2+, which results in an increase in
intracellular Ca2+ concentration following channel opening. It is postulated that the Ca2+
then acts as a second messenger resulting in the appropriate cell response to
mechanical stimulation [9]. Other classes of MS ion channels exhibit selective
permeability for Na2+, K+ or Cl- ions. It is therefore apparent that changes in the
mechanical environment initiate electrical and chemical changes in the cell which
define the cellular response.
MS channels are also modulated by cellular components, such as the contractile
actin cytoskeleton. The cytoskeleton provides support to the plasma membrane and
forms part of the coupling mechanism to the extracellular matrix. It reacts and
rearranges its configuration in response to alterations in the extracellular environment.
The tensegrity model [2] proposes that the cytokeleton is the architectural basis for
cellular mechanotransduction, and application of physical stress promotes changes in
actin polymerization to control cell shape and realignment [10]. A direct coupling may
exist between MS channels and the cytoskeleton since agents that disrupt the
organization of actin filaments, such as colchicine and cytochalasin-D, affect MS
channel activity. Studies suggest that the cytoskeleton exerts a tonic inhibition on MS
channel activity and support the contention that MS ion channels are functionally
linked with the dynamics of the intracellular cytoskeleton [11, 12].

1.2. Transient Receptor Potential Channels

Members of the transient receptor potential (TRP) superfamily of membrane-associated

proteins are recognized to have role in mechanosensation. These ion channels are Ca2+-
permeable and lead to an increase in intracellular Ca2+ concentration upon their
activation, either by facilitating Ca2+ influx, or by promoting the release of Ca2+ from
intracellular stores [13]. TRP channels are widely expressed in the CNS and peripheral
cell types and exhibit substantial evolutionary conservation. The TRP channels are
critically involved in sensory function having an essential role in vision, hearing, taste,
olfaction and mechnosensation. A number of possibilities for mechanical activation of
TRP channels exist [4]. The channels may be activated directly by mechanical
stimulation via forces conveyed through lipid tension or through structural proteins.
Alternatively, these channels may be activated indirectly via another mechano-sensitive
protein that is distinct from the TRP channel, or may be regulated by a diffusible
second messager, such as arachidonic acid, generated by a mechano-sensitive protein
linked to phospholipase A2. The TRP channels are classified in to seven subgroups;
TRPC, TRPM, TRPV, TRPN, TRPA, TRPP and TRPML. TRPV1, the vanilloid
receptor, is responsible for sensing heat [14], yet is also proposed to have a role in
mechanosensation associated with bladder filling [4]. TRPV4 is activated in response
to cell swelling via an indirect mechanism involving second messenger pathways, such
as arachidonic acid [15], whereas the TRPC1 channel is gated directly by the tension
204 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics

that is generated in the plasma membrane [16]. In the nematode, C. elegans, the
mechanosensitive TRPN channel has recently been found to be involved in stretch
receptor-mediated proprioception to regulate sensory-motor integration during
locomotion [17]. Thus, the TRP family of receptors are mechanically regulated via
distinct mechanisms, and are involved in sensing changes in lipid tension precipitated
by stretch and a hypotonic environment.

Figure 1. Cell pathways associated with mechanotransduction. Cells respond to mechanical stimulation via
activation of membrane-associated mechanoreceptor proteins which undergo a conformation change resulting
in the activation and recruitment of a cascade of intacellular signaling molecules. The signaling molecules
may be released into the extracellular matrix (e.g nitric oxide) to influence neighbouring cells, or may target
organelles such as the nucleus and cytoskeleton to evoke changes in gene transcription or cell motility,
respectively.

1.3. Membrane –Associated Enzymes

Enzymes associated with the intracellular domain of the plasma membrane, such as
phospholipase A2, phospholipase C and tyrosine kinases have also been reported to
have a mechanosensitive role [18]. Membrane stretch increases the production of
prostaglandins via phospholipase A2-mediated release of arachidonic acid [19]. In a
number of cell types, flow and shear have been shown to induce activation of
phospholipase C with subsequent activation of the phosphatidylinositol pathway. In
glial cells exposured to pressure, cell proliferation is increased and this effect is
blocked by genistein, the tyrosine kinase inhibitor, but not by blockers of
mechanosensitive channel blockers, implicating tyrosine kinases as mechanosensitive
elements in some cell types [20].
 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics 205

1.4. Intracellular Enzymes and Diffusible Messengers

Mechanical force has been shown to activate the mitogen-activated protein kinase
(MAPK) superfamily [21]. The extracellular-regulated protein kinase (ERK) has been
definitively implicated in mechanotransduction, but a role for p38 and c-jun N-terminal
kinase (JNK) has also been reported [22]. ERK activation has been observed in
response to steady and pulsatile stretch in the vasculature [23], renal mesangial cells
[24], osteoblasts [25] and gingival fibroblasts [26] where downstream consequences of
ERK activation include proliferation, immediate early gene expression and
extracellular matrix accumulation. Furthermore, the strain-induced activation of ERK is
thought to play a role in ‘mechanical strain memory’ in airway smooth muscle,
whereby a mechanically-induced deformation of the smooth muscle cell produces a
structural change which modulates mechanical performance in subsequent contractions
[27].
Mechanical strain activates nitric oxide synthase in endothelial cells [28], stromal
cells [29], chondrocytes [30], and osteocytes [31] leading to the formation of nitric
oxide (NO), a gaseous diffusible second messenger that regulates vascular remodeling,
increases expression of bone-related genes, regulates protoeoglycan synthesis, and
alters cytoskeletal fibre alignment, respectively. The strain-induced changes in NO
formation are coordinated by ERK signaling [32], demonstrating substantial crosstalk
between these intracellular pathways.

1.5. Integrins and Cadherins

Integrins are transmembrane proteins that link components of the intracellular

cytoskeleton to the extracellular matrix. These cell-matrix adhesion proteins can
transmit information about membrane deformation, shear across the membrane, and
pressure transients to the cytoskeleton and associated signal transduction pathways.
Integrins faciltate the rearrangement of the cytoskeleton, and its associated organelles
and nuclei, in order that the cell may be strengthened against mechanical distortion.
Integrins also play a role in the functional adaptation of cells to matrix rigidity where
the Dvß3 integrin has been demonstrated to activate a ridigity response to regulate cell
spreading [33]. Furthermore, ß1/3 integrins are upregulated in ventricular myocytes
exposed to stretch, and the downstream signaling pathways, involving src and focal
adhesion kinase (FAK), results in secretion of vascular endothelial growth factors
which stimulates the induction of other gap junction proteins [34]. This pathway
influences the pattern of cardiac cell-cell communication which is critical for optimal
contractility of cardiac tissue. An integrin-dependent mechanotransduction pathway
involving src signaling and induction of osteogenic proteins has also been reported in
bone cells exposed to mechanical force, providing further evidence that mechanical
force influences osseous regeneration [35]. Thus, mechnical activation of integrin
receptors can regulate dynamic events such as reorganization of the cytoskeleton [33,
36] or the expression of genes encoding proteins associated with cell adhesion and
differentiation [34, 37].
Cell-cell adhesion sites composed of cadherins also provide a mechanism for
mechanotransduction. Cadherins mediate a force-induced increase in Ca2+ influx
through fibroblast mechanosensitive channels [19] and can directly infuence gene
transcription since components of the cadherin junctional complex, have deen
demonstrated to translocate to the nucleus in osteoblasts [38].
206 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics

2. Biomechanics of Differentiation and Maintenance of Tissue Phenotype

2.1. Stem Cells

A number of lines of evidence support the hypothesis that stem cells are responsive to
mechanical stimulation, and that stem cell differentiation may be regulated by strain.
Certainly during development the embryo is exposed to stresses imposed by the
external environment of the uterus, as well as to stresses imposed by neighbouring cells.
These growth-generated strains and pressures can regulate expression of a number of
developmental genes [39]. Administration of a neuromuscular blocker to a chick
embryo causes paralysis, and interferes with the development of the musculoskeletal
system [40], thereby reflecting the importance of the local mechanical environment to
tissue differentiation. At a cellular level, there is substantial evidence demonstrating
that mechanical stresses exert a direct effect on stem cells. For example, mouse
embryonic stem cells generate reactive oxygen species in response to mechanical strain
and this pathway initiates differentiation into cells of the cardiovascular system [41].
Adult mesenchymal stem cells (MSCs) isolated from the marrow are multipotent cells
that have the proclivity to differentiate along osteogenic, chondrogenic and adipogenic
lineages [42]. Human MSCs exposed to cyclic uniaxial strain in the presence of
osteoinductive factors increase the formation of bone-associated extracellular matrix
[43], whilst exposure to low-intensity ultrasound promotes chondrogenesis [44]. MSCs
isolated from adipose tissue respond to mechanical loading applied via pulsating fluid
flow by increasing nitric oxide production and undergoing osteogenesis [45]. As well
as controlling stem cell differentiation potential, mechanical signals are also important
for regulating MSC proliferation via Ca2+ -dependent mechanisms [46]. Thus
mechanical stimulation regimes may have potential to modulate stem cell proliferation
and differentiation profiles for tissue engineering applications (Figure 2).

2.2. Vascular System

The walls of blood vessels react to the mechanical stimuli of pressure and shear stress
evoked by the flowing blood. The responses to such mechanical events are necessary to
control blood flow, or to adapt the vessel structure to its requirements. Shear stress is a
critical biophysical phenomenon for determining vascular homeostasis, vascular
remodeling and development of the cardiac system, as well as for governing the
development of atherosclerotic lesions. The endothelial cells which line blood vessels
are responsible for vascular responses to the shear stress that is evoked by fluid flow.
Shear stress is transmitted from the apical surface of the endothelial cell through the
cytoskeleton to points of attachments at cell-cell and cell-matrix junctions. These
junctions, composed of proteins, such as platelet endothelial cell adhesion molecule
(PECAM), transmit mechanical force via interaction with the intracellular signaling
proteins, Src and P13kinase, leading to activation of the transcription factor, NFNB [47].
In laminar shear, this integrin pathway promotes endothelial cell alignment in the
direction of flow, and a transient activation of the NFNB pathway. In contrast, in the
face of disturbed shear, such as may occur at vessel branch points, bifurcations and
regions of high curvature, the PECAM pathway is activated in a sustained manner. The
sustained activation of NFNB leads to prolonged induction of NFNB-regulated genes,
(e.g adhesion molecules, ICAM, VCAM, E-selectin) which are molecular hallmarks of
 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics 207

atherosclerosis-prone sites. The mechanosensory pathway involving the PECAM

complex is therefore described as one of the earliest-known events in the development
of atherosclerotic lesions. Thus, while this mechanosensory complex is important for
physiological adaptation to fluid flow, emerging evidence suggests that overactivation
of this pathway can lead to cardiovascular disease.

Figure 2. Bone-associated gene expression in stem cells is modulated by constraining the substrate or the
application of uniaxial strain.Effects of one-dimensional clamping (Clamped) and 5% uniaxial cyclic strain
(Clamped & Stretched) (1 Hz) on levels of bone-associated RNAs from mesenchymal stem cell-seeded
collagen glycosaminoglycans scaffold constructs incubated for 7 days in the presence of osteogenic factors,
when compared to unconstrained constructs (Unconstrained). BSP; bone sialoprotein, OCN; osteocalcin.
*p<0.05, **p<0.01. Data courtesy of Dr. Elaine Byrne, RCSI

2.3. Skeletal System

The skeletal system can regulate bone density and architecture in order to meet the
body’s strucural and metabolic demands. While the metabolic demands of the skeleton
are controlled by the calcitropic hormones, the responsiveness of the skeleton to
mechanical load governs skeletal strucure. The cells of the skeleton – osteoblasts,
osteocytes, osteoclasts, chondrocytes, as well as their progenitor bone marrow-derived
cells – display mechnosensitivity which influences how these cells function, and thus
the skeletal characteristics. At a macro-level it is clear that application of load is
anabolic to bone [48], while un-weighting, such as occurs in the absence of gravity,
induces bone loss [49]. At the micro-level a number of cells involved in the
maintenance of bone may be regulated by mechnical force. The marrow stromal
progenitor cells display alterations in proliferation, gene expression, and nitric oxide
formation in response to mechnical load [32], and RANKL expression, a key factor in
osteoclast formation, is increased in response to fluid shear stress [50].
Osteocytes are the principal cell in cortical bone and by extending cellular
processes through a network of fluid-filled cannaliculi, they are ideally placed to
monitor mechanical strain as it is transmitted to the cell surface. Application of
mechanical load increases the production of bone-related proteins [39]. In relation to
208 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics

bone mechanoreceptors, stretch-activated ion channels have been identified in bone

cells [51] and L-type Ca2+ channel have a mechanosensitive role [52]. Mechanical
stimuli across the cell membrane can be transmitted to the cytoskeleton via integrin
receptors [53]. Osteoblasts, osteocytes and chondrocytes express integrin receptors, and
fluid flow upregulates ß1 integrin expression and induces downsteam intracellular
signaling pathways such as src [54, 55]. An additional signaling molecule involved in
bone cell responsiveness to mechanical stimulation is focal adhesion kinase (FAK)
which, upon phosphorylation, activates the mitogen-activated protein kinase (MAPK),
one of the effectors of oscillatory flow in bone cells [56]. Activated MAPK can
translocate to the nucleus to regulate gene transcription, and this represents a
mechanism for long-term adaptation to mechanical stimulation since it involves de
novo protein synthesis. The responsiveness of bone cells to mechanical strain, and thus
bone remodeling, is likely to be affected by disease. In support of this, bone cells
isolated from osteoarthritic donors displayed a reduced metabolic response to fluid
shear stress compared to cells isolated from a patient with osteoporosis (see Figure 3)
[57].
Cartilage homeostasis is also regulated by mechanical stress during embryonic
limb development, as well as during fracture repair and skeletal remodeling in the adult
[58]. Chondrocytes respond to mechanical strain that is evoked by deformation of the
extracellular matrix, as well as to compressive and shear flow forces. Chondrocytes
maintained in a 3-dimensional in vitro environment to mimic the cartilage matrix
environment, and exposed to cyclic matrix deformation, exhibit changes in
proliferation rate and synthesis of cartilage-specific proteins [59]. This stretch-induced
change in chondrocyte proliferation and differentiation was affected differentially by
the Ca2+ channel blocker, nifedipine, and the stretch-activated channel blocker,
gadolinium, suggesting that stretch-induced matrix deformation regulates chondrocyte
proliferation and differentiation via two distinct pathways. Integrin signaling has a role
in the response of chondrocytes to dynamic compression [60], and chondrocytes are
also responsive to shear stress, since flow-induced shear stress increases type II
collagen deposition and tensile mechanical properties [61].

Figure 3. Comparison of osteoporotic and osteoarthritic cell responses to shear fluid flow Effect of pulsating
fluid flow (PFF) at low (0.4±0.1 Pa), medium (0.6± 0.3 Pa), or high shear stress (1.2±0.4 Pa) on nitric oxide
(NO) and prostaglandin E2 (PGE2) production by osteoporotic (OP, black squares) and osteoarthritic (OA,
white circles) bone cells. Data are expressed as PFF-treated over static culture ratios calculated per individual
donor. A; NO response, B; PGE2 response. PFF/Stat=1 means no effect. a implies a significant effect of PFF,
b implies significantly different from OA cells, P<0.05. Reproduced from [57].
 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics 209

Figure 4. Generation of nitric oxide (NO) in a single MC-3T3 osteoblast cell in response to stimulation by
atomic force microscopy. Time-series images of NO fluorescence intensity from 0 s to 1600 s. Periodic
indentation with peak force of ~22nN was applied from 30 s until 60 s at the location indicated (white arrow).
Reproduced from [65].

3. Experimental Models of Cellular Biomechanics

In order to understand the various mechanical transduction pathways mentioned above,

various models have been developed at the molecular, cellular, organ and animal levels
[62]. It is often helpful to use a number of complementary approaches to elucidate both
what a molecule can do as well at what it does do [63]. In general, the study of isolated
molecules provides precise information regarding their properties but may not provide
the context or real life effect of their activation. On the other hand, whole animal
models can give valuable clues to how cellular biomechanics plays out, but they are
subject to a high level of background noise and sometimes difficult to interpret.
The mechanics of some molecules, such as actin and helicase, have been studied
extensively in isolation so that their biological behaviour is fairly well understood.
However, biomechanical responses involve many other molcules that are better
examined in conditions that more closely resemble their native environment. For
example, patch clamping is able to measure currents in membrane channels that are
maintained in a lipid bilayer. Delevopments in atomic force microscopy mean that the
mechanical properties of individual molecules can be explored on a cell’s surface and
the response of a living cell observed in real time. For example, we have characterized
the geodesic F-actin structures that appear in rat mesenchymal stem cells [64] and have
demonstrated that mechanical stimulation of the cytoskeleton in a single bone cell can
lead to the mobilization of nitric oxide (Figure 4) [65].
For practical reasons, most biomechanical responses have been measured in
cultured cell populations, rather than individual cells. This is because a cell response
210 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics

(second messenger production, ion flux, protein phosphorylation) is often more readily
and reliably measured from a large number of cells. Similarly, it may be more feasible
to expose a population of cells to a particular stimulus (stretch, shear, pulsating flow)
than a single cell. However, there are drawbacks with this approach, including the need
to ensure a homogeneous cell population and the assumption that all cells in the
experiment will experience precisely the same strain. Nonetheless, the observation of
biomechanical responses in cells has been very useful in undertanding how cells are
likely to behave in a variety of conditions, for example in coronary heart disease, bone
fractures, during skeletal development or orthodontic tooth movement. An extension of
the purely cell-based experiment is to use an entire tissue or organ in culture to study
biomechanical responses. Isolated muscle, bone, cartilage, tendon or blood vessel may
be maintained long enough in vitro to study cellular responses to applied strain over a
period of hours to perhaps a few days. Improved systems for maintaining tissue
viability, ‘bioreactors’, are discussed elsewhere in this text and they have been used, for
example, to study the response of bone to the mechanical stimulation of artificial
implants [66], the apoptosis of vertebral disc cartilage cells caused by compression [67]
and the stimulation of growth at cranial sutures [68].
Biomechanical studies may be extended further, to the whole animal, to observe a
wide variety of responses, particularly those that depend on the function of entire
systems. For instance, rat hindlimb suspension has been used to compare the cellular
(and matrix) response of bone to the loss of mechanical loading as it would be difficult
to model this effect in an in vitro system [69]. Similarly, cellular changes have been
studied resulting from shear stresses in arterial remodelling [70] and gene expression
caused by mechanical stresses generated following lung injury [71]. Such data may
clarify transduction pathways or even inform clinical decisions [72]. It is important to
note that studies of cellular biomechanics do not necessarily follow the paradigm of
application of stress followed by an observation of altered cell function. The advent of
transgenic and knockout animal models means that key molecules in biomechanical
pathways can be modified or ablated and any change in function revealed [73]. In this
way, valuable new insights into cell function and disease may be gained.

4. Cellular & Molecular Biomechanics: State of the Art and Future Directions

The mechanical environment in which cells exist dictates the physiological attributes of
the cell, and this in turn regulates the function and maintenance of the tissue in which
the cell resides. It is clear that a number of cellular elements serve roles as
mechanoreceptors, and that the intracellular signaling pathways that become activated
by mechanical strain are diverse. Understanding exactly how cells respond to strain is
an emerging area of research, and the knowledge gained from investigations in this area
will be relevant to the tissue engineering field as we strive to develop paradigms with
which to control cellular differentiation and the maintenance of tissue phenotype.

References

[1] B. Martinac, Mechanosensitive ion channels: molecules of mechanotransduction, J Cell Sci. 117
(2004), 2449-60.
 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics 211

[2] D.E. Ingber, Cellular mechanotransduction: putting all the pieces together again, Faseb J. 20 (2006),
811-27.
[3] B. Martinac and A. Kloda, Evolutionary origins of mechanosensitive ion channels, Prog Biophys Mol
Biol. 82 (2003), 11-24.
[4] S.Y. Lin and D.P. Corey, TRP channels in mechanosensation, Curr Opin Neurobiol. 15 (2005), 350-7.
[5] F.B. Kalapesi, J.C. Tan, and M.T. Coroneo, Stretch-activated channels: a mini-review. Are stretch
activated channels an ocular barometer? Clin Experiment Ophthalmol. 33 (2005), 210-7.
[6] A.J. el Haj, L.M. Walker, M.R. Preston, and S.J. Publicover, Mechanotransduction pathways in bone:
calcium fluxes and the role of voltage-operated calcium channels, Med Biol Eng Comput. 37 (1999),
403-9.
[7] M. Casado and P. Ascher, Opposite modulation of NMDA receptors by lysophospholipids and
arachidonic acid: common features with mechanosensitivity, J Physiol. 513 (1998), 317-30.
[8] T. Kawakubo, K. Naruse, T. Matsubara, N. Hotta, and M. Sokabe, Characterization of a newly found
stretch-activated KCa,ATP channel in cultured chick ventricular myocytes, Am J Physiol. 276 (1999),
H1827-38.
[9] J. Lammerding, R.D. Kamm, and R.T. Lee, Mechanotransduction in cardiac myocytes, Ann N Y Acad
Sci. 1015 (2004), 53-70.
[10] M.J. Cipolla, N.I. Gokina, and G. Osol, Pressure-induced actin polymerization in vascular smooth
muscle as a mechanism underlying myogenic behavior, Faseb J. 16 (2002), 72-6.
[11] F. Guharay and F. Sachs, Stretch-activated single ion channel currents in tissue-cultured embryonic
chick skeletal muscle, J Physiol. 352 (1984), 685-701.
[12] X.Wan, P. Juranka, and C.E. Morris, Activation of mechanosensitive currents in traumatized
membrane, Am J Physiol. 276 (1999), C318-27.
[13] S.F. Pedersen, G. Owsianik, and B. Nilius, TRP channels: an overview, Cell Calcium. 38 (2005), 233-
52.
[14] M.J. Caterina, T.A. Rosen, M. Tominaga, A.J. Brake, and D. Julius, A capsaicin-receptor homologue
with a high threshold for noxious heat, Nature 398 (1999), 436-41.
[15] H. Watanabe, J. Vriens, J. Prenen, G. Droogmans, T. Voets, and B. Nilius, Anandamide and
arachidonic acid use epoxyeicosatrienoic acids to activate TRPV4 channels, Nature 424 (2003), 434-8.
[16] R. Maroto, A. Raso, T.G. Wood, A. Kurosky, B. Martinac, and O.P. Hamill, TRPC1 forms the stretch
activated cation channel in vertebrate cells, Nat Cell Biol. 7 (2005), 179-85.
[17] W. Li, Z. Feng, P.W. Sternberg, and X.Z. Xu, A C. elegans stretch receptor neuron revealed by a
mechanosensitive TRP channel homologue, Nature 440 (2006), 684-7.
[18] G. Apodaca, Modulation of membrane traffic by mechanical stimuli, Am J Physiol Renal Physiol. 282
(2002), F179-90.
[19] K.S. Ko and C.A. McCulloch, Partners in protection: interdependence of cytoskeleton and plasma
membrane in adaptations to applied forces, J Membr Biol. 174 (2000), 85-95.
[20] Y. Oishi, Y. Uezono, N. Yanagihara, F. Izumi, T. Nakamura, and K. Suzuki, Transmural compression
induced proliferation and DNA synthesis through activation of a tyrosine kinase pathway in rat
astrocytoma RCR-1 cells, Brain Res. 781 (1998), 159-66.
[21] L.C. Martineau and P.F. Gardiner, Insight into skeletal muscle mechanotransduction: MAPK activation
is quantitatively related to tension, J Appl Physiol. 91 (2001), 693-702.
[22] C. Yan, M. Takahashi, M. Okuda, J.D. Lee, and B.C. Berk, Fluid shear stress stimulates big mitogen
activated protein kinase 1 (BMK1) activity in endothelial cells. Dependence on tyrosine kinases and
intracellular calcium, J Biol Chem. 274 (1999), 143-50.
[23] S. Lehoux, B. Esposito, R. Merval, and A. Tedgui, Differential regulation of vascular focal adhesion
kinase by steady stretch and pulsatility, Circulation. 111 (2005), 643-9.
[24] J.C. Krepinsky, Y. Li, D. Tang, L. Liu, J. Scholey, and A.J. Ingram, Stretch-induced Raf-1 activation
in mesangial cells requires actin cytoskeletal integrity, Cell Signal. 17 (2005), 311-20.
[25] J.P. Hatton, M. Pooran, C.F. Li, C. Luzzio, and M. Hughes-Fulford, A short pulse of mechanical force
induces gene expression and growth in MC3T3-E1 osteoblasts via an ERK 1/2 pathway, J Bone Miner
Res. 18 (2003), 58-66.
[26] T.E. Danciu, E. Gagari, R.M. Adam, P.D. Damoulis, and M.R. Freeman, Mechanical strain delivers
anti-apoptotic and proliferative signals to gingival fibroblasts, J Dent Res. 83 (2004), 596-601.
[27] H.R. Kim and C.M. Hai, Mechanisms of mechanical strain memory in airway smooth muscle, Can J
Physiol Pharmacol. 8 (2005), 811-5.
[28] R.J. Hendrickson, C. Cappadona, E.N. Yankah, J.V. Sitzmann, P.A. Cahill, and E.M. Redmond,
Sustained pulsatile flow regulates endothelial nitric oxide synthase and cyclooxygenase expression in
cocultured vascular endothelial and smooth muscle cells, J Mol Cell Cardiol. 31 (1999), 619-29.
212 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics

[29] X. Fan, J.A. Rahnert, T.C. Murphy, M.S. Nanes, E.M. Greenfield, and J. Rubin, Response to
mechanical strain in an immortalized pre-osteoblast cell is dependent on ERK1/2, J Cell Physiol. 207
(2006), 454- 60.
[30] M. Matsukawa, K. Fukuda, K. Yamasaki, K. Yoshida, H. Munakata, and C. Hamanishi, Enhancement
of nitric oxide and proteoglycan synthesis due to cyclic tensile strain loaded on chondrocytes attached
to fibronectin, Inflamm Res. 53 (2004), 239-44.
[31] J.G. McGarry, J. Klein-Nulend, and P.J. Prendergast, The effect of cytoskeletal disruption on pulsatile
fluid flow-induced nitric oxide and prostaglandin E2 release in osteocytes and osteoblasts, Biochem
Biophys Res Commun. 330 (2005), 341-8.
[32] J. Rubin, T.C. Murphy, L. Zhu, E. Roy, M.S. Nanes, and X. Fan, Mechanical strain differentially
regulates endothelial nitric-oxide synthase and receptor activator of nuclear kappa B ligand expression
via ERK1/2 MAPK, J Biol Chem. 278 (2003), 34018-25.
[33] G. Jiang, A.H. Huang, Y. Cai, M. Tanase, and M.P. Sheetz, Rigidity sensing at the leading edge
through alphavbeta3 integrins and RPTP_, Biophys J. 90 (2006), 1804-9.
[34] K. Yamada, K.G. Green, A.M. Samarel, and J.E. Saffitz, Distinct pathways regulate expression of
cardiac electrical and mechanical junction proteins in response to stretch, Circ Res. 97 (2005), 346-53.
[35] S.T. Rhee and S.R. Buchman, Colocalization of c-Src (pp60src) and bone morphogenetic protein 2/4
expression during mandibular distraction osteogenesis: in vivo evidence of their role within an
integrinmediated mechanotransduction pathway, Ann Plast Surg. 55 (2005), 207-15.
[36] D. Riveline, E. Zamir, N.Q. Balaban, U.S. Schwarz, T. Ishizaki, S. Narumiya, Z. Kam, B. Geiger, and
A.D. Bershadsky, Focal contacts as mechanosensors: externally applied local mechanical force induces
growth of focal contacts by an mDia1 dependent and ROCK-independent mechanism, J Cell Biol. 153
(2001), 1175-86.
[37] M.E. Chicurel, R.H. Singer, C.J. Meyer, and D.E. Ingber, Integrin binding and mechanical tension
induce movement of mRNA and ribosomes to focal adhesions, Nature 392 (1998), 730-3.
[38] S.M. Norvell, M. Alvarez, J.P. Bidwell, and F.M. Pavalko, Fluid shear stress induces beta-catenin
signaling in osteoblasts, Calcif Tissue Int. 75 (2004), 396-404.
[39] H. Henderson and D.R. Carter, Mechanical induction in limb morphogenesis: the role of growth
generated strains and pressures, Bone. 31 (2002), 645-53.
[40] B.K. Hall and S.W. Herring, Paralysis and growth of the musculoskeletal system in the embryonic
chick, J Morphol. 206 (1990), 45-56.
[41] M. Schmelter, B. Ateghang, S. Helmig, M. Wartenberg, and H. Sauer, Embryonic stem cells utilize
reactive oxygen species as transducers of mechanical strain-induced cardiovascular differentiation,
Faseb J. 20 (2006), 1182-4.
[42] M.F. Pittenger, A.M. Mackay, S.C. Beck, R.K. Jaiswal, R. Douglas, J.D. Mosca, M.A. Moorman,
D.W. Simonetti, S. Craig, and D.R. Marshak, Multilineage potential of adult human mesenchymal
stem cells, Science 284 (1999), 143-7.
[43] A. Wiesmann, H.J. Buhring, C. Mentrup, and H.P. Wiesmann, Decreased CD90 expression in human
mesenchymal stem cells by applying mechanical stimulation, Head Face Med. 2 (2006), 8.
[44] J.H. Cui, K. Park, S.R. Park, and B.H. Min, Effects of low-intensity ultrasound on chondrogenic
differentiation of mesenchymal stem cells embedded in polyglycolic acid: an in vivo study, Tissue Eng.
12 (2006), 75-82.
[45] M. Knippenberg, M.N. Helder, B.Z. Doulabi, C.M. Semeins, P.I. Wuisman, and J. Klein-Nulend,
Adipose tissue-derived mesenchymal stem cells acquire bone cell-like responsiveness to fluid shear
stress on osteogenic stimulation, Tissue Eng. 11 (2005), 1780-8.
[46] R.C. Riddle, A.F. Taylor, D.C. Genetos, and H.J. Donahue, MAP kinase and calcium signaling mediate
fluid flow-induced human mesenchymal stem cell proliferation, Am J Physiol Cell Physiol. 290 (2006),
C776-84.
[47] E. Tzima, M. Irani-Tehrani,W.B. Kiosses, E. Dejana, D.A. Schultz, B. Engelhardt, G. Cao, H.
DeLisser, and M.A. Schwartz, A mechanosensory complex that mediates the endothelial cell response
to fluid shear stress, Nature 437 (2005), 426-31.
[48] C.T. Rubin and L.E. Lanyon, Regulation of bone mass by mechanical strain magnitude, Calcif Tissue
Int. 37 (1985), 411-7.
[49] G. Carmeliet, L. Vico, and R. Bouillon, Space flight: a challenge for normal bone homeostasis, Crit
Rev Eukaryot Gene Expr. 11 (2001), 131-44.
[50] M. Mehrotra, M. Saegusa, S.Wadhwa, O. Voznesensky, D. Peterson, and C. Pilbeam, Fluid flow
induces Rankl expression in primary murine calvarial osteoblasts, J Cell Biochem. 98 (2006), 1271-83.
[51] N. Kizer, X.L. Guo, and K. Hruska, Reconstitution of stretch-activated cation channels by expression
of the alpha-subunit of the epithelial sodium channel cloned from osteoblasts, Proc Natl Acad Sci U S
A. 94 (1997), 1013-8.
 V.A. Campbell and B. O’Connell / Cellular & Molecular Biomechanics 213

[52] J. Li, R.L. Duncan, D.B. Burr, and C.H. Turner, L-type calcium channels mediate mechanically
induced bone formation in vivo, J Bone Miner Res. 17 (2002), 1795-800.
[53] J. Rubin, C. Rubin, and C.R. Jacobs, Molecular pathways mediating mechanical signaling in bone,
Gene 367 (2006), 1-16.
[54] S. Kapur, D.J. Baylink, and K.H. Lau, Fluid flow shear stress stimulates human osteoblast proliferation
and differentiation through multiple interacting and competing signal transduction pathways, Bone 32
(2003), 241-51.
[55] F.A. Weyts, Y.S. Li, J. van Leeuwen, H. Weinans, and S. Chien, ERK activation and alpha v beta 3
integrin signaling through Shc recruitment in response to mechanical stimulation in human osteoblasts,
J Cell Biochem. 87 (2002), 85-92.
[56] T. Ishida, T.E. Peterson, N.L. Kovach, and B.C. Berk, MAP kinase activation by flow in endothelial
cells. Role of beta 1 integrins and tyrosine kinases, Circ Res. 79 (1996), 310-6.
[57] A.D. Bakker, J. Klein-Nulend, E. Tanck, I.C. Heyligers, G.H. Albers, P. Lips, and E.H. Burger,
Different responsiveness to mechanical stress of bone cells from osteoporotic versus osteoarthritic
donors, Osteoporos Int. 17 (2006), 827-33.
[58] Y.J. Kim, R.L. Sah, A.J. Grodzinsky, A.H. Plaas, and J.D. Sandy, Mechanical regulation of cartilage
biosynthetic behavior: physical stimuli, Arch Biochem Biophys. 311 (1994), 1-12.
[59] Q.Q. Wu and Q. Chen, Mechanoregulation of chondrocyte proliferation, maturation, and hypertrophy:
ion-channel dependent transduction of matrix deformation signals, Exp Cell Res. 256 (2000), 383-91.
[60] [60] T.T. Chowdhury, R.N. Appleby, D.M. Salter, D.A. Bader, and D.A. Lee, Integrin-mediated
mechanotransduction in IL-1 beta stimulated chondrocytes, Biomech Model Mechanobiol. 5 (2006),
192-201.
[61] C.V. Gemmiti and R.E. Guldberg, Fluid flow increases type II collagen deposition and tensile
mechanical properties in bioreactor-grown tissue-engineered cartilage, Tissue Eng. 12 (2006), 469-79.
[62] T.P. Lele, J.E. Sero, B.D. Matthews, S. Kumar, S. Xia, M. Montoya-Zavala, T. Polte, D. Overby, N.
Wang, and D.E. Ingber, Tools to study cell mechanics and mechanotransduction, Methods Cell Biol. 83
(2007), 443-72.
[63] D. Stamenovic and N. Wang, Invited review: engineering approaches to cytoskeletal mechanics, J Appl
Physiol. 89 (2000), 2085-90.
[64] [64] P. Maguire, J.I. Kilpatrick, G. Kelly, P.J. Prendergast, V.A. Campbell, B.C. O’Connell, and S.P.
Jarvis, Direct mechanical measurement of geodesic structures in rat mesenchymal stem cells, HFSP
Journal Epub ahead of print (2007).
[65] J.G. McGarry, P. Maguire, V.A. Campbell, B.C. O’Connell, P.J. Prendergast, and S.P. Jarvis,
Stimulation of nitric oxide mechanotransduction in single osteoblasts using atomic force microscopy, J
Orthop Res, 26 (2008), 513-21.
[66] V. Mann, C. Huber, G. Kogianni, D. Jones, and B. Noble, The influence of mechanical stimulation on
osteocyte apoptosis and bone viability in human trabecular bone, J Musculoskelet Neuronal Interact. 6
(2006), 408-17.
[67] K. Ariga, K. Yonenobu, T. Nakase, N. Hosono, S. Okuda, W. Meng, Y. Tamura, and H. Yoshikawa,
Mechanical stress-induced apoptosis of endplate chondrocytes in organ-cultured mouse intervertebral
discs: an ex vivo study, Spine 28 (2003), 1528-33.
[68] S.S. Tholpady, T.F. Freyman, D. Chachra, and R.C. Ogle, Tensional forces influence gene expression
and sutural state of rat calvariae in vitro, Plast Reconstr Surg. 120 (2007), 601-11; discussion 612-3.
[69] D.A. Hardiman, F.J. O’Brien, P.J. Prendergast, D.T. Croke, A. Staines, and T.C. Lee, Tracking the
changes in unloaded bone: Morphology and gene expression, Eur J Morphol. 42 (2005), 208-16.
[70] Y.H. Li, C.Y. Hsieh, D.L. Wang, H.C. Chung, S.L. Liu, T.H. Chao, G.Y. Shi, and H.L. Wu,
Remodeling of carotid arteries is associated with increased expression of thrombomodulin in a mouse
transverse aortic constriction model, Thromb Haemost. 97 (2007), 658-64.
[71] B.A. Simon, R.B. Easley, D.N. Grigoryev, S.F. Ma, S.Q. Ye, T. Lavoie, R.M. Tuder, and J.G. Garcia,
Microarray analysis of regional cellular responses to local mechanical stress in acute lung injury, Am J
Physiol Lung Cell Mol Physiol. 291 (2006), L851-61.
[72] M.C. Meikle, Remodeling the dentofacial skeleton: the biological basis of orthodontics and dentofacial
orthopedics, J Dent Res. 86 (2007), 12-24.
[73] J.P. Schmitt, E.P. Debold, F. Ahmad, A. Armstrong, A. Frederico, D.A. Conner, U. Mende, M.J.
Lohse, D. Warshaw, C.E. Seidman, and J.G. Seidman, Cardiac myosin missense mutations cause
dilated cardiomyopathy in mouse models and depress molecular motor function, Proc Natl Acad Sci U
S A. 103 (2006), 14525-30.
214 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-214

IV.3. Bioreactors in Tissue Engineering

Niamh PLUNKETT a,b and Fergal J. O’BRIEN a,b
a
Department of Anatomy, Royal College of Surgeons in Ireland, 123 St. Stephen’s
Green, Dublin 2, Ireland
b
Trinity Centre for Bioengineering, Department of Mechanical Engineering,
Trinity College Dublin, Dublin 2, Ireland

Introduction

What is a Bioreactor?

Bioreactors have been used for many years in areas other than tissue engineering. They
have been used in diverse areas such as in fermentation, in water treatment, in food
processing and in the production of pharmaceuticals [1]. All of these bioreactors are
devices in which biological or biochemical processes develop under a closely
monitored and tightly controlled environment. Bioreactors have been used in animal
cell culture since the 1980s in order to produce vaccines and other drugs and to culture
large cell populations. Bioreactors for use in tissue engineering have progressed from
such devices.
A tissue engineering bioreactor can be defined as a device that uses mechanical
means to influence biological processes [2]. In tissue engineering, this generally means
that bioreactors are used to stimulate cells and encourage them to produce extra-cellular
matrix (ECM). There are numerous types of bioreactor which can be classified by the
means they use to stimulate cells. A number of these will be discussed below.

Why are Bioreactors Needed in Tissue Engineering?

Tissue engineering technologies are based on the biological triad of cells, signalling
mechanisms and extracellular matrix. To simulate the development of tissues in vitro,
tissue engineering aims to optimise cell growth, by providing regulatory signals in the
form of growth factors and a regeneration template in the form of scaffold. Bioreactors
may be used as an alternative to or in conjunction with growth factors in the signalling
part of the triad. Defects requiring tissue-engineering solutions are typically many
millimetres in size [3]. Scaffolds in such a size range are easily fabricated. However,
problems arise when culturing cells on these scaffolds. Static culture conditions result
in scaffolds with few cells in the centre of the construct [4]. This heterogeneous cell
distribution is a major obstacle to developing any three-dimensional tissue or organ in
vitro. It has been shown that despite homogeneous cell seeding, after long periods in
culture, more cells are found on the periphery of demineralised trabecular bone
constructs [4]. This is due to cell necrosis and cell chemotaxis. Necrosis occurs at the
centre of the scaffold due to a lack of nutrient delivery to, and waste removal from that
area. The only mechanism by which nutrients and waste can move when a scaffold is in
 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering 215

static culture is by diffusion. As the size of the scaffold increases, diffusion to the
centre of the construct becomes more difficult. In addition, as cells on the periphery
grow and secrete extracellular matrix, movement of fluid to the interior of the scaffold
is further impeded. Chemotaxis of the cells from the interior towards the periphery
occurs because of the concentration gradient in nutrients that has been set up [3].
Nutrient concentration is greater at the periphery so cells move along this gradient
towards the periphery in order to obtain the nutrients they require. It has been found
that mineralised bone matrix reaches a maximum penetration depth of 240 μm on
poly( DL-lactic-co-glycolic acid) scaffolds seeded with stromal osteoblasts, which is
far thinner than an ideal bone graft replacement [5].

Figure 1. Histological sections of tissue engineered vascular tissue cultured using a bioreactor (on left) and
statically (on right). Sections stained with Verhoff’s elastin stain [6].

Figure 1 shows the distribution of smooth muscle cells in tissue engineered

vascular tissue. Homogeneous distribution has been achieved by using a bioreactor [6].
In order to increase cell viability throughout a scaffold, fluid transport needs to be
enhanced. A bioreactor may be used to achieve this aim. In addition to enhancing cell
distribution, another important aspect of bioreactor use is cell stimulation. Cells
respond to mechanical stimulation and bioreactors can be used to apply such
stimulation. This can encourage cells to produce ECM in a shorter time period and in a
more homogeneous manner than would be the case with static culture. For example, in
comparisons between ECM protein levels after 5 weeks in culture, scaffolds cultured
under hydrostatic pressure showed significant improvements over scaffolds cultured in
static medium [7]. A benefit of ECM production is the increase in mechanical stiffness
that it provides to the construct. A six-fold increase in equilibrium aggregate modulus
(an intrinsic property of cartilage which is a measure of stiffness) was found after 28
days of culture in a compression bioreactor compared to free swelling controls [8].
Another important application of bioreactors is in cellular differentiation.
Mechanical stimulation can be used to encourage stem cells down a particular path and
hence provide the cell phenotype required. Bioreactors can provide biochemical and
physical regulatory signals that guide differentiation [9]. There is great potential for
using mesenchymal stem cells and other multipotent cells to generate different cell
types and bioreactors can play an important role in this process.
216 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering

Bioreactor Design Requirements

In general, bioreactors are designed to perform at least one of the following five
functions: 1) providing uniform cell distribution, 2) maintaining the desired
concentration of gases and nutrients in the medium, 3) providing mass transport to the
tissue, 4) exposing tissue to physical stimuli, or 5) providing information about the
formation of 3D tissue [10].
While the detailed requirements for bioreactor design are tissue- and/or
application- specific, there are a few general principles which have to be adhered to
when developing a bioreactor. The material selection is very important as it is vital to
ensure that the materials used to create the bioreactor do not elicit any adverse reaction
from the cultured tissue. Any material which is in contact with media must be
biocompatible or bioinert. This eliminates the use of most metals, although stainless
steel can be used if it is treated so that chromium ions do not leach out into the medium.
Numerous plastics comply with this constraint but there are further limitations on
material selection that must also be kept in mind. Materials must be usable at 37°C in a
humid atmosphere. They must be sterilisable if they are to be re-used. Bioreactor parts
can be sterilised by autoclaving or disinfected by submersion in alcohol. If they are to
be autoclaved, materials that can withstand numerous cycles of high temperature and
pressure must be used in bioreactor manufacture. Alternatively, some non-sterilisable
disposable bioreactor parts may be used which can be replaced after each use of the
bioreactor. Other material choices are between transparent or opaque and flexible or
inflexible materials. Materials with different properties are needed for various
components in the bioreactor. For example, transparent materials can be of benefit in
allowing the construct to be monitored in the bioreactor during culture while flexible
tubing can help with assembly of the bioreactor.
The design of the bioreactor should be as simple as possible, avoiding the
introduction of, for example, machined recesses which could collect condensed steam
during autoclaving and become breeding grounds for micro-organisms. Simplicity in
design should also mean that the bioreactor is quick to assemble and disassemble.
Apart from being more efficient, this ensures that cell-seeded constructs inserted into
the bioreactor are out of the incubator for the minimum amount of time possible. This
minimises the risk to the cells and the experiment being undertaken.
The specific application of the bioreactor must be kept in mind during the design
process to ensure that all the design constraints are met. If various parameters such as
pH, nutrient concentration or oxygen levels are to be monitored, these sensors should
be incorporated into the design. If a pump or motor is to be used, it must be small
enough to fit into an incubator and also be usable at 37°C and in a humid environment.
The forces needed for cellular stimulation are very small so it is important to ensure
that the pump/motor has the capability to apply small forces accurately. In any design
involving fluids, problems can arise with leaking fluid seals and, if possible, the need
for seals should be removed. However, in most cases, fluid seals are necessary and
good design should decrease the problems with them. If a prototype bioreactor is being
designed, it is worthwhile thinking about scale up opportunities for the bioreactor from
the outset. This may mean designing a device that is relatively easy to enlarge without
changing its characteristics or designing a simple device of which many more can be
made so that numerous scaffolds can be cultured at one time.
 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering 217

1. Bioreactors in Tissue Engineering

1.1. Spinner Flask Bioreactor

Continuous stirred-tank reactors are commonly used in bioprocessing, for example in

solid-state fermentation of organisms such as yeast [11] and spinner flask bioreactors
for use in tissue engineering progressed from these devices. In a spinner flask (Figure
2), scaffolds are suspended at the end of needles in a flask of culture media. A magnetic
stirrer mixes the media and the scaffolds are fixed in place with respect to the moving
fluid. Flow across the surface of the scaffolds results in eddies in the scaffolds'
superficial pores. Eddies are turbulent instabilities consisting of clumps of fluid
particles that have a rotational structure superimposed on the mean linear motion of the
fluid particles. They are associated with transitional and turbulent flow. It is via these
eddies that fluid transport to the centre of the scaffold is thought to be enhanced [3].
Typically, spinner flasks are around 120 ml in volume (although much larger flasks of
up to 8 litres have been used), are run at 50-80 rpm and 50% of the medium used in
them is changed every two days [12]. Cartilage constructs have been grown in spinner
flasks to thicknesses of 0.5 mm [13]. While this is an improvement on cartilage grown
in static medium, it is still too thin for clinical use. Mass transfer in the flasks is not
good enough to deliver homogeneous cell distribution throughout scaffolds and cells
predominantly reside on the construct periphery [3].

Figure 2. A spinner flask bioreactor. Scaffolds are suspended in medium and the medium stirred using a
magnetic stirrer to improve nutrient delivery to the scaffold (Kim et al., 2005)

1.2. Rotating Wall Bioreactor

The rotating wall bioreactor was developed by NASA [14]. It was originally designed
with a view to protecting cell culture experiments from high forces during space shuttle
take off and landing. However, the device has proved useful in tissue engineering here
218 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering

on earth. In a rotating wall bioreactor (Figure 3), scaffolds are free to move in media in
a vessel. The wall of the vessel rotates, providing an upward hydrodynamic drag force
that balances with the downward gravitational force, resulting in the scaffold remaining
suspended in the media. Fluid transport is enhanced in a similar fashion to the
mechanism in spinner flasks and the devices also provide more homogeneous cell
distribution than static culture [3]. Gas exchange occurs through a gas exchange
membrane and the bioreactor is rotated at speeds of 15-30 rpm. Cartilage tissue of 5
mm thickness has been grown in this type of bioreactor after seven months of culture
[15]. As tissue grows in the bioreactor, the rotational speed must be increased in order
to balance the gravitational force and ensure the scaffold remains in suspension.

Figure 3. A rotating wall vessel bioreactor. Scaffolds are suspended in medium due to opposing gravitational
and drag forces (www.che.msstate.edu/research/MDERL/demos/DEMoBioRctrExperiment)

1.3. Compression Bioreactors

Another widely used type of bioreactor is the compression bioreactor. This class of
bioreactor is generally used in cartilage engineering and can be designed so that both
static loading and dynamic loading can be applied. This is because static loading has
been found to have a negative effect on cartilage formation while dynamic loading,
which is more representative of physiological loading, has provided better results than
many other stimuli [16].
In general, compression bioreactors consist of a motor, a system providing linear
motion and a controlling mechanism. An example of such a system is shown in Figure
4, where a cam-follower system is used to provide displacements of different
magnitudes and frequencies. A signal generator can be used to control the system and
load cells and linear variable differential transformers can be used to measure the load
response and imposed displacement respectively [8, 17]. The load can be transferred to
the cell-seeded constructs via flat platens which distribute the load evenly, however in a
device for stimulating multiple scaffolds simultaneously, care must be taken that the
constructs are of similar height or the compressive strain applied will vary as the
 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering 219

scaffold height does. Mass transfer is improved in dynamic compression bioreactors

over static culture (as compression causes fluid flow in the scaffold) and dynamic
compression can also improve the aggregate modulus of the resulting cartilage tissue to
levels approaching those of native articular cartilage [8].

Figure 4. A compression bioreactor. Constructs are housed in a petri dish and subjected to compressive
forces exerted on them via a cam and follower mechanism [41].

1.4. Strain Bioreactors

Tensile strain bioreactors have been used in an attempt to engineer a number of

different types of tissue including tendon, ligament, bone, cartilage and cardiovascular
tissue. Some designs are very similar to compression bioreactors, only differing in the
way the force is transferred to the construct. Instead of flat platens as in a compression
bioreactor, a way of clamping the scaffold into the device is needed so that a tensile
force can be applied. Tensile strain has been used to differentiate mesechymal stem
cells along the chondrogenic lineage. A multistation bioreactor was used in which cell-
seeded collagen-glycosaminoglycan scaffolds were clamped and loaded in uniaxial
tension [18]. Alternatively, tensile strain can also be applied to a construct by attaching
the construct to anchors on a rubber membrane and then deforming the membrane. This
system has been used in the culture of bioartificial tendons with a resulting increase in
Young’s modulus over non-loaded controls [19].

1.5. Hydrostatic Pressure Bioreactors

In cartilage tissue engineering, hydrostatic pressure bioreactors can be used to apply

mechanical stimulus to cell-seeded constructs. Scaffolds are usually cultured statically
and then moved to a hydrostatic chamber for a specified time for loading. Hydrostatic
pressure bioreactors consist of a chamber which can withstand the pressures applied
and a means of applying that pressure (Figure 5). For example, a media-filled pressure
220 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering

chamber can be pressurised using a piston controlled by an actuator [16]. For sterility,
the piston can apply pressure via an impermeable membrane so that the piston itself
does not come into contact with the culture media. Variations on this design include a
water-filled pressure chamber which pressurises a media-filled chamber via an
impermeable film and is controlled using a variable backpressure valve and an actuator
[20].

Figure 5. A hydrostatic pressure bioreactor. Constructs are placed in a chamber which is subsequently
pressurised to the required level [16].

1.6. Flow Perfusion Bioreactor

Flow perfusion bioreactors generally consist of a pump and a scaffold chamber joined
together by tubing. A media reservoir may also be present. The scaffold is kept in
position across the flow path of the device (Figure 6). Media is perfused through the
scaffold, thus enhancing fluid transport.
Culture using flow perfusion bioreactors has been shown to provide more
homogeneous cell distribution throughout scaffolds. Collagen sponges have been
seeded with bone marrow stromal cells and perfused with flow. This has resulted in
greater cellularity throughout the scaffold in comparison to static controls, implying
that better nutrient exchange occurs due to flow [21]. Using a biphasic calcium-
phosphate scaffold, abundant ECM with nodules of CaP was noted after 19 days in
steady flow culture [22].
In comparisons between flow perfusion, spinner flask and rotating wall bioreactors,
flow perfusion bioreactors have proved to be the best for fluid transport. Using the
same flow rate and the same scaffold type, while cell densities remained the same using
all three bioreactors, the distribution of the cells changed dramatically depending on
which bioreactor was used. Histological analysis showed that spinner flask and static
culture resulted in the majority of viable cells being on the periphery of the scaffold. In
 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering 221

contrast, the rotating wall vessel and flow perfusion bioreactor culture resulted in
uniform cell distribution throughout the scaffolds [3, 23]. After 14 days in culture, the
perfusion bioreactor had higher cell density than all other culture methods [3].
In our laboratory, a flow perfusion bioreactor has been developed to examine the
effects of different flow profiles on cell-seeded collagen-glycosaminoglycan scaffolds
[24]. The scaffold chamber was specifically designed to ensure that the compliant
scaffold was under perfusive flow. This involved using a scaffold of larger diameter
than the flow path and using spacers to ensure the scaffold was under 10% compression
during culture. A programmable syringe pump was used in order to stimulate the cell-
seeded constructs using different flow profiles. As discussed below, this device
demonstrated that intermittent flow perfusion is advantageous for mechanically
stimulating osteoblasts while maintaining cell viability.

Figure 6. A flow perfusion bioreactor (A). Media is forced through the scaffold in the scaffold chamber (B)
by the syringe pump [24].

1.7. Combined Bioreactors

In addition to the bioreactors mentioned thus far, numerous combinations of the

different types of bioreactor have been used in order to better mimic the in vivo
environment in vitro. In most cases, these more complicated bioreactors involve adding
a perfusion loop on top of a standard bioreactor. Examples include compression, tensile
strain or hydrostatic bioreactors with added perfusion [7, 20, 25]. These different
bioreactors allow for nutrient exchange to take place due to perfusion while stimulation
occurs due to a different mechanical stimulus. Bioreactors for engineering very
specialised tissues have also been developed. One example of such a bioreactor is a
combined tensile and vibrational bioreactor for engineering vocal fold tissue [26]. This
bioreactor mimics the physiological conditions that human vocal folds experience in an
222 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering

attempt to develop the tissue in the laboratory (Figure 7). Vocal fold tissue experiences
vibrations of 100-1000 Hz at an amplitude of approximately 1 mm, which is a unique
stimulus in the human body [27].

Figure 7. A bioreactor for vocal fold tissue engineering. Constructs can be stimulated using both tensile and
vibrational forces [26].

2. Tissue Formation in Bioreactor Systems

Tissue engineering of all three dimensional tissues require homogeneous cell

distribution in order for homogeneous tissue to develop. Therefore, there is a need for
bioreactor culture in numerous different disciplines in tissue engineering. Bioreactor
culture has been used in a diverse range of applications including skin, bladder and
liver tissue engineering [1, 28]. Here, the use of bioreactor culture in bone, cartilage
and cardiovascular tissue engineering will be discussed.

2.1. Bone

Bone is comprised of a mineral phase, collagen and cells. The mineral phase of bone is
a hard, brittle material (hydroxyapatite) that is strong in compression but cannot
withstand large shear or tensile loads. In contrast, collagen fibres are strong in tension
and flexible in bending but weak in compression. Bone grafts are required to aid bone
defect and non-union healing. The aim in using them is to restore function to the
damaged area as quickly and completely as possible [29]. They are required in a
number of procedures including, for example, replacing diseased bone, filling bone
voids after non-unions or cyst removal, reconstructive surgery and in spinal fusion
operations. The most commonly used graft in bone replacement is the autograft.
Autografts are grafts taken from the patients themselves, usually from the iliac crest,
 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering 223

although the distal femur, the greater trochanter, and the proximal tibia can also be used
[29]. This process is expensive and the size of graft that can be obtained is limited.
Morbidity of the site the graft is acquired from is another problem and complications
can arise due to infection and chronic pain [29]. An alternative to the autograft is the
allograft. The term allograft is used for bone grafts which are obtained from an organ
donor. A drawback to this option is the danger of infection. Xenografts, acquired from
animal bone, are another alternative. However, the morphology of the bone is different
to that of humans and the risk of cross-species infection exists. Coral has also been
used but its structure is very different to that of bone and hence osteointegration of this
type of graft has proved difficult [30].
The use of natural bone grafts has proved problematic and therefore attention has
turned to tissue engineering. Engineered tissue must be strong enough to allow load
bearing after implantation. Resorbability of the scaffold is another important issue.
Ideally, as the scaffold is resorbed, bone should be deposited in its place, thus ensuring
no loss in mechanical strength. A morphology that allows movement of cells and
supports vascularisation is important for a scaffold material. A high porosity and
controllable pore size can provide this. Cells must be able to penetrate into the core of
the scaffold so that they are homogeneously distributed throughout the graft. There
must be space to allow for transport of nutrients to and waste removal from cells, so the
pores must be interconnected [31]. Thus, a balance must be struck between mechanical
strength and porosity to develop an ideal scaffold for bone tissue engineering. The use
of a bioreactor with such a scaffold should provide a homogeneous distribution of
stimulated cells.
It has been shown that fluid flow can stimulate bone cells to increase levels of
bone formation markers [4, 32-37] and its use could improve mineralisation of the
scaffold on which cells are seeded. Flow perfusion bioreactors increase alkaline
phosphatase (ALP) expression after 7 and 14 days more than spinner flasks or rotating
wall vessels [3] and are more commonly used than any other bioreactor for use in 3-D
stimulation studies. In one study, MC3T3-E1 cells were seeded on decalcified human
trabecular bone, the flow rate of perfusion altered and the mRNA expression of Runx2,
OC and ALP measured [4]. It was found that using a steady flow rate of only 1 mL/min
killed nearly all the cells on the scaffold after 7 days in culture. However, a flow rate of
0.01 mL/min led to a high proportion of viable cells both on the surface and inside the
scaffold. This compared favourably to static culture, where cells were predominantly
on the periphery [4].
Using a CaP scaffold and a flow rate of 0.025 mL/min in a flow perfusion
bioreactor, PGE2 levels were found to increase over static controls. When a stimulus of
30 minutes of oscillatory flow at 1 Hz with a 40 mL/min peak was superimposed on the
steady flow, PGE2 levels increased further. The number of cells left residing on the
scaffolds decreased due to this large dynamic stimulus but this decrease was not found
to be statistically significant [38]. When different flow profiles were used intermittently
to stimulate cells on a highly porous collagen-glycosaminoglycan scaffold in our
laboratory, it was found that intermittent flow caused greater stimulation than a
continuous low flow rate without a loss in cell number [39]. Cyclyoxygenase-2 and
osteopontin expression increased due to culture in the bioreactor, as did prostaglandin
E2 production. This lends further backing to the hypothesis that the combination of a
perfusion period (for nutrient delivery and waste removal) and a stimulation period
may deliver enhanced fluid transport with enhanced stimulation of cells and may yet
prove to be the optimum regime for bioreactor culture of bone cells.
224 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering

2.2. Cartilage

Cartilage is a supporting tissue containing chondroitin sulphates, collagen and elastic

fibres and cells. The cells present in cartilage are known as chondrocytes and they are
situated in lacunae in the cartilage matrix. Cartilage is avascular and nutrient and waste
product exchange occurs purely by diffusion through the cartilage matrix. When minor
damage occurs to cartilage, it can repair by appositional growth, but when severe
damage occurs, the body cannot replace the cartilage [40]. Cartilage tissue engineering
may offer the solution to this problem. There are three types of cartilage: hyaline
cartilage, elastic cartilage and fibrocartilage. Joints can contain both hyaline cartilage
and fibrocartilage, with the more flexible hyaline cartilage covering the bone and the
more durable fibrocartilage acting as a shock-absorber between bones. As a joint
moves, there is motion between two articulating layers of cartilage. This deforms the
cartilage, causes fluid flow within it and induces a hydrostatic pressure load on it.
These mechanical forces affect the chondrocytes in the cartilage. The force applied,
along with the length of time it is applied for and the frequency of application modifies
the response of chondrocytes [41]. This is useful in bioreactor design and for use in
cartilage tissue engineering; if the correct stimulation pattern is used, chondrocytes can
be induced to produce more extracellular matrix and this can result in more cartilage-
like tissue being formed.
After twenty weeks in static culture, the aggregate modulus of tissue-engineered
cartilage was 179±9 kPa. This is 40% of the value reported for native cartilage [42]. At
twenty five weeks, the modulus had not increased further so this may be the closest
approximation to cartilage that can be cultured without the aid of a bioreactor. The
most commonly used bioreactors in cartilage tissue engineering are compression
bioreactors. When free swelling controls were compared to dynamically loaded agarose
gels, after 28 days in culture, there was a six-fold increase in the equilibrium aggregate
modulus for the loaded gels [8]. A sinusoidal strain of 10% at 1 Hz was applied to the
gels for five days per week for a total of three hours per day with a rest period of one
hour between each hour of loading. This complex loading pattern was deemed to be
physiological and it resulted in increased glycosaminoglycan content over free swelling
controls after 21 days in culture. The combination of increased modulus and increased
glycosaminoglycan formation over free swelling controls after only four weeks in
culture demonstrates the benefits of bioreactor culture in cartilage tissue engineering.
Compression bioreactors have also been used to examine the effect of loading on
the differentiation of bone marrow mesenchymal stem cells down the chondrocytic
lineage. Growth factors such as transforming growth factor (TGF-E) can also be used to
encourage differentiation. In a study comparing the use of compressive loading, the use
of TGF-E and the use of a combination of loading and TGF-E, it was found that
compressive loading alone was just as effective at inducing chondrogenic
differentiation as TGF-E or TGF-E plus loading [17].

2.3. Cardiovascular Tissue

There are three main types of tissue that are encompassed by the heading
cardiovascular tissue, namely: vascular, cardiac muscle and heart valve tissue. These
different tissues experience different mechanical environments in the body and
therefore, in tissue engineering, they are cultured using different bioreactors and under
 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering 225

different loading conditions in an attempt to recreate physiological loading conditions

in the laboratory.

Figure 8. Structure of an artery. The three layers of the tissue (the tunical intima, the tunica media and the
tunica externa) can be clearly seen [40].

2.3.1. Vascular tissue

Blood vessels must be flexible enough to move during physiological activities and yet
tough enough to withstand the pressure changes that occur as the heart beats. Blood
vessels consist of three layers of tissue: the tunical intima, the tunica media and the
tunica externa. The intima (innermost part of the blood vessel) consists of an
endothelial lining and a layer of connective tissue containing elastic fibres. The media
consists of smooth muscle tissue and loose connective tissue and in arteries, the externa
consists of connective tissue and collagen fibres (Figure 8). Arteriosclerosis is
associated with half of all deaths in the United States each year [40]. It is a thickening
and toughening of the arterial walls and can lead to coronary heart disease if it occurs
in the coronary arteries or stroke if it occurs in the arteries supplying the brain with
blood. Coronary heart disease can be treated by bypass surgery, when a section of
artery or vein is removed from elsewhere in the body and used to bypass the blockage.
Drawbacks with this procedure are that it creates another trauma site in the body and
often, sufficient tissue for bypass is not available [6]. Research into tissue engineering
of vascular tissue is proving promising for growing arteries in vitro that may be used
instead of the patient’s own blood vessels.
The ideal graft for vascular bypass consists of a confluent endothelium at the
lumen and a smooth muscle layer surrounding it with sufficient mechanical integrity
for suture retention and tolerance of arterial pressures. The endothelial layer is
important as a confluent layer of endothelial cells inhibits the proliferation of smooth
226 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering

muscle cells. The smooth muscle cells are therefore more likely to be quiescent and this
decreases the risk of thrombosis formation and luminal occlusion [6]. In order to
recreate the structure of blood vessels in vitro, co-cultures of different cell types must
be used. Scaffolds can be cultured with smooth muscle cells initially and then a layer of
endothelial cells can be seeded as the culture period nears its end.
The arteries nearest the heart experience the highest pressures and they have
resilient walls in order to cope with these changing pressures. They are known as
elastic arteries and their tunica media contains a high proportion of elastic fibres. This
enables them to expand as the pressure inside the lumen increases and recoil when the
pressure decreases, thus allowing them to deal with pressure changes but also to damp
the pressure oscillations to make blood flow continuous rather than oscillatory in nature
[40]. It is this expansion and recoil of blood vessels that is the inspiration for a number
of bioreactors for use in vascular tissue engineering.
Some success has been achieved when culturing vascular tissue in both rotating
wall vessels [13] and biomimetic bioreactors [6, 25]. Biomimetic bioreactors use
combinations of strain and perfusion to stimulate vascular tissue development. In one
such bioreactor, a tube of biodegradable polymer was seeded with smooth muscle cells
and silicone tubing placed through the centre of the tube. The construct was maintained
in a bath of stirred media and the highly distensible silicone tubing was pulsed at 165
beats per minute so that the scaffold experienced 5% radial strain over an eight week
culture period [6]. The silicone tubing was then removed, an endothelial cell lining
applied to the inner layer of the construct and culture medium perfused through the
centre of the scaffold. This set up resulted in vessels with a rupture strength similar to
that of native tissue and with a morphology that showed migration of smooth muscle
cells throughout the construct. In contrast, scaffolds kept in static culture over the same
culture period showed no migration of smooth muscle cells and therefore an
inhomogeneous structure. A bioreactor has also been designed to apply both radial and
axial strain to cell-seeded constructs, to mimic the environment in the body more
closely [25]. The bioreactor has two perfusion loops that feed the internal and external
parts of the scaffold. The internal perfusion can be pulsatile in nature and is of variable
frequency.
2.3.2 Cardiac Muscle Tissue
Cardiac muscle cells are called cardiocytes or cardiomyocytes. They form extensive
connections with each other and are almost totally dependent on aerobic metabolism
for energy. When a myocardial infarction (heart attack) occurs, part of the blood supply
to the heart muscle is blocked and cardiac muscle cells die due to lack of oxygen
delivery [40]. Cardiac muscle only has a limited ability to regenerate and scar tissue
formed may create additional problems by constricting other vessels in the damaged
region [40, 43]. Cardiac muscle tissue engineering is therefore of interest in developing
a method for myocardial repair. Oxygen delivery to cardiac cells is vital in order that a
sufficiently thick layer of tissue can be grown in vitro. This means that the use of
bioreactors that increase oxygen delivery to cells is important in cardiac muscle tissue
engineering. The ideal tissue-engineered cardiac muscle tissue graft should have a
dense, uniform distribution of cardiomyocytes, should contract in response to an
electrical stimulus and have the mechanical integrity to allow implantation [43].
Polyglycolic acid (PGA) scaffolds have been seeded with cardiomyocytes and
cultured in spinner flasks and rotating wall bioreactors [44, 45]. In both these culturing
systems, a peripheral region densely populated with cells and a central region with a
 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering 227

sparse population of cells was apparent. Spontaneous contractions of the tissue

occurred during culture using the spinner flask and impulse propagation occurred due
to electrical stimulation, but only in the peripheral region [44]. Use of a perfusion
system, however, delivered a much more uniform distribution of cells with higher cell
number on perfused scaffolds than scaffolds cultured in an orbitally mixed dish at all
time points up to 7 days [43]. Spontaneous contractions were also observed in the
perfusion system but only up to 5 days after seeding. Perfused scaffolds contracted
synchronously and at constant frequency upon electrical stimulation after seven days,
while dish-cultured scaffolds exhibited an arrhythmic contraction pattern [43].
2.3.3 Heart Valves
There are four valves in the heart that prevent backflow of blood from the ventricles to
the atria or from the pulmonary trunk and aorta into the ventricles. Problems with these
valves decrease the heart’s ability to maintain circulatory flow [40]. In severe cases,
valve replacement is the only option. However, thromboembolism is a substantial risk
when mechanical valves are used to replace native valves and xenografts or other non-
viable tissue-based grafts often fail due to calcification [46, 47]. Patients with
congenital malformations of the valves often require a durable replacement that can
grow as the patient does, and a tissue-engineered valve may therefore be the best option.
Heart valves experience a complex mechanical environment including high
bending stress and high shear stress in vivo [10]. When grown in static culture, valve
leaflets are fragile, have a rough surface and a low suture retention strength. In
bioreactor culture under pulsatile flow, however, leaflets that are intact, mobile, pliable,
competent during closure and have a tensile strength greater than that of native tissue
are formed after 14 days [46]. In this case, the pulse was applied to the scaffolds by
pumping air into a chamber connected to the scaffold chamber via a silicone diaphragm.
Scaffolds were seeded with myofibroblasts initially and after a four day culture period,
with endothelial cells. The pressure at which the pulse was applied was increased over
time, as was the flow rate of the media. Thus, as the scaffold became stronger, the
forces applied to it were increased. The resulting valve leaflets were implanted into
lambs and there was no evidence of thrombus formation up to 20 weeks after valve
replacement. There was some evidence of pulmonary regurgitation at 16 and 20 weeks,
however. Optimisation of many aspects of the culturing process remains to be
completed but bioreactor culture of valve leaflets has enabled a functioning heart valve
to be developed in vitro [46].

3. Bioreactors: State of the Art and Future Directions

The use of bioreactors has brought us a step closer to engineering numerous tissue
types in the laboratory. At present, most bioreactors are specialised devices with low
volume output. Their assembly is often time consuming and labour intensive. Many
also exhibit operator dependent variability. While scaled-up versions of some devices
may be useful for developing larger amounts of tissue, problems with process
consistency and process contamination may persist. A better understanding of the
different effects of mechanical stimulation on cell signalling and mechanotransduction
is also needed. This can be achieved through the use of existing simple bioreactors in
conjunction with numerical simulation of culture conditions to minimise the number of
experiments needed.
228 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering

In the future, ways of minimising the time and effort needed in order to form tissue
must be found if costs are to be minimised and the use of engineered tissue is to
become routine clinically. One way to do this is to automate the process. The ideal
bioreactor would need autologous material and a scaffold as inputs and, after a defined
culture period, would output the required amount of the required tissue. Automated
systems for culturing cells already exist and work has begun on extending them to
incorporate mechanical stimulation into the culturing process. Aastrom has developed a
system that takes bone marrow as an input and expands the stem and progenitor cell
population. The Automation Partnership has numerous systems for expanding cells. If
systems such as these are used in conjunction with monitoring systems and a feedback
loop, so that factors such as the temperature, oxygen level and pH can be regulated, cell
culture can be optimised. Closed bioreactor systems for seeding and culturing skin
grafts under perfusion were developed by Advanced Tissue Sciences before the
company’s liquidation in 2002. As this system demonstrated, the technology exists to
enable the incorporation of mechanical stimulus into an automated cell culture system
and this may be the future for bioreactors in tissue engineering.

References

[1] I. Martin, D. Wendt and M. Heberer, The role of bioreactors in tissue engineering, Trends Biotechnol
22 (2004), 80-6.
[2] E. M. Darling and K. A. Athanasiou, Biomechanical strategies for articular cartilage regeneration, Ann
Biomed Eng 31 (2003), 1114-24.
[3] A. S. Goldstein, T. M. Juarez, C. D. Helmke, M. C. Gustin and A. G. Mikos, Effect of convection on
osteoblastic cell growth and function in biodegradable polymer foam scaffolds,Biomaterials 22 (2001),
1279-88.
[4] S. H. Cartmell, B. D. Porter, A. J. Garcia and R. E. Guldberg, Effects of medium perfusion rate on cell-
seeded three-dimensional bone constructs in vitro, Tissue Eng 9 (2003), 1197-203.
[5] S. L. Ishaug, G. M. Crane, M. J. Miller, A.W. Yasko, M. J. Yaszemski and A. G. Mikos, Bone
formation by three-dimensional stromal osteoblast culture in biodegradable polymer scaffolds, J.
Biomed. Mater. Res. 36 (1997), 17-28.
[6] L. E. Niklason, J. Gao, W. M. Abbott, K. K. Hirschi, S. Houser, R. Marini and R. Langer, Functional
arteries grown in vitro, Science 284 (1999), 489-93.
[7] S. E. Carver and C. A. Heath, Semi-continuous perfusion system for delivering intermittent
physiological pressure to regenerating cartilage, Tissue Eng 5 (1999), 1-11.
[8] R. L. Mauck, M. A. Soltz, C. C. Wang, D. D. Wong, P. H. Chao, W. B. Valhmu, C. T. Hung and G. A.
Ateshian, Functional tissue engineering of articular cartilage through dynamic loading of chondrocytes
seeded agarose gels, J Biomech Eng 122 (2000), 252-60.
[9] G. H. Altman, R. L. Horan, I. Martin, J. Farhadi, P. R. Stark, V. Volloch, J. C. Richmond, G. Vunjak-
Novakovic and D. L. Kaplan, Cell differentiation by mechanical stress, Faseb J 16 (2002), 270-2.
[10] V. Barron, E. Lyons, C. Stenson-Cox, P. E. McHugh and A. Pandit, Bioreactors for cardiovascular cell
and tissue growth: A review, Ann Biomed Eng 31 (2003), 1017-30.
[11] Y. Martin and P. Vermette, Bioreactors for tissue mass culture: Design, characterization, and recent
advances, Biomaterials 26 (2005), 7481-503.
[12] R. I. Freshney, Culture of animal cells, Wiley-Liss, New York, 2000.
[13] L. E. Freed and G. Vunjak-Novakovic, Tissue engineering bioreactors. in: R. P. Lanza, R. Langer and
J. Vacanti, (Eds.), Principles of tissue engineering, Academic Press, San Diego, CA, 2000, pp. 143-
156.
[14] R. P. Schwarz, T. J. Goodwin and D. A. Wolf, Cell culture for three-dimensional modeling in rotating
wall vessels: An application of simulated microgravity, J Tissue Cult Methods 14 (1992), 51-7.
[15] L. E. Freed, R. Langer, I. Martin, N. R. Pellis and G. Vunjak-Novakovic, Tissue engineering of
cartilage in space, Proc Natl Acad Sci U.S.A 94 (1997), 13885-90.
[16] E. M. Darling and K. A. Athanasiou, Articular cartilage bioreactors and bioprocesses, Tissue Eng 9
(2003), 9-26.
 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering 229

[17] C. Y. Huang, K. L. Hagar, L. E. Frost, Y. Sun and H. S. Cheung, Effects of cyclic compressive loading
on chondrogenesis of rabbit bone-marrow derived mesenchymal stem cells, Stem Cells 22 (2004), 313-
23.
[18] L. A. McMahon, A. J. Reid, V. A. Campbell and P. J. Prendergast, Regulatory effects of mechanical
strain on the chondrogenic differentiation of mscs in a collagen-gag scaffold: Experimental and
computational analysis, Ann Biomed Eng 36 (2008), 185-94.
[19] J. Garvin, J. Qi, M. Maloney and A. J. Banes, Novel system for engineering bioartificial tendons and
application of mechanical load, Tissue Eng 9 (2003), 967-79.
[20] S.Watanabe, S. Inagaki, I. Kinouchi, H. Takai, Y. Masuda and S. Mizuno, Hydrostatic
pressure/perfusion culture system designed and validated for engineering tissue, J Biosci Bioeng 100
(2005), 105-11.
[21] J. Glowacki, S. Mizuno and J. S. Greenberger, Perfusion enhances functions of bone marrow stromal
cells in three-dimensional culture, Cell Transplant 7 (1998), 319-26.
[22] F. W. Janssen, J. Oostra, A. Oorschot and C. A. van Blitterswijk, A perfusion bioreactor system
capable of producing clinically relevant volumes of tissue-engineered bone: in vivo bone formation
showing proof of concept, Biomaterials 27 (2006), 315-23.
[23] X. Yu, E. A. Botchwey, E. M. Levine, S. R. Pollack and C. T. Laurencin, Bioreactor-based bone tissue
engineering: The influence of dynamic flow on osteoblast phenotypic expression and matrix
mineralization, Proc Natl Acad Sci U S A 101 (2004), 11203-8.
[24] M. J. Jaasma, N. A. Plunkett and F. J. O’Brien, Design and validation of a dynamic flow perfusion
bioreactor for use with compliant tissue engineering scaffolds, J Biotechnol. 133 (2008), 490.
[25] K. Bilodeau, F. Couet, F. Boccafoschi and D. Mantovani, Design of a perfusion bioreactor specific to
the regeneration of vascular tissues under mechanical stresses, Artif Organs 29 (2005), 906-12.
[26] I. R. Titze, R. W. Hitchcock, K. Broadhead, K. Webb, W. Li, S. D. Gray and P. A. Tresco, Design and
validation of a bioreactor for engineering vocal fold tissues under combined tensile and vibrational
stresses, J Biomech 37 (2004), 1521-9.
[27] I. R. Titze, On the relation between subglottal pressure and fundamental frequency in phonation, J
Acoust Soc Am 85 (1989), 901-6.
[28] M. C. Wallis, H. Yeger, L. Cartwright, Z. Shou, M. Radisic, J. Haig, M. Suoub, R. Antoon and W. A.
Farhat, Feasibility study of a novel urinary bladder bioreactor, Tissue Eng (2007).
[29] C. R. Perry, Bone repair techniques, bone graft, and bone graft substitutes, Clin Orthop Relat Res 360
(1999), 71-86.
[30] J. Y. de la Caffiniere, E. Viehweger and A. Worcel, [long-term radiologic evolution of coral implanted
in cancellous bone of the lower limb. Madreporic coral versus coral hydroxyapatite], Rev Chir Orthop
Reparatrice Appar Mot 84 (1998), 501-7.
[31] F. J. O’Brien, B. A. Harley, I. V. Yannas and L. Gibson, Influence of freezing rate on pore structure in
freeze-dried collagen-gag scaffolds, Biomaterials 25 (2004), 1077-86.
[32] N. N. Batra, Y. J. Li, C. E. Yellowley, L. You, A. M. Malone, C. H. Kim and C. R. Jacobs, Effects of
short-term recovery periods on fluid-induced signaling in osteoblastic cells, J Biomech 38 (2005),
1909-17.
[33] M. R. Kreke, W. R. Huckle and A. S. Goldstein, Fluid flow stimulates expression of osteopontin and
bone sialoprotein by bone marrow stromal cells in a temporally dependent manner, Bone 36 (2005),
1047-55.
[34] Y. J. Li, N. N. Batra, L. You, S. C. Meier, I. A. Coe, C. E. Yellowley and C. R. Jacobs, Oscillatory
fluid flow affects human marrow stromal cell proliferation and differentiation, J Orthop Res 22 (2004),
1283-9.
[35] K. M. Reich and J. A. Frangos, Effect of flow on prostaglandin E2 and inositol trisphosphate levels in
osteoblasts, Am J Physiol 261 (1991), C428-32.
[36] J. You, G. C. Reilly, X. Zhen, C. E. Yellowley, Q. Chen, H. J. Donahue and C. R. Jacobs, Osteopontin
gene regulation by oscillatory fluid flow via intracellular calcium mobilization and activation of
mitogen activated protein kinase in MC3T3-E1 osteoblasts, J Biol Chem 276 (2001), 13365-71.
[37] J. Klein-Nulend, E. H. Burger, C. M. Semeins, L. G. Raisz and C. C. Pilbeam, Pulsating fluid flow
stimulates prostaglandin release and inducible prostaglandin g/h synthase mrna expression in primary
mouse bone cells, J. Bone Min. Res. 12 (1997), 45-51.
[38] J. Vance, S. Galley, D. F. Liu and S. W. Donahue, Mechanical stimulation of mc3t3 osteoblastic cells
in a bone tissue-engineering bioreactor enhances prostaglandin E2 release, Tissue Eng 11 (2005), 1832-
9.
[39] M. J. Jaasma and F. J. O’Brien, Mechanical stimulation of osteoblasts using steady and dynamic fluid
flow, Tissue Eng Part A 14 (2008), 1213-23.
[40] F. Martini, Fundamentals of anatomy and physiology, Prentice Hall, 2002.
230 N. Plunkett and F.J. O’Brien / Bioreactors in Tissue Engineering

[41] R. L. Mauck, B. A. Byers, X. Yuan and R. S. Tuan, Regulation of cartilaginous ecm gene transcription
by chondrocytes and mscs in 3d culture in response to dynamic loading, Biomech Model Mechanobiol
6 (2007), 113-25.
[42] P. X. Ma and R. Langer, Morphology and mechanical function of long-term in vitro engineered
cartilage, J Biomed Mater Res 44 (1999), 217-21.
[43] M. Radisic, L. Yang, J. Boublik, R. J. Cohen, R. Langer, L. E. Freed and G. Vunjak-Novakovic,
Medium perfusion enables engineering of compact and contractile cardiac tissue, Am J Physiol Heart
Circ Physiol 286 (2004), H507-16.
[44] N. Bursac, M. Papadaki, R. J. Cohen, F. J. Schoen, S. R. Eisenberg, R. Carrier, G. Vunjak-Novakovic
and L. E. Freed, Cardiac muscle tissue engineering: Toward an in vitro model for electrophysiological
studies, Am J Physiol 277 (1999), H433-44.
[45] M. Papadaki, N. Bursac, R. Langer, J. Merok, G. Vunjak-Novakovic and L. E. Freed, Tissue
engineering of functional cardiac muscle: Molecular, structural, and electrophysiological studies, Am J
Physiol Heart Circ Physiol 280 (2001), H168-78.
[46] S. P. Hoerstrup, R. Sodian, S. Daebritz, J. Wang, E. A. Bacha, D. P. Martin, A. M. Moran, K. J.
Guleserian, J. S. Sperling, S. Kaushal, J. P. Vacanti, F. J. Schoen and J. E. Mayer, Jr., Functional living
trileaflet heart valves grown in vitro, Circulation 102 (2000), III44-9.
[47] R. J. Levy, F. J. Schoen,W. B. Flowers and S. T. Staelin, Initiation of mineralization in bioprosthetic
heart valves: Studies of alkaline phosphatase activity and its inhibition by AlCl3 or FeCl3
preincubations, J Biomed Mater Res 25 (1991), 905-35.
Basic Engineering for Medics and Biologists 231
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-231

IV.4. Characterisation and Testing of

Biomaterials
Sebastian DENDORFERa, Joachim HAMMERb and Andreas LENICHc
a
AnyBody Technology A/S, Niels Jernes Vej 10, Aalborg, Denmark
b
Laboratory for Materials Science, University of Applied Sciences Regensburg,
Galgenbergstrasse 30, 93053 Regensburg, Germany
c
Department of Trauma Surgery, Klinikum Rechts der Isar, Technical University of
Munich, Ismaninger Str. 22, 81675 Munich, Germany

1. Introduction

The investigation of the structural mechanics of the entire body system is despite it’s
rather long history in science, compared to more recently developed research field like
genetics, still extremely important and by no means outdated. Improved innovative
treatments of the musculoskeletal system will require specifically designed implants
and implant materials to guarantee optimal function of the injured segment and fast
rehabilitation. As the actual loading situation of the implant-biomaterial compound is
highly sensitive to implant-induced changes in the specific in vivo situation, the
mechanical properties of living biological materials still require intense research
activities, especially due to the large scatter in terms of biological variation.
Structural failure of biomaterials by plastic deformation or fracture is observed if
the material dependent threshold strength is exceeded. This load inducing failure can be
the result of a single load case or a superposition of different load combinations (i.e.
monotonic or static loading). Furthermore, material failure can also be induced by
repeated (cyclic) loading at loads which are significantly below the material or
compound stability characterizing static strength. This failure type is generally the most
common (already repeated load application in the order of 20 cycles can induce
dynamic failure). It should be noted that the characteristic material or compound
parameters are strongly affected and directly related to the material (micro-) structure.
For biological materials this is generally described by biological scatter in terms of the
mechanical morphogenesis. Therefore, any analysis concerning mechanical parameters
or properties has to be interpreted in terms of the specific microstructure of the material.
The focus of this chapter is to give an introduction into the basic principles of load
induced material behavior and failure under static and dynamic conditions with respect
to specific material structure. Further information can be found in standard text books
on biomechanics and solid mechanics [4], [14], [12],[13], [19], [20], [10], [24], [15],
[16].
232 S. Dendorfer et al. / Characterisation and Testing of Biomaterials

2. Mechanical Properties and Material Structure

Under in vivo conditions, materials and structures are exposed to a variety of different
superimposed forces. Unfortunately, these in vivo loading conditions are rather
complex and cannot be applied reproducibly in laboratory experiments. Therefore, it is
common scientific practice to investigate the load bearing capability of materials or
structures under idealized and standardized experimental conditions. These results are
transferred to real components or segments with individual geometries by specific
equations and computational methods.
How do biomaterials respond to specific types of loading due to external forces?
Generally, all materials initially react on external forces (i.e. mechanical loads) by
deformation followed by damage due to increasing deformation. The strength of
materials is thus defined as the resistance of a material against deformation or fracture
due to its atomic composition and its specific microstructure. The ratio of strength and
acting forces has two aspects: The main function of hard materials (e.g. bones) is load
bearing and thus they have to withstand high loads. For this purpose maximum stability
is essential. On the other hand, soft materials (e.g. muscles) may react on external
forces by high forming work and deformation without any risk of fracture. In this
aspect maximum deformation potential is of pronounced interest.
To compare different materials concerning their mechanical properties (e.g.
stability or ductility) and to exclude geometrical size effects, it is essential to normalise
external forces and deformations: A bolt doubled in cross sectional area can carry
doubled loads – therefore, it is appropriate to describe the material property strength by
impingement forces, defined as stresses: Stress, MPa (N/mm2) = Force, N per unit of
area, mm2; σ = F/A. Stresses are directed and therefore, act as vectors affecting the
planes of the loaded structure (Figure 1a). Axial stresses react normal to the surface
while shear stresses are acting as tangential paired forces (Figure 1b).
Equivalent to the forces appropriately normalized by the corresponding area,
deformations are referred to the initial dimensions, e.g. the elongation Δl to the initial
length l0. The obtained relation for the deformation is non-dimensional: Strain =
Elongation, mm per initial length, mm; ε = Δl/l0.
With respect to the direction of the induced forces the following types of
mechanical loading are differentiated (and Table 1).

Table 1: Definitions of mechanical loading conditions.

Loading condition Material response

Tensile stress σ Dilatation ε = Δl/l0
Compressive stress σ Compression ε = - Δl/l0
Shear stress τ Shear strain γ = Δγ/h0
Torsional moment MT Torque Φ = Δu/r0
Bending moment MB Deflection Δh/l0
Hydrostatic pressure p Compaction −ΔV/V0
 S. Dendorfer et al. / Characterisation and Testing of Biomaterials 233

Figure 1: Systematic of mechanical loading classified according to the acting load vectors: axial stresses, a);
shear stresses, b); bending and torsion loading, c); hydrostatic pressure, d).

2.1. Elastic Deformation

Under small loads, each solid body at first deforms in a purely elastic manner. This
implicates that for subsequent unloading the deformation is completely reversible. In
most technical cases a linear relationship between the acting stress and the
corresponding deformation is observed. This constant of proportionality between stress
and strain is defined as Young´s modulus (Equation 1), for planar shear stress and the
corresponding shear strain as shear modulus (Equation 2) and for hydrostatic pressure
and the resulting reduction in volume as compression modulus (Equation 3). As strain,
shear and volume reduction have no dimension, the corresponding moduli have the
dimensions of a stress and, analogous to Hooke´s law can be considered as the spring
constant of the material.
234 S. Dendorfer et al. / Characterisation and Testing of Biomaterials

Figure 2: Inelastic deformation: rectangular stress application (load signal) and delayed deformation
response ε, a) and time compensation of σ (t), ε (t) in terms of a stress-strain diagram, b).

Δσ
= E (1)
Δε e

Δτ
= G (2)
Δγ e

Δp
= K (3)
− ΔV / V 0

In general, for crystalline materials, elastic elongation is directly related to a

reduction in cross-section. (Just think of pulling on a rubber band.) This effect is
defined as transverse contraction and expressed by the Poisson´s ratio ν = εtrans / εlong.
Totally, among the elastic constants, (E, G, ν) only two are independent, i.e. if two
variables are known the third can be calculated, E = 2(1+ν) G.

2.2. Viscoelasticity

Elastic materials which are deformed to a certain strain ε = σ/E (comp. Equation 1) by
a stress σ immediately recedes to ε = 0 after unloading. In many cases and also for
biomaterials, deformation under small applied stresses can be regarded as elastic in
 S. Dendorfer et al. / Characterisation and Testing of Biomaterials 235

terms of reversibility, but is associated with a phase shift (i.e. retarded reaction)
between loading and unloading. These cases where elastic deformation is retarded to a
measurable extent are defined as viscoelastic deformations (Figure 2a). This delay in
material response ε(t) with respect to loading σ(t) is also acting if the loading function
σ(t) is not rectangular but has a sinusoidal shape with the latter being of pronounced
technical importance (as in- or decreases in load never happen infinitely fast). This
phase shift between loading and deformation can be expressed in time-compensated
stress-strain hysteresis (Figure 2b). For strictly elastic behaviour each stress value
would corresponds to exactly one equivalent strain. In terms of viscoelastic material
behaviour, for each stress, two strain values (one for loading and one for unloading) are
equivalent. The inclination of this mechanical hysteresis corresponds to E and the
width describes the phase shift between stress and strain.

2.3. Plastic Deformation and Fracture

Each deformation induced by external forces acting is in a first step basically elastic.
With further increase in loading, changes in the material response occur. In this case,
two basic phenomena are of pronounced importance: plastic flow and fracture.
a) Above a certain threshold stress σy the material deforms by plastic flow (yield
stress, related strain εy = σy/E). This strain contribution is irreversible and
characteristic for ductile materials, e.g. most biomaterials, most metals and
some polymers (Figure 3a).
b) Exceeding a specific threshold stress σr, failure of the material occurs without
any previous plastic deformation (rupture stress, corresponding strain εr =
σr/E). This brittle behaviour is observed especially for glasses, ceramic
materials and hardened metals (Figure 3a).
To characterise plastic behaviour experimentally the monotonic stress-strain test is
suitable (Figure 1). For the analysis of extremely brittle materials (glasses, ceramic
materials) bending experiments are more appropriate.
Plastic deformation can be generally expressed in terms of a complex function
(Equation 4), which describes the material behaviour as a relation of five parameters:
applied stress σ, deformation ε, time t, microstructure S and temperature T, respectively
(please do not worry about the complexity and the mathematical solution of this
“cactus” equation! Things are easier than expected):

Ψ = Ψ (σ ,ε , t , S , T ) (4)

Unfortunately, a closed mathematical solution of this function has not been

developed yet, and therefore, a differentiated procedure in experimental analysis is
performed. The solution of Ψ is obtained in a way that two of the functional parameters
are experimentally measured while the others are kept constant.
Mostly, the mechanical variables σ, ε and the time t in the sense of a deformation
or strain rate dε/dt are of pronounced interest. Therefore, three basic static experiments
are of superior interest:
In the monotonic stress-strain test, the stress is continuously increased and
registered as a function of the resulting strain while the parameters time t in the sense
of strain rate dε/dt, S and T remain constant (Figure 3a). Measuring the deformation
236 S. Dendorfer et al. / Characterisation and Testing of Biomaterials

ε versus time t under constant stress or load is defined as creep experiment (Figure 3b),
and finally, constant deformation ε leads to a functional relation between stress σ and
time t which is generally defined as stress relaxation test (Figure 3c).

Figure 3: Basic types of results describing the mechanical behaviour under monotonic loading: Stress-strain
curves for ductile and brittle materials and corresponding fracture morphology (schematic), a); creep
experiment, b); stress relaxation test with time dependent stress decrease until a constant relaxed stress is
reached (internal stress), c).

2.4. Summary

Elastic Material Behaviour

Purely elastic material behaviour can be generally defined by three basic
characteristics:
• complete reversibility after load reduction
• linear relation between applied load and resulting deformation
• limitation to small deformations (i.e. ≤ 1 % for crystalline materials)
 S. Dendorfer et al. / Characterisation and Testing of Biomaterials 237

Viscoelastic Behaviour
The material response to loading or unloading is basically reversible, but deformation
is retarded with respect to the actual load value.
• characterizes time dependent elastic deformation
• results in damping of externally induced oscillations
• strength and modulus depend on the strain rate at which the material is
deformed
Plastic Deformation
• occurs after a material dependent threshold stress is exceeded
• is always combined with the irreversibility of deformation
• no linear relation between stress and deformation, characteristic line for
the plasticity of materials is the stress-strain curve
• characteristics of the stress-strain curve are:
o work hardening and thus more stress is required for further
deformation
o maximum defining the tensile strength
o decrease until fracture
Ductile/Brittle Behaviour
• ductile materials exhibit measurable (visible) plastic deformations
• brittle material behaviour is defined for fracture without (practical) plastic
deformation
• fracture occurs directly in the elastic region
Creep
As often (falsely) assumed, creep does not necessarily implicate deformation for very
long times, but only basically describes plastic deformation under constant stresses.
• time dependent plastic deformation, usually under constant load or stress

3. Experimental Methods

Additional to the in the following reviewed standard experimental tests, other

techniques have been developed to address specific research questions. For example,
indentation tests are widely used to analyse surfaces or layers e.g. [2] and various
multi-axial testing methods are used to characterise material properties e.g. [18]. But
these methods are not within the scope of this paper.

3.1. Stress-Strain-Experiments

The information obtained from the standardized stress-strain experiment is of basic

superior interest for the characterisation of materials under monotonic loading. Figure 4
shows a typical test rig for mechanical testing.
238 S. Dendorfer et al. / Characterisation and Testing of Biomaterials

Figure 4: Schematic test rig for analysing mechanical properties, a) and standardized specimen geometry for
tensile experiments, b).

In most cases tensile experiments are performed. With respect to the use of
different specimen geometries and dimensions, the ratio of cross section to gauge
length is specified to: l0 = 5.65 √A0 (Figure 4b). (This value is used in traditional
material science. Even so, it may not be possible to create arbitrary specimen
geometries with biomaterial it is of high importance to use standardised sizes.) The
specimen is deformed under a constant extension rate dε/dt while the applied load F
and the corresponding deformation Δl are continuously recorded and plotted in force vs.
displacement curves. Normalizing the data by the initial specimen cross section A0 and
the initial length l0, respectively, gives the geometry independent stress-strain curve
σ = σ(ε), e.g. Figure 3. This curve represents the characteristic deformation of the
specific material. It depends not only on the chemical composition of the material, but
also on microstructure (and processing conditions).
How to interpret stress-strain diagrams ? Starting from the origin, the first section
of the curve corresponds to the linear increase in stress (slope E). The transition from
elastic to plastic material behaviour is not clearly obvious in many cases, but is more or
less continuous. Due to this, it is often advantageous to define an additional material
parameter: The flow stress Rp0.2 (Yield Strength YS) is defined as the stress under
which a plastic deformation ε = 0.2 % is obtained after complete unloading. Strain
hardening is characterized by an increase in stress for further deformation (strain). This
can be interpreted that for additional strain, the internal resistance against this
deformation increases (work hardening, characteristic for the deformation of metals).
Consequently, this hardening reaches a peak value Rm (Ultimate Tensile Strength, UTS).
In the following, the force needed for further strain decreases and signalizes “that the
end is near”. The material loses strength until fracture occurs. Strictly spoken, it is
impossible to load a material with loads exceeding Rm. It should also be mentioned that
in the case of work hardening effects, the elastic regime of the material is also
 S. Dendorfer et al. / Characterisation and Testing of Biomaterials 239

prolonged. In a second load cycle the maximum stress/strain of the first load cycle may
be reached.
The area below the stress-strain curve provides another important parameter.
Mathematically, the area corresponds to the integral and is equivalent to the work
required for the material deformation and thus, can be regarded as energy density,
usually measured in J/m2:

εf l
1
W = ∫ σ dε
0
=
A0 ⋅ l0 ∫ F dl
l0
(5)

The practical sense of the integral, W, is the work required for a deformation
process, e.g. in terms of the energy absorption, i.e. compensation of kinetic energy due
to impact loading. Further characteristic parameters which can be obtained from the
stress-strain curve are: uniform deformation, describing homogeneous elongation of a
specimen under simultaneous reduction in cross section until Rm. After Rm necking is
initiated, leading to a local reduction in cross section and all subsequent deformation is
confined at this neck. Finally, fracture occurs at the neck.
Testing of certain biomaterials may not result in as nicely shaped curves. E.g. the
tensile testing of cartilage results in a stress-strain curve which exhibits a smooth
transition from initial deformation to the elastic regime with rather large deformations
[11]. This transition zone is often referred to as the toe region and is mostly affected
due to the un-crimping of collagen fibres and elasticity of elastin.

3.2. Creep

Creep is generally defined as time controlled plastic deformation at constant stress. The
initial deformation rate dε/dt continuously decreases, although, the deformation is not
stopped (it does not approach zero) and the material still deforms (creeps) with slow
deformation rates. For constant stress, deformation ε follows a function of time t and
for describing the material behaviour the stress-strain curve is replaced by a strain vs.
time curve ε(t) (creep curve, Figure 5).
Generally, creep can be characterized by three classic regions: Immediately after
mechanical loading and elastic deformation equivalent to the magnitude of the applied
stress εelsat = σ/E, a transition to plastic deformation is observed (primary creep).
During the second stage a constant, stationary creep rate is acting. In the third stage
creep rates increase again and failure is initiated (Figure 5). Secondary creep is
characterized by a constant microscopic structure of the material. This steady state can
be regarded as dynamic equilibrium: While the material reacts on the applied load by
continuous work hardening (i.e. (dσ/dε)dε > 0) competitive softening effects are acting
(dσ/dε)dε < 0).
Steady state creep is therefore observed, if the increase in deformation resistance is
compensated by softening effects of the material (relaxation). After sufficient
deformation (tertiary creep), an acceleration in creep rate is observed until fracture
occurs. This increase in creep is substantially driven by rate controlling damaging
mechanisms which already take place during secondary creep but become more
pronounced in the tertiary creep regime, indicating that the end is near. Creep fracture
occurs at sufficient high strains which only slightly depend on loading and temperature.
240 S. Dendorfer et al. / Characterisation and Testing of Biomaterials

Figure 5: Typical creep curve for loading under constant stress: Strain vs. time, a) and differentiated form
strain rate vs. strain.

3.3. Fatigue Testing

Strength is generally defined as deformation resistance against monotonic (static) loads.

A further and practically more relevant load condition is the material behaviour under
loads varying with time (cyclic deformation). This effect of fatigue is of pronounced
importance as many materials or structures exhibit significant reductions in load
bearing capability for stresses far below the monotonic strength when exposed to cyclic
loading. Fatigue failure is always brittle-like in nature even for ductile materials. It is
catastrophic and insidious, occurring very suddenly and without warning. The process
of fatigue failure is driven by the initiation and propagation of short cracks and
ordinarily the fracture surface is perpendicular to the direction of the loading axis.
Cyclic Stresses
The applied stress may be axial (tension-compression), flexural (bending) or torsional
(twisting) in nature. The material behaviour is analysed in most cases for sinusoidal
loading.
This is done either in force or in deformation control while the number of load
cycles until fracture is counted. In addition to the oscillating stress amplitude, a
constant static stress can be superimposed. With respect to this mean stress, different
forms of fatigue loading are defined. For example tensile fatigue loading is
characterised by the peak and valley loads being both tensile. In the range of stresses
where the maximum load is tensile while the minimum load is compressive, the ratio
between peak and valley load is defined by the stress ratio R = σu /σo.
 S. Dendorfer et al. / Characterisation and Testing of Biomaterials 241

Figure 6: Schematic S-N curve and determination of fatigue life and fatigue strength.

The S-N Curve

For specifically defined variations in the load amplitudes (Δσ) applied in experiments
the number of cycles (N) is counted until fracture occurs and the results are typically
plotted in a semi-logarithmic S-N curve (Figure 6). If stresses are sufficiently low, a
horizontal section is observed for which the number of cycles is no longer a function of
the applied stress. This threshold is defined as endurance limit S e. This means that for
stress amplitudes below Se no fracture will occur. It should be noted that the S-N curve
is a purely statistical average indicating that 50 % of the experiments exceed the
number of cycles, but 50 % fail at lower cycle numbers, if not specified differently.
Generally, two regions of the S-N curve are defined: In cases of small stresses and high
cycle numbers (usually in the order of 106 or more) high cycle fatigue (HCF) is
dominating, whereas for stresses in the order of the Yield Strength and load cycles
between 102 to 105 , low cycle fatigue (LCF) is present.
Failure is not always as well defined as in tension fatigue test where the specimen
finally ruptures. For example in compression fatigue tests, a failure criterion has to be
defined. This criterion could be, for example, a certain reduction in the material
stiffness compared to the initial, unharmed stiffness.
Crack initiation and Propagation
Fatigue failure is characterized by three distinct phases: Crack initiation (phase I),
wherein microscopically small cracks form (usually at the surface) at points of high
stress concentration; crack propagation (phase II) where small cracks grow
incrementally with each applied stress cycle (normally preferred during the tensile
phase of the load cycle) and failure (phase III), initiated very rapidly as soon as one
crack has reached a critical size. It should be emphasised that all geometries which
cause stress concentrations can be regarded as sites of crack nucleation, i.e. threads,
sharp corners, drill holes, surface roughness or scratches and (micro-) porosity.
242 S. Dendorfer et al. / Characterisation and Testing of Biomaterials

4. In-Vitro Characterisation

In order to transfer the experimental procedure for the characterisation of materials to

the in vitro characterisation of biomaterials, specific emphasis has to be focussed on the
related boundary conditions. This aspect is of utmost importance in order to derive
reliable and comparable results in the sense of a clear scientific interpretation and
comparison to data in the literature.

4.1. Boundary Conditions

In Table 2, a rough summary of relevant boundary conditions which can significantly

influence the results of in vitro experiments is demonstrated. For practical experimental
analysis and with respect to the variety of influencing parameters, well defined
planning is essential for obtaining clear reliable results, from which further precise
conclusions can be made, concerning the transfer of the experimental data to in vivo
tissue / material behaviour or interaction. Referring to Equation 4, this implicates that
the experiment should be kept as simple as possible and as few parameters as necessary
should be varied. Furthermore, the experimental set up must be carefully assembled
and optimized in terms of the required accuracy. Thus, danger of the occurrence of
artefacts and a misinterpretation of the results can be minimized. Although, in many
cases some parameters of interest, for example, the specimen structure, can only be
determined with a large experimental effort, it is always better to follow this way than
to let parameters unclear and interpret uncertain results by means of statistical methods.
Strictly speaking, for experimental in vitro analysis, all required parameters should be
defined and controlled during the tests. This guarantees precise and transferable
conclusions and finally will lead to minimized deviations in the concluding statistical
view which is indispensable in any case.

4.2. Characterisation of the microstructure

Almost all biomaterials exhibit a highly inhomogeneous and anisotropic microstructure.

Therefore, the material behaviour will be highly dependent on the orientation of the
microstructure with respect to the acting load. This implies the necessity of examining
the structure and careful alignment of the specimen axis and the load axis [8]. The
analysis of the microstructure is often accomplished with image based procedures.
Depending on the scale level, these methods could be light microscopy, x-ray
quantitative computed tomography, micro-magnetic resonance imaging or high
resolution micro-CT. Numerous researchers studied the structure of materials with
these methods and derived measures for quantification e.g. [22], [1], [3].

4.3. Example of In Vitro Behaviour

In order to provide an example of the requirements described above, in vitro

experiments on cancellous bone will be demonstrated and discussed in the following
section. As physiological loading of cancellous bone structures is mostly compressive,
the experiments reported are also performed in compression.
The response of a bovine cancellous bone specimen exposed to monotonic
compressive loading is shown in Figure 7. Typical for a cellular solid [14], the
monotonic loaded cancellous bone specimens’ behaviour was found to be linear (in
 S. Dendorfer et al. / Characterisation and Testing of Biomaterials 243

stress-strain space) at the beginning of the compression. The linear regime is followed
by a decreasing slope, indicating failure and plastic collapse, and an increase
(densification) at higher strain values. While the magnitude of the maximum
compressive stress varies up to tenfold between the groups, the characteristics of the
deformation behaviour are similar. Even if the initial deformation behaviour appears to
be rather linear, non-linearities can be found by analysing the deformations with a
higher local resolution. For this reason optical surface deformation measurements are
applied. The surface deformations of the specimen reveal these inhomogeneous
components. Localised strain concentrations can be found already at a small percentage
of the maximum stress. These strain concentrations do further localise and increase in
value until failure occurs in the same region. Macroscopic damage appears in form of a
slip line across the whole specimen cross section.

Table 2: Important boundary conditions to be considered for in vitro testing.

Type of Boundary Condition

Physiological Biological Mechanical Analytical /

Preparation

Temperature Biological Preload Definition of experiment

Isotonic / wet / scatter: (Micro-) Strain rate and limitations
dry structure Specimen fixing
Load application
Strain rate Human / animal Integral / differential strain
Transient effects / arte-
Loading state Male / female facts measurement
Age Fixing quality Sample rate
Cell vitality Specimen dimension Preparation/pre-damage
Orientation Set up stiffness Pre-conditioning
Test with / Stress state Initial state characterisation
without (density, morphology,
surrounding Specimen alignment image resolution ..)
tissue Friction effects Signal drift

The early appearance of these strain localisations indicates that structural damage
is already induced at rather low stress levels, even at the beginning of the elastic regime.
Figure 7 shows also data for other types of cancellous bone, human vertebral bone
cored in the main physiological axis and human vertebral bone cored perpendicular to
the main axis. A remarkably different behaviour is obvious. An explanation of these
differences can be found by looking at the underlying structure. Firstly, there is a large
difference in density between bovine and human specimens. Furthermore, the bovine
specimens contain a combination of plate and rod like trabeculae (Figure 8), whereas
the human bone is strictly composed of rods (Figure 9). A rather large influence of the
orientation of the main material axis with respect to the applied load vector is also
244 S. Dendorfer et al. / Characterisation and Testing of Biomaterials

observed. Stiffness and strength decrease enormously by changing the load axis from 0
degree (perfectly aligned with the main material axis) to 90 degree (perpendicular).
This difference can be explained with the high anisotropy of the trabecular network.
Furthermore, the influence of non-optimal boundary conditions is shown in Figure 7.
The dotted line shows a distinctive toe region, but in contrast to tests on cartilage where
this behaviour is a result of the underlying microstructure, here it is an experimental
artefact due to early trabecula failure on the boundaries and small misalignments. In all
the other shown curves, the specimens were embedded on both sides in order to
achieve a homogeneous load transfer from the test rig to the specimen.

Figure 7: Stress-strain curves for different types of cancellous bone, a). The dotted line shows a specimen
without embedding at the ends and therefore experimental artifacts in the beginning of the test. Von-Mises
surface strains (analyzed with an optical deformation measurement system) are shown at different stages
during the experiment, b)-d). The angle value refers to the deviation between load vector and main
physiological axis. [7]

Only taking into account all the experimental and specific material-structure
properties results in comparable, reproducible and valuable data.
 S. Dendorfer et al. / Characterisation and Testing of Biomaterials 245

Figure 8: Bovine vertebral cancellous bone, Figure 9: Human vertebral cancellous bone, main
main physiological axis in vertical direction. [7] physiological axis in vertical direction. [7]

5. State of the art and future directions

Nowadays, image based methods are increasingly used to characterise the materials
microstructure. For example, micro-CT based methods set a new standard in the
quantitative evaluation of the structure. These techniques provide high resolution
images of the structures and image processing methods can create a 3D computational
model from the data. An active research field is the derivation of quantitative measures
for both, the description of the structure but also for the linkage between structure and
mechanical properties [25], [5]. Furthermore, the damage and deformation behaviour of
biomaterials can be studied using high resolution scans which results in a much deeper
understanding of the ongoing processes. Imaging in combination with numerical
methods like Finite Element Analysis and additional data from mechanical testing has
proven to be a very promising way forward. [21], [17].
With ongoing research on the microstructures of the materials a basis can be set to
improve the relationships, originally derived for engineering materials to the specific
need and challenges of biomaterials [23]. Recently, novel methods in musculoskeletal
simulations allow insight in the in-vivo forces acting on many biomaterials [6]. This
information can be used to adjust experimental conditions to approach the relevant in-
vivo scenario.
Besides the technological advances, a standardisation of load cases and
experimental conditions will be a challenge for the future. The need for an improved
comparability of data from various sources as well as the increasingly complex man-
made biomaterials has to be drawn to standards, comparable to engineering design
codes.

Acknowledgements

The authors would like to thank the staff at the Laboratory for Materials Science at the
University of Applied Sciences Regensburg. Special thanks are due to R. Mai for
technical and experimental support, to N. Fischer for preparing the images and to H.-P.
Bräu for providing the electron microscopic images.
246 S. Dendorfer et al. / Characterisation and Testing of Biomaterials

References

[1] O. Beuf, D. C. Newitt, L. Mosekilde, and S. Majumdar. Trabecular structure assessment in lumbar
vertebrae specimens using quantitative magnetic resonance imaging and relationship with mechanical
competence. J Bone Miner Res, 16(8):1511–9, 2001.
[2] C. Yanping, M. Duancheng, and D. Raabe. The use of flat punch indentation to determine the
viscoelastic properties in the time and frequency domains of a soft layer bonded to a rigid substrate.
Acta Biomater, 5(1):240–248, Jan 2009.
[3] D. Chappard, P. Guggenbuhl, E. Legrand, M. F. Basle, and M. Audran. Texture analysis of X-ray
radiographs is correlated with bone histomorphometry. J Bone Miner Metab, 23(1):24–9, 2005.
[4] S.C. Cowin. Bone mechanics handbook. CRC Press Inc., Boca Raton, 2001.
[5] S.C. Cowin. The relationship between the elasticity tensor and the fabric tensor. Mechanics of Materials,
4(2):137–147, 1985.
[6] M. Damsgaard, J. Rasmussen, S. Tørholm, E. Surma, and M. deZee. Analysis of musculoskeletal
systems in the anybody modeling system. Simulation Modelling Practice and Theory., 14:1100–1111,
2006.
[7] S. Dendorfer. Cyclic deformation and fatigue behaviour in cancellous bone. PhD thesis, University
Paderborn, 2008.
[8] S. Dendorfer, H. J. Maier, D. Taylor, and J. Hammer. Anisotropy of the fatigue behaviour of cancellous
bone. J Biomech, 41(3):636–641, 2008.
[9] S. Dendorfer, H. J Maier, and J. Hammer. How do anisotropy and age affect fatigue and damage in
cancellous bone? Stud Health Technol Inform, 133:68–74, 2008.
[10] [edited by] Mow, Van C. Basic orthopaedic biomechanics. Lippincott Williams and Wilkins,
Philadelphia, 2nd edition, 1997.
[11] J. Fierlbeck, J. Hammer, C. Englert, and R. L. Reuben. Biomechanical properties of articular cartilage
as a standard for biologically integrated interfaces. Technol Health Care, 14(6):541–547, 2006.
[12] Y. C. Fung. Biomechanics: motion, flow, stress, and growth. Springer-Verlag, New-York, 1990.
[13] Y. C. Fung and Pin Tong. Classical and computational solid mechanics. World Scientific, 2001.
[14] L. J. Gibson and M. F. Ashby. Cellular solids: structure & properties. Cambridge University Press,
Cambridge, 2nd, 1997.
[15] R.W. Hertzberg. Deformation and fracture mechanics of engineering materials. John Wiley and Sons,
Inc., 1995.
[16] R. B. Martin, D. B. Burr, and N. A. Sharkey. Skeletal Tissue Mechanics. Springer Verlag, New York,
1998.
[17] A. Nazarian, J. Muller, D. Zurakowski, R. Mueller, and B. D. Snyder. Densitometric, morphometric
and mechanical distributions in the human proximal femur. J Biomech, Jan 2007.
[18] G. L. Niebur, M. J. Feldstein, and T. M. Keaveny. Biaxial failure behavior of bovine tibial trabecular
bone. J Biomech Eng, 124(6):699–705, 2002.
[19] M. Nordin, Oe. Nihat. Fundamentals of Biomechanics. Springer-Verlag New York, Inc., 1999.
[20] S. Suresh. Fatigue of Materials. University Press, Cambridge, 2 edition, 1998.
[21] P. J. Thurner, P. Wyss, R. Voide, M. Stauber, M. Stampanoni, U. Sennhauser, and R. Mueller. Time-
lapsed investigation of three-dimensional failure and damage accumulation in trabecular bone using
synchrotron light. Bone, 39(2):289–299, Aug 2006.
[22] W. J. Whitehouse. The quantitative morphology of anisotropic trabecular bone. Journal of Microscopy,
101(2):153–168, 1974.
[23] G.M Williams, K. R Gratz, and R. L Sah. Asymmetrical strain distributions and neutral axis location of
cartilage in flexure. J Biomech, Dec 2008.
[24] H. Yamada. Strength of biological material. Williams & Wilkins, Baltimore, 1970.
[25] P. K. Zysset. A review of morphology-elasticity relationships in human trabecular bone: theories and
experiments. J Biomech, 36(10):1469–1485, 2003.
Basic Engineering for Medics and Biologists 247
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-247

Introduction to
Chapter V: Medical Imaging
Peter NIEDERER and T. Clive LEE (eds.)

According to a German proverb, a picture says more than a thousand words. This
statement is certainly true in clinical diagnostics. With the advent of the X-ray
projection technique, somewhat more that 100 years ago, it became for the first time
feasible to obtain insight into the human body without opening it. Since then, various
methods of medical imaging have developed into major tools in clinical diagnosis
which cover an enormously wide field of pathologic situations. The various modalities
are still in partly spectacular development.
Ultrasound, X-ray and magnetic spin resonance are the most often applied
modalities in medical imaging performed on the intact human body. The basic physics
underlying the procedures derived thereof are well known and established, and they are
outlined in the next three sections. The aim is such that the reader is in a position to
understand and appreciate present and novel developments and clinical applications for
which there exists a vast literature.
Mathematicl methods, in particular Fourier analysis, along with elementary
physics are essential prerequisites in the understanding of the methods under
consideration in the following. An attempt has nevertheless been made in agreement
with the goals of this book such that the basics can be understood without a thorough
mathematical background. The relevant mathematics are given in appendices.
There are a number of further imaging modalities with an increasing importance in
medical imaging, viz, Positron Emission Tomography (PET), Single Photon Emission
Computed Tomography (SPECT) and methods based on the application of infrared, to
mention the most important. In part, similar (in particular mathematical) methods are
applied in these technologies. Since this book is limited to the treatment of the basic
elements of the most important imaging methods, these procedures are not included.
A further important area in medical imaging is devoted to image analysis with the
aid of the computer. At least some numerical treatment is necessary to present the
results of measurements made with ultrasound, X-ray, or magnetic spin resonance on a
computer-driven screen in a clinically useful form. However, medical image analysis,
regardless whether it is performed automatically or with human intervention, is a wide
field in its own right and is not approached here.
This page intentionally left blank
Basic Engineering for Medics and Biologists 249
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-249

V.1. Ultrasound Imaging and Doppler

Flow Velocity Measurement
Peter F. NIEDERER
Institute for Biomedical Engineering
ETH Zurich and University of Zurich
Gloriastrasse 35
CH-8092 Zurich, Switzerland

Abstract. High frequency ultrasound (2 – 8 MHz typically) has established itself

as a major medical imaging method associated with a wide range of clinical
applications. Advantages include real-time applicability, lower cost compared with
other medical imaging technologies, possibility of measuring blood flow velocities
and desk-top instrumentation. Disadvantage is associated with lower image quality
than is obtained with x-ray or magnetic resonance methods.

Keywords. Ultrasound, medical imaging, Doppler flow measurement, color

Doppler

Introduction

Clinical applications of ultrasound include diagnostic as well as therapeutic procedures.

From a technical point of view, the difference between diagnostic and therapeutic
applications firstly consists of the amount of energy thereby delivered to the tissue: In
case of diagnostic ultrasound, the power density at the body surface is limited to
100mW/cm2, except for the eye where a lower limit of 20 mW/cm2 is set. Below these
limits, any effect on the tissue can be excluded with sufficient certainty. Therapeutic
ultrasound, in contrast, is intended to induce a tissue reaction, mostly in the form of
heating and increase of blood perfusion, therefore irradiated energies are higher.
Secondly, ultrasound therapy is essentially limited to the delivery of energy, whereas in
diagnostic applications the analysis of the backscattered signal (“echo”) is to the fore.
Ultrasound frequency bands used extend from kHz (mostly therapeutic ultrasound)
to MHz (imaging and Doppler technique) and GHz (ultrasound microscopy). The
wavelengths accordingly are in the cm to sub-μm range (corresponding to the
conditions prevailing in water; wave speed about 1500 m/sec).
The first technical applications of ultrasound were made during the first World
War where SONAR methods (Sound Navigation and Ranging) were developed in order
to detect submarines (the French physicist Paul Langevin built the first ultrasound
oscillator in 1914). A breakthrough was however achieved only after electrical short-
pulse techniques became available. During the twenties, material testing methods based
on ultrasound were introduced (S. Sokolov, F. Firestone and others).
The first use of ultrasound for medical imaging was made by the Austrian
neurologist Karl Dussek together with his brother Friedrich Dussek who presented their
250 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

results in 1938 (Figure 1). Doppler methods for the measurement of blood flow
velocities were finally developed by the Japanese physicist S. Satomura in 1957.

Figure 1. “Hyperphonogram” of the brain, Dussek 1938

This primer is devoted to ultrasound imaging and Doppler methods. In both

techniques an electromechanical transducer is positioned on the body surface from
which short ultrasound pulses in the form of a wave package with a defined center
frequency are propagated into the tissue. The backscattered echo signal is then recorded
by the same transducer in order to synthesize a cross section of the body part under
consideration and/or to investigate the local perfusion of the tissue by way of an
analysis of the Doppler-shifted frequency components of the backscattered signal. As
soon as all the echoes generated by the pulse in the tissue and backscattered within the
predetermined range of the instrument have reached the transducer, the next pulse is
emitted, etc.
In comparison with other medical imaging modalities such as computed
tomography or magnetic resonance imaging, ultrasound has three major advantages,
viz.,
x Real-time operation
x Inexpensive
x Availability of small and light-weight instruments which can easily be moved
around in a clinic
A disadvantage derives from the generally limited image quality, however.
Ultrasound waves as used in medical imaging consist in essence of compression
waves; other types of waves such as shear or surface waves are of minor importance
and can be neglected. The propagation of such waves in an inhomogeneous biological
medium is however highly complex. Concurrent coherent and incoherent scattering
processes along with strong damping cause the generation of complicated echo signals
from which it is difficult to extract the signal components of interest. Mirror-like
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 251

reflections occurring at layered structures or speckle formation are therefore among the
major difficulties.
Typical pressure amplitudes associated with diagnostic ultrasound applications in
the MHz range are of the order of a few MPa. Such pressures are sufficiently high to
induce nonlinear scattering processes involving, e.g., mode conversion. This
phenomenon can be used for image enhancement which is usually denoted as 2nd
harmonic imaging.
Bony or air-filled structures are, for practical purposes, non-transparent for
diagnostic ultrasound waves because of almost complete reflection (an exception is the
transcranial Doppler method). Accordingly, one cannot “see” behind bones, the lung
or air bubbles in the intestine. Due to the inhomogeneous composition of typical
biological materials, furthermore, the propagation of ultrasound waves is not directed
along exactly straight lines (such as x-rays). For this reason, amongst others, all
attempts to develop ultrasound CT imaging have not been successful to date.
According to a rule of thumb, diagnostic ultrasound in the MHz range is damped
by about 1 db per MHz and cm penetration depth (Figure 2). In applications such as
cardiology where distances up to some 25 cm should be covered to create a useful
image, frequencies higher than about 8 MHz cannot be used because of signal/noise
limitation. The resolution, in turn, should at least reach around 1 mm for practical
applications such that frequencies below 2 MHz are not useful because the resolution is
essentially given by the wavelength which increases with decreasing frequency. In
applications requiring a short penetration distance only (eye, skin), frequencies up to 20
MHz can be used. Even higher frequencies are applied in ultrasound microscopy,
where GHz waves with a wavelength below 1 μm are utilized. Penetration depth is
however limited to a few tens of μm because of damping.

d d : damping (db)
z·f z : distance (cm)
f : frequency (MHz)
2.5

2.0 muscle

1.5

liver
1.0
brain

0.5 fat

0.1 0.2 0.4 0.6 0.8 1.0 2 4 6 8 10

frequency (MHz)

Figure 2. Damping of ultrasound waves in selected biological tissues

252 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

From the point of view of signal generation and analysis, medical ultrasound
technology has much in common with RADAR (Radio Detection and Ranging) used in
general aviation; major differences derive from the fact that RADAR waves are
electromagnetic waves and operate in the GHz range.

1. Ultrasound Transducers

Ultrasound waves are mechanical by nature whereas the preparation of suitable pulses,
as well as the signal analysis and representation of measured results, is entirely based
on electronics. Accordingly, electromechanical transducers are necessary for the
generation of ultrasound waves and for the recording of echoes. Piezoelectricity
provides the most useful tool for this purpose: When certain crystals are exposed to an
electric tension, they deform because of rearrangement of electric charges in the
interior; in turn, upon deformation, charges are produced at the surface. This effect is
denoted as piezoelectricity, and it can be applied because of its reciprocity for emitting
as well as for receiving ultrasound waves.
A piezoelectric ultrasound transducer consists of an arrangement where one or
more small bars or plates made of piezoelectric material are placed on a rigid,
electrically inactive backing material mounted in a hand-held application device. Both
surfaces of the piezo-elements are electrically connected, much as an electric condenser,
to an oscillator for transmission and switched to a receiver for recording echoes. While
the oscillator sets the transducer into vibration such that ultrasound waves (mostly in
the form of short pulses) are emitted, the returning echoes deform the piezo-material
passively whereby electric signals are generated that can be recorded. Typical materials
used for medical ultrasound transducers are based on sintered lead zirconate titanate
(PZT) where the piezoelectric effect is particularly strong.

1.1. Generation of Ultrasound Waves, Beam Characteristics of Transducers, Phased

Array Technology

In order to analyse the characteristics of wave fields produced by oscillating transducer

surfaces theoretically, Huygens’ principle (see Appendix 2) is most useful1. According
to this principle, a wave field can be synthesized from a continuous distribution of a
dense set of vibrating point sources located on a surface enclosing the area of interest.
Each point is thereby the origin of a spherical wave.
As an example, we consider the wave field produced by a circular transducer of
radius a which is set into vibration in a thickness mode at a given frequency (Figure 3).
The transducer is assumed to be located in a uniform fluid-filled cartesian space
(coordinates x , y , z ).
Only the field along the axis of symmetry, x , can be calculated analytically; off-
axis values have to be determined numerically (see later). As outlined in the
introduction and worked out in Appendix 1, compression waves in the fluid are only
considered. For the solution of such wave propagation problems, it is advantageous to

1
The principle is valid for linear problems only. The nonlinearities mentioned in the introduction,
however, are of importance only in the interaction of the waves with scattering centers and not in the
generation and propagation of the wave field itself.
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 253

introduce a scalar velocity potential, < ( x , y , z ,t ) , from which the velocity field
&
v ( x , y , z , t ) to be calculated is obtained as the gradient ( t denotes the time)
& &
v ( x , y , z ,t )  ’ < ( x , y , z ,t ) (1)

. z

r
K P x

vibrating surface

Figure 3. Vibrating disk on a rigid backing support. The frontal surface is assumed to vibrate uniformly with
a constant amplitude while the one in the back is at rest.

According to Huygens’ principle, the velocity potential has to be prescribed over a

closed surface containing the space of interest. This surface includes the transducer
surface along with the y - z plane and an enclosure at infinity over the right half space.
Since only the transducer surface vibrates, < = 0 outside. The transducer furthermore
vibrates sinusoidally in x – direction with a constant amplitude such that only one
&
component of the gradient, ’ < , namely w < w x is z 0 . We therefore have as
boundary condition

w<
v0 e i Z t for K d a
wx (2)
0 outside

whereby v0 denotes the (constant) amplitude of the vibration,

Z 2S f the circular frequency (frequency f ),
K the radial coordinate on the transducer surface.
(Since linearity is assumed, a complex notation can be used).
The velocity potential at an arbitrary point P located on the x  axis is now
obtained according to Huygens’ principle as the integral of a continuous distribution of
point sources over the transducer surface, all of them producing the same spherical
v0  i ( k r  Z t )
wave w( r ,t ) e with r denoting the distance between the point
r
254 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

source and P ( r 2 K 2  x 2 ) and k the wave number, k Z c 2S O (wave speed

c , wave length O ). (The amplitude of a spherical wave is proportional to the square-
root of its energy per unit area, therefore, it is proportional to 1 r .)
The integral extends over the surface with radius a (single integration due to
symmetry with the circular surface element 2SKdK and the factor 2S according to
Appendix 2)

a
v0 e i Z t ª1  i k r º
<P ³ «r e » 2 S K dK (3)
2S 0 ¬ ¼
dK
a
After integration the relative intensity (energy K
Density I at point P can be calculated from
the real part of <P squared. One arrives at

ªS º
I v sin 2 « §¨ a 2  x 2  x ·¸» (4)
¬ O © ¹ ¼

Figure 4 exhibits this result qualitatively. It is important to note the typical near-field
and far-field characteristics; from a practical point of view, the useful area of such a
transducer is at a distance of a 2 / O . This aspect will further be discussed below.

Intensity

x
a 2/O - O /4

near field far field

Figure 4. Near field and far field characteristics of a circular ultrasound transducer. The intensity varies
rapidly close to the transducer.

The off-axis intensity can be determined by numerical integration (Figure 5; also

analytical approximations have been presented in the literature). It is found that besides
a main lobe, several side lobes exist. The larger the diameter of the transducer is in
relation to the ultrasound wavelength, the narrower is the beam, furthermore, the
number of side lobes increases.
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 255

2a

Figure 5. Main and side lobes of a circular ultrasound transducer. Note the influence of the a/O ratio on the
beam characteristics.

Note: Since linearity is assumed, these results hold for ultrasound propagation in
continuous wave or pulsed mode. As we shall see, most medical ultrasound
applications use pulsed mode techniques however; the similarity with RADAR has
been mentioned before. A typical ultrasound pulse as used in medical applications is
shown in Figure 6. (Schlieren imaging method).

Figure 6. Schlieren image of a 5 MHz ultrasound pulse consisting of a package of 8 sinusoidal waves,
wavelength 0.3 mm.
256 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

One of the methods adopted from RADAR is the “phased array” technique which
is in widespread usage in pulsed ultrasound devices. Thereby, the Huygens’ principle is
approximated by an array of small transducers which produce each a quasi-spherical
wave package having the envelope of a short pulse. The timing of each pulse along
with the amplitude and phase of the center frequency are controlled such that, upon
superposition, a combined wave forming a main lobe with a defined spatial orientation
results (Figure 7). The problem how the amplitudes and phases of the individual
transducer elements of a phased array have to be controlled in order to obtain a
predefined far field is addressed in Appendix 3.

wave front

transducer elements

Figure 7. Phased array technique. Each individual transducer element is excited independently with a pulse
whose timing is such that by superposition a beam with a defined direction of propagation is formed. By
dynamic variation with time, e.g., an oscillating beam can be produced which scans an area of interest much
like a windshield wiper. This technique is used for producing real-time imaging of selected cross-sections of
the body. Most phased arrays presently in use are linear, one-dimensional arrangements; two-dimensional
arrays which allow scanning of a volume can also be made, however.

The more transducer elements are used, the better a desired wave form can be
obtained. In turn, transducer elements have to be made smaller when more elements are
implemented which may cause mechanical sensitivity and electrical impedance
matching problems. Furthermore, individual transducer elements cannot be positioned
too close to each other because of mechanical crosstalk. If too far away, however, the
Huygens’ principle my not be properly approximated in that the distance of adjacent
elements has to be smaller than O 2 , otherwise, the sampling theorem is violated (not
worked out here).
In the analysis of propagating ultrasound waves considered so far, we have
assumed a vibrating surface consisting theoretically of an infinite number of point
sources according to Huygens’ principle. Yet, the transducer is a mechanical oscillator
in its own right associated with its particular mechanical characteristics, viz., mass,
elasticity and damping. In the simplest approximation, a transducer can be modeled as
a linear oscillator. Its eigenfrequency is given by the mass and deformation parameters,
while the width of the resonance curve is determined by the damping. Transducers are
usually operated at their basic eigenfrequency and the bandwidth of the response
corresponds to the resonance curve. Without going into mathematical detail, one finds
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 257

that the resonance curve becomes higher and narrower as the damping is smaller. Such
a transducer exhibits on the one hand a high sensitivity; on the other, undesired
transient vibration behavior (“ringing”) renders a transducer with insufficient damping
unsuitable. The higher the damping, in turn, the larger the bandwidth becomes and the
better a pulse form is reproduced mechanically, this at the cost of sensitivity. In
conclusion, the fabrication of ultrasound transducers requires a high amount of
experience, precision work and skill and is a true industrial specialty.

1.2. Impedance Matching

The acoustic impedance U˜c (density of the material times speed of sound, see
Appendix 1) of transducers is typically a few orders of magnitude higher than the one
of biological tissues (the density of water is 1 g/cm3, of PZT around 10 g/cm3, the
speed of sound in water is 1500 m/sec, in PZT ten times higher). At interfaces between
different materials, waves are reflected and refracted much as in optics. The amount of
reflexion and the angle of refraction thereby depends on the ratio of the impedances of
the adjacent materials (corresponding to the indices of refraction in optics). The
solution of the wave equation (Appendix 1) shows that in case of large impedance
jumps, waves are mostly reflected. Accordingly, ultrasound waves cannot be
propagated through the skin from a transducer without appropriate impedance
matching measures; in particular, if there is air (impedance four orders of magnitude
smaller than that of water) in between. It has to be kept in mind, thereby, that in
medical ultrasound echo techniques each impedance jump occurs twice, first during
emission, second during backscatter.
Impedance matching is achieved by positioning suitable interface materials
between transducer and skin in the form of surface layers on the transducer and
coupling gels. The rule according to which the properties of the ideal matching material
are determined can be found on the basis of a simplified model (Figure 8). We consider
three layers, viz., the transducer, the matching layer (thickness D ) and the medium
into which the sound wave is to be propagated:
Z1 , Z 2 , Z 3 acoustic impedances ( Ui ci , see Appendix 1), Z1 !! Z 3
<1 velocity potential in the transducer, primary wave
<1' velocity potential in the transducer, reflected wave
<2 , <2' same for the matching layer
<3 medium, outgoing wave

At each interface, there are two physical conditions to be met:

st
&
1 There is no cavitation, i.e., the velocities,  ’ < , on both sides are the same.
2nd The interface itself is massless, implying that the pressures, U w < w t  (see
appendix 1), on both sides are the same.
258 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

transducer matching medium

layer

Z1 Z2 Z3

<1
<2
<3
<2’
<1’

Figure 8. Simplified model of sound propagation through an interface layer of thickness D

The matching is ideal, when <1' 0 (no reflexion into the transducer, all of the
energy is transmitted into the medium). Upon solving the set of linear equations
associated with these conditions, one finds that the thickness of the interface layer, D ,
has to be equal to
D O / 4 while the impedance Z 2 Z1 ˜ Z 3  . (5)
The first result (D = O/4) can be obtained without calculation: Taking into account the
phase jump of S when the wave is reflected at an interface with higher impedance,
considering positive and negative interference of the forward and backward wave will
lead to the desired result.
The refraction characteristics of acoustic waves can be used to design ultrasound
lenses for focusing purposes (Figure 9). If the speed of sound in the lens material is
higher than in biological tissue, a concave lens will produce focusing (in contrast to
optical lenses). With phased arrays, focusing can be achieved by suitable timing of the
individual pulses.

z
Lens Focal region

x Focusing with lens

Transducer

Focusing with phased array

Figure 9. Focusing of ultrasound beams

 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 259

2. Medical Echo technique

In clinical ultrasound applications, ultrasound pulses in the form of wave packages

consisting of 4 to 16 sinusoidal waves are typically utilized (Figure 6). At a frequency f
of 4 – 8 MHz this corresponds to pulses with a duration of 0.5 μsec – 4 μsec (as
mentioned in the introduction, higher frequencies cannot be used because of damping,
at lower frequencies, the spatial resolution would be insufficient). The length of such
pulses is 0.75 mm - 6 mm while the width is determined by the transducer
characteristics and focusing methods. In all of these numbers, it is assumed that the
pulse rise and decay times are short in comparison with the pulse duration. It was
discussed earlier that these properties, i.e., the transient behavior of the transducer is
strongly related to the damping characteristics; the transducer is usually operated at its
lowest resonance frequency.
The pulses are emitted repetitively with the pulse repetition frequency, fPR . Since
all echoes of interest have to return to the transducer before the next pulse is emitted,
the pulse repetition frequency is determined by the operating range of the instrument. If,
e.g, a maximal distance from the transducer of 25 cm has to be reached, the pulse
repetition frequency is limited to 3 kHz corresponding to the return time at a speed of
sound of 1500 m/sec. In case of shorter distances, fPR can be increased which may be
of importance for Doppler application as will be seen later.
With the aid of a phased array, a cross section of the human body can be obtained
without moving the transducer. To this end, the beam is deflected in real time like a
windshield wiper (Figure 10). If 25 frames/sec are made at a pulse repetition frequency
of 3 kHz, say, each frame consists of 190 individual rays. Parallel beam technique can
also be used if the anatomical location allows the application of an extended transducer
array.

Phased array
Body surface transducer

Figure 10. Scanning of a 10 week old foetus in a typical obstetrical application.

260 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

An echo image exhibits in essence all impedance jumps which a beam encounters
as it propagates through the tissue of interest. If the propagation speed of the sound
package can assumed to be constant (1500 m/sec such as in water) the conversion of
the echo sequence into an image is straightforward. Damping is compensated by
exponential amplification.
As mentioned earlier, backscatter may be noncoherent (advantageous for imaging)
or coherent (image deterioration due to mirroring and speckle formation which is due
to interferences). Nonlinear effects during scattering cause the formation of higher
harmonics of the basic ultrasound frequency. The evaluation of the 2nd harmonic can be
used to reduce speckle noise (“2nd harmonic imaging”).
All body parts which are not covered by bone or air volumes where the large
impedance jump usually prevents sufficient wave transmission can be imaged. Primary
applications are in cardiology (heart) and obstetrics (foetus). Physiological fluids
containing minute air bubbles (μm diameter) are sometimes injected systemically to
serve as contrast media to enhance image contrast.

3. Medical Doppler technique

We consider again a pulsed ultrasound method where pulses are emitted with a pulse
repetition frequency fPR. Pulse characteristics (frequency, length, amplitude) are largely
the same as those implemented in echo (imaging) applications.
If ultrasound waves are scattered by moving targets, Doppler shifts are observed
which depend on the speed of the target and the direction of incidence and observation.
Accordingly, this effect can be used to measure blood flow velocities since clouds of
red blood cells act as scattering objects. (The diameter of red blood cells, 8 μm, is very
small with respect to the ultrasound wavelength of around 0.5 mm. The wave
backscattered from one single cell is therefore much too weak to be determined. Blood
cells are inhomogeneously distributed within a blood vessel such that a net effect
remains from the myriads of backscattered waves which would cancel themselves by
interference if the scattering centers would be arranged homogeneously. Nevertheless,
this net effect that makes up the Doppler signal, is usually small.)
Upon transmission of a pulse, the echoes return continuously to the transducer
which, after completion of the transmission cycle, is immediately switched to receiving.
Each echo is processed sequentially whereby it is subdivided into contiguous constant
time steps of duration G ' t , numbered as m 1, 2 ,..., mmax . Accordingly, each
individual step corresponds to a fixed distance from the transducer (Figure 11).
We consider a pulse, ni (i = 1, …) from the sequence of pulses regularly emitted
by the transducer with the pulse repetition frequency fPR, which is emitted at time
i
ti ti 1  . ti 1 thereby denotes the instant in time when the measurement
f PR
begins. The echo associated with this pulse is processed in consecutive channels m
according to the sampling function (Figure 12)

ª n G n Gº
:m
n t  1 for t  « ti 1   m˜ G  , ti 1   m˜ G  »
¬ f PR 2 f PR 2¼ (6)
0 otherwise
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 261

Figure 11. Principle of pulsed Doppler measurement

v blood flow velocity

'x = v / fPR distance traveled by a typical erythrocyte (red blood cell) configuration between the
passage of two consecutive pulses
M angle of the beam axis with respect to the axis of the blood vessel, i.e., the direction of
flow

As mentioned in Figure 11, a particular blood cell cloud travels a distance

v
'x between the passage of two consecutive pulses. We assume that this
f PR
configuration remains constant during the time that it passes through the volume
examined by the beam, i.e., each pulse “looks” at the same cloud such that differences
between echoes which are due to small, flow-induced changes of the configuration can
be neglected. Between each two consecutive pulses, the emitted pulse as well as the
backscattered echo from this cloud have to propagate by an additional distance
'x · cos(M) on their way from and to the transducer. The associated time delay is

'x v 1
'W 2 cos M 2 ˜ cos M (7)
c f PR c

causing a phase shift denoted as Doppler shift of

v 1
'D 2 ˜ Z0 cos M (8)
f PR c
between consecutive echoes arriving at the transducer (Z0 denotes the frequency of the
ultrasound, c the speed of sound).
262 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

Pulse (n) and channel (m) sequence

m = 123 ... mi . . . mmax

Time

n = 1 2 3 4 ...

Sampling function 6 :nm(t)

n

G
m·G + G/2

m·G - G/2

Time

t1 t2 t3 t4
1/fPR

Figure 12. Echo processing sequence and sampling function. G denotes the duration of each channel, m

Starting with pulse n = 1 , the cloud under consideration is now assumed to be

within the volume covered by the channel mk of the beam, located within the vessel
lumen. The signal produced by the transducer in channel mk is

S mk ¦ n t  > A1 cos Z0 t   A2 cos Z0 t  n ˜ 'D @

:m k (9)
n

A1 cos (Z0 t) echo from stationary objects, e.g., the vessel wall within
the channel volume m,
A2 cos (Z0 t + n·'D) Doppler-shifted signals,

i.e., the signal consists of stationary, non-shifted contributions as well as of Doppler-

shifted parts. In practical applications, it turns out that for the amplitudes A1 >> A2
usually holds so that signal/noise considerations are particularly important in Doppler
methods.
In order to obtain the flow velocity “hidden” in S mk , the signal first has to be
demodulated by multiplication with cos (Z0 t). Upon application of the theorem
2 cos J cos G = cos (J + G) + cos (J - G)
and low-pass filtering (contributions with the frequency 2 Z0 t can easily be removed)
one obtains
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 263

ª § n v ·º
S ' mk ¦ :m
n t  « A1  A2 cos ¨¨ 2
k ˜ Z0 cos M¸¸» (10)
n «¬ © f PR c ¹»¼

n
The step function can be Time

f PR t

approximated by the time t and

n/f
v PR

the expression Z D 2 Z0 cos M

c
corresponds to the usual Doppler
formula. The (mostly strong) 1/f PR

stationary echo A1 has to be

eliminated by high-pass filtering. Time

This is a nontrivial task because, as 1 2 3 4 5 6 Number of pulses

can be seen from the Doppler formula,

the Doppler shift goes to zero with the flow velocity. If a filter is used that does not
exhibit an extremely steep fall-off above zero, important contributions to the flow
signal are eliminated along with the stationary echo. In turn, radial vessel wall motions
which may reach up to 1/10 of the average flow velocity under pulsatile flow
conditions (arteries) and produce a strong echo may adversely contaminate the desired
signal from the blood and may have to be suppressed by choosing appropriate filter
characteristics.
An important consequence is associated with the fact that the Signal Sm is sampled
with the sample frequency 1/fPR according to the sample function :. Assuming, say,
Z0 = 4 MHz , v = 1 m/sec and cos M = 0.5 , one finds ZD = 2.7 kHz. If the pulse
repetition frequency is 3 kHz in order to reach a desired distance, the Nyquist limit
(sampling theorem) is already close. Higher flow velocities which are often reached in
arteries will induce aliasing. Although there are several methods to extend the Nyquist
limit, the problem remains. It can be added that in continuous-wave Doppler
instruments (not described here) this problem is absent. If the channels are sufficiently
small and the blood vessel sufficiently large, finally, the pulsed Doppler method allows
flow profiles to be determined.

4. Duplex Scanner, Color Scanner

A combination of echo-imaging and Doppler analysis is of interest in many medical

applications, in particular in cardiology. In duplex scanners both modalities are
integrated independently while in color scanners the blood flow information obtained
from the Doppler signals are overlaid in color over a black and white echo image.
These modalities are especially attractive because they operate in real time. Since the
imaging process involves a continuous scanning of the entire region to be imaged the
area covered by a blood vessel or a heart chamber from where the Doppler information
is desired is “seen” only during short fractions of the scanning sequence (Figure 13).
As has been demonstrated above, the phase shift between consecutive pulses is needed
in order to evaluate the Doppler signal quantitatively, however. In general, therefore,
duplex or color scanners exhibit qualitative flow information only, based on limited
measurements. (The imaging frame rate, typically 25 Hz, is too small, by orders of
264 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

magnitude, to allow for a Doppler evaluation.) More sophisticated scanning schemes

have nevertheless been devised (the imaging beam is e.g. delayed at the area of interest
for a more precise Doppler measurement) which allow for more detailed analyses.
Figures 14 and 15 exhibit typical examples.

5. Further Developments

It has been mentioned previously that nonlinear effects associated with the propagation
and scattering of ultrasound involve „mode conversion“, i.e. the generation of higher
harmonics. This effect can be used mainly to reduce the influence of speckle noise (2nd
harmonic imaging or tissue harmonic imaging).
Intravasal ultrasound imaging using circular transducers emitting beams in a radial
direction have been developed. This method allows us to document, e.g., pathologic
vessel wall conditions such as intimal thickening, fatty streaks or stenoses.
Catheterization is of course necessary. Full 3D noninvasive flow imaging is presently
only possible with the aid of expensive and partially non-real time MR methods.
Two dimensional phased arrays allow the real-time imaging in volumes as, with
the aid of such transducers, a volume can be scanned without moving the transducer.
Figure 16 shows an application in cardiology.

Body surface (skin) Transducer

Heart chamber
(schematically) Single beams

Figure 13. Principle of duplex scanning (schematically). The area of interest for a Doppler flow
measurement (heart chamber) is covered only by three beams during each imaging cycle. Accordingly, the
Doppler signal has to be obtained from two phase measurements only unless a more sophisticated scanning
scheme is applied.
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 265

Figure 14. Imaging of the heart with overlaid Doppler information in color

Figure 15. Imaging of a carotic artery bifurcation with stenosis. A recirculation zone behind the stenosis
manifests itself.
266 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

Figure 16. Three-dimensional real-time imaging of the heart valves by utilizing a two dimensional phased
array transducer

Further Reading

[1] P.N.T. Wells, Ultrasound Imaging, Phys Med Biol 51 (2006), R83-R98
[2] M.F. Hamilton, D.T. Blackstock, Nonlinear Acoustics, Academic Press, New York, 1998
[3] F.A. Duck, Nonlinear Acoustics in Diagnostic Ultrasound, Ultrasound Med Biol, 28 (2002), 1-18
[4] A. Kurjak, S. Kupesic, Clinical Application of 3D Sonography, Parthenon Publ., New York, 2000

Appendix 1: Wave Propagation

We consider linear wave propagation in an ideal fluid (simplified acoustic model)
which can serve as a sufficiently accurate approximation for most conditions prevailing
in medical ultrasound modalities. The two physical quantities of importance within the
framework of this approximation are the density of the material where the sound
& & & *
propagates Ur ,t  and the velocity field therein v r ,t  as function of space r  and
time t  . The two equations that are needed to describe this model are the
& & wU wp
x continuity equation ’ ˜ U v   U p and the (1)
wt wt
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 267

& &
wv
x momentum equation U ’ p (2)
wt
dU
p thereby denotes the (hydrostatic) pressure and U p the compressibility of the
dp
fluid. The second equation represents a linear approximation of the Navier-Stokes
equation for an ideal (frictionless) fluid.
&
Upon introducing a scalar velocity potential < r ,t  (irrotational flow field, i.e.,
& & & &
no eddies and no turbulence) such that the velocity field v r ,t   ’ \ r ,t  one
w<
finds p U , furthermore the wave equation
wt

w2 < 1
'< Up with the wave speed c2 (3)
2 Up
wt
&
In this derivation, a term containing ’ p has been neglected which is in agreement
with the acoustic approximation.
We now introduce a Cartesian coordinate system such that the velocity field
& & & &
v r ,t  can be represented as v  
v x , v y , v z (the independent variables r and t
are thereby omitted for brevity). Forward and backward running plane waves
propagating in direction z are solutions of equation (3)

vx vy 0
vz v0 e i k z r Z t  (4)
v0 i k z r Z t 
<  e
ik

with the frequency Z and the wave number k = 2S/O (wavelength O).
In analogy with electric systems, the acoustic impedance of a medium is defined as
the ratio between the pressure (analogous to electric tension) and the flow (analogous
to electric current)
p
I Uc (5)
v
At locations where the impedance changes, scattering, reflexion and refraction set
in as can be found from general solutions of the above equations. If the impedance
change occurs over a well-defined surface (e.g., a blood vessel wall) mirror effects and
coherent scattering are also observed. The latter leads to interference phenomena which
are associated with speckle formation. Further effects (not included in the foregoing
equations) are damping and nonlinear contributions.
268 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

Appendix 2: Huygen’s Principle

We consider an arbitrarily shaped surface of limited extent which is set into motion,
e.g., by the piezoelectric effect. We intend to calculate the wave field in the adjoining
space. The boundary of this space contains the vibrating surface and is closed by
extending this surface to enclose a virtual volume. Mathematically, the problem
consists of the wave equation according to Appendix 1, to be solved for the virtual
space whereby the motion of the boundary (enclosing surface) is prescribed.
Green’s formula is useful to approach this problem

§ wu ww· &
³  u ˜ 'w  w ˜ 'u  dV ³ ¨¨ w w n&  u & ¸ dB
w n ¸¹
(1)
V B©

u, w denote arbitrary scalar

differentiable functions,
V an arbitrary volume with B n
closed boundary B dB
&
n the outer normal
&
( dB has the direction of the V
w
outer normal and & denotes the
wn
derivative in the direction of the
outer normal.)
With the aid of this formula, a volume integral is transferred into an integral over
its (closed) surface. It is applied here such that u is identified with the solution of the
wave equation to be determined, \ , and for w a special function (Green’s function)
is inserted, namely a spherical wave (distance from the center r , wave number k ,
frequency Z , time t )
1  i k r  Z t 
w e (2)
r
(The choice of the Green’s function is crucial and problem-dependent; for a related
theory see Courant and Hilbert, Methods of Mathematical Physics). Since \ and w
are supposed to satisfy the same wave equation (harmonic waves with frequency Z
and wave number k , the left side of equation (1) vanishes.
After this specialization, equation (1) reads

³
1 i k r w \ &
e & dB  ³<
> &
@
w 1 r  e  i k r &
dB 0 (3)
B
r wn B
wn

(The term ei Z t is omitted for brevity and only included at the end again.) In order to
determine the function \ in point P (Figure 1), we proceed as follows: For simplicity
(without loss of generality), we assume that the vibrating surface (transducer) is plane,
bounded, part of the boundary B and located in the y – z plane according to Figure 1.
The virtual space with boundary B is then extended to include the entire right half
space.
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 269

z
B P

J
r

y x
transducer

Figure 1. Derivation of Huygens’ principle. The closed surface B includes the vibrating surface (transducer)
and is closed such as to include the entire right half space (the broken line is extended to infinity.

Next, the point P is excluded from the volume by a small sphere (radius J ). The
surface B now includes also the surface of this sphere. Later, we let J o 0. (That this
procedure leads to the desired result, is in essence due to the particular properties of the
Green’s function.) We can assume (since we are seeking a physically realistic solution)
that \ is finite and almost constant (in particular, for J o 0 ) within the small sphere
w< & ªw < º 2
such that the term & d B equals « » J cos M dM d-
wn ¬ w r ¼J
( J 2 cos M dM d- is the surface element on the sphere with polar angles M and - ).
Upon evaluation of equation (3) for the sphere, we obtain, for J o 0,

1 i k J ªw < º 2 ª w § 1  i k r ·º
³ e « » J cos M dM d-  < J « ¨ e ¸» 4 S J 0  4 S <P (4)
B
J ¬ w r ¼J ¬w r © r ¹¼ J

Accordingly,

4 S <P ³
1 i k r w\ &
e & dB  ³ < &
>
w 1 r  e  i k r &
dB
@ (5)
B
r w n B
w n

whereby the integral now extends over the outer boundary only.
w< &
& vn& denotes the component of the velocity to be prescribed along the surface
wn
B which is perpendicular to the surface. Since only the surface of the transducer is in
vibratory motion, edge effects and components of the velocity which are directed along
w<
the surface are neglected and < is assumed to go to zero at infinity, & is different
wn
270 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

from zero only on the transducer surface. The integral (5) therefore extends over the
transducer surface only,

4 S <P ³
1 i k r w\ &
e & dB  ³<
> &
@
w 1 r  e  i k r &
dB (6)
Tr
r wn Tr
wn

The integral (6) can further be simplified by making the special assumption that
w< &
the gradient & vn& is symmetric with respect to the surface, i.e., both sides of
wn
the transducer (the surface facing the + x direction and the one oriented towards - x ,
Figure 1) vibrate in phase but in opposite directions. The same procedure can now be
applied to the negative half space except for the small sphere which is absent since the
point P of interest is on the right side. We therefore obtain a similar expression (6) for
the left half space, but the left hand side is zero,

0 ³
1 i k r w\ &
e & dB  ³<
> &
@
w 1 r  e  i k r &
dB (7)
Tr
r wn Tr
wn

Again, the integral extends only over the transducer surface. However, comparing
the two sides of the transducer, we observe that
w w & &
& o  & and d B o d B
wn wn
w< w<
while we assume that & o & since both sides exhibit the same vibration
wn wn
component. Accordingly, the relative sign in equation (7) is different from the one in
equation (6). Both integrals are therefore equal, and we obtain the usual formulation of
Huygens’ principle,
1 1  i ( k r  Zt ) w \ &
<P ³ e & dB (8)
2S B r wn

The wave field can be represented as a superposition of spherical waves emanating

from the transducer surface. From a numerical point of view, this formulation is far
more efficient for calculating sound fields that the integration of the wave equation. An
application is shown in Figure 2.
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 271

Transducer

Transducer
3 cm

15 cm 1 cm

Figure 2. Intensity distribution of a circular transducer measured (top) with the aid of a hydrophone and
calculated using Huygens’ principle (bottom).

Appendix 3: Beam Formation

Phased array transducers are used to produce ultrasound beams according to a pattern
in space and time that is shaped for a particular application. To this end, amplitude and
phase of each individual transducer element have to be chosen appropriately. In the
following, a strategy is derived according to which these characteristics can be
determined when a particular sound field is prescribed.
We confine ourselves to a one-dimensional array (Figure 1) for simplicity and
apply Huygens’ principle as shown in Appendix 2
1 1  i ( k r  Zt ) w \ &
<P ³ e & dB (1)
2S B r wn
The array consists of elements with width b and infinitesimal height dK. (The latter
is assumed for ease of calculation. This corresponds to an unrealistic amount of
infinitely many transducer elements. The question with respect to the number of
elements necessary to approximate Huygens’ principle with sufficient accuracy is
addressed later.)
272 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement

z
z’
P (x0 , z0 )
r
dK
K

2a x

b
Linear Phased Array (length 2a, width b)

Figure 1. Linear phased array. The elements (width b and height dK) are aligned along the coordinate z.

In the problem to be solved here, <P is prescribed while in turn

ªw\ º
« » v0 K e i MK (2)
¬ w x ¼x 0

is to be determined. Thereby, v0(K) denotes the amplitude and exp [iM(K)] the phase
function. Outside z = r a we have \ = 0 , and within z = r a there is only the
derivative with respect to x different from zero. K denotes a coordinate along the array
in direction z.
The field \P in point P (x0 , z0 ) (Figure 1) is therefore given by
a
b ei Z t § i k r
i MK ¨ e
·
< x0 , z0 , t  ³ 0
v K  e ¸ dK (3)
2S ¨ r ¸
a © ¹

with r 
x02  z 0  K 2 
Upon expansion of the square root  K, z 0  r  one arrives at Fresnel’s
approximation,
 P.F. Niederer / Ultrasound Imaging and Doppler Flow Velocity Measurement 273

ª § k z2 · º
b ˜ exp «  i ¨ k x0  0  Z t ¸ »
«¬ ¨© 2 x0 ¸
¹ »¼
<  x 0 , z 0 ,t  x
2 S x0
§ ª 2º · (4)
¨ i k «z K  K » ¸
a ¨ « 0 ¸
2 ¼»
³ v0 K e i MK exp ¨ ¬ ¸ dK
a
¨ x0 ¸
¨ ¸
¨ ¸
© ¹
In addition, in the far-field, we have x >> K , such that a further step can be performed
to arrive finally at the Fraunhofer approximation, viz.,
i k z0K
b ˜ e  i k x0  Z t 
a
< x0 , z0 ,t  ˜ ³ v0 K e i MK e x0
dK (5)
2 S x0 a

Accordingly, the farfield and the distribution of the amplitude and phase on the
transducer are connected by a Fourier-transform, i.e., the excitation v0 Ke i MK to
be determined can be obtained from a Fourier-transform of the desired field.
In a practical application, the integral in equation (5) is replaced by a sum
according to the size, shape and number of individual transducer elements making up
the phased array. The smaller the elements, the more elements can be accommodated
within a given transducer surface and the better the integral can be approximated, yet,
mechanical and electrical impedance problems along with manufacturing procedures
prevent the implementation of too small elements. In order to avoid crosstalk,
furthermore, the individual elements have to be separated by a certain distance which
has to be determined empirically. It can be shown (not worked out here, but easily
understandable qualitatively) that the distance between adjacent transducer elements
must not be larger than half the wavelength of the desired ultrasound wave (Nyquist
limit). The larger the transducer elements (however observing the Nyquist limit), the
less elements can be accommodated; it turns out that the undesired side lobe activity
increases at the same time. Manufacturing ultrasound transducers is a great specialty,
scientific considerations and practical manufacturing aspects are both important.
274 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-274

V.2. X-Ray-Based Medical Imaging:

X-Ray Projection Technique
Image Subtraction Method
Direct Digital X-Ray Imaging
Computed Tomography (CT)
Peter F. NIEDERER
Institute for Biomedical Engineering
ETH Zurich and University of Zurich
Gloriastrasse 35
CH-8092 Zurich, Switzerland

Abstract. X-ray projection imaging, introduced in the early 20th century, was for a
long time the only and still is a major routine diagnostic procedure that can be
performed on the intact human body. With the advent of Computed Tomography
in 1973, for the first time, a true 3D imaging of anatomical structures became
feasible. Besides, digital recording and processing allowed further developments
such as subtraction angiography in real-time or direct x-ray imaging without wet
photographic methods.

Keywords. X-ray, Computed Tomography, digital imaging

Introduction

In 1896 W.C. Röntgen announced the discovery of the potential of an at that time yet
unknown type of radiation to penetrate materials and to cause opacification of
photographic plates (after development). It turned out that various biological tissues
attenuated this radiation differently such that the generation of projections of the
human body onto a photographic plate was possible (Figure 1). As the physical nature
of the radiation was unknown, it was said to consist of „X-rays“ – a designation which
is still in use today.
A further milestone was the introduction of computed x-ray tomography (CT) by
G.N. Hounsfield in 1973 which for the first time allowed to produce cross sections of
the human body for diagnostic purposes in vivo and noninvasively apart from the
application of a radiation dose (Figure 2).
The x-rays used for diagnostic purposes in medicine are electromagnetic waves
with a photon energy between 30 - 200 KeV typically. “X-ray” was until a few years
ago the most often applied medical imaging technology. Since then, methods on the
basis of ultrasound (in contrast to x-ray non-ionizing) have become more often used. In
addition, MRI (Magnetic Resonance Imaging) is of rapidly increasing significance in
medical imaging, in particular as with the combination MRI/MRS (MR Imaging and
 P.F. Niederer / X-Ray-Based Medical Imaging 275

Spectroscopy) also functional examinations can be made. The combination of x-ray CT

or MRI with PET (Positron Emission Tomography), in turn, has in recent years
fostered significant and novel clinical applications for functional investigations.

Figure 1. X-ray of the hand of W.C. Röntgen’s wife

The radiation dose which is delivered to the patient is always of concern; it should
be as small as possible. An enormous progress has been achieved through the years
such that today a “simple” x-ray image can be produced with the aid of a radiation dose
which can be considered harmless (Figure 3).

Figure 2. First CT scanner (1973) and first image of a brain tumor.

276 P.F. Niederer / X-Ray-Based Medical Imaging

energy dose
(mGy)

8 intensifying
universal intesifying screen
screen (CaWO4)
storage screen
(luminescence)
6
image intensifyer
specialized intesifying + TV imaging
screen (CaWO4)
4 rare earth intensifying screen
Ti2Y2O2S Gd2O2S

2 image intensifyer
storage screen

analog digital imaging

1960 1970 1980 1990
year in which technology was introduced

Figure 3. Typical radiation dose delivered for a x-ray projection image

1. Generation of X-Rays

X-rays for diagnostic purposes in clinical medicine are usually generated with the aid
of a x-ray tube (in order to produce the much higher-energy therapeutic radiation for
tumor treatment in the MeV-range, linear accelerators are mostly used). In such a tube
whose interior is highly evacuated (the pressure is typically 10-6 Torr ~ 10-4 Pa),
electrons are „evaporated“ from a wolfram cathode by heating (the melting temperature
of wolfram is 3400° C) and exposed to an anode voltage, V, of some 30’000 to 200’000
Volt for acceleration towards the anode. The electrons thereby obtain a kinetic energy
of 30 to 200 keV (electron charge, e, times potential difference, V). The geometric
arrangement of the cathode along with the spatial characteristics of the accelerating
field are designed such that the electron beam hits the target (anode) in a carefully
outlined distribution (focal spot, see later).
When the accelerated electrons are being absorbed by the target material, x-ray is
generated due to two effects, viz.,
• bremsstrahlung
• characteristic radiation
Besides that, most of the kinetic energy of the electrons is converted into heat (see
below).
 P.F. Niederer / X-Ray-Based Medical Imaging 277

Bremsstrahlung is emitted, when fast electrons interact with the strong (positive)
electric field of atomic nuclei. The electrons are deflected in this field and radiate
electromagnetic waves in the form of photons. The electrons thereby loose kinetic
energy, i.e., are braked down (therefore the German expression „bremsstrahlung“). The
spectrum of this radiation is continuous because the effect depends essentially on the
distance at which the electron passes the nucleus and each distance has the same
probability. The spectrum therefore extends from (theoretically) zero up to the maximal
possible photon energy, Emax , which occurs when the electron looses its entire kinetic
energy in one step.
The spectrum as function of energy associated with a thin target is constant,
because most electrons, while traversing the thin layer, are involved in only one
scattering process, thereby loosing part of their energy such that each portion of energy
from zero to the maximum has the same probability. In a thick target, in turn, the
spectrum is linearly decreasing with E → Emax , because except for a few electrons
which are backscattered at the surface layer all electrons always loose their entire
kinetic energy in various single steps of different intensity (there are more small steps
possible than large ones). According to
Emax = e V = h νmax (1)
(h: Planck’s constant, ν: photon frequency, e: electron charge), the minimal wavelength,
λmin , which is associated with the maximal photon energy Emax , is (speed of light
c = λν )
λmin = h c / e V (2)
with h = 6.63⋅10-34 W sec2
c = 3⋅108 m/sec
V = 100 kV
e = 1.6⋅10-19 Clb
one obtains λmin = 12.4⋅10-2 Å = 1.24 nm . Typical spectra are shown in Figure 4.
The decrease at low energies is on the one hand due to self-absorption in the target
material, and on the other to prefiltering (low energy photons are removed purposely
because they do not contribute to image formation but just increase the radiation dose
imparted to the patient, see below and 2. Absorption of X-Rays).
The electrons hitting and penetrating the target not only interact with the nuclei of
the target material, but also with the electrons thereof. Electrons may thereby be ejected
from their orbit, and radiation quanta are emitted during subsequent recombination.
Since the orbits have defined energy levels, the quanta have energies corresponding to
the various orbits (K, L, ..., Figure 4) and the spectrum consists accordingly of lines
(characteristic x-ray).
Most of the kinetic energy is however converted into heat (∼99%) in the target
material (electromagnetic waves with long wavelengths, largely due to interactions
with free and weakly bound electrons of the target material). As mentioned above, the
low-energy part of the spectrum is usually removed with the aid of prefilters (Al or Cu
sheet metal plates) in order to reduce the radiation dose.
The radiation beam is strictly speaking not continuous as it is composed of a large
number of incoherent electromagnetic wave packets (photons; the scattering processes
in the target material occur independently from one another). The number of photons is
278 P.F. Niederer / X-Ray-Based Medical Imaging

so large, however, that quantum noise is not of importance (exception: image

subtraction, see later).

a K-lines of
wolfram

without filter characteristic

x-ray
photon flux

an
od
e
vo
l ta
ge
b

photon energy

Figure 4. Typical spectrum of an x-ray tube (photon flux as function of energy)

a: Bremsstrahlung (theoretical, without prefilter)
b: Spectrum at different anode voltages; wolfram target,
1 mm Al prefilter (bremsstrahlung + superimposed characteristic x-ray)

There is a great variety of x-ray tubes available for different medical applications;
yet, the basic physical effects are always the same. Material problems are prominent
(hot cathode, extreme local heating of the anode). At a typical anode current ( =
electron current) of I = 100 mA and a voltage U = 100 kV the power deposited in
the anode is
E = I⋅ U = 10 kW (3)
A projection image with a good resolution is only obtained if the focal spot (Figure
5) is small (another possibility is to locate the x-ray tube as far away from the object as
possible; this is however limited due to practical reasons). X-rays propagate along
straight paths, there exist no x-ray „lenses“ for practical medical imaging purposes 1.
The electrons are focused on the focal spot by the suitably shaped accelerating field.
Since all the heat is produced in the focal spot, a small spot is difficult to achieve.
Therefore, the anode is usually inclined with respect to the electron beam such that the
focal spot can be extended (Figure 6) and the heat production is distributed accordingly.
In the direction of the projection, the focal area still has a localized aspect (for special
1
New developments in x-ray technology include the usage of diffraction (phase contrast imaging,
much as in optical microscopy); likewise, Bragg reflection can be used for special purposes.
 P.F. Niederer / X-Ray-Based Medical Imaging 279

purposes, also tubes with a line focus are used). In addition, in high power tubes the
anode rotates (5000 min-1 typ.) such that the focal spot is not static.

focal spot

focussing and
accelerating field
electron emitter
(heated wolfram wire)

Figure 5. Focusing of electrons, originating from the cathode (-) onto the focal spot of the anode (+)

Figure 6. Schematic view of static (top) and rotating (bottom) anode tub
280 P.F. Niederer / X-Ray-Based Medical Imaging

The heat is disposed of by diffusion through the anode as well as by radiation. The
latter obeys the Stefan-Boltzmann law according to which a black body radiates the
energy per unit area, W, as
W = σ T4 (4)
(σ = Stefan-Boltzmann constant, 5.7⋅10-12 W/cm2K4, T = temperature)
At T = 2500° K and an emission efficiency of 0.7 (graphite; the emission
efficiency of a black body is 1) the radiated power per unit area is W ∼ 200 W/cm2.
Even under normal operating conditions, the anode may be extremely loaded thermally
(Fig.ure 7).

Figure 7. Red hot rotating anode under typical working conditions

(courtesy: Comet AG, Bern)

The further away the tube is positioned from the object, the sharper the images
become since the focal spot has always a certain size. In contrast, the recording
medium, e.g., a x-ray film cassette should be located as closely as possible behind the
object.
For certain applications, in particular computed tomography (see later), the high
voltage used in the tube for the acceleration of the electrons has to be extremely stable
in order that the x-ray spectrum is constant in time. For the same reason, the stability of
the electron beam hitting the target is an important aspect.
 P.F. Niederer / X-Ray-Based Medical Imaging 281

2. Absorption of X-Rays

In the energy range which is used for diagnostic medical imaging (ca. 30 - 200 KeV)
two absorption effects occur when the photons interact with material, viz.,
• photo effect (Figure 8, a)
• Compton effect (Figure 8, b).

a b
Figure 8. Photo effect ( a ), Compton effect ( b )

The photo effect is a process where a photon is absorbed by an electron which is

ejected from its shell to a higher energy level within the atom or completely removed
from the atom (ionisation). The photon energy must at least correspond to the energy
difference between the two electron states in question, i.e., to the difference between
the initial and final state of the excitation process (multiphoton processes are not
observed with the photon fluxes used in medical imaging). When the absorption
spectrum of a material which consists only of one type of atoms is considered, “edges”
are therefore observed which correspond to the energy levels associated with the
various electron shells (K-, L-, M, etc.). As the K-electrons exhibit the highest binding
energy, the K-edge is encountered first on the energy scale when approached from
above (see Fig. 9, a). This property is being used in the energy subtraction method (see
later).
With the Compton effect (Figure 9, b), occurring at higher energies, a photon is
likewise absorbed by an electron which is thereby excited, but there is excess energy
which is emitted at the same time in the form of a lower energy secondary photon.
These secondary photons represent scattered radiation since their direction does in
general not coincide with the one of the original photon. They have a blurring effect
and may lead to such a high noise level in an image that countermeasures have to be
taken (collimators, anti scatter grids, see below).
282 P.F. Niederer / X-Ray-Based Medical Imaging

cm 2 /g r
cm 2 /g r
10

100
w a te r
1
10

0 .1
w ater
0 .1

0 .0 1
lead 0 .0 1

0 .00 1

10 keV 1 00 k e V 1 M e V 10 M eV 10 keV 1 0 0 k eV 1 M eV 10 M eV 100 M eV

p h o to n en erg y p h o to n en erg y

Figure 9. a: Mass absorption coefficient (photoeffect alone) for lead and water.
b: Total mass absorption coefficient μ’ = μ /ρ (the definition of μ is given below, Beer-Lambert’s law) as
function of the photon energy, E . It is composed of the contribution of the photoeffect ( τ / ρ ), of the
Comptoneffect ( σ / ρ ) as well as the one of pair production ( κ / ρ ). The latter effect is not considered here
because for the production of an electron-positron pair an energy of at least 2 m0 c2 = 1.022 MeV is
necessary (m0 = electron mass, 9.11⋅10-31 kg). Since this energy is much higher than what is used in
medical x-ray imaging, this effect is not of importance here (the additional curve, η / ρ , in the graph, is
related to the energy conversion).
In total, the relation holds μ’ = μ /ρ = τ /ρ + σ /ρ + κ /ρ

Since the Compton effect requires on the average a higher amount of photon
energy, at lower energies the photoeffect prevails while at higher energies the
Comptoneffect sets in.
The cross section of an atom with respect to a certain interaction process is defined
as the (theoretical) area σ (in cm2) perpendicular to the incoming radiation which –
much like the target in a shooting range – is available for the initiation of the process in
question. If the photon hits this area, the effect is triggered, if not, no interaction takes
place. In order to obtain the specific absorption capacity (per unit of mass) of a material,
the cross section has to be multiplied by the number of atoms per unit of mass.
Accordingly, the mass absorption coefficient is defined as the total cross section per
gram, i.e.,
μ’ = σ NA / A ( NA Avogadro’s number, A atomic weight) (5)
Since NA indicates the number of atoms per A grams of a substance (6.023⋅1023),
the quotient NA / A corresponds to the number of atoms per gram, i.e., μ’ has the
dimension cm2/g.
With regard to medical imaging, the dependence of the mass absorption coefficient
on the atomic number, Z (= number of protons in the nucleus), is decisive. Because of
this dependence, the various biological tissues in the body exhibit different x-ray
absorption characteristics such that their appearance in the projection image (grey
level) varies according to their chemical composition. In case of the photo effect, σ is
 P.F. Niederer / X-Ray-Based Medical Imaging 283

about proportional to Z5 , while for the Compton effect it is about proportional to Z .

For both effects, however, the cross section decreases with increasing energy, i.e., the
penetrating power of the x-rays increases (Figure 9). According to the object to be
imaged, therefore, the voltage of the x-ray tube is set.
Because of the stochastic nature of the occurrence of the various effects, the
attenuation of an x-ray beam when penetrating a thin sheet (thickness dx) is
proportional to dx ,
I (x + dx) - I (x) = - μ dx
which yields
I(x) = I0 exp (-μ x) (6) I(x) I(x + dx)
(Beer-Lambert’s law).
The linear attenuation coefficient,
μ, is determined from the
mass absorption coefficient
introduced above according
to μ = μ’ ρ (ρ = density), dx
because the entire mass of the
material in the layer dx has I(x) : Number of particles
to be taken into account. In per area and sec
case of composite materials or
mixtures (which is always given
in biological tissues) the total mass attenuation coefficient μ’tot can be determined
from a weighted average (according to the density) from the individual coefficients of
the various components.
As mentioned above, biological materials are always composed for various
constituents in different concentration such that they can be distinguished according to
their x-ray absorption properties (Figure 10). These effects are systematic and
reproducible.

bone
muscle
fat

Figure 10. Dependence of the linear attenuation coefficient on the energy for various tissues
284 P.F. Niederer / X-Ray-Based Medical Imaging

3. X-Ray Projection Imaging

3.1. Resolution of Imaging Procedures

From a technical point of view the resolution provided by an entire x-ray imaging chain
(x-ray tube → image presentation) has to be considered. One has thereby to
distinguish between
a) spatial
b) contrast
c) temporal resolution.
Only linear mappings are considered here.
a) Spatial Resolution: The spatial resolution of an imaging chain is often given in terms
of the normalized intensity amplitude as a function of line pairs/mm (lp/mm), denoted
as Modular Transfer Function (MTF). The practical procedure to measure the MTF for
x-ray systems consists of imaging lead grids of variable spacing (Figure 11). The closer
the lines are together, the smaller is the amplitude of the image and the average grey
level increases until the lines cannot be discerned any more. The maximal amplitude
(arbitrary unit 1) is defined as the maximal intensity step that can be represented.
Typical MTF curves are shown in Figure 12.
Instead of the MTF, the line spread or point spread function (LSF, or PSF),
respectively, are sometimes given. These functions indicate onto what image a line or
point is mapped, in particular, how a point is blurred (Figure 13).

Figure 11. Imaging of a lead grid. With increasing spatial frequency (lp/mm)
the amplitude of the image (grey level) decreases.
 P.F. Niederer / X-Ray-Based Medical Imaging 285

1: 25 cm image intensifyer,
CsJ screen

2: 25 cm image intensifyer,
15 cm setting
1
2 3: 17 cm image intensifyer,
MTF in
3 CsJ screen
lp/mm 4
5 4: 17 cm image intensifyer,
ZnS or CdS screen

5: Film- intensifying screen

combination

Figure 12. MTF of various x-ray systems (system definition see later)

imaging

object image

Figure 13. Principle of the point spread function (PSF)

In the Appendix 1, a mathematical derivation of the MTF is given. It can be shown

that the MTF and LSF or PSF (in two dimensions) are related by a Fourier transform.
Since every object (one or two dimensional) can be thought to be composed of lines
(one-dimensional image) or points (two-dimensional image), in case of a linear
mapping it is sufficient to know the LSF or PSF, respectively, to determine images.
This procedure is often used in medical imaging.
b) Contrast Resolution: In order to assess the contrast resolution of x-ray projections,
the contrast in the radiation behind an object has to be considered. As an example, we
determine the intensity of an x-ray beam after passing through d = 20 cm of typical
biological tissue:
I = I0 exp (-μ d) , or I / I0 = exp (-μ d) (7)
286 P.F. Niederer / X-Ray-Based Medical Imaging

With μ = 0.35 cm-1 (average for biological tissue devoid of bone and air) we
obtain a contrast of exp (-7), i.e., about 1/1000. In the presence of bone and air (lung,
intestine) the contrast is even considerably higher. Neither with photographic film nor
with conventional TV such a contrast range can be recorded (film, as shown below, has
a linear gray scale range of about 2.5 - 3, corresponding to a contrast range of about
1/300 to 1/1000, while conventional TV has less). Depending on whether, e.g., the lung,
other soft tissues or bones are to be imaged, the radiologist has to choose the
parameters anode voltage, anode current, exposure time in order to keep the desired
contrast within the available range.
Digital imaging techniques allow to record images with an extended contrast range
(12 – 16 bit resolution). Typical TV monitors, however, visualize only a limited grey
level range of about 6 bit. Since images are stored in digital form, the visible contrast
range can be selected and altered up and down in real time („windowing“).
c) Temporal Resolution: The temporal resolution is of importance whenever dynamic
processes are to be investigated (phlebogram, ventriculogram, stomach-intestine
passage, etc.). Cassette changers (see later) reach a frequency of up to about 6
images/sec which is sufficient, e.g., for phlebograms. In order to resolve a heart cycle
adequately, up to 50 images/sec are necessary.

3.2. Photographic X-Ray Film

Diagnostic x-ray cause precipitation of silver bromide in a photographic emulsion such

that x-ray intensity is reflected as grey level after development. A sharp image is only
obtained, if the photographic layer is thin because of the mostly oblique incidence of x-
rays in a projection arrangement. Besides, there is always undesired scattered radiation.
Since the absorption in the photographic emulsion increases exponentially with the
thickness (Beer-Lambert’s law), in a typical x-ray film only about 2 % of the incident
radiation is absorbed. If not highest resolution is necessary (e.g., mammography or
dental applications) double-sided film is used in order to increase the thickness.
Accordingly, a typical photographic x-ray film consists of a middle polyester carrier
sheet (thickness 0.15 mm) to which on both sides a silver emulsion layer (thickness 7 –
20 μm) is attached by way of a thin (1μm) adhesive coating. An outer covering layer
on both sides of the film serves as protection..
The low efficiency (requiring a correspondingly high radiation dose delivered to
the patient) can be increased by adding intensifying screens on both sides of the film
which is then called a cassette. This increases the absorption but decreases the
resolution. The intensifying screen contains e.g. CaWO4 , a substance which exhibits
blue fluorescence under the incidence of x-ray which increases film exposure (other
fluorescent substances include Y2O2S, Gd2O2S, BaFCl). The efficiency can so be
increased up to a factor of about 5.
The opacity 1/T (inverse transmittance, dimensionless) of a (processed)
photographic sheet is defined as
1/T = ( I0 / I ) (8)
( I0 : incident, I : behind the sheet densitometrically measured intensity or brightness
of a normalized white light source, unit Lumen [Lm])
 P.F. Niederer / X-Ray-Based Medical Imaging 287

The density S [ = log (1/T) ] as a function of the exposure (characterizing the

quantity of light delivered to the emulsion) of photographic film follows the diagram
shown in Figure 14.

2
α γ = tan α

0
-3 -2 -1 0 log E

Figure 14. Typ. characteristic curve of photographic film [ S : log(opacity),

E : exposure (in meter-candle-sec) ]

The γ value of a film is a measure of the maximum change of film density for a
certain change of exposure. This corresponds to the part of the characteristic curve with
the steepest slope, i.e., at the point where the maximum derivative is found ( = tangent
of the inclination angle α). Typical photographic film has a linear range of about
Δ S = 2.5. The “steepness” γ is also used for other imaging modalities.
To eliminate scattered radiation, anti-scatter grids (Figure 15) are used. Such grids
are characterized by the distance and thickness of the lamellae (typ. 25 – 50
lamellae/cm) as well as the grid ratio (height of the lamellae/distance, typ. 5 - 15).
X-ray films are mostly used for single-shot static exposures. If slow processes are
to be examined (e.g., in case of a phlebogram), cassette changers are applied which
allow to take up to about 6 images/sec. The advantage of this procedure consists of the
high image quality which corresponds to the one of x-ray film. If higher frame rates are
required (e.g., for imaging of the heart), other procedures have to be applied, in
particular in order to avoid excessive radiation doses (see later, image intensifier).

source

object

scattered
radiation

anti scatter grid

lead
lamellae

film

Figure 15. Anti-scatter grid arrangement

288 P.F. Niederer / X-Ray-Based Medical Imaging

3.3. Direct Digital X-Ray → Image Conversion

Photographic film is an analog storage medium and preparation for computer analysis
and archiving including networking necessitates off-line digitization. In addition, film
development requires chemistry (cost and waste). A number of methods have therefore
been developed which allow for a direct digital image acquisition and which are mostly
in use today.
Digital Luminiscence Radiography (DLR), Storage Screen Radiography,
Computer Radiography (CR, Figure 16): On a screen made of BaFBr:Eu2+ (Europium-
doted barium halegonide) a latent image is created by irradiation (excited electron
states) which is converted off-line into visible fluorescence by laser scanning. One of
the advantages of this technology consists of its large dynamic range.
Selenium-based detectors: Prior to the exposure to x-ray, the surface of the
detector is charged electrically. The photons are converted into electrical signals in that
they create charges in the interior of the selenium layer which neutralize the surface
charges. Again, a latent image is formed which is scanned off-line.
Flat screen detectors on the basis of amorphous silicium: On a silicium wafer a
pixel matrix containing the necessary electrical components (transistors) is
implemented which allows for a direct digital conversion of the charges created by the
incident x-rays Because the absorption of x-ray by silicium is only weak, additional
layers containing heavy atoms have to be overlaid on the silicium screen.
Scintillator-fiberoptic-CCD chip arrangement (in the future also the much simpler
and less expensive CMOS technology will be used instead of CCD): For particular
applications, (e.g., μCT) a scintillator screen serves for spatially resolved photon
acquisition which is imaged on a CCD chip by fiber optic connection. The possibility
thereby exists for image enlargement by cutting the fiber optic bundle obliquely.

incident x-ray
exposure

laser beam
stimulated
light luminiscence

excited state
reset read-out
ground state
remaining excitation is
removed by exposure
to intensive light

Figure 16. Principle of digital luminescence radiography

 P.F. Niederer / X-Ray-Based Medical Imaging 289

3.4. Image Intensifier

In a x-ray image intensifyer (Figure 17) the radiation image is converted into an
equivalent electron image in the entrance screen. The conversion occurs in two steps:
First, light flashes are created by x-ray photons in thin CsJ needle crystals (scintillator)
which are about 0.5 - 1 cm long (good absorption of x-ray) and densely arranged
perpendicular to the screen in order to suppress scattered radiation as well as lateral
diffusion of light (veiling glare). These light flashes cause the creation of free electrons
in the photocathode which is connected to the scintillating screen. In addition, on the
entrance side of the scintillator a thin aluminum foil is located which reflects light back
onto the photocathode. The electrons whose lateral diffusion in the photocathode adds
to the veiling glare and has therefore to be suppressed, are accelerated by 25 kV
typically and directed towards the exit screen. Thereby, an electrostatic field acts as an
imaging electron-“optics” such that a bright image on the exit screen appears. The
intensifying effect is due to the kinetic energy which the electrons obtain during
acceleration. The exit screen is recorded digitally; in the past, video or cinéfilm was
used (up to 50 images per sec). Thanks to the good absorption properties of the
entrance screen, the efficiency of present intensifiers is about 40 - 50 %, i.e.,
substantially larger than film/intensifying screen combinations. The efficiency is given
by the conversion factor G in (cd/m2) / (mR/sec) whereby cd (candela) denotes the
light intensity dI/dΩ (light intensity or light flux per solid angle), mR (milli-Röntgen),
the radiation dose (old unit).

incident
radiation

observation
screen

Figure 17. Cross section through image intensifyer showing electron paths and field lines
290 P.F. Niederer / X-Ray-Based Medical Imaging

4. Image Subtraction Techniques, Digital Subtraction Angiography (DSA)

The lumina of blood vessels, stomach, intestine, ventricles of the heart or of the brain
are only faintly or not visible on a routine x-ray projection image. In order to obtain a
diagnostically useful visualization, the lumen in question has to be filled with an
appropriate contrast agent (e.g., bariumsulfate in case of stomach-intestine, or a
compound containing iodine for blood vessels or the heart which is well tolerated, i.e.,
which causes no immune reactions).
To image a section of an arterial vessel (e.g., if an aneurysm, a stenosis or a
dissection is suspected) or of a ventricle of the heart it is necessary to administer the
contrast agent at the very location under consideration (selective catheterization) in
order to obtain sufficient image contrast. This procedure is associated with a
considerable expense (catheter laboratory) since an arterial vessel has to be opened,
furthermore, there is always the danger of an afterbleeding. In contrast, a venous
infusion of a contrast agent is largely without problems. Although part of the contrast
agent arrives at the location to be imaged on the arterial side also if a venous
administration is made (e.g., through the vena cava), the resulting image contrast is
insufficient because of the dilution of the contrast agent after the lung passage
necessary to reach the arterial side and the subsequent distribution in the entire
vasculature of the body. In addition, not only the location of interest contains therefore
contrast agent, but all adjacent, overlying and underlying vessels including capillaries
and veins such that the image of the vessel section in question is embedded and
covered by other vessels.
With the aid of image subtraction techniques, however, an increase of contrast can
be reached such that at least in the case of not too small arterial vessels a sufficiently
good representation for diagnostic purposes can be obtained also with venous
administration of contrast agent. Since for this technology on-line digital image
treatment is applied, the method is called Digital Subtraction Angiography (DSA).
(Note: Image subtraction can also be made easily with photographic film in that a
negative can be obtained from contact exposure; for medical purposes, this procedure
has been in use since about 1935). With DSA, an image called “mask” is first made (a
digital projection image before the contrast agent is administered) and stored in the
computer. Then, the contrast agent is applied transvenously, mostly through the vena
cava. After about 20 sec, the exact time delay thereby depending on the location to be
imaged, the contrast agent appears in diluted form at the desired section on the arterial
side. Image sequences can now be acquired whereby the mask is subtracted in an on-
line fashion (video, Figure 18). Providing that the patient has not moved (among other,
breathhold is necessary) between the acquisition of the mask and the later images, only
the difference between the images, i.e., the shadow of the contrast agent should be
visible.
Overlay is still present, but above all, noise amplification occurs since images
characterized by small differences associated with noise are subtracted and
subsequently amplified. Since minimal exposure to radiation is desired, the discrete
nature of x-ray may become apparent (“quantum noise”). This can partially be
compensated by using more contrast agent and/or a higher radiation dose, both of
course undesired (on the one hand, the incidence of immune reactions increases with an
increasing amount of contrast agent, on the other, a higher radiation dose is undesired
anyway). In spite of these drawbacks, the advantage of avoiding an arterial
catheterization is by far dominating such that DSA is applied whenever possible.
 P.F. Niederer / X-Ray-Based Medical Imaging 291

kidney

Iliac artery

Ilio-femoral
bypass

femoral
artery

Figure 18. DSA image (real-time video subtraction) of a highly pathologic abdominal vessel situation

5. Computed Tomography (CT)

The „classical“ tomographic procedure in radiology was performed in such a way that
the x-ray source (tube) and the photographic film were shifted during the exposure
continuously in opposite directions. Only one plane is thereby imaged sharply,
structures which are located underneath or above this plane are smeared. The images
have therefore a low contrast, however. Computed tomography (derived from greek
τομειν, to cut), in turn, was introduced in 1973 by G. N. Hounsfield in England (Figure
2). A. M. Cormack in South Africa had in fact already earlier investigated methods
which allowed to reconstruct objects from a series of suitably chosen x-ray projections
(Hounsfield made his later work however independently from Cormack’s results).
Since at the time when Cormack made his investigations there were no reasonably
priced computer-assisted imaging methods available, he could not represent the results
of his calculations as true tomograms, i.e., images of cross sections, but just in the form
of numbers and curves; accordingly, his work had no practical consequences
(nevertheless, both researchers received the Nobel prize in 1979). What both
researchers did not know, however, was the fact that the mathematical problem of
292 P.F. Niederer / X-Ray-Based Medical Imaging

calculating a n-dimensional object from n-1 dimensional projections had been solved
much earlier by the Austrian mathematician Johann Radon (1887 – 1956) already in
1917. This was recognized only after CT had been introduced into clinical routine. Yet,
this was not of importance because Radon’s formal mathematical solution was found to
be unsuitable for practical applications for numerical reasons.
The method of CT consists of the acquisition of a sequence of x-ray projections of
a predetermined cross section of the body under different angles. In scanners of the first
generation, a single projection was taken by translating a collimated x-ray beam
(“pencil beam”) over the chosen cross section. This procedure required a synchronous
movement of the source (x-ray tube) and the detector. Later, the “fan” beam technology
was introduced whereby an entire projection is taken at the same time with the aid of a
detector array (consisting, e.g., of 512 detectors) without translation of the x-ray source.
Subsequent rotation of the x-ray source and detector array around 180° in equally
spaced angle intervals (e.g., 180°/512) further projections are taken (Figure 19). A total
set of (one dimensional) projections then allows to reconstruct the interior of the cross
section, i.e., the two-dimensional distribution of the attenuation coefficients μ (x,y) (x,
y are Cartesian coordinates in the cross section) The image is represented in the form
of a grey level distribution D(x,y). In present scanners, the fan beam – detector array is
rotated and advanced in a spiraling motion around the patient such that entire spatial
body sections can be imaged within a few seconds (Figure 20).

Quelle
x-ray source

Objekt
object

Projektion
projection

Detektor

detector

Figure 19. Principle of first generation CT scanner: A narrow x-ray beam (pencil beam) is generated by
collimators at the x-ray source as well as at the detector (the latter primarily suppresses scattered radiation).
One-dimensional projections are obtained by translating the unit (embedded in a gantry) over the object. By
successive rotation of the gantry after each translation by, e.g., π/512, a complete set of projections is
recorded.
 P.F. Niederer / X-Ray-Based Medical Imaging 293

For simplicity, we consider the situation of a first-generation translation-rotation

scanner (Figure 21). The beam intensity, Iϕ(r), which is recorded in a direction
perpendicular to a fixed angle ϕ as function of the linear coordinate, r , is obtained as
s
I ϕ (r ) = I 0 exp ⎡− ∫ 0 μ (s{r ,ϕ}) ds ⎤ (9)
⎢⎣ − s0 ⎥⎦

where I0 denotes the intensity of the unattenuated beam (before impinging on the
object) and s[r,ϕ] is the line defined by r and ϕ between source and detector in the
x - y plane. The integration is performed along the line s from -s0 to +s0 (distance
between source and detector).

base (a)

atlas (b)

b carotid
arteries (c)

c vertebral
d arteries (d)
c
d

Figure20. Spiral CT image of the base of the skull (after segmentation)

As projection Pϕ (r) the quantity

[
Pϕ (r ) = log I ϕ (r ) I 0 ] = −∫
s0
− s0
μ[s(r , ϕ) ds ] (10)

is defined. The task now consists of the determination of the linear attenuation
coefficients, μ(x,y), from a complete set of projections Pϕ (r).
294 P.F. Niederer / X-Ray-Based Medical Imaging

Detector [ Iϕ(r) ]
(+s0)
y
s
r

x
ϕ
pencil beam
object
μ(x,y)
collimator source [ I0 ]
(-s0)

Figure 21. Principle of 1st generation translation-rotation scanner

As mentioned above, Radon found an analytic solution of the problem, which is

unsuited for a numerical application, however. Among the various methods which have
been evaluated for the calculation of the grey scale image D(x,y) derived from the
attenuation coefficients μ(x,y), the method outlined in the following called convolution
- backprojection has proven to be particularly useful.
A simple backprojection without further pre- and postprocessing of the data is
performed by distributing the entire value of the integral Pϕ (r) uniformly along the
projection path, s (Figure 22). As an example, we consider the projection of a
cylindrical object which is scanned perpendicularly to its long axis. If it is assumed that
(i) the radius of the object → 0, (ii) the number of projections (angles) → ∞ , and (iii)
each one-dimensional projection is continuous, i.e., it consists of infinitely many points,
theoretically, the image of a point results (point spread function, PSF). The PSF is
found to be proportional to 1/r , since along the perimeter of every concentric circle
around the point under consideration the same amount of backprojection is
accumulated.
According to chapter 3.1. it suffices to know the PSF associated with the imaging
procedure in order to be able to determine the image (i.e., the reconstruction) of any
two dimensional object. The reconstruction problem would therefore be solved, if the
“corrected” PSF was known, more precisely, if the mathematical procedure to be
applied in order that a point appears as a point without artifacts was known.
Accordingly, the artifacts present in the simple backprojection procedure, in particular
the 1/r dependence, have to be compensated.
This is achieved by applying a prefilter (mathematical convolution) to the
projections. It can be shown (Appendix 2) that a convolution of the form
1 ∞
P' ϕ (r ) = ∫ P *ϕ (u ) e iur u du (11)
2 −∞
2π
 P.F. Niederer / X-Ray-Based Medical Imaging 295

yields the desired result. The function |u| thereby compensates the 1/r dependence as
well as the fact that the Fourier space associated with the scanned area is not uniformly
sampled (Appendix 2, this follows from the Fourier-slice theorem): High spatial
frequencies are sampled less densely than low frequencies.
Particular problems which have to be given careful attention in x-ray CT machine
design are related to:
Beam hardening: X-ray sources as used in clinical CT are not monochromatic (see
paragraph 1, Generation of X-Ray). Since x-ray absorption is energy-dependent, the
spectrum of the radiation changes continuously when traversing an object. This effect
is nonlinear and object dependent and can therefore not be corrected mathematically.
Careful, object-dependent calibration procedures are necessary for highest precision.
Partial volume effect: For practical computational reasons, the area of interest has to be
subdivided into (usually rectangular) volume elements. Oblique rays cover most such
elements only partially, however. For each direction and ray, the partial volumes
covered by the rays have to be determined and taken into account in the reconstruction
process.

object
Objekt

projections
Projektionen

Figure 22. Schematic representation of four projections (top) of a cylindrical, homogeneous object and
corresponding simple backprojection (bottom). The measured projection values are thereby distributed
homogeneously (“backprojected”) over the projection path. The resulting “reconstruction” is characterized by
star-like artifacts, moreover, the image of the cylindrical cross section extends over the entire scanned area. If
the radius of the cylinder is assumed to approach zero and the number of projections is increased to infinite,
the point spread function (PSF) is theoretically obtained.

Scatter: Secondary radiation due to Compton scattering occurs when an object is

irradiated. This secondary radiation appears as noise on the detector. Collimators can
be used to prevent most scattered rays from reaching the detector because Compton
scattering is omnidirectional in contrast to the primary radiation.
First generation scanners are no longer in use, a number of developments were
made in order to improve speed and resolution. In particular, various filters (instead of
u ) are being implemented depending on the application, e.g., according to
Ramachandran and Lakshminarayanan (see the literature listed in Further Reading).
Speed was increased, first, by using a fan-beam x-ray geometry, later, cone-beam
irradiation and 2D detectors were introduced in order to allow for scanning volumes.
With helical (sometimes also referred to as spiral) scanning technology, finally,
extended volumes can be scanned within a few seconds. High-precision micro-CT
296 P.F. Niederer / X-Ray-Based Medical Imaging

scanners are furthermore being applied in the analysis of biological samples (Figure
23). It should be noted in this context that there are also significant non-medical
applications of CT, in particular for material testing purposes.

Figure 23. Micro-CT measurement of cortical bone. The long cleft appearing in the middle has a width of 3
μm. The lacunae (openings where the osteocytes are housed) can furthermore be seen.

Further Reading

[1] Elliott K. Fishman and R. Brooke Jeffrey Jr., Spiral CT: Principles, Techniques and Clinical
Applications (2nd edition), Lippincott-Raven Publishers, Philadelphia, PA (1998)
[2] Marc Kachelriess, Clinical X-Ray Computed Tomography, in: Wolfgang Schlegel, Thomas Bortfeld
and Anca-Ligia Grosu, New Technologies in Radiation Oncology, Springer, Berlin, 2006
[3] Herman GT, Correction for Beam Hardening in Computed Tomography, Phys. Med. Biol. 24 (1979),
81-106
[4] Yu Zou, Emil Y Sidky, Xiaochuan Pan, Partial Volume and Aliasing Artefacts in Helical Cone-Beam
CT, Phys. Med. Biol. 49 (2004), 2365-2375
[5] Thorsten M. Buzug, Computed Tomography: From Photon Statistics to Modern Cone-Beam CT,
Springer, Berlin 2008
 P.F. Niederer / X-Ray-Based Medical Imaging 297

Appendix 1: The Modular Transfer Function (MTF)

The spatial resolution of imaging systems is often given in the form of the Modular
Transfer Function (MTF, i.e., grey-level contrast in the image space as function of the
number of line pairs/mm in the object space; we assume for simplicity the one
dimensional case, object space coordinate x). This concept can be introduced as
follows: Each (one dimensional) object, characterized by a grey-level or intensity
function O (x) can be thought to be composed of lines
∞
O( x ) = ∫−∞ O( X ) δ(x − X ) dX (1)

by making use of the Dirac function δ(x - X). We assume now that the Line Spread
Function (LSF), L(x’ ), which denotes the mapping of a line at location x = 0 in the
object space, i.e., of δ(x) , onto the image space with coordinate x’ , is known. A line
at location X in the object space with intensity O(x) will then yield an image
O(x ) δ(x − X ) → O( X ) L(x' − X ) (2)

In case of an optical imaging system with lenses, e.g., the LSF will be a diffraction
pattern, for x-ray projection images, the line will be blurred2.
A general object, O(x), can be decomposed into Fourier components (providing
that it fulfils the usual mathematical conditions) according to
∞
O(x ) = ∫−∞ [a(k ) sin(kx ) + b(k ) cos(kx)] dk (3)

with a(k), b(k) amplitudes

k = 2π / λ wave number, wavelength λ
One component
∞
J k (x ) = a(k ) sin(kx ) = a(k ) ∫ δ (x − X ) sin(kX ) dX (4)
−∞

is imaged by way of the LSF into the image space as Jk’(x’) in the form
∞
J k ' (x' ) = a(k ) ∫ L (x' − X ) sin(kX ) dX (5)
−∞

Upon performing the transformation x’’ = x’ – X and making use of the addition
formulas for circular function, one arrives at
∞ ∞
J k ' (x' ) = a(k ) ⎡ sin(kx' ) ∫ L(x' ' ) cos(kx' ' ) dx' ' − cos(kx' ) ∫ L(x' ' ) sin(kx' ' ) dx' ' ⎤
⎢⎣ −∞ −∞ ⎥⎦
= a(k ) [sin(kx' ) a1 − cos(kx' ) a 2 ] (6)

with

2
For two dimensional images, the Point Spread Function (PSF) is used instead of the line spread
function, in that a two-dimensional image can be decomposed into points, using the Dirac function in both
dimensions.
298 P.F. Niederer / X-Ray-Based Medical Imaging

1/ 2
2 2
⎧⎪
η(k ) = ⎨ ⎡
∞ ⎤ + ⎡
∞ ⎤ ⎫⎪
⎪⎩ ⎣⎢
∫−∞ cos(kx' ' ) L(x' ' ) dx' ' ⎥⎦ ⎢⎣ ∫−∞ sin(kx' ' ) L(x' ' ) dx' ' ⎥⎦ ⎬⎪ (7)
⎭
and using the addition property of circular functions once more, the result can be
written as
J k ' (x' ) = a(k ) η(k ) sin(kx' − ϕ) (8)

The Fourier component Jk(x) = a(k) sin(kx) is therefore imaged onto the image
space as harmonic function of the same frequency, k , but displaced by the (spatial)
phase ϕ and with the amplitude a(k) η(k) . The relation holds for every k .
The function η(k) is denoted as Modular Transfer Function (MTF), and it is seen
that the relation holds 0 ≤ MTF ≤ 1 . This function describes the dependence of the
amplitude of the original intensity function in the image space (which is always
reduced) as function of the spatial frequency k . Since the relation holds for every k ,
one can conclude that
F (image) = MTF ⋅ F (original) (F : Fourier transformation) (9)
respectively, that the image corresponds to the convolution of the original with the
MTF according to the convolution theorem.
The relation with the LSF (for two-dimensional images with the PSF) is seen from
the definition of the MTF.

Appendix 2: The Convolution / Backprojection Method

It was seen in Chapter 5, that a simple backprojection leads to characteristic artifacts; in

particular, if the projections of a point object are backprojected, a 1/r dependence
results. By prefiltering the projections, the artifacts can be compensated (or at least
substantially reduced). This aim is best reached in an indirect way, since on the one
hand the function 1/r exhibits a singularity and on the other it has to be noted that that
the real scanning procedure (translation - rotation) is made in finite steps which – as
will be seen below – leads to an inhomogeneous sampling in the Fourier space, or, in
the real space, there are significant deviations from the theoretical 1/r - dependence.
First, the projection Pϕ(r) is Fourier transformed
∞ ∞
{
F Pϕ (r )} = P *ϕ (u ) =
∞
∫−∞ Pϕ (r ) e −iur dr = ∫ ∫ μ(s[r ,ϕ]) e
−iur
dr ds (1)
−∞ −∞

(As mentioned above, the integration over s is extended to infinite for formal reasons,
as outside the source - detector area we can assume μ = 0.) Upon transformation of the
variables r = x cos ϕ + y sin ϕ , s = - x sin ϕ + y cos ϕ one obtains
 P.F. Niederer / X-Ray-Based Medical Imaging 299

∞ ∞
− P * (u ) = ∫ ∫ μ(s[r , ϕ]) e
−iur
dr ds
−∞ −∞
∞ ∞
−iu [x cos ϕ + y sin ϕ]
= ∫ ∫ μ (x , y ) e dx dy
−∞ −∞
(2)
∞ ∞
−ix [u cos ϕ] −iy [u sin ϕ ]
= ∫ ∫ μ(x , y ) e e dx dy
−∞ −∞
∞ ∞
∫ ∫ μ(x , y ) e
−ix p
= e −iy q dx dy p = u cos ϕ , q = u sin ϕ
−∞ −∞

This implies that the Fourier transform of the projection Pϕ(r) corresponds to the
two-dimensional Fourier transform of μ(x,y) on a straight line which runs (in the two-
dimensional Fourier space (p, q) under the angle ϕ through the origin (Fourier - slice -
theorem). The Fourier space is therefore sampled inhomogeneously in a star-like
fashion in case of translation-rotation scanning (Figure 24). In particular, the high
frequencies (image sharpness ! ) are covered less densely than the low frequencies. In
order to recover the Fourier transform P* (p,q) of μ(x,y) with uniform density in the
entire Fourier space of interest from the Fourier-transformed projections P*ϕ(u),
interpolations are necessary.
From the relation (1) can be seen that the reconstruction problem, i.e., the
calculation of μ(x,y) can be solved by a reverse Fourier-transformation (after having
determined the Fourier-transformed projections and interpolated P*(p,q) ). A two
dimensional reverse Fouriertransform has however to be executed numerically for this
purpose which is an unsuitable procedure. This can be prevented in the following
fashion.
From the formula for the reverse Fourier-transformation
∞ ∞
1
− μ(x , y ) = ∫ ∫ P* ( p, q) e
ixp
e iyq (3)
2
2π −∞ −∞

one obtains after having transformed the variables according to

p = u cos ϕ , q = u sin ϕ , dp dq = u du dϕ (4)
2π ∞
1 iu [x cos ϕ + y sin ϕ]
− μ(x , y ) = ∫ ∫ P * (u ,ϕ) e u du dϕ (5)
2
2π 0 0

Upon making use of the symmetry of P* (u, ϕ), viz.,

P * [u , ϕ + π] = P * [− u , ϕ] (6)

since the same projection is recorded after a rotation of 180°), this expression can be
written as
π ∞
1 iu [ r ]
− μ(x , y ) = ∫ ∫ P * ϕ (u ) e u du dϕ (7)
2π 2 0 −∞
300 P.F. Niederer / X-Ray-Based Medical Imaging

with x cos ϕ + y sin ϕ = r. By writing

∞
1
P' ϕ (r ) = ∫ P * ϕ (u ) e
iur
u du dϕ (8)
2
2π −∞

and
π π
− μ(x , y ) = ∫ P' ϕ(r ) dϕ = ∫ P' ϕ( x cos ϕ + y sin ϕ) dϕ (9)
0 0

one finds that this corresponds to a filtering or a weighting of the Fourier-transformed

projections P*ϕ(u) with |u| before execution of the reverse transform, which now is
restricted to one dimension. The subsequent integration over the angle ϕ is a
summation over all weighted reverse-transformed projections and is equivalent to a
backprojection.

u
ϕ
p

Figure 24. Fourier-slice theorem; sampling of Fourier space

On the basis of (2) and (3) the question relating to the PSF can finally be answered.
According to chapter 3.1. the convolution theorem reads ( “ * ” denotes a convolution)

Image = PSF * Object , resp. F {Image} = F {PSF} ⋅ F {Object} (10)

Here, “Image” denotes the reconstruction which is provided by the PSF and
“Object” the true distribution of the linear absorption coefficients μ(x,y). The (two-
dimensional) Fourier-transform of μ(x,y) is known, if measurements are made under
all angles and the entire two-dimensional Fourier-space (p,q) is interpolated (it is
however inhomogeneously covered by the one-dimensional Fourier-transforms
P*ϕ(u) ). Eq. (2) indicates that the factor |u| compensates the artifacts. The inverse
transform of |u| would therefore correspond to the corrected PSF. Upon performing
 P.F. Niederer / X-Ray-Based Medical Imaging 301

an inverse transform of the function |u| exp (-ε |u| ) for ε → 0 ( |u| itself cannot be
transformed !), a dependence ∼1/r2 is found. This can be interpreted such that the 1/r
– dependence of the PSF (in simple backprojection) is additionally compensated by a
further 1/r factor which is due to the inhomogeneous coverage of the Fourier space
which also is characterized by a 1/r dependence.
If therefore for the backprojection (3) not the original projections Pϕ(r), but the
corrected “projections” P’ϕ(r) are used, the desired results are obtained. In the object
space, the procedure involves a convolution such that the method is denoted as
convolution-backprojection method. It is the method of choice in computed
tomography technology. If further effects which cause image deterioration are taken
into account by adaptation of the kernel function |u| still further improvements can be
achieved.
302 Basic Engineering for Medics and Biologists
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-302

V.3. Basic Elements of Nuclear Magnetic

Resonance for Use in Medical Diagnostics:
Magnetic Resonance Imaging (MRI)
Magnetic Resonance Spectroscopy (MRS)
Peter F. NIEDERER
Institute for Biomedical Engineering
ETH Zurich and University of Zurich
Gloriastrasse 35
CH-8092 Zurich, Switzerland

Abstract. Magnetic Resonance Imaging (MRI) has established itself as a major

imaging modality in life science research and clinical practice. It is characterized
by high spatial resolution, high soft tissue contrast, non-invasiveness, and
universal applicability in terms of orientation and location of imaging areas. The
procedure allows furthermore the investigation of physiological and
pathophysiological processes, in particular in combination with magnetic
resonance spectroscopy (MRS). MR methodology is not exhausted, new
procedures and areas of application develop widely in life science and medicine.
This article is limited to basic physical aspects.

Keywords. Magnetic resonance imaging, magnetic resonance spectroscopy,

diagnostic imaging

Introduction

Nuclear Magnetic Resonance (NMR) was discovered and documented concurrently by

F. Bloch (Stanford University) and E. Purcell (Harvard University, both Nobel
laureates in 1952) in the year 1946. Since then, analytic laboratory methods which are
based on NMR have become indispensable tools in solid state physics as well as in
chemistry. Thanks to the availability of large and extremely strong magnets (super
conductivity), powerful spectrum analyzers and computers the NMR-technology could
be introduced in clinical medicine during the early seventies. Of particular importance
in this context is the pulse-spectrometry developed by R Ernst (Nobel laureate in 1991).
Further Nobel laureates were P.C. Lauterbur and P. Mansfield (2003) who made major
contributions in the development of medical MR, in particular of imaging methods.
The first cross section of a human thorax was however presented by R. Damadian et al.
in 1977 (Figure 1).
The principal application of NMR in clinical medicine is imaging (MRI, Magnetic
Resonance Imaging). For the purposes of biological and medical research,
spectroscopy (MRS, Magnetic Resonance Spectroscopy) in combination with imaging
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 303

is also of interest. In comparison with other clinical imaging methods (CT, ultrasound),
MRI/MRS has a number of distinct advantages and some disadvantages.
Compared to CT, MRI yields a better contrast of soft tissues, allows the imaging of
cross sections which exhibit an arbitrary spatial orientation and there is no ionizing
radiation - the strong constant and time-dependent magnetic fields have not been found
to cause negative effects in humans. Some disadvantages, however, derive from the
high cost of the technology and installation, the access surveillance (no ferromagnetic
objects must come close to the magnet, patients with pacemakers, surgical clips and
some types of implant cannot be treated or, at least, extreme care has to be exercised).
Claustrophobia is sometimes a problem. For the investigation of the skeleton (bone) CT
remains the imaging method of choice, since mineralized, hard tissue cannot be imaged
directly with MRI. Likewise, in emergency units where rapid examinations are required,
CT scanners are advantageous because of speed, accessibility as well as the fact that
hemorrhages show a good CT contrast due to the iron content of hemoglobin.

Figure 1. First in vivo human MR image (cross section, thorax) made by Damadian et al. 1977

Ultrasound is even more rapid and real-time imaging can be performed. In addition,
ultrasound equipment is considerably less expensive and smaller that MRI or CT
installations. Yet, MR image quality is largely superior compared to ultrasound,
moreover, ultrasound can only be applied through acoustic windows and bony
structures and air-filled cavities are almost intransparent to ultrasound.
Of great future interest is real-time MR imaging which is under development and
may become feasible within the next few years. A further interesting development is
associated with “open” systems which allow direct patient access and provide the
possibility for MRI-guided surgery.
304 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

MRI and MRS have established themselves as significant research tools also in a
wide range of non-clinical applications relating to human behavior in general and
psychology in that brain activity can be monitored under various real and artificial
exposure situations. Due to its noninvasive character, MR technology is furthermore
widely used in animal research.
At this time, MR technology in the life sciences is by far not exhausted and new
applications emerge constantly. Accordingly, the literature on MRI/MRS is vast; this
introductory text is limited to basic physical aspects which are presented in a partly
simplified form. Quantum mechanics and thermodynamics are only used as far as
necessary.

1. Nuclear Spin Resonance

1.1. Intrinsic Angular Momentum (Spin)

Elementary particles (protons, neutrons, electrons, etc.) have an intrinsic angular

momentum (spin), whose size is specific for each particle and may in particular be zero.
It has no classical analogon, i.e., it cannot be ‘explained’ on the basis of classical
mechanics. In fact, relativistic quantum field theory is necessary to fully appreciate spin
systems.
Atomic nuclei consist of protons and neutrons and therefore also possess a spin.
This spin depends on the type of nucleus and derives in a nontrivial manner from the
internal nuclear structure. Of particular interest for medical applications are the stable
nuclei 1H, 4N, 19F, 23Na, 31P, as well as the isotope 13C, which occur naturally in the
human body in various concentrations and which all have a spin different from zero.
According to the laws of quantum mechanics the values of the modulus of angular
r
momentum vectors, J , can in stationary states assume only integer multiples of
Planck’s constant h 2π = h ( h = 6.63⋅10-34 W⋅sec2 ) with respect to a fixed axis.
This includes nuclear spins as well. As a reference axis, one usually chooses the z –
component of a Cartesian coordinate system x, y, z , such that for the z - component
of the angular momentum, Jz , follows
Jz = hm ( m integer, denoted as magnetic quantum number) (1)

The possible values of |m| are furthermore limited; the relation holds
m = I , I - 1 , I - 2 , ... , - I (2)
I represents the spin quantum number. It is specific for each nucleus and can be integer
or half-integer. A nucleus can therefore assume 2I + 1 stationary states with respect to
angular momentum. In the following, the nucleus 1H with I = 1/2 will mainly be
considered (this nucleus is abundant in the human body), such that in this case
Jz = ±h 2 (3)

As atomic nuclei are electrically charged, the spin is associated with a magnetic
dipole moment. This amounts to
r r
μ = γJ (4)
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 305

γ is denoted as gyromagnetic ratio and is likewise specific for each type of nucleus.
For 1H , γ = 4.26⋅107 (T⋅sec)-1 (the unit T stands for „Tesla“, the unit used for the
r
magnetic B - field).
r
In a magnetic field with flux density B (also denoted as magnetic induction) the
axis of reference is given in a straightforward way by the direction of the vector-field
which is identified with the z – axis. Accordingly,
μ z = γ h m ; or, in case of 1H , we have μz = ± γh 2 . (5)

Following the laws of quantum mechanics, physical quantities assume a defined

value only if there is an interaction with the surroundings. Here, this interaction is
r
provided by the magnetic field B . In the absence of such a field, there is no a priori
direction defined, the magnetic quantum number undefined and all corresponding states
are degenerated, i.e., all states possess the same energy. (This also explains the term
“magnetic quantum number” for m : The spin manifests itself only in the presence of a
sufficiently strong interacting field. The gravitation of the earth or its magnetic field are
far too weak to have a noticeable effect.)
r r
The energy of a magnetic dipole, μ , in the field B reads (Figure 2)
r r
( )
E = − μ, B , resp. E m = − γ h B m (6)

in the states m = ± 1 2 , if the potential energy of the system in the field-free space is
normalized to zero

E
m = - 1/2

ΔE = γh B

m = + 1/2

Figure 2. Energy of the states m = +/- 1/2 as function of the field B

In thermal equilibrium, the state associated with lower energy (m = +1/2, spin
parallel and oriented along the same direction as the field) is more populated than the
state with higher energy. The population density thereby obeys the Boltzmann
distribution according to the probability
306 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

⎡ −E ⎤ ⎡γ h B m⎤
W = N exp ⎢ ⎥ (general), resp. W m = N exp ⎢ ⎥ (7)
k T
⎣ B ⎦ ⎣ kB T ⎦
with Wm probability for the system being in state m ,
N normalization factor (sum of all probabilities = 1),
kB Boltzmann constant ( 1.38 ⋅ 10-23 W sec / ° ),
T absolute temperature
In the case considered here, the system has only two states (m = ± 1/2). Upon
⎡γ h B m⎤
calculating the Boltzmann factor, exp ⎢ ⎥ , one finds a value around
⎣ kB T ⎦
[ ]
exp 10 −6 ⋅ B m (at T = 300°K) , i.e., the populations of the two states differ little
even in the presence of high fields. The normalization factor, N, can therefore be
approximated by 0.5, the population difference Δn can furthermore be determined by
linearization as
Δn = n +1 2 − n −1 2 ≈ n γ h B [2 k B T ] (8)

n thereby denotes the number of magnetic moments per unit volume. The total nuclear
r
magnetization M results as the sum of all dipole moments per unit volume
r r r r
M = Σ μ i , M = Δn μ z = n γ 2 h 2 B [4 k B T ] = χ 0 (T ) B (9)

χ 0 is denoted as nuclear magnetic susceptibility.

1.2. Bloch’s Equations

r
Systems of identical charged particles i with angular momentum J i , that are exposed
r r r
to a magnetic field B , are subjected to a moment according to μ i × B . The
r
macroscopic magnetisation M executes a motion which can be described by the
classical angular momentum equation,
r
d Σ Ji r r r r
= Σ μi × B = M × B (10)
dt
r r r
d Ji d μi dM r r
respectively, γ Σ = Σ = = γM ×B (11)
dt dt dt
r r
In the most simple case, i.e., for M = const. and B = ( 0, 0, B0 = const. ) ,
one obtains

d Mx dMy d Mz
= γ B0 M y , = − γ B0 M x , = 0 (12)
dt dt dt
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 307

The solution of this differential equation, ie.,

M x = m x 0 cos (ω 0 t ) , M y = − m y 0 sin (ω 0 t ) , M z = m z 0 (13)

(the values for mi 0 are thereby given by the initial conditions) represents a precession
r
(i.e., a rotation around the z-axis) of M with frequency ω0 = γ B0 . The direction of
r
the vectorial rotation is opposed to B . ω0 is denoted as Larmor frequency. According
to
ΔE = hν = h ω 0 = γ h B0 (14)

This corresponds to the frequency, which is equivalent to the energy difference

between the two states with m = ± 1/2 .
Typical Larmor frequencies at a field strength of 1 T and for protons ( 1H, spin 1/2)
are 42.6 MHz, for 23Na (spin 3/2) 11.3 MHz, for 31P (spin 1/2) 17.2 MHz, i.e., the
intrinsic angular momentum vectors execute a precession with these respective
frequencies around the z-axis. If one excites the individual atomic spins from their
equilibrium states and synchronizes their precession, the macroscopic magnetization
r
M will precess with the same frequency and emit a recordable response.
In medical MR applications, the object to be examined is positioned in a stationary,
r
strong magnetic field B0 , whose orientation coincides with the z – axis of the
r
reference (laboratory) system. A magnetization M 0 as strong as possible is attempted
to be obtained in this fashion. In a typical medical MR scanner today the stationary
field strength is 1.5 T or 3 T, respectively. (higher field strengths are used
experimentally as well as in small animal scanners). The exciting fields mentioned
above which induce the desired effects (see later) are time-dependent and have the
r
components Bx(t) resp. By(t) , perpendicular to B0 . The total magnetic field is
r
[
therefore of the form B = B x (t ), B y (t ), B0 . ]
The time-dependent fields cause an excitation of the system from its equilibrium
state characterized by Mx = 0 , My = 0 , Mz = |M0| = M0 ; the latter value being
given by the population differences as calculated above under the influence of the field
B0 . After termination of the time-dependent exciting fields the system relaxes back to
its equilibrium state because of interactions with the environment associated with
energy transfer. There are in essence two relaxation phenomena (explained in more
detail later):
1. Through interaction (i.e., energy exchange) with the surrounding atoms the
system relaxes back to its thermodynamic equilibrium state. Since the early NMR
experiments were made in solid state (lattice) materials, this effect is denoted as spin-
lattice interaction. It is characterized by a time constant T1 .
2. The spins carry a small magnetic field of their own which causes interactions
in that the Larmor frequencies of the individual nuclei differ by a minute amount
depending on the arrangement of the surrounding spins. This causes the components of
the magnetization perpendicular to the z-axis, viz. Mx and My to disappear because of
a gradual loss of synchronization or dephasing (spin-spin-interaction, time constant T2 ).
The relaxation phenomena can be modeled with sufficient accuracy as linear first
order processes. Accordingly, the angular momentum equations, denoted as Bloch’s
equations, reads
308 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

dM x
dt
= [
γ M y B0 − M z B y (t ) − ] Mx
T2
dM y My
= γ [M z B x (t ) − M x B0 ] − (15)
dt T2
dM z
dt
= [
γ M x B y (t ) − M y B x (t ) − ] M z − M0
T1

We calculate the solution of these equations for the time after the excitation is
terminated (for simplicity, we assume that at t = 0 both fields Bx(t) and By(t) are
shut down). The equations then are
dM x Mx
= γ M y B0 −
dt T2
dM y My
= − γ M x B0 − (16)
dt T2
dM z M z − M0
= −
dt T1

Upon elimination of My using the first two equations, one arrives at

d 2M x 2 dM x ⎛ 1 ⎞
+ ⋅ + ⎜ ω0 2 + ⎟M = 0 (17)
dt 2 T2 dt ⎜ T2 2 ⎟ x
⎝ ⎠
The solution of this equation is a damped oscillation with frequency ω0 and damping
constant 1 / T2

⎡⎛ 1 ⎞ ⎤
M x (t ) = M x 0 exp ⎢⎜⎜ i ω 0 − ⎟ t⎥ (18)
⎢⎣⎝ T2 ⎟⎠ ⎥⎦

Mx0 represents the magnetization in x-direction at time t = 0 (initial condition,

induced by the excitation). For t → ∞ it is seen that Mx goes to zero. A
corresponding solution is obtained for My.
The solution for Mz , in turn, reads

⎛ t ⎞
M z (t ) = (M z 0 − M 0 ) exp ⎜⎜ − ⎟⎟ + M 0 (19)
⎝ T1 ⎠
and shows that Mz goes towards M0 for t → ∞ ( Mz0 is the magnetization in z-
direction at time t = 0 ).

1.3. The Free Induction Decay (FID)

Nuclear spin- “resonance” is reached when the components Bx(t) and By(t) of the
time-dependent excitation field initializing a measurement oscillates harmonically such
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 309

r
[ ] r
that the resulting field B1 = B x (t ), B y (t ), 0 rotates with the frequency ω 0 , i.e. with
the Larmor frequency. This can be shown by solving Bloch’s equations as follows.
r
First, we remember that the Larmor frequency ω 0 has a sense of rotation which is
r
opposite to the one of the field direction B0 . In order to avoid unnecessary calculations,
we assume that the exciting field with for the time being arbitrary frequency ωz has a
sense of rotation which likewise is antiparallel to this field (otherwise no resonance is
r r
possible), i.e. ω z = (0, 0, − ω z ) . The associated field B1 reads
r
[ ]
B1 = B x (t ), B y (t ), 0 = B1 (cos ω x t , − sin ω x t , 0) (20)

This corresponds to a field rotating around the z-axis counterclockwise with constant
r
amplitude B1 = B1 which at time t = 0 has x-direction.
r
r r r r
In order to solve Bloch’s equation,
dM
dt
[ ]
= γ M × B0 + B1 (t ) + R (the relaxation
r
terms are included for brevity in the vector R ), a transformation is now made into a
r
system rotating with ω z . This procedure will lead to a quite illustrative solution
r
because the time dependence of the field B1 disappears: In the rotating system, both
axes x’ and y’ (primes denote the new axes in the transformed system) rotate
counterclockwise around the z-axis with frequency - ωz , while the z-axis rotates on
r
itself (z’ = z). Since the amplitude of the field B1 is constant, a time-independent set
of equations results. Upon application of the general formula for a transformation into a
rotating system,
d' d r
= − ω× (21)
dt' dt
Bloch’s equation read
r r
r r r r r ⎡r r ⎤ r
d' M
[ ]r ω
= γ M × B0 + B1 − ω z' × M + R' = γ M × ⎢ B0 + z' + B1 ⎥ + R'
γ
(22)
dt' ⎣ ⎦
r
The transformed relaxation terms, R' , are first not considered in a simplified analysis,
they are therefore not calculated in their transformed form. Note that the term
r r r
B0 + ω z γ has only a component in the z’-direction, while the vector B1 is constant
after the transformation and is perpendicular to z’ and oriented in the x’- direction. The
r r r r
quantity Beff = B0 + ω z' γ + B1 represents a constant, “effective” field in the
rotating system such that the equation can be written as
r
d' M r r r
= γ M × Beff + R' (23)
dt'
310 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

z’ Beff

M0 B0 + ω z’ / γ

M α y’

nutation cone

B1
x’
Figure 3. Nutation cone in rotating system

r r
Before the field B1 is activated, the magnetization M has only a z’-component,
r
M0 , which is constant in time. Once B1 is turned on, the equations then valid can
r
easily be solved if the relaxation terms R' are disregarded. The solution describes a
r
nutation or secondary precession (i.e., rotation) of the vector M around the axis
r
defined by Beff with the angular velocity ω1 = γ Beff (Figure 3). The opening angle of
r r
the nutation cone, α, can be changed at fixed B0 and B1 by variation of ωz’. The
surface line M0 remains thereby unchanged, i.e., the cone always includes the z = z’ -
axis. If one now chooses for - ωz’ the Larmor frequency, - ωz’ = γ B0 , resonance is
r r
obtained. The opening angle α becomes 90°, since Beff coincides with B1 (note
r r
ω
that B0 + z' = 0 ). The cone becomes a plane lying in the y’ - z’ – plane; the tip of
γ
r
the magnetization vector M follows a circle. Accordingly, at resonance a virtual
r r
observer in the rotating system “sees” a constant field B1 = Beff in x’- direction
r
because he rotates with the same angular velocity as B1 . He furthermore can observe
the magnetization vector to rotate with constant angular velocity in the y’- z’- plane
(Figure 4).
The angular velocity of nutation (or secondary precession) ω1 at resonance equals
r
γB1, since Beff is equal to B1 . If the field B1 is therefore activated during the time
r
period Δt , the macroscopic magnetization M will nutate or precess around the angle
θ = Δt ω1 = Δt γ B1. Accordingly, by an appropriate choice of Δt and B1 it is
possible to reach any desired flip angle. In practice, often 90°- or 180°- pulses are
r
applied, which cause a re-orientation of M around 90° into the -y’ – axis, respectively
around 180° into the -z’ – axis. The term “pulse” is used in this context because Δt is
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 311

typically on the order of a few msec while the Larmor frequency is in the MHz-range.
The measurement fields are therefore high frequency (radiofrequency, RF) pulses.

z’

M0
y’

Beff = B1
M

α x’

α: flip angle

Figure 4. Secondary precession (nutation) at resonance

After termination of the excitation pulse the magnetization vector precesses with
constant angular frequency ω0 around the z’-axis. Since this vector is a macroscopic
quantity in non-equilibrium conditions, it emits a radio signal with the same frequency
which is sufficiently strong that it can be measured in a receiving coil. In order to
analyze this signal, the relaxation terms have now to be taken into account.
The spin-spin interaction derives from the fact that each spin represents a magnetic
dipole associated with a minute magnetic field which is superimposed over the external
magnetic field. In typical biological tissues, the individual spins are inhomogeneously
distributed. This causes small local inhomogeneities of the total field implying that the
elementary spins have slightly different Larmor frequencies. The spins which are
contained in a macroscopic volume element execute initially (i.e., immediately after the
excitation) a synchronous precession. Yet, with increasing time, the spins get gradually
out of phase because of the slightly different Larmor frequencies such that the
component of the magnetization vector which is perpendicular to the z’-axis disappears.
This can be described with sufficient accuracy as a linear process and the associated
relaxation decay is of the first order with a characteristic time constant T2. This effect
is in practice significantly amplified because the external ‘homogeneous’ and constant
magnetic field exhibits small local inhomogeneities itself for technical reasons in spite
of correction measures (shim coils), thereby in addition causing locally varying Larmor
frequencies. Accordingly, a relaxation time constant, T2*, is found which is
significantly smaller than T2 and which in particular is machine-dependent. For
biological tissues and field strengths around 1-3 T, T2* is typically on the order of a
few tens of msec (see later).
The spin-lattice relaxation, in turn, is due to interactions (collisions) with
surrounding atoms whereby energy is exchanged and dissipated. These processes cause
the magnetization vector to relax back into its thermodynamic equilibrium orientation
in z’-axis. This can also be described by a first order relaxation process with an
312 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

associated time constant T1. The latter is usually much larger than T2* (a few hundred
msec).
Due to the relaxation phenomena and field inhomogeneities, a macroscopic
volume induces a signal in a receiving coil, which oscillates almost harmonically, has a
frequency ω close to the theoretical Larmor frequency ω0 and whose amplitude decays
exponentially with the characteristic time constant T2* . (Since the time constants T2
and T1 are significantly larger than T2* they manifest themselves only marginally.)
This signal is denoted as Free Induction Decay (FID) and has the form, as shown above
(only real part),
T2*
U = U 0 e −t cos ωt (24)

The demodulation of this signal is performed as usual by multiplication with

cos (ω 0 t ) ,

U’ = U0 exp ( -t / T2* ) cos ωt cos ω0t , after low-pass filtering

U’’ = ( U0 / 2 ) exp ( -t / T2* ) cos ( ω0 - ω ) t (25)
results. The function describes a sinusoidal curve with exponentially decreasing
amplitude (FID, Figure 5, left).

U" F

~ exp (-t/T2*)

t
ω0 - ω p

Figure 5. FID signal and Fourier transform

The real part of the Fourier-transform (independent variable p) of this signal reads

T2*
RT [F (U ' ' )] ∝ (26)
(
1 + T2* 2 ω p − ω − p )2
and approaches under optimal conditions (signal-to-noise ratio) a line (Figure 5, right).
The determination of such lines in the spectrum (NMR - spectroscopy) is of importance
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 313

likewise in medical imaging applications as well as in chemical analysis (see later). In

particular, the primary signal strength, characterized by the area under the „line“, is
proportional to the amount of spins contributing to the signal, i.e. the spin density, ρ.
As we shall see, this quantity is one of the main imaging parameters.

1.4. Measurement of T2, Spin-Echo Procedure

Once the magnetization has been rotated into the y’-axis by a 90° pulse (Figure 6a), the
spins start immediately to de-phase because of the slightly different precession
velocities, such that the measured signal – as previously shown – decays exponentially
with the time constant T2* . In the rotating system this means that the spins slowly fan
out in both directions from the y’-axis (b). If after a time interval TI a 180° pulse is
applied (Figure 7), all spins are rotated in a plane parallel to the y’-z’-plane around
180° (c).

Figure 6. Spin-echo: De-phasing and re-phasing of spins

There are two reasons for the de-phasing of the spins, viz., (1) the spin-spin
interaction which is in essence of a stochastic nature, (2) the inhomogeneities of the
static magnetic field which are constant and systematic (but machine-dependent). By
application of a 180° pulse the second effect can be compensated in that spins that
rotate slower or faster due to field inhomogeneities continue to do so after the 180°
pulse (d). Accordingly, the spins will rephrase after the time period 2TI (e). In the
receiving coil a signal will therefore build up and subsequently decay (f, Figure 7). The
stochastic spin-spin interaction is not affected by this maneuver, such that the
characteristic signal times are T2*. T2 can e.g. be determined from repeating the
procedure (repetition time TR).
314 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

90° 180°

U" t

~ exp (-t/T2 )
~ exp (-t/T2*)

TI TI
TR
// t

Figure 7. Spin-echo sequence

1.5. Measurement of T1, Inversion-Recovery-Method

By application of a 180° pulse the magnetization is turned in -z’ – direction (Figure 8).
Due to the spin-lattice interaction it relaxes back to the original orientation. The
momentary size of the z’- component of the magnetization during the relaxation
process can be determined by triggering FID signals that are obtained from 90° pulses.
T1 is determined by fitting the exponential curve.

180° 90°

t
~ [1 – 2 exp ( -t/T1 )]
M

U"

TI
TR
// t
Figure 8. Inversion-recovery method
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 315

1.6. Chemical Shift

The spin-spin-interaction has a further important consequence. Depending on the

number, type and location of the atoms surrounding a nucleus, the local spin
configuration that a nucleus is exposed to is different and, along with this, the local
field strength. The surroundings of a nucleus are however determined by the chemical
compound in which it is located and by the position it has within this compound.
Accordingly, the Larmor frequency of a nucleus is shifted according to its exact
location within a chemical compound (“chemical shift”). An analysis of NMR spectra
therefore allows us to elucidate the sterical configuration of a compound. This
procedure is of particular value in protein analysis (Nobel laureate Kurt Wüthrich
2002).

2. Basics of Medical MR Imaging (MRI)

In MR imaging, the three parameters ρ (spin density), T1 (spin-lattice relaxation time)

and T2 (spin-spin relaxation time) are measured and imaged in various combinations.
These parameters exhibit characteristic differences for different tissues, e.g., for
B0 = 0.3 T :

Table 1. Typical relaxation times for biological tissues (B0 = 0.3 T)

T1 (msec) T2 (msec)
Brain: Pons 445 75
Brain: Cerebellum 585 90
Bone marrow 320 80

A spatial resolution is achieved by the application of three mutually perpendicular

gradient fields. These fields are gated and constant during the duration of application.
Their direction is typically oriented along the z – axis such that the stationary field, B0
has a contribution depending on the spatial location. With the linear gradients
∂B ∂i = Gi (i = x , y , z ) the magnetic field inducing magnetization is
r
(
B0 = 0 , 0, B0 + Gx ⋅ x + G y ⋅ y + G z ⋅ z ) (27)

Due to the gradient fields, the Larmor frequency is dependent on the spatial location.
The gradients can furthermore be combined arbitrarily, such that arbitrary cross
sections through the object can be examined
In the following, the most often applied Fourier-imaging method is considered in a
basic, simplified form. For further simplicity, a cross section perpendicular to the z –
axis is chosen. The imaging procedure consists of three steps (Figure 9).
In a first step, the gradient Gz is activated during the duration TP which is
necessary to induce a 90° flip angle (excitation field B1, frequency ωP ). The stationary
field during this period of time is
316 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

r
B0 = ( 0, 0, B0 + G z ⋅ z ) (28)

The resonance condition is only fulfilled at that location z, where

γ (B0 + G z z ) = ω P (29)

This implies that only a layer z = z0 = const. is excited. The thickness of the layer,
Δz, depends on the gradient strength and the bandwidth of the applied pulse TP. Outside
this layer the resonance condition is not fulfilled such that associated volume elements
will not contribute to the signals to be recorded later.
After the interval TP, the excitation B1 and gradient Gz fields are shut down
concurrently. The magnetization in the chosen layer Δz which has been flipped around
90° starts to precess in the x - y – plane around the z – axis. Now, in a second step, the
gradient Gy is activated. Due to this gradient, the local Larmor frequencies depend on
the coordinate y. For a given yi , however, the Larmor frequency does not depend on
the coordinate x, i.e., in a slab extending in x - direction at location z0, yi , the
precession frequencies are the same. The size of the slab perpendicular to the x – axis,
Δy Δz, depends on the gradient strength and the bandwidth of the read-out system
active in the third step.
The local precession frequency is
ω ( yi ) = γ ( B0 + Gz z + Gy yi ) = ωP + γ Gy yi (30)
After the time interval Ty the orientation of the local magnetisation vector has an
orientation which exhibits a relative phase δ depending linearly on yi :
δ ( yi ) = [ω ( yi ) - ωP ] Ty = γ Gy yi Ty (31)
The y – coordinate is therefore „packed“ into a phase; accordingly, the gradient
Gy is often denoted as phase-encoding gradient.

Figure 9. Principle of image acquisition, in particular, the phase encoding along the y-axis
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 317

After the interval Ty the gradient Gy is turned off and the measurement is started
(third step). In order to achieve a resolution also in the x – direction, the gradient Gx
is activated. The signal recorded in the measurement coil represents then a
superposition of FID-pulses with different frequencies since the frequency of the FID
signal ω(xj) depends on x because of the gradient Gx (Figure 10). The recoded
signal S, originating from the planar area at z0 (without taking into account the
relaxation terms), reads
[( ) ]
∑ ρ(x j , yi , z 0 )e
i ω x j t + δ( yi )
S (z 0 , t ) ∝ with (32)
i, j

ω ( xj ) = γ ( B0 + Gx xj ) = ω0 + γ Gx xj (33)

Gz
t
B1
90o
t
Gy
t
Gx
t
measurement
Messung

Figure 10. Sequence of gradients in the Fourier imaging scheme described here

The resolution Δx is again determined by the strength of the gradient and the
bandwidth of the measurement system. After demodulation (multiplication with cos ω0
and low-pass filtering) the signal reads
[ ]
∑ ρ(x j , yi , z 0 )e
i γ Gx x j t + γ G y yi Ty
S' (z 0 , t ) ∝ (34)
i, j

The signal is sampled in time steps Txm ( m = 1 , ... , M ) in M channels. The

measurement is subsequently repeated with n = 1 , ... , N different phase-encoding
gradients Gyn . The total signal is therefore finally composed of contributions of the
form
318 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

S ijmn (z 0 ) ∝ ∑ ρ(x j , yi , z0 )e
[
i γ Gx x j Txm + γ G yn yi Ty ] (35)
i, j

The summands represent the components of a discrete Fourier-transform such that the
quantity ρ ( xj , yi , z0 ) , i.e., the image, can be obtained by application of Fourier
methods.
Once the entire Fourier space (denoted as k-space in MR-jargon) is sampled, the
image can be calculated. By an appropriate choice of the parameters Gx , Txm , Gyn , Ty
as well as of their temporal sequence the k-space can be sampled in various fashions
which are adapted to the analysis to be performed.

3. MR Spectroscopy (MRS) and Further Methods

Due to the abundance of hydrogen nuclei in biological materials, 1H imaging is used

for routine examinations as the best signal/noise ratio is obtained with this element.
Most routine applications are limited to morphology where spatial resolution is of
primary importance. There are other, physiologically interesting elements however
with non-zero spins that lend themselves for analysis and imaging, in particular also for
functional imaging. 23Na (spin 3/2), Larmor frequency (at 1 T) 11.3 MHz, and 31P (spin
1/2), Larmor frequency 17.2 MHz, mentioned previously, are among these elements.
Since the concentration of these elements is much lower that the one of hydrogen,
signal/noise is correspondingly low. Image resolution is by far inferior such that a
combination with 1H imaging is usually made.
The chemical shift phenomenon allows for further analyses based on spectroscopy
(MRS). A typical example is associated with 31P as phosphor is a major component in
physiological energy turnover. During this cycle, the phosphor atom appears in various
chemical surroundings giving rise to associated chemical shifts (Figure 11).
MRI and MRS methods are still developing rapidly. A large number of advanced
methods and medical applications are available; some of them are mentioned in the
following. In particular, MR brain imaging is increasingly being applied in non-
medical research, mostly related to the examination of human behavior patterns and
psychology. For details and explanations the reader is invited to consult the literature,
in particular the journal Magnetic Resonance in Medicine, where most advanced
research results are being published.
With Echo-Planar Imaging (EPI) an entire image, i.e., the entire k-space is scanned
during one single excitation. Imaging is extremely rapid, however, the method is
associated with several drawbacks: The gradient coils have to be switched rapidly
which is technically demanding and may furthermore induce adverse
electrophysiological effects in the patient, susceptibility artifacts (caused by local
variations in susceptibility due to the influence of the biological tissues) degrade image
quality.
Sensitivity-Encoding (SENSE) is an efficient method to reduce imaging time.
Thereby, the physical coil field characteristics are used to enable a parallelization of the
k-space sampling.
Angiography is made by the application of pulse sequences that are sensitive to
movement.
 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics 319

Figure 11. Phosphor spectrum. α, β, γ correspond to the three different positions of the phosphor atom in the
ATP molecule. Due to the different chemical surrounding, Larmor frequencies are different (in ppm, parts
per million deviation from frequency). PCr denotes phosphocreatinine, where the phosphor atom is again in
different surroundings. The frequency of this strong line is chosen as reference (zero deviation).

By the application of pulse sequences which cause local saturation (no signal is
thus obtained), patterns can be overlaid on biological structures (tagging, Figure 12).
Muscular deformations can be visualized in this fashion.
Small-bore, high-field systems (up to 19 T) are used in small animal research
(mice, rats). In combination with MR spectroscopy, fluorescence diffusion optical
tomography (FDOT) or bioluminescence recording (both latter technologies are based
on infrared diffusion) physiological and pathological processes are studies in animal
models.

Figure 12. Tagging: The pattern induced by an appropriate pulse sequence allows the monitoring of tissue
deformations as function of time, here of the heart (left: short axis section; right: long axis section).
320 P.F. Niederer / Basic Elements of Nuclear Magnetic Resonance for Use in Medical Diagnostics

Most biological tissues exhibit an anisotropic fiber structure. In Diffusion Tensor

Imaging (MR-DTI) the components of the diffusion tensor of water molecules is
measured. The primary eigenvector characterizes the average fiber direction in each
voxel since water molecules diffuse faster along membranes than across. By
concatenation of such vectors (fiber tracking), the fiber architecture of organs can be
highlighted (Figure 13).

RV LV

Figure 13. Fiber architecture of the human ventricles as determined by MR-DTI (ex vivo). RV: right
ventricle, LV: left ventricle.

Further Reading

[1] C.N. Chen, D.H. Hoult, Biomedical Magnetic Resonance Technology. Medical Sciences, Taylor &
Francis, 1989. ISBN 978-0852741184.
[2] A. Oppelt, Imaging Systems for Medical Diagnostics: Fundamentals, Technical Solutions and
Applications for Systems Applying Ionizing Radiation, Nuclear Magnetic Resonance and Ultrasound,
Wiley-VCH, 2006, ISBN 978-3895782268.
[3] M. Blaimer, F. Breuer, M. Mueller, R.M. Heidemann, M.A. Griswold, P.M. Jakob, SMASH, SENSE,
PILS, GRAPPA: How to Choose the Optimal Method, Topics in Magnetic Resonance Imaging 15
(2004), 223–236. http://cfmriweb.ucsd.edu/ttliu/be280a_05/blaimer05.pdf.
[4] A.G. Filler, The history, development, and impact of computed imaging in neurological diagnosis and
neurosurgery: CT, MRI, DTI, Nature Precedings. doi:10.1038/npre.2009.3267.2.
Basic Engineering for Medics and Biologists 321
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-321

Introduction to
Chapter VI: Rehabilitation
Engineering
Tadej BAJD and T. Clive LEE (eds.)

Rehabilitation engineering is a rapidly evolving field that can be defined as the

application of science and engineering to the development, design and application of
assistive technologies and neurorehabilitation techniques that need to be matched with
the specific needs of a person with a particular disability or a particular condition. This
definition in its core assumes and requires a substantial degree of technical competence
on the side of rehabilitation engineers that needs to be complemented with clinical
competence of physicians and therapists all of whom form a team that assesses the
needs of and determine the goals of a rehabilitation process for a particular client. This
requires from all the rehabilitation team members not only specialized knowledge in
their own field of competence but also the ability to successfully communicate with the
other team members. In a successful rehabilitation team this may only be possible if
rehabilitation engineers possess basic knowledge on anatomy, physiology,
rehabilitation medicine and therapy while on the other hand the medical professionals
understand basics of rehabilitation engineering. Rehabilitation of mobility and
manipulation is the domain where a harmonic co-operation of the team members is
critical for a successful application of an assistive device or a specific regime of
training.
This contribution is divided into three standalone sections that nicely complement
each other. In the first section we present basics of gait analysis and synthesis including
orthotics and prosthetics that require an understanding of underlying biomechanics.
The second section is devoted to the functional electrical stimulation (FES) of
extremities, a technique which may be used as orthotic means or a temporary training
aid, where the basic knowledge on electrical and biomechanical parameters associated
with FES is presented. The third section is devoted to rehabilitation robotics, a field
that saw a rapid development in the last decade, mainly due to introduction of virtual
reality and haptic interaction in the rehabilitation of various neurological disorders. All
three sections jointly provide basic concepts of biomechanics, kinesiology and
technology needed to comprehend possible benefits when applied correctly to the
patients.
This page intentionally left blank
Basic Engineering for Medics and Biologists 323
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-323

VI.1. Gait Analysis and Synthesis:

Biomechanics, Orthotics, Prosthetics
Zlatko MATJAČIĆ
University Rehabilitation Institute, Republic of Slovenia
University of Ljubljana
Linhartova 51
1000 Ljubljana
Slovenia

Abstract. This contribution presents and establishes the biomechanical principles

that underlie human walkin. This is done by using a range of simplified
biomechanical models of bipedal walking to explain the laws of movement and
associated energetic requirements. Based on these simplified models, the
measurements of normal walking are described. Selected pathological cases are
used to illustrate the changes that occur in abnormal walking patterns. Finally, the
basic design principles used when applying orthotics and prosthetics to enhance or
restore impaired or missing function in walking are described for these case
studies.

Keywords. Walking, bipedal locomotion, kinematics, kinetics, orthotics

Introduction

The analysis of human bipedal locomotion is of interest to many different disciplines,

including biomechanics, human movement science, rehabilitation and medicine in
general. Gait is the collective term for the two types of bipedal locomotion, walking
and running. In this paper, we will deal only with walking. The objective is to alter
pathological walking by means of orthotic or prosthetic aids.
Gait analysis involves basic measurements, such as walking speed, step length and
cadence, and more detailed measures of the relative motion of body segments and
joints, the patterns of forces applied to the ground and the sequence and timing of
muscular activity. The movements of joints and segments, including position, velocity
and acceleration, are termed kinematics. Kinematic data are obtained by contactless
movement tracking systems that use video cameras for tracking reflective markers
attached to predetermined anatomical positions on the selected body parts.
The pattern of forces is obtained from force platforms, sensors that are imbedded
in the floor of a laboratory, that measure the forces exerted by the foot when in contact
with the ground. These ground reaction forces, are used together with kinematic and
anthropometric data, segment sizes, masses and inertial properties, to calculate net joint
torques produced by muscular activity during gait. This calculation requires the
324 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

application of general equations of motion in the inverse dynamics model of the human
body (Koopman paper, Chap1). These models constitute a part of the software of
modern movement tracking systems. The ground reaction forces and the net joint
torques are termed kinetics.
The patterns of muscular activity are termed electromyographs and can be
obtained from surface electrodes applied over the skin of the muscle of interest.
Kinematics, kinetics and electromyography represent important variables that help the
clinicians to understand a particular gait pattern and to discriminate primary gait
abnormalities, from the compensatory changes that are the consequence of the former.
Based on an understanding of a pathological gait, a treatment procedure or application
of suitable orthotics and prosthetics can be devised. However, understanding complex
gait patterns is a challenging task for clinicians, as walking is largely defined by
biomechanical laws. While understanding normal gait patterns can be learned through
descriptions of events and observations during walking, the comprehension of
individual pathological cases is challenging because they can be very diverse. In order
to correctly interpret measured gait patterns, a basic knowledge of the biomechanics of
walking is required.
The aim of this paper is to establish the biomechanical principles that underlie
human walking. This is done by using a range of simplified biomechanical models of
bipedal walking to explain the laws of movement and associated energetic
requirements. Based on these simplified models, the measurements of normal walking
are described. Selected pathological cases are used to illustrate the changes that occur
in abnormal walking patterns. Finally, the basic design principles used when applying
orthotics and prosthetics to enhance or restore impaired or missing function in walking
are described for these case studies.

1. Bipedal walking in humans

1.1. Basic Definitions and Terminology

Walking is a cyclical activity that is composed of several succeeding cycles. The gait
cycle is most frequently defined as a period from heel contact of one leg until the next
heel contact of the same leg. Figure 1 illustrates the sequence of silhouettes of a
walking person that constitute a gait cycle. The gait cycle consists of two distinctive
phases: the stance phase in which the foot of the leg is placed on the ground and the
swing phase in which the leg is in the air and advances to the position of the next heel
contact. Each stance phase has a period of time at the beginning and at the end, termed
double stance, in which both legs are on the ground. In normal walking, the stance
phase extends over 60% of the gait cycle, while the swing phase occupies the
remaining 40%. The duration of each double stance is approximately 10%. The stance
phase is further sub-divided into: initial contact, loading response, mid-stance, push-off
and pre-swing. The swing phase is sub-divided into: initial swing, mid-swing and
terminal swing.
Step length is defined as the distance between the same points on each foot during
double stance. Stride length is the distance between two successive heel strikes by the
same foot and therefore represents the distance travelled within the gait cycle. Walking
speed is the average speed over multiple strides, while cadence is the number of steps
per time unit.
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 325

1.2. Simple Biomechanical Models of Walking

Walking is a cyclical activity where one of the important features is energy

conservation. In the following section, we will derive a simple biomechanical model of
bipedal walking that accounts for, and explains, all the major characteristics of bipedal
walking.

0% Stance phase 60% Swing phase 100%

Double stance

Figure 1. The gait cycle is divided into a stance phase, when the leg is in contact with the ground, and a
swing phase, when the leg is in the air. Within the stance phase there are periods of double stance when both
legs are on the ground.

Figure 2 shows a simple pendulum composed of a mass particle and a light,

weightless rod. When the mass particle is displaced from the vertical position, work is
done that increases the potential energy of the system. When the pendulum is released
it starts moving along the circular arc. The level of potential energy is decreasing while
the kinetic energy is increasing. When the mass particle passes the vertical position, all
of the initial potential energy is transformed into kinetic energy. Later, when the
pendulum continues moving along the arc, the velocity is decreasing as well as the
level of kinetic energy, while the level of potential energy is increasing. If no
dissipative force, such as friction, is present in the system, the swinging of the
pendulum would continue indefinitely in accordance with Newton’s First Law of
Motion. In the ideal case presented, the level of mechanical energy, being a sum of
potential and kinetic energies, remains constant at every time instant.
Figure 3 shows a simple inverted pendulum, where similar energetic
considerations are applicable as in the case of the pendulum in Figure 2. Here, the mass
particle initially possesses a certain velocity and, consequently, kinetic energy. When
rotating upwards towards the vertical position, the level of kinetic energy is decreasing,
due to a deceleration action of the gravitational force, while at the same time the level
of potential energy is increasing, thus conserving the total mechanical energy of the
326 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

system. When the vertical position is reached, only a minimal horizontal velocity
needed for passing the vertical position is present. Afterwards, when the gravitational
force accelerates the mass particle along the arc, the potential energy level is decreasing
while the kinetic energy level is increasing. The inverted pendulum model from Figure
3 can be considered as the simplest model of bipedal walking during the stance phase.
The model is completed by adding another weightless rod articulated with the first one
through a simple hinge joint at the location of the mass particle which constitutes a hip
joint.

r
Wp = m ⋅ g ⋅ h
r2
m⋅ v
Wp WK =
m 2
h r
v
WK W P + W K = const .

Figure 2. Qualitative model of a simple pendulum. Pendulum movement is illustrated with a sequence of
three positions. When the pendulum is displaced from equilibrium it possesses the largest value of potential
energy WK , velocity of movement equals zero. When passing vertical position the pendulum possesses the
r
largest velocity v WK , which equals potential energy at the initial displaced
and thus also kinetic energy
r
position. (m = mass of pendulum bob; g = acceleration due to gravity, v indicates that velocity, v, is a
vector quantity, as it has both magnitude and direction)

Figure 4 shows such a bipedal model, where one leg is in stance, providing a
support to the mass particle on the top, while the other leg is in swing, in preparation
for the next stance phase of the gait cycle. In the preceding treatment of bipedal
walking we have considered only the gravitational force, which is considered an
external force. Another external force that acts upon the inverted pendulum is a ground
reaction force (GRF), that acts between the tip of the leg and the ground. In the simple
inverted pendulum model, the GRF is entirely passive and acts along the leg
throughout the circular motion. While the model from Figure 4 is conceptually clear, it
misses a very important aspect of bipedal walking - the impact of the swinging leg and
the ground at the moment of initial contact.
Figure 5 shows the situation at the moment of impact. In this passive inverted
pendulum model, the impact time is very short and can be regarded as instantaneous.
r
Just before impact, the mass particle possesses certain velocity v− that is perpendicular
with respect to the stance leg. When the leading leg impacts the ground a GRF L ,
which is directed along the leg, acts on the leg contacting the ground. In the short time
of impact the force impulse, which equals the product of the magnitude of GRF L and
the time duration of the impact, acts in such a way that the direction as well as the
amplitude of the velocity of the mass particle is changed. This change is depicted by a
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 327

r
vector composition, which results in a new velocity of the mass particle v+ , which is
directed perpendicularly with respect to the leg that just entered stance in order to be
able to perform circular motion.

WP
WK

r
v

Figure 3. Qualitative model of an inverted pendulum. Movement of the inverted pendulum is illustrated with
a sequence of five positions with indicated velocity vectors. Squares denote a share of potential and kinetic
energy in each of five positions. Note that the sum of both energies remains constant in each case.

Figure 4. The simplest model of bipedal walking presented by two inverted pendulums articulated with a
simple rotational joint.

Due to the geometry of both legs in double support phase, the consequence of the
force impulse of the GRF L at impact always decreases the total mechanical energy of
the system. The change in the mechanical energy before and after the impact is
proportional to the squared change in the velocities before and after the impact. The
lost energy is transformed into a sound that we hear at the time of the impact and the
deformation (which is so small that can not be noticed with a naked eye) of the leg that
impacted the ground. The situation displayed on Figure 5 demonstrate one of the most
important features of the bipedal walking; in every step there is a loss of mechanical
energy due to the impact of the swinging leg with the ground. We can easily imagine
that due to this loss of energy the simple mechanism would not be able to make another
328 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

step, because the kinetic energy of the system after the impact would be insufficient to
reach the vertical position. It is clear that lost energy needs to be recovered within each
step; this means that another important feature of bipedal walking is that in every step,
work needs to be performed to recover the level of mechanical energy that is necessary
for alternating circular motion of both legs.

r
v+ ΔW−
r
v−

GRF L

Figure 5. Energetics in double stance. A force impulse of GRF L changes the direction of movement of
mass particle m as well as reducing kinetic energy.

One possibility for efficient generation of mechanical energy in bipedal walking is

shown in Figure 6. The geometry of the mechanism is identical to that in Figure 5.
However, just before the impact of the swinging leg with the ground, an impulse of the
GRFT under the trailing leg is performed. Such an impulse could, for example, be
achieved by releasing a pre-compressed spring (ideally, compressed through a prior
impact, thus storing energy rather than dissipating it). Releasing the elastic energy from
the spring would change the direction and the magnitude of the velocity of the mass
particle in such a way that the mechanical energy of the system would be increased due
to the increased magnitude of the velocity of the mass particle. Consequently, the
impulse of the GRF L of the leading leg impacting the ground would further change
the direction and the magnitude of the mass particle velocity, similar to the case
described in Figure 5. The overall effect of both force impulses would be such that the
velocity of the mass particle before and after the impact would be exactly the same in
magnitude while the direction after the impact would be perpendicular allowing for a
passive rotation of the stance leg as an inverted pendulum as shown in Figure 3. Also,
when comparing Figures 5 and 6, we can notice that a smaller amount of energy is
dissipated when a push-off impulse preceded the impact.
A more realistic model of the double support phase is obtained if both legs are not
rigid, but can retract and extend, i.e. change their length. In this way, changes in the
energetics of walking need not be instantaneous, which is associated with high forces,
but can be achieved in a more continuous manner. This is illustrated in Figure 7 where
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 329

the three snapshots are used to show the gradual changes in the direction of the velocity
of the mass particle throughout the double support phase. We can see that, instead of
the rather abrupt change in the trajectory of the mass particle as shown in Figure 4, a
more fluid and continuous trajectory curve is achieved when simultaneous extension of
the trailing leg and retraction of the leading leg are utilized. Also, the changes in the
mechanical energy are not instantaneous but depend on i) the absorption of power by
the leading leg ( PL ), which in every time instant equals the vector dot product of the
r
GRF L and the v and ii) the generation of power by the trailing leg ( PT ), which in
r
every time instant equals to the vector dot product of the GRFT and the v . The time
integral of both powers relate to changes in the mechanical energy of the system.

r r
v+ v+ ΔW−

r
v− r ΔW+
v−

GRFT GRFL

Figure 6. Active push-off is achieved by a force impulse of the trailing leg just before the impact of the
leading leg with the ground. The overall effect of both force impulses is an appropriate change in the
direction of velocity of the mass particle.

To summarize the above derivation of the simple bipedal walking model, the
essential components are the two legs that can change their length, a mass particle
representing the mass of the whole body and the two external forces; the gravitational
force acting directly on the mass particle and the GRF acting on the mass particle
through both legs. Most of the movement in the steady state is passive while energy
absorption is performed by the leading leg following impact and energy generation is
performed by the trailing leg just prior to entering the double support phase.
While this conceptual simple model of bipedal walking captures all the essential
features of bipedal walking, it is only a first step toward understanding human walking,
since our legs are not simple rods that extend and retract; rather, our legs are made of
segments (pelvis, thighs, shanks and feet) that are articulated with joints (hips, knees
and ankles). It is therefore necessary to extend our treatment to human-like
biomechanical model in order to fully comprehend the mechanisms of bipedal walking.
However, a proper and rigorous treatment of the biomechanics of such a model from
the aspects of mathematics and physics is rather challenging and requires detailed
knowledge of several engineering disciplines. Still, we can make certain simplifications
330 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

that will allow us a simplified treatment leading to useful results that relate to human
walking.
r
v

GRFT GRFL

= r
PT GRFT ⋅ v
= r
PL GRFL ⋅ v

Figure 7. Realistic presentation of a double stance, where both ground reaction forces act in a way that
gradually changes the direction of velocity of the mass particle. The leading leg is retracting and thus
absorbing power and energy, while the trailing leg is extending and thus generating power and energy.

To summarize the above derivation of the simple bipedal walking model, the
essential components are the two legs that can change their length, a mass particle
representing the mass of the whole body and the two external forces; the gravitational
force acting directly on the mass particle and the GRF acting on the mass particle
through both legs. Most of the movement in the steady state is passive while energy
absorption is performed by the leading leg following impact and energy generation is
performed by the trailing leg just prior to entering the double support phase.
While this conceptual simple model of bipedal walking captures all the essential
features of bipedal walking, it is only a first step toward understanding human walking,
since our legs are not simple rods that extend and retract; rather, our legs are made of
segments (pelvis, thighs, shanks and feet) that are articulated with joints (hips, knees
and ankles). It is therefore necessary to extend our treatment to human-like
biomechanical model in order to fully comprehend the mechanisms of bipedal walking.
However, a proper and rigorous treatment of the biomechanics of such a model from
the aspects of mathematics and physics is rather challenging and requires detailed
knowledge of several engineering disciplines. Still, we can make certain simplifications
that will allow us a simplified treatment leading to useful results that relate to human
walking.
Figure 8 shows a model that resembles human skeletal system in a sequence of
five snapshots that illustrate the conditions throughout the stance phase from initial
contact until push-off. The leg segments are weightless as was the case in our previous
models, while the mass particle is replaced by a large ellipsoid representing the head,
arms and trunk. In the middle of this ellipsoid is a small circle that represents a centre
of mass (COM). Since the legs have no weight, the COM remains in the same relative
position within the ellipsoid. We shall first observe the sequence of humanoids from
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 331

Figure 8 in the same way as was presented in previous figures. We need to concentrate
only on the COM, GRF L and a “virtual” leg, which is represented by the dashed line
connecting the point on the ground where GRF L acts on the leg; this point is called the
centre of pressure (COP). We can see that there is considerable resemblance with a
simple inverted pendulum.

r
r r MH r r
MH MH MH MH
r r
MK MK
r r r
MA MA MA

r r
MA MA
GRF L GRF L
GRF L GRF L GRF L

a.) b.) c.) d.) e.)

Figure 8. Human-like simple biomechanical model of walking in stance phase. Depending on the external
moment in each joint produced by the GRF, the necessary balancing muscular moments are shown in the
r r r
ankle ( M A ), the knee ( M K ) and the hip ( M H ) in all sub phases of stance: a.) initial contact, b.) loading
response, c.) midstance, d.) push-off, e.) pre-swing.

The GRF always acts in the direction of the COM, while the ‘virtual’ leg rotates
similar to the inverted pendulum in Figure 3. The length of the ’virtual’ leg depends on
the position of the thighs, shanks and feet of both legs and changes throughout the
whole stance phase. In energetically efficient walking, these changes are such that the
oscillations of the COM in the vertical direction are minimal. So far, the forces that we
associated with bipedal walking were the external forces (gravity and GRF). Here, we
need to introduce also the so-called internal forces that are produced by the leg muscles
and result in equivalent net joint torques. The relationship between the external and
internal forces in our model is rather simple. In each of the lower extremity joints, the
net joint torques must equal to the product of the magnitude of GRF L and the
perpendicular distance from the line of action of the GRF L and each individual joint.
For example, at the initial contact, the GRF L acts posterior to the ankle joint, thus
producing the moment around the ankle that tends to plantarflex the foot onto the
ground. Thereby, ankle dorsiflexors must be active to produce a net joint torque that
nearly balances the external ankle moment produced by the GRF L . Similarly, in the
332 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

hip the GRF L passes in front of the hip joint thus producing external moment that
tends to flex the hip. Therefore, hip extensors must produce a net joint torque that
nearly balances the external hip moment produced by the GRF L . In the knee joint no
net joint torque is needed from either knee extensors or flexors, as the GRF L passes
through the joint, thereby producing no external moment. In the subsequent phases of
stance, different demands arise for each joint muscles, depending on the requirements
imposed by the GRF L . Here we need to stress that the only external force that is
independent is gravitational force, while the GRF L depends on gravity, mechanism
dynamics and internal forces and moments produced by the muscles and other soft
tissue. Through loading response, the greatest demand is on the knee extensors that act
eccentrically (producing tension while being lengthened), thereby absorbing the energy
due to the impact of the leading leg with the ground. Also, considerable activity is
needed from the gluteal muscles to maintain the trunk erect. In midstance, there are
rather small net joint torques needed, as the ‘inverted’ pendulum passively rotates
around the ankle joint and the GRF L passes in the vicinity of all three joints. During
push-off, a strong concentric activity (producing tension while shortening) of
plantarflexors is needed (to extend the ‘virtual’ leg), thereby generating the power and
energy needed to compensate for the energy loss that will follow during the impact of
the other leg. Here also, activity of hip flexors is needed to counteract the external
GRF L moment. During pre-swing, the activity of hip flexors, knee extensors and
plantarflexors are needed due to the orientation of the GRF L
The double support phase is illustrated in Figure 9. The left side of Figure 9 shows
both ‘virtual’ legs, both GRFs and all the moments in hips, knees and ankles of both
extremities. The right side of Figure 9, similar to Figure 7, shows the evolution of the
COM trajectory that results from shortening of the leading leg (flexion of all three
joints) and the trailing leg extending (extension of all three joints). The relationships
for power absorption ( PL ) in the leading leg and power generation in the trailing leg
( PT ) in the model from Figure 7 (simple rod shortening and lengthening) and in the
model from Figure 9 (‘virtual’ leg shortening and lengthening resulting from changes
in the angular positions of the leg segments) are given in the following equations:
r r r r r r r
PT = GRF T ⋅ v = M HT ⋅ ω HT + M KT ⋅ ω KT + M AT ⋅ ω AT (1)

r r r r r r r
PL = GRF L ⋅ v = M HL ⋅ ω HL + M KL ⋅ ω KL + M AL ⋅ ω AL (2)

The generated power/absorbed power in each joint depends on the net joint torques
produced by the muscles and the angular velocity. If the direction of net joint torque
and angular velocity is the same as is the case in Figure 9 for the trailing leg, the power
and energy is generated in a particular joint, if the two directions are opposite as is the
case in Figure 9 for the leading leg the power and energy is absorbed in a particular
joint. Therefore, while the requirements that the leading leg is absorbing power and
energy during loading while the trailing leg is generating power and energy during
push-off is an inherent, non-negotiable feature of bipedal walking, this can be achieved
in many different ways (with different joints contributing differently depending, for
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 333

example, on pathological conditions) resulting in different kinematic and kinetic

patterns.

r
ω HT r r r
M HL ω HL
r M HT
ω KT r r
r ω KL
M KT M KL
r
ω AT r r
r ω AL
M AT M AL

GRF L
GRF T GRF T GRF L

a.) b.)
Figure 9. Human-like simple biomechanical model of walking in double stance phase. a.) Depending on the

external moment at each joint produced by the ground reaction forces ( GRF L , GRF T ), also the
r r
necessary balancing muscular moments and angular velocities are shown in the ankle ( M AT , M AL and
v v r r v v r r v v
ω AT , ω AL ), the knee ( M KT , M KL and ω KT , ω KL ) and the hip ( M HT , M HL and ω HT , ω HL ).
b.) Realistic presentation of double stance where both ground reaction forces act in a way that gradually
changes the direction of velocity of the mass particle.

The swing phase, which completes our theoretical treatment of bipedal walking, is
depicted in Figure 10. For clarity, only the swinging leg is shown. Immediately after
the leg leaves the ground (initial swing) net joint torques are needed in the hip (flexor
moment) to propel the leg forward, while knee extensors need to eccentrically arrest the
movement of the knee in the flexion that resulted from preceding push-off. During
mid-swing, almost no muscular activity is needed as the leg segments are moving
balistically. Terminal swing is characterized with eccentric activity of hip extensors and
knee flexors to decelerate hip flexion and knee extension in order to prepare the leg for
the next stance phase.

1.3. Normal Walking in Humans

In the previous section, we derived a conceptual biomechnical model of bipedal

walking that was based on certain simplifications. In order to validate this model, we
need to compare it with real measurements of walking.
334 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

r r r r r
M H ωH ωH MH ωH
r r
ωK r ωK
r r MK
MK ωK
r r
MA MA

a.) b.) c.)

Figure 10. Human-like simple biomechanical model of walking in swing phase.
Muscular moments and angular velocities are shown for each joint in the a.) initial
swing, b.) mid-swing, c.) terminal swing.

Figure 11, which is similar to Figure 8, shows five snapshots of stance phase of
walking in a neurologically and orthopaedically intact individual at a velocity of 1 m/s.
Figure 11 shows the displacements of body segments and the magnitude and direction
of GRF along the dashed line pointing into the COM. Figure 12 shows the horizontal
and vertical components of GRF throughout the gait cycle. Consistent with our simple
biomechanical model, we can notice that the first 30% of the stance phase is
characterized by a negative value of horizontal GRF component (braking the
movement of an inverted pendulum) while the remaining 30% is characterized by a
positive value of horizontal GRF (propelling the movement of the inverted pendulum
forward in the direction of progression). The first 20% of stance is characterized by a
first peak value of vertical GRF component that exceeds the gravitational force, which
is associated with the deceleration of the downward movement of COM and its
redirection into an upward movement, as predicted by our simple energetic model of
the double support phase. In the next 20% of stance, the inverted pendulum is almost
passively rotating around the ankle joint, therefore the magnitude of vertical GRF
component is lower than the gravitational force (body weight). In the last 20% of the
stance the vertical GRF component again exceeds the gravitational force which is
associated with strong forward and upward propulsion (leg is extending and generating
positive power).

Figure 11. Representation of pelvis and lower extremities movement together with a GRF in five
consecutive snapshots within the stance phase in normal walking.
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 335

Figure 12. Graphs of horizontal component of GRF ( Rx ) and vertical component of GRF ( Rz )
throughout the gait cycle. Stars above the curves denote time instants that correspond to the snapshots from
Fig. 11. Mean values and standard deviations are for normal walking (adapted from Winter, 1991). Both
periods of double stance are also shown.

Figure 13 shows the joint kinematics, kinetics and energetics throughout the whole
gait cycle. Consistent with the simple biomechanical model, we can observe that the
ankle undergoes dorsiflexion (rotation of the inverted pendulum) throughout the
majority of stance, ankle plantarflexion net joint torque slowly increases due to the
requirements imposed by the GRF. During the loading response, the knee performs
controlled movement into flexion and then returns to a neutral position, which is
associated with leg shortening. This shortening is opposed by a strong knee extensor
moment, resulting in considerable power absorption, which agrees with our simple
biomechanical model. The hip undergoes a movement from flexed position at the initial
contact into extension to facilitate the vertical position of the trunk. This is
accompanied by a noticeable hip extensor moment required by the GRF passing the hip
anteriorly. In this period there is a small burst of positive power generated in the hip
that propels the COM forward. At the end of the stance phase, within the second double
support phase the ankle goes rapidly into plantarflexion, which generates considerable
positive power and energy as predicted by our simple biomechanical model. With a
slight delay the power generation in the ankle is accompanied by a considerable power
generation in the hip, which propels the leg into swing phase.
Figure 14 of the recorded EMG activity in the major lower limb muscles, shows
timing and intensity of activity that corresponds well with the magnitude of joint
moments throughout the gait cycle.
The comparison of the simple biomechanical model of bipedal walking and the
measurement of human walking shows close resemblance. The magnitude of joint
moments closely relates to the external moments generated around each joint by the
GRF, while power generation and absorption just before and during the double support
phase corresponds to energetic considerations of the inverted pendulum model.
336 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

Figure 13. Graphs of joint goniograms, muscular moments and powers in the ankle, knee and hip joints
throughout the the gait cycle of normal walking. Negative values in goniograms represent extension; negative
values of moments represent flexor muscular moments; negative values of power represent absorption of
energy. Stars above the curves denote time instants that correspond to the snapshots from Fig. 11. Mean
values and standard deviations for normal walking were adapted from Winter (1991). Both periods of double
stance are also shown.

Figure 14. Graphs of electromyograms from the selected muscles of lower extremity throughout the gait
cycle. Mean values and standard deviations are for normal walking were adapted from Winter, 1991. Both
periods of double stance are also shown.
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 337

2. Orthotics

In pathological cases where normal walking function is either severely impaired or

impossible due to diseases, injuries or other pathological changes (such as amputations
of a part of the lower limb) of a neurological or musculo-skeletal nature, various
orthoses and prostheses can be effectively used to regain adequate functional abilities.
Let us consider the stance phase of walking in a child with a cerebral palsy that
resulted in pathological changes of the musculo-tendon structures, predominantly
around the knee joint (Figure 15). We can see that the child is performing the initial
contact with foot flat, rather than with the heel. This is primarily because the knee is
forced to a flexed position due to muscle contracture, which occurs due to imbalance
between the spastic hamstring muscles and the weakened knee extensors. The loading
response is also very different from normal walking. The knee remains in a pronounced
flexed position; therefore the child uses a considerable ankle joint moment to shift the
centre of pressure (COP) where the GRF acts. By doing that, the GRF passes through
the knee joint, thus minimizing the requirements for the weakened knee extensors.
However, we can notice that this results in a considerable external hip flexor moment
that needs to be balanced by increased output from the hip extensors that are also rather
weak in such pathology. Therefore, this child is doing his best to find a feasible way to
control the GRF and COP in such a way that enables him to walk. However, such
walking is difficult; it requires considerably more energy, because throughout the
whole of stance, the muscles need to produce considerable joint moments. In this
situation one can think of applying a suitable knee orthosis that would externally and
passively generate a knee extensor moment having two orthotic effects: 1.) the orthosis
would stretch the knee flexor muscles that are spastic and in contracture, which brings
a beneficial therapeutic effect and 2.) during the loading response, the orthosis would
be able to provide more knee extensor moment that would reduce the requirements for
ankle plantarflexion moment and hip extensor moment. The overall effect would be a
more normal walking pattern. The knee orthosis could be a mechanical knee splint with
an artificial joint approximately aligned with the physiological point of rotation in the
knee and preloaded with a suitable spring.

Figure 15. Presentation of pelvis and lower extremities movement together with a GRF in five consecutive
snapshots within the stance phase of walking in a child with cerebral palsy.
338 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

By providing the above example, we can state the primary function of an orthosis
is to change the biomechanics of the lower extremity. Application of an orthosis can
have the immediate effect of substituting for the lost muscle function (in our case the
weakened knee extensors). Also, an orthosis can bring about improvement indirectly by
providing an opportunity to enhance or acquire motor skills. In our example, this means
that the child whose walking pattern is enhanced could make initial contact on the heel,
followed by the gradual weight transfer over the entire sole of the foot, which is not the
case in his regular walking. Walking on tiptoes is rather unstable, therefore by enabling
a more normal weight transfer the orthosis would also give the opportunity to the child
to practice and improve his balance skills.
When designing an optimal orthosis for a given pathological case, one first needs
to identify the primary cause for the observed changes in the walking patterns in order
to determine what kind of orthosis would be appropriate, i.e. foot orthosis, ankle-foot
orthosis, ankle orthosis etc. When this is determined, the mechanical properties of an
orthosis need to be determined; the form, resting position-angle, the stiffness. Two
major design principles need to be considered when making an orthosis. The first
principle relates to GRF control. Basically, this principle derives directly from our
simple biomechanical model of walking where we have shown that the GRF
determines the required joint moments in the hip, knee and ankle joints. Alternatively,
we can say that the joint moments developed by muscle activity shift the COP and GRF.
Therefore, when designing an orthosis we need to anticipate what kind of changes the
mechanical action of an orthosis will introduce in the kinematics and kinetics of
walking in each particular case.
The second design principle relates to a so called three-point pressure control. The
purpose of this design principle is to influence joint stability and/or mechanical
properties by providing the first point of pressure above the axis of rotation, the second
point of pressure below the axis of rotation and finally to provide the third point of
pressure that acts in the opposite direction at or near the axis of rotation. This is
illustrated in the Figure 16 where a KAFO (knee-ankle-foot orthosis) and AFO (ankle-
foot orthosis) are shown with the indicated three points of pressure in each case.
Special care must be given to the mentioned three points of pressure where the orthotic
forces are transmitted onto the skeletal system. The areas needs to be large enough and
shaped such as to enable even pressure distribution over the skin in contact with an
orthosis. The mechanical action in each of the artificial joints of an orthosis may be
different. For example, a KAFO may be designed in such a way as to lock the
movement in the ankles and the knees thus providing stability during the stance phase
to a person who is paralyzed but with some remaining function in the hips which are
used together with the crutches for propulsion during ambulation. On the other hand,
the action of a knee orthosis in the above example of crouch walking in a CP child
must be such as to provide just enough stiffness to adequately modify the COP and
GRF during walking.
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 339

Figure 16. Three-point pressure examples in KAFO (knee-ankle-foot orthosis) and AFO (ankle-foot
orthosis).

3. Prostethics

Unlike orthoses, which are generally used to augment function of the existing
biological structures, prostheses are used to completely replace, not only lost function,
but also a lost part of the biological structure – part of a limb. Although some
similarities exist between the design principles for orthoses and prostheses, there are
also major differences. Let us consider the application of an above-knee prosthesis,
which is also called a trans-femoral prosthesis. Such a prosthesis is fitted after the
amputation of the leg at the level between the hip and the knee joint. The remaining
thigh is called a stump. Figure 17 illustrates the remaining leg and a trans-femoral
prosthesis. The interface between the stump and the prosthesis is the socket, which is a
critical element of prosthesis. It needs to provide a firm contact between the stump and
prosthesis as well as even distribution of the pressure over the entire stump-socket
contact area in order to avoid damage to the skin of the stump. Below the socket there
is an artificial knee joint that can be a simple hinge joint or a more sophisticated
polycentric knee that resembles the movement in a human knee joint. Below the knee
joint there is a pylon that connects the ankle-foot complex with the knee.
Figure 18 shows the swing phase of walking with a trans-femoral prosthesis. Since
the knee and the ankle-foot complex are passive, the user of a trans-femoral prosthesis
needs to learn to control the movement appropriately. First, the hip flexor muscles
accelerate the hip in flexion as in normal walking. However, this also initiates the
acceleration of the artificial knee joint into flexion. The artificial knee joint has certain
damping, which can be mechanical friction or viscous damping of the hydraulic fluid in
the joint, that is set-up such to assure appropriate swing time of the artificial lower leg.
Immediately after the leg is in the air, a brief activity of hip extensors is needed to
reverse the motion in the knee joint in order for the leg to be fully extended when
impacting the ground, which is vital for the stability of the limb in the following stance
phase.
340 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

pelvis

stump

hip socket

knee

pylon

foot

Figure 17. Schematic presentation of various components of a trans-femoral prosthesis – sagittal plane view.

Figure 19 shows the stance phase of walking with a trans-femoral prosthesis. At

the moment of impact, the knee becomes locked. This is achieved by the mechanical
design of the artificial knee, which incorporates a weight-dependent brake mechanism.
The foot-ankle complex must be designed in such a way as to allow for a plastic
deformation, which enables the power and energy absorption that is an integral part of
bipedal walking. The control of GRF must be such that the GRF vector always passes
in front of the artificial knee joint to guarantee stability of the artificial leg. During
mid-stance, the leg rotates around the passive ankle-foot complex where some of the
movement is enabled by a compliant ankle and most of the rotation is enabled by the
circular shape of the foot sole. During this roll-over, the foot undergoes elastic
deformation, thereby storing mechanical energy, which is released at push-off to
provide some limited propulsion power. Most of the propulsion power comes from the
hip joint, which generates positive power and energy during the first half of stance.
This is possible because, at the beginning of the stance phase, the trunk is inclined
anteriorly, which also shifts the GRF more anteriorly, thereby reducing the duration of
the deceleration period in which the horizontal force is directed opposite to the walking
direction. The biomechanics of walking with a trans-femoral prosthesis requires that a
user learns a new motor control scheme for safety and efficiency, which is due to a
fixed damping confined only to one, selected walking velocity. Computer controlled
artificial legs also exist, where damping of the hydraulic knee joint is adaptively
controlled in order to accommodate different walking speeds. Walking with a trans-
femoral prosthesis is characterized by higher energy consumption (from 30% to 70%
more than in normal walking) and higher vertical GRF (up to 30% more than in normal
walking).
 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics 341

r r
MH MH

acc knee
damping damping locked
acc

acc
acc

Figure 18. Control of a trans-femoral prosthesis during swing phase.

Walking with a trans-tibial or below-knee prosthesis poses similar gait

characteristics as with a trans-femoral prosthesis, however they are less pronounced, it
is more energy-efficient and walking appears more normal.

r r
r MH MH
MH

knee knee
locked locked

impact

plastic roll over elastic push

deformation deformation off

Figure 19. Control of a trans-femoral prosthesis during stance phase.

342 Z. Matjačić / Gait Analysis and Synthesis: Biomechanics, Orthotics, Prosthetics

References

[1] Gage, J., R., (editor), The Treatment of Gait Problems in Cerebral Palsy, Mac Keith Press, London, UK,
2004
[2] Inman, V. T., Ralson, H. J., Todd F., Human Walking, Williams and Wilkins, Baltimore, USA, 1981
[3] Perry, J., Gait Analysis: Normal and Pathological Function, SLACK Incorporated, Thorofare, USA,
1992
[4] Winter, D. A., The Biomechanics and Motor Control of Human Gait: Normal, Elderly and Pathological,
University of Waterloo Press, Waterloo, Canada, 1991
[5] Zatsiorsky, V., M., Kinematics of Human Motion. Human Kinetics, Champaign, USA, 2002
[6] Zatsiorsky, V., M., Kinetics of Human Motion. Human Kinetics, Champaign, USA, 2002
[7] D. Popović and T. Sinkjaer, Control of Movement for the Physically Disabled, Springer, 2000
[8] Seymour, R., (editor), Prosthetics and Orthotics: Lower Limb and Spinal, Lippincott Williams &
Wilkins, Baltimore, USA, 2002
Basic Engineering for Medics and Biologists 343
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-343

VI.2. Basic Functional Electrical

Stimulation(FES) of Extremites
– an Engineer's View
Tadej BAJD and Marko MUNIH
Faculty of Electrical Engineering
University of Ljubljana
Slovenia

Abstract. The historical development of electrical stimulators producing

contraction of paralyzed muscles is briefly presented. The influence of electrical
stimulation parameters (amplitude of pulses, frequency, pulse duration, and
duration of a pulse train) is explained. Special attention is paid to the description of
the muscle recruitment curve. The phenomenon of reversed recruitment order,
resulting in fatiguing of electrically stimulated muscle, is presented. The properties
of surface electrodes (electrode size, polarity, resistance, and distance between
electrodes) are examined. The use of surface electrodes made of metal plate or
wire mesh, silicone impregnated with rubber, and conductive adhesive gel are
discussed. The design of electrical stimulator circuits is also presented.

Keywords. Electrical stimulation parameters, muscle recruitment, muscle fatigue,

electrodes, stimulators

Introduction: Brief History of FES

Medical applications of the discharge of electric fish for the treatment of headache and
arthritis were first recorded by Scribonius Largus in the second century AD. However,
the first electric devices making use of static electricity were developed in the
eighteenth century. These were based on the principle of using friction of insulators,
such as glass, to develop high voltages due to charge separation. The ninteenth century
brought a more practical stimulator with an induction coil. The so-called 'Faradic
stimulator' consisted of a mechanical interrupter in series with a battery and the primary
winding of a transformer. The output from the secondary winding of the transformer
was a series of pulses which were similar to the stimuli of the present day electrical
stimulators. The Faradic stimulator was the first device that could produce controlled
and sustained muscle contractions. The output frequency could be adjusted to easily
exceed the stimulation rate of 100 pulses per second and the output level was also
controllable. Finally, the twentieth century brought transistors, integrated circuits, and
microprocessors permitting sophisticated electronic circuits to be incorporated into
very small devices.
344 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites

1. FES Parameters

Functional electrical stimulation (FES) is a rehabilitation technology that uses electrical

currents applied to peripheral nerves. When a stimulating current is applied to the
electrodes placed on the skin overlying sensory-motor structures, an electric field is
established between two electrodes and ions will create a current in the tissue (Figure
1).

Figure 1. Electric field between a positive and negative electrode.

The ionic flow across the nerve influences transmembrane potential and can
generate an action potential. The action potential propagates along the nerve causing
contraction of a paralyzed muscle. In this way, FES provides restoration of movement
or function, such as standing or walking by a person with a spinal cord injury.
FES is performed in a series of rectangular monophasic or biphasic (symmetrical
or asymmetrical) electric pulses described by the following parameters: amplitude or
intensity of pulses, frequency or pulse repetition rate, duration of single pulse, and
duration of a pulse train. In most cases of surface FES applications, periodic
monophasic or unidirectional pulses are used. Biphasic or bidirectional pulses prevent a
slow deterioration of the electrodes, while the chemical conditions on the skin and in
the muscular tissue remain unchanged.

1.1. Stimulator Circuit

With respect to the stimulator output circuit, electric pulses are either voltage- or
current- controlled. Stimulators providing a constant output voltage can maintain the
voltage desired, irrespective of resistance changes in the stimulated muscle tissue.
Stimulators with current output stages make constant current pulses possible. An
important difference between the two stimulator types becomes evident in the case of
an improper contact between the electrode and the skin. In the case of a constant
current stimulator, a smaller effective electrode surface results in greater current
density, which can cause skin burns. With a constant voltage source, the resistance
 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites 345

increases due to insufficient contact, which results in a decrease of current and,

consequently, of the muscle response, but causing no skin damage.

1.2. Muscle Recruitment

A muscle recruitment curve represents the dependence of isometric (measurement

performed at a constant muscle length) joint torque upon the FES amplitude or pulse
duration (Figure 2). The joint torque is not linearly dependent upon the stimulation
intensity. Two nonlinearities occur - threshold and saturation. The increase in joint
torque due to an increasing amplitude of electrical stimulation occurs as a result of
activating new fibers in a nerve bundle laying in an electric field between the electrodes.
The main reason why all nerve fibers do not react to the same stimulation amplitude is
found in the differences in the stimulation threshold and various distances from the
stimulation electrodes. First, the fibers closest to the electrodes are stimulated. In
addition, the fibers with a greater diameter respond earlier. Beyond a certain
stimulation intensity, the force of contraction no longer increases. At such a stimulation
amplitude, all nerve fibers are excited, and a further increase of the stimulus does not
increase contraction. In surface FES of knee extensors for example, the values of the
stimulation threshold range between 20 and 60 V, while the saturation value is between
100 and 150 V.

Figure 2. Muscle recruitment curves assessed in 10 paraplegic subjects after a FES restrenghtening
programme.
346 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites

A single stimulation pulse provokes only a short-lived muscle twitch of no more

than 0.2 s. If electrical stimuli are repeated every second, a twitch occurs every second,
between which the muscle relaxes. If the frequency of stimulation pulses increases up
to 10 pulses per second (10Hz), between two twitches there is no time left for muscle
relaxation. When measuring isometric contraction, we get twitching responses. This
twitching is considerably reduced at stimulation frequencies between 15 and 20 Hz
(Figure 3). At higher frequencies, the response is already smooth: this is known as
tetanic contraction. The frequency at which the tetanic contraction occurs is called
fusion frequency. It is not the same for all muscles and depends on properties of muscle
fibers. Changes in stimulation frequencies also affect the intensity of the response. As
regards the response intensity, slight losses are observed at lower stimulation
frequencies. On the whole, the changing of frequency between 40 and 100 Hz causes
small differences in the isometric torque as measured in the joint.

Figure 3. Influence of stimulation frequency on muscle response.

1.3. Muscle Fatigue

A low stimulation frequency results in less pronounced fatigue of the neuromuscular

system. An electrically stimulated muscle fatigues more quickly than in the case of
voluntary contraction. The main reason is the reversed recruitment order. It is not too
serious an oversimplification to think of human muscle as a combination of slow fibers,
capable of sustaining low levels of contractile activity without fatigue for prolonged
periods, and fast fibers, capable of developing large forces, but fatiguing so rapidly that
they can be used only in intermittent activities. Some muscles are predominantly made
up of slow fibers, some predominantly of fast, and some of a given mixture of the two.
All muscle fibers innervated by the same motoneuron have been found to be of the
same type. Motoneurons innervating predominatly slow muscles have axons of small
 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites 347

diameter, and those supplying fast muscles have axons of a larger diameter. In a
voluntary contraction of a normally innervated muscle, the slow fibers are recruited
first, and as increased muscle force is required, the fast fibers are recruited. Slow fibers
are, therefore, activated frequently, while fast fibers are employed only infrequently,
during a burst of intense activity. When applying electrical stimulation, fibers with a
greater diameter respond earlier. This are motoneurons innervating fast muscles. The
normal order of recruitment is, therefore, reversed resulting in an increased fatiguing of
electrically stimulated muscle. In addition, by electrical stimulation, the same nerve
fibers are stimulated all the time, whereas with a healthy muscle the work is divided
among different motor units of the same muscle. Also, due to relatively high
stimulation frequency, the transmitter in the neuromuscular junctions is being
exhausted, so the muscle stimulated soon shows signs of fatigue.

1.4. Duration of Stimulus

In a fashion similar to amplitude, pulse duration also exerts a direct effect upon the
intensity of contraction. Here again, this is determined by threshold value and response
saturation. When applying surface stimulation electrodes, the accompanying unpleasant
sensation (when preserved) or even skin damage is due mainly to an excessively long
duration of a stimulus, therefore short durations are used (0.1 to 0.5 ms), while the
force of a paralyzed extremity is controlled by increasing the stimulus amplitude.
Changing the pulse duration has little or no effect on stimulated muscle fatigue.
Functional movement of a paralyzed extremity cannot be obtained by a single
electric stimulus, but requires a series of stimuli of a certain duration, following one
another at an appropriate frequency. Such a series of stimuli is called a stimulation
pulse train. In FES training of atrophied muscles, a stimulation pulse train is followed
by a pause, and then by another stimulation train. The relationship of train duration and
pause is often called the duty cycle and exerts an influence upon the fatigue of a
stimulated muscle.

2. FES Electrodes

2.1. Anode & Cathode

A surface stimulation electrode is a terminal through which electrical current passes

into the underlying tissue. At the electrode-tissue interface, a conversion occurs
between the current of electrons driven through the wires coupled to the stimulator and
the current of ions in the tissue. An electrode is usually made of metal. However, it
may be made of a nonmetal, commonly carbon. The electrode through which current
passes from the metallic or nonmetallic conductor to the tissue is called the anode and
that through which current passes from the tissue to the conductor is the cathode. In
electrical circuits, the current flows from the terminal at higher electrical potential to
the terminal at lower electrical potential. In this way the anode is the positive electrode
and the cathode the negative electrode.
348 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites

2.2. Unipolar & Bipolar

Besides distinguishing between positive and negative electrodes, we also speak about
unipolar and bipolar electrical stimulation techniques. With unipolar stimulation, one
electrode is often considerably smaller than the other, whereas the electrodes used in
bipolar stimulation both have the same size. In unipolar stimulation the smaller
electrode is negative and is also called an active electrode due to the fact that in its
vicinity there occurs depolarization of the membrane of nerve fibers. In motor nerve
stimulation, the active electrode is positioned as closely to the motor point of the
muscle as possible. The motor point is a site on the skin, where the amplitude of the
stimulus required to fully activate the muscle is at a minimum, and where all of the
motor nerve fibers are closest to the stimulating electrode. In multichannel electrical
stimulation systems, it is possible to have a single anode and several independent
cathodes or to have anodes and cathodes that are galvanically separated.
Let us examine four properties of surface stimulation electrodes and electrodes
positioning, which influence the effectiveness of electrical stimulation: electrode size,
polarity of electrodes, resistance, and distance between the electrodes.
(i) Electrode size: Electrical stimulation is applied to a nerve fiber, since muscle
fibers have a considerably higher stimulation threshold. Thus we can say that larger
electrodes are used to stimulate the nerve endings spreading all over the underlying
tissue, whereas smaller electrodes are applied to influence the nerve when the latter
come closer to the skin. Using larger electrodes, stronger contraction is obtained along
with a reduced current density and a less pronounced unpleasant sensation on the skin.
However, large electrodes permit no selective choice of a desired movement of the
stimulated paralyzed extremity. The active areas of electrodes range between 2 cm2 and
50 cm2. Electrodes of 2 cm2 to 4 cm2 are used to stimulate the nerves near the surface,
those of about 8 cm2 for the stimulation of smaller muscles, while electrodes of 25 cm2
or more are used in case of larger muscles.
(ii) Polarity:A positive and a negative electrode are placed along the muscle to be
stimulated. Considering their polarity, the electrodes are positioned so as to provoke an
optimal movement from the functional point of view. Stronger movement is usually
obtained by placing the positive electrode distally.
(iii) Resistance:It is desirable that the resistance should be as low as possible in
order to avoid energy losses before the stimulation has reached the neuromuscular
tissue. The impedance between the electrode and the skin is frequency dependent. The
DC (or low frequency) impedance tends to be several orders of magnitude larger than
the impedance at higher frequencies. Nominal values of 2 kȍ are encountered. Contact
conduction is increased by moistening the electrodes with water or special conductive
electrode gels. Adipose tissue offers high resistance to electrical currents and so higher
stimulation amplitudes must be used, causing pain in the skin as a side effect. Bones,
too, are very bad conductors of electric current; electrical stimulation cannot reach
muscles which are behind them.
(iv) Distance between electrodes:The greatest current density appears at the skin-
electrode contact and tends to decrease with distance from the electrodes as the flow
spreads out over a larger area. Closely spaced, small electrodes generally make the
effective area of stimulation rather superficial due to the lower impedance of the
current path through proximal tissue. Deeper tissues will be stimulated by using a
greater distance between the electrodes. Increasing the electrode separation leads in
general to an increase of the maximal achievable force. If the skin between the
 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites 349

electrodes is moist, this causes the current between the electrodes to flow to the skin
which results in a burning sensation and a slight or no muscle contraction at all.

2.3. Electrode Design Criteria

The design criteria for surface stimulation electrodes are: physical comfort to the skin,
electrical surface area greater than four square centimeters to prevent skin irritation, use
of hypo-allergenic materials, flexibilty to follow body surface, ease of attachement and
ability to remain in position for a duration of at least one active day, reusable, low cost,
reliable means of connection to stimulator, resistant to medical solvents and electrode
gels, low and stable electrical resistance.
The simplest surface electrodes consists of a metal plate or metal wire mesh coated
with fabric or sponge. Common materials used are stainless steel, silver-silver chloride,
platinum, or gold. For safety purposes, the upper part of the electrode is covered with a
non-conductive material. The electrode is applied after having been moistened with
water. Such electrodes are usually fixed on the extremity by means of velcro or elastic
bands. With the development of longer term surface electrodes, these electrodes are
often used for site evaluation either in pain treatment or stimulation provoking muscle
contraction. Small, button-shaped electrodes of similar design are highly suitable for
the stimulation of a single nerve. Here, the metal plate is coated with several layers of
gauze so that the electrode might retain the moisture required for as long as possible.
Surface electrodes made of silicone impregnated with carbon are applied to the
skin surface with conductive gels and held in place with adhesive patches. A too thinly
or unevenly spread gel increases current density at certain points, thereby bringing
about a danger of burns. Electrodes may be left on the skin for several days a time.
Another important property of electrodes made of conductive rubber is their flexibility
making them adaptable to any part of the body. These electrodes can be shaped by
cutting, so as to adapt them as much as possible to a proper stimulation site.
Conductive adhesive gel electrodes provide self-adhesion when applied (Figure 4).
Karaya gum was the first of these adhesive gels. Later a variety of electrically
conductive polymers were developed, enabling good contact with the irregularities of
the skin surface and a more uniform current density at the electrode-skin interface.
These electrodes can be used for extended periods with minimal skin irritation. An
important breakthrough in electrode design was made by Axelgaard. This is a flexible
electrode which adheres well to the patient's skin, is easily removed therefrom, and is
able to move with the patient's skin ensuring proper placement of the electrode. The
electrode is in the form of a knit conductive fabric. Conductive fibers include a blend of
stainless steel and polyester. The fabric can be stretched up to about 20%. A conductive
adhesive fills interstitial areas of the knit fabric and adheres the electrode to the
patient's skin. A non-conductive sheet on the other side of the knit fabric prevents
undesired electrical contacts.

2.4. Problems with Electrodes

When improperly handled, electrodes can damage the skin in the contact area. Burns
typically occur underneath the anode, but not the cathode, when using identical surface
electrodes. Another problem resides in a precise electrodes positioning along a muscle.
Sometimes a displacement of a few millimeters completely changes the muscle
response. This happens when a selected nerve (e.g. peroneal nerve) ought to be
350 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites

stimulated by surface electrodes. Surface electrodes may excite pain receptors in the
skin, although patients sensibility may be reduced to such an extent that the sensation
of pain is not critical. Another problem is undesired motion of the skin with respect to
the neuromuscular tissue. Even though an electrode seemingly occupies the same place
all the time, its distance from the nerve is not constant. This is one of the reasons why
the movements caused by electrical stimulation cannot be easily repeated. Another
limitation is that small muscles usually cannot be selectively activated and deep
muscles cannot be stimulated without first exciting the superficial muscles. Relatively
high voltages, sometimes in excess of 100V, between electrode pairs cause hazards for
the patients and the personnel that treat them. Finally, the applicability of the surface
stimulation electrodes depends on fixation problems. Stretchable garments with
electrodes already mounted in appropriate locations have been developed by several
manufacturers to simplify the application of electrodes to the skin surface. In the case
of lower limb stimulation, fixation problems can be overcome by specially designed
trousers carrying stimulation electrodes and cables. Such stimulation equipment is
comfortable and easy to handle. In the non-invasive, upper limb neuroprosthesis, the
surface stimulation electrodes were built into an elegant, self-aligning, and flexible
splint. The splint provides additional fixation of the wrist joint and allows the entire
electrode array to be positioned within a few seconds.

Figure 4. A conductive adhesive gel electrode with a portion of nonconductive sheet peeled back showing
the knit fabric.

It is not difficult to realize that most of the inconveniences of surface stimulation

electrodes can be overcome by the use of implanted electrodes. Nevertheless, because
of their simple non-invasive application, surface electrodes will remain of use in
therapeutic treatments.

3. Electrical timulators

Electrical stimulators comprise an input circuit, pulse generator, output stage, and
power supply (Figure 5). The input into the electrical stimulator is represented by a
control signal automatically switching electrical stimulation pulses on or off or may be
under the patient's voluntary control. The output electric pulse current is led, via
 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites 351

electrodes, to the selected stimulation site. The type of stimulation pulse is determined
by a pulse generator. The output pulses provided by the circuit are considerably lower
than the surface stimulation pulses required for a functional movement of an extremity
but are of an appropriate frequency and duration. The output stage ensures energy for
the electrical stimulation of paralyzed muscle in either constant current or the constant
voltage outputs. In case of current stimulation, the internal resistance of the end stage is
considerably higher than the tissue resistance between the electrodes. The current
source of stimulation pulses provides a constant current irrespective of the resistance of
the skin and the tissue between the electrodes. In the case of constant voltage output
stages, skin resistance is lower than that between the electrodes, and the stimulator
provides a constant voltage independent of the skin and tissue resistance. A power
supply provides the energy necessary for the operation of particular electronic circuits
(low voltage) and the electrical stimulation itself (high voltage). Electrical stimulators
are usually battery powered with stimulation pulses with an amplitude of more than
100 V. A high voltage for the output stage is obtained from low battery voltage by
means of a voltage convertor.
The development of electronics made it possible for Wladimir Liberson to develop
a stimulator to preventing foot-drop in hemiplegic patients. It was triggered by a heel
switch in the shoe of the affected leg. Stimulation electrodes were positioned above the
peroneal nerve in the popliteal fossa, behind the knee joint. Each time the patient raised
the heel, the heel switch triggered the stimulator, causing the nerve to cause the
extensor group of muscles to contract and dorsiflex the ankle and so lift the foot.
Advanced versions of these dropped foot stimulators were developed in Ljubljana,
Slovenia and Salisbury, UK. Both stimulators were applied to a large group of stroke
patients. In the WalkAide peroneal nerve stimulator, a tilt sensor triggers ankle
dorsiflexion.

Input Pulse Output

Circuit Generator Stage

Control Electrodes

Power
Supply

Figure 5. Block scheme of an electrical stimulator.

A minimum of four channels of FES were used for synthesis of a simple reciprocal
gait pattern in completely paralyzed paraplegic subjects. During reciprocal walking, the
stimulator must be controlled through three different phases of walking: right swing
352 T. Bajd and M. Munih / Basic Functional Electrical Stimulation (FES) of Extremites

phase, double stance phase, and left swing phase. This is achieved by two hand-push
buttons built into the handle of the walker or crutches. When niether of the push
buttons is pressed, both knee extensors are stimulated providing support to the body.
On pressing the push button in the right hand, the peroneal nerve is stimulated in the
right leg eliciting the flexion reflex presented by simultaneous hip and knee flexion and
ankle dorsiflexion. The same is true for the left leg. The Ljubljana FES walking system
consists of two small two-channel stimulators attached to each leg. Only three
electrodes are applied to a single leg in order to produce knee extension and flexion
responses. As both activities never occur simultaneously, the distal electrode placed
over the knee extensors represents the common electrode for both stimulation channels.
Using the same principles as the Ljubljana FES system, and adding two channels of
stimulation to both hip extensors, the FDA-approved Parastep surface stimulation
system was developed. A new multipurpose programmable transcutaneous electric
stimulator, Compex Motion, was developed to allow users to design various custom-
made neuroprostheses, neurological assessment devices, muscle exercise systems, and
experimental setups for physiological studies. The Compex Motion stimulator can
generate any arbitrary stimulation sequence, which can be controlled in real-time using
any external sensor.
Despite the fact that correct application of a stimulator implies no danger either for
the patient or therapist, we wish to emphasise some important safety points. When
buying a new electrical stimulator, be sure that it carries the CE mark. Every stimulator
has its own characteristics which must be indicated in the instructions of use, an
obligatory accompaniment of any commercially-available stimulator. Special attention
should be paid to stimulation parameters: pulse duration, frequency, and maximal
current or voltage. Also of utmost importance is the information on the proper use of
surface electrodes. The use of electrical stimulators might be dangereous in case of
patients with an implanted pacemaker. Simultaneous use of electrical stimulators and
high frequency surgical devices is prohibited. Electrical stimulators might not work
properly in close proximity to microwave devices. Transthoracic positioning of FES
electrodes may cause fibrillation of the heart. The stimulator should not be switched on
in case of short-circuit of the electrodes. It is true, that electrical stimulators are made
in such a way that a short-circuit of several minutes does not damage them. However,
one should not take unfair advantage of this property.

Further Reading

[1] L.A. Benton, L.L. Baker, B.R. Bowman, R. Waters, Functional electrical stimulation – A practical
guide, Rancho Los Amigos Hospital, 1980
[2] L. Vodovnik, T. Bajd, A. Kralj, F. Graþanin, and P. Strojnik, Functional electrical stimulation for
control of locomotor systems, CRC Critical Rev. Bioeng. 6 (1981), 63-131
[3] A.R. Kralj and T. Bajd, Functional electrical stimulation: Standing and walking after spinal cord
injury, CRC Press Inc., 1989
[4] L.S. Illis (ed.), Neurological Rehabilitation, Blackwell Scientific Publications, 1994
[5] D. Popoviü and T. Sinkjaer, Control of movement for the physically disabled, Springer, 2000
[6] P.J. Rosch and M. Markov (eds.), Bioelectromagnetic Medicine, Marcel Dekker Inc., 2004
[7] K.W. Horch, GS Dhillon (eds.), Neuroprosthetics – Theory and Practice, World Scientific, 2004
[8] J.D. Bronzino (ed.), The Biomedical Engineering Handbook, CRC Press and IEEE Press, 2000
[9] M. Akay (ed.), Wiley Encyclopedia of Biomedical Engineering, John Wiley&Sons, 2006
[10] L. R. Sheffler, J Chae, Neromuscular electrical stimulation in neurorehabilitation, Muscle Nerve 35
(2007), 562-590
Basic Engineering for Medics and Biologists 353
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-60750-527-3-353

VI.3. Rehabilitation Robotics

Marko MUNIH and Tadej BAJD
Faculty of Electrical Engineering
University of Ljubljana
Slovenia

Abstract. The paper presents the background, main achievements and components
of rehabilitation robotics in a simple way, using non-technical terms. The
introductory part looks at the development of robotic approaches in the
rehabilitation of neurological patients and outlines the principles of robotic device
interactions with patients. There follows a section on virtual reality in
rehabilitation. Hapticity and interaction between robot and human are presented in
order to understand the added value of robotics that cannot be exploited in other
devices. The importance of passive exercise and active tasks is then discussed
using the results of various clinical trials, followed by the place of upper and lower
extremity robotic devices in rehabilitation practice. The closing section refers to
the general importance of measurements in this area and stresses quantitative
measurements as one of the advantages in using robotic devices.

Keywords. Robot, haptic interface, virtual reality, measurement

Introduction

The application of robotic approaches in neurological patient rehabilitation was

introduced almost two decades ago [1]. Even though the number of robotic
rehabilitation systems is large, the number of clinical trials remains quite limited. In
fact, it is not yet clear what characteristics should be incorporated in a therapeutic
robotic assistant platform.
Conventional therapeutic techniques and robot assisted techniques must not be
perceived as two opposing modalities, but rather as two complementary approaches.
Two very positive aspects of robotic therapy are high repeatability and automatic
measurement during exercise. In contrast, the activities of a therapist unavoidably
include many subjective elements, but an experienced therapist has an in-depth
understanding of the individual patient which no high-tech device can ever possess. In
future, robotic therapy will complement existing clinical practice: by reducing a
therapist's workload, providing less costly and more extensive therapeutic programmes;
by using quantitative measures of an intervention or injury and last, but not least, by
new insights into the treatment process.
One of the natural common points of conventional and robotic therapy is related to
haptic interaction between a patient and therapist. This approach gains specific
importance for instance in the Bobath concept, which is based on an holistic approach
to a patient, together with the International Classification of Functioning, Disability and
Health (ICF) of the World Health Organization. At present, the emphasis is directed to
problem solving, on predefining some intermediate goals and then adjusting theraphy
354 M. Munih and T. Bajd / Rehabilitation Robotics

in gradual steps to finally learn complete movement, for example, in spasticity. This
approach highlights and relies on the plasticity of the nervous system. Solving
problems includes activities focused at a specific task, motivation, planning, interaction
with the environment, selective placement of attention by stimuli and biomechanical
aspects for effective and functional direction of movement. Thus, haptic interaction
between a patient and a therapist must not be invasive and unilateral. On the contrary, it
must evoke all the capabilities of a person leading to functional movement by
minimum intervention.
To enable the best possible imitation of natural human interaction, while using a
machine as a rehabilitator assistant, a robotic therapist must be highly consistent with
human interaction. It is no coincidence that the original ideas which brought these
systems to life did not arise in the rehabilitation field, but rather in neurophysiology
and haptic perception, including sensimotor learning [7-9]. Robotic therapist devices
differ in a number of ways from industrial robots. These are traditionally moved
between two known points along a defined trajectory. Maintaining a known and highly
accurate position is essential. In contrast, the robotic therapist must be programmable
and adjustable. This can be achieved by impedance control algorithms or by passive
manual control as in backdrivable robot manipulators including MIT-Manus and
Braccio di Ferro. Other devices designed for robotic therapy might use industrial
manipulator technology and have high manipulator arm mechanical impedance due to
high gearing ratio and admittance control scheme.
Robotic devices may be used in positional trajectory mode for enforcing passive
movements, for simple positional control of extremities, which, does not involve active
participation of a patient on either neuromuscular or sensory levels. Such passive
exercise, quite the opposite of the Bobath concept, using a robot contributes to
rehabilitation, at least in specific clinical conditions [10-12].
Other studies [13-15] indicate better results with the application of techniques that
consider the adaptive nature of the nervous system by solving problems with a suitable
adaptability level. These techniques include active assisted exercises in which the robot
moves the extremity along a predetermined trajectory. In the active constrained
exercises, the robot provides higher opposing forces. The movement may be limited
inside some 3D virtual space, with opposing forces applied when the subject tries to
move outside this region. Similarly, one could imagine a ball object that limits
movement toward the center (inside), with completely empty space all around.
Furthermore, in active resistive exercises, the robot opposes the intended movement.
In adaptive exercises, the robot is providing a previously unknown dynamic
environment to which the subjects have to react and adapt.
Passive exercises need no input from the patient, while active constrained
exercises require at least residual movement cababilities or subconscious sensory-
motor co-operation. Active resistive and adaptive exercises require cooperation with
sufficient volitional motor activity. Active resistive and adaptive exercises are therefore
not a suitable choice for persons with larger movement deficits, since they usually
cannot use them independently. Still, such persons may exploit robot-therapeutic
exercises in which minimum co-operation on their side will help them in using and
strengthening their remaining abilities.
 M. Munih and T. Bajd / Rehabilitation Robotics 355

1. Virtual Reality in Rehabilitation

Most of us are familiar with virtual reality (VR) technology from entertainment (e.g.
games) and military simulations. Lately, its use has expanded to other areas, for
instance computer-aided design (CAD), architecture, general virtual presentation of
data and to medical applications. Medical applications of virtual reality include training
in medicine and surgery, modelling of hard and soft tissue, image displays, remote
surgery, ergonomy and rehabilitation. In rehabilitation, virtual reality is used for
training and measuring the motor ability (e.g. in hemiplegia, paraplegia, Parkinson's
disease).
Virtual reality is usually understood as a three-dimensional computer model which
primarily defines the geometrical model (kinematic or dynamic model) of different
virtual objects and their environment. It is possible to define not only the image of
objects, but also their inherent physical characteristics. Such a virtual world can change
in time - not only the position and orientation of particular items, but also dynamic
characteristics of the surroundings, such as friction and gravity. Realistic visualisation
and deep presence of person in VR are mostly dependent on high-quality graphics
which take the user into the virtual world. The impression of virtual reality depends on
the realistic appearance of the scene and activity. Simpler non-immersive
visualisations include a 2D computer screen, a projection screen and 3D projection
techniques in the environment. The immersive method is more realistic, enabling the
user to percieve a full 3D visual field by use of special glasses.
The dynamic characteristics of objects, including mass, moment of inertia, forces,
torques, and torques and forces resulting from the interaction of objects and, for
instance, compliance (stiffness) of objects, roughness and smoothness of objects cannot
be visually detected or presented. These require the sense of touch.

2. Hapticity

The verb ‘ĮʌIJȦ’ - hapto is originating from Greek for reaching, holding and touching.
If a person exerts a force F onto a mass m, which is integrated in the environment via
the damper h and the spring k, the movement x is determined by the differential
equation:
§ d2 d ·
F ¨m  h  k¸x m x  h x  k x (1)
¨ dt 2 dt ¸
© ¹
If the coefficients m, h and k are constant, the force depends on the position
(distance to origin), velocity and acceleration. The first force term in the equation
depends on mass and acceleration m x , with m representing the mass of e.g. a cube on
icy surface. Significant force will be needed only during acceleration and deceleration
with high m, while constant velocity movement requires no force. Only a small force is
needed for the acceleration and deceleration of a light styrofoam cube with a small m.
The force can, as a second example, depend on the term h x , where h stands for
damping of e.g. of an oar in water. If the velocity x is low, only a small force is
needed, whereas at higher velocities, a considerably greater force is required. Another
medium, like oil or a spoon in honey or air, represents a different damping coeficient
356 M. Munih and T. Bajd / Rehabilitation Robotics

value h and thus a different force at the same velocity x . The damping force is small
when an oar moves in air, but if the velocity is high, in the case of an airplane propeller,
the pull/push force becomes significant. The impact of stiffness k is presented by
means of a spring. The greater the deviation x, the greater the force. Also, a higher
spring stiffness coeficient k value, with x unchanged, requires greater force.
A general body in a virtual environment, onto which a person exerts a force, is not
a point mass, pure damper or spring but rather a combination of all three terms in the
above equation that becomes more complex. This is typical for haptic touch in a virtual
environment, where none of the three coeficients of the equation is constant. The
values of m, h and k change locally in the virtual environment. Actually, one of the
three parameters, e.g. k, has six values at the same point (or position) in space: kx, ky
and kz along individual axes of space, and kRx, kRy and kRz around individual axes of
rotation – thus it has six degrees of freedom (DOF) (Fig. 1). An empty space has nil or
small values of all six k. A flat wall in a virtual environment is represented as a high k
horizontally. A wall can also be slippery as ice or non-slippery if coated with rubber,
which is set with the other two (transversal) k parameters of the wall. The remaining
three k elements stand for the three rotational stiffness coeficients. Similar six-
dimensional understanding applies to the other two parameters - h and m.

hZ kZ
hY

kRZ
hRZ kY
hRY
zO kRY
yO
O
xO
FO kRX
hRX
y hX

kX
x

Figure 1. Second-order mechanical system with six DOF, three translations and three rotations
and m, h and k

In rehabilitation haptic rendering is often used to denote a virtual tunnel that

connects two extreme points of movement, the starting and the final one. The trajectory
between them can run over a straight line or a curve having a different shape. All
 M. Munih and T. Bajd / Rehabilitation Robotics 357

elements m, h and k along the direction of movement equal zero, whereas,

perpendicular to the direction of movement, the stiffness coefficient k increases by
selected functions. This is reflected in the virtual pipe or tunnel that forces the user to
the central curve line, where this force component does not exist.
In a series of attempts, should the user appear at exactly the same coordinate of
space, carrying the same coefficients, the current values of x and x (current
acceleration and velocity) are very likely to differ, meaning that the force F felt by the
user in a virtual environment with a haptic interface is also different. A haptic interface
is, therefore, a robot capable of functioning in line with the above equation, contrary to
the classic understanding of robot functioning. These are usually position-controlled, as
referred to under passive exercise. Haptics in a virtual environment is achieved in real
time by setting the interaction force in the point between the user and interface,
considering all the numerous parameters mentioned above. It is understandable why
this technology has become accessible only in the last two decades.
The connection of quantities F and x represents an energy contact, in reality the
transfer of power between man and environment Fx and thus controlled transfer of
energy via man-machine interaction. In addition to the vision sensory pathway, the
user participating in haptic interaction in a virtual environment also engages and uses
touch, position, force and texture sensation in their body as well as the mechanisms of
visual and tactile recognition, along with reflex and volitional control mechanisms,
including the entire motor chain.
Further to vision and haptic modalities, further information can be supplied to the
user via sound. Particular cases would be sound produced during movement along
rough or smooth surfaces, along a ladder, or when typing on virtual keyboards.

3. Passive Exercise

CPM (Continuous Passive Motion) devices have been used in the postoperative
rehabilitation of joints since approximately 1960. It was later established that joint
immobilisation in a test group of rabbits led to great problems in restored mobility [16].
After five to six weeks of immobilisation, most joints developed moderate to extreme
changes, including those in joint cartilage and bone mass, with changes already
apparent in the second week. In clinical practice, this presented as joint contractures
and reduced range of motion (ROM). To avoid this, they tried to maintain a good ROM
by simple CPM devices which moved the limb over an arc corresponding to joint
movement. Such devices consist of a simple motor with a mechanically or
electronically adjustable range and velocity of movement. Most modern CPM devices
adopt ideal joints with a fixed point of rotation, and can provide movement in one
vertical plane (2D). Usually they cannot be reprogrammed and are controlled using the
open loop principle, which means that their movement angles and forces are not
measured immediately but are corrected and reset a few 100-times in a second.
Such CPM devices are very efficient in preserving range of movement, they reduce
stiffness in joints, decrease the need for drug administration and shorten the length of
hospitalization. Comparison of CPM and physiotherapy reveals no major differences as
regards the above parameters, however in the majority of cases, CPM results in weaker
muscles, delays in activation of extensors and stiffness of flexors [17]. It would be
358 M. Munih and T. Bajd / Rehabilitation Robotics

difficult to expect anything else, since the CPM devices only move a person's
extremities without activating the muscles.

4. Active Exercises

Several research groups developed robotic devices for rehabilitation of the upper and
lower extremities. These devices, designed for exercise, use various methods of
operation. Active constrained movements are made possible by most of the mentioned
devices, while movements with active resistance are provided by MIT-Manus, Bi-
Manu-Track and MIME [18-20]. Adaptive exercises are possible using Bi-Manu-Track
and MIME [19, 20], in which case the healthy arm leads and the injured arm imitates
movement, both moving simultaneously.

Finteraction

Virtual track in computer

Finteraction

Robot Tip and Human Arm

Figure 2. Trajectory in active constrained exercise

Figure 3. Example of environment in active constrained exercise, GENTLE/S project

 M. Munih and T. Bajd / Rehabilitation Robotics 359

Figure 4. Example of active exercise with active resistance (torus), HIFE project

Figure 5. Example of active exercise with active resistance (pipe with a spring), I-Match project

The literature often contains descriptions of technical approaches and solutions.

However, there is a limited number of clinical trials with clear goals and methodology
that would systematically present the influence of robotic devices and existing modes
of operation on the rehabilitation process. Among 17 known clinical trials, for a variety
of reasons, only a handful can be used for direct comparison. None of them
systematically investigates the influence of the method of applied exercise. Most trials
have so far employed all three modes of operation, including: passive, active
constrained and active resistive, regardless of the probability that one of these methods
360 M. Munih and T. Bajd / Rehabilitation Robotics

might be more successful than others. Only Fasoli et al. [27] and Stein et al. [28] tried
to specifically evaluate particular exercise versions. The findings of the Fasoli trial
disclose that active resistive exercises are more useful than active constrained exercises
as regards the upper extremities. Still, the repeated trial covering the same group did
not reveal any differences between the exercises involving active constrained
movements and those involving active resistive movements, perhaps because of the
manner of data calculation. In the first trial Fasoli et al. [27] included in the group of an
active constrained movement those who did not have sufficient motor function for the
active resistive group. This was not the case in the trial conducted by Stein et al. [28].
Thus, the results of active resistive exercises were overrated in advance. These findings
point to the fact that some robot-assisted therapies are less suitable for specific groups
of patients. Therefore, the precise contribution of individual exercise methods to the
rehabilitation of upper extremities in a hemiplegic person remains unexplained.
Other methods of exercise are perhaps equally or even more important for robotic
rehabilitation, but they have not been investigated sufficiently, or at all. One such
method is gravity compensation of the upper extremity. Most robotic devices provide
some sort of support for the arm. The evolution of gravity compensation devices has
been going on for decades – for example Sanches [29]. Beer et al. specifically
investigated the implementation and reach of gravity compensation for the upper
extremity [30-32]. Their preliminary research showed that poor coordination of muscle
activation led to unexpected torques in the hemiplegic arm joints (e.g. shoulder
abduction prevented the ability of elbow extension). Further trials revealed immediate
improvement of motor abilities in persons after a stroke, if gravity did not influence the
upper extremity. The active abduction of the shoulder was reduced and consequently
extension in the elbow increased in static conditions. The latest results point to a
similar mechanism functioning in dynamic circumstances. Such findings indicate
improved motor abilities due to the use of gravity compensation. The second trial
involving exercise with a sling revealed no significant difference. Even though the
principle of gravity compensation is at present rather unexploited, faster recovery can
be expected along with new trials. Recently, three arm gravity compensators were put
on the market: Armon1, Dynamics Arm Support2 (DAS) and Armeo3. The first two
devices provide lower arm support, while the third is an exoskeleton type mechanism
allowing free arm movement for people with little muscular arm strength, using passive
principles, thus providing for static and not accounting for dynamic compensation.
More research in this field can clarify the mechanisms of the compensation impact, as
well as indicate where it would be reasonable to apply these mechanisms.
Besides the above approaches, other methods of rehabilitation can additionally
promote fast recovery in hemiplegic patients. Interesting assertions were published by
Kahn et al. [35], claiming that equally positive changes in motor function were
observed in a group undergoing robot-assisted reaching exercise (target is always
reached) and in a group undergoing robot-assisted reaching exercise without obligatory
reaching of the target. Some other possible techniques deserve mention: functional
electrical stimulation, pharmacology, intensive exercise, including many repetitive
exercises, loading therapy - automated use of this method - and sensor-motor exercise.

1
www.mginside.info
2
www.exactdynamics.com
3
www.hocoma.ch
 M. Munih and T. Bajd / Rehabilitation Robotics 361

A systematic review of the effect of robot-aided therapy on recovery of the

hemiplegic arm after stroke, collecting results from eight selected studies is provided
by Prange et al. [36]. This indicates that robot-aided therapy of the proximal upper limb
can improve short- and long-term motor control of the paretic shoulder and elbow. This
statement is supported by quantitative analysis of short-term pooled data in chronic
stroke patients and indicates that increased motor recovery of chronic patients is
possible after robot-aided therapy. However, no consistent effect on improvement in
functional ability has been reported, although the training modalities were not directly
designed for this. Restoration of motor control appears greater after robot-aided therapy
than conventional tehrapy. It was not possible to establish which aspects of robot-aided
therapy (e.g. increased intensity of movements, more effective training modalities)
were responsible for the beneficial influence on recovery. Clinical experience seems to
show that robot-aided therapy can improve motor control of hemiplegic upper limbs,
perhaps even more than conventional therapy, in both sub-acute and chronic stroke
patients.

5. Devices for Robot-Aided Therapy

A number of research or commercial platforms exist today designed specifically for the
tasks of rehabilitation robotics, while in some other platforms the primary design goal
has been aimed at something else, but the resulting device also suits the needs of
rehabilitation. Regardless of their origin, the devices can be of either the exoskeleton or
the end-effector type. An exoskeleton is an external framework, close to, or in contact
with, human limbs.
Exoskeletons contain rigid and resistant components for each of their segments,
being able to providing passive or active translation or rotation in a particular joint. The
rotation axes of the robot must correspond to the rotation axes of the human skeleton,
and the limb may be connected to the exoskeleton at several points. Powered
exoskeletons may be designed, for example, to assist and protect soldiers and
construction workers, or to aid the survival of people in other dangerous environments.
Wider medical markets exist for providing mobility assistance and rehabilitation for
aged and infirm people. A weak point in exoskeleton devices is the limited amount of
Z-bandwidth and the transparency of the haptic interface – the parameters that tell how
good is the touch (impedance) compared to the ideal model in the virtual environment.
A strength is the good fit to each human body segment and the ability to move each
limb and interact with supervised known parameters.
Perry et al. [39] is using an exoskeleton robot with seven degrees of freedom in the
most important joints of human arm. L-Exos is a tendon-driven, wearable haptic
interface with 5 DoF, optimized to obtain a solution with reduced mass and high
stiffness, by employing special mechanical components and carbon fiber structural
parts [41]. Neural control of an upper limb powered exoskeleton system has 8 DoF
[44]. ARMin represents an interesting later design that at the moment enables
movements in 6 DoF [26].
In most cases, robotic haptic arms that interact with the environment only via end-
effectors are in contact with the human arm via the wrist or with the human leg via the
foot only. As a consequence, the trajectory of each body segment under motion is not
supervised and determined by a machine, as above, but also depends on the person.
This can be good, but may also lead to a less determined environment. Good points in
362 M. Munih and T. Bajd / Rehabilitation Robotics

the end-effector approach are better values for Z-bandwidth and transparency of the
haptic device.
Examples of end-effector upper extremity devices include MIT Manus [18],
Assisted Rehabilitation and Measurement (ARM) Guide [19], Mirror Image Motion
Enabler (MIME) [20], Bi-Manu-Track [21], GENTLE/S [22], Neurorehabiltation
(NeReBot) [23], REHAROB [24], Arm Coordinating Training 3-D (ACT3D) [25],
Braccio di Ferro, [42], NEDO project device [43] and Baltimore Therapeutic
Equipment Co. (BTE), produce several training devices.
Most devices have been designed for training of the proximal joints (shoulders and
elbows) of the hemiplegic arm. Some allow for two-dimensional plane movement only,
while in most cases, movement is possible in three dimensions within the limited
section of the entire working area of the arm. On the contrary, Bi-Manu-Track includes
the forearm and wrist, which is enabled also by the recent version of the MIT-Manus
device. Over and over again we encounter new robotic devices and evolutional
upgrades of the existing ones (e.g. Furusho et al. [33] and Colombo et al. [34]).
There are also devices available for the lower extremities, Gait Trainer GT I [37]
and HapticWalker [38] most frequently in combination with treadmills, and also
separately for ankles. In the lower extremities, the end-effector devices only guide foot
motion, while the multiple degrees of freedom of the rest of the body (e.g. leg, hip)
remain completely unrestricted. In the case of a patient with an unstable knee joint, the
physiotherapist has to stabilize the knee manually. Studies on the Gait trainer GT I
revealed comparable muscle activation patterns as in treadmill walking.
Research devices are also available for exercising some other joints of the body,
like ankle, wrist, individual fingers [45] or several fingers simultaneously.

Figure 6. ARMin robot is one of the recent exoskeleton devices (ETH Zürich)
 M. Munih and T. Bajd / Rehabilitation Robotics 363

Figure 7. GENTLE/S project rehabilitation environment as an example of end effector approach

6. Measurements in Rehabilitation

In neurology, standard rating scales are used for measurements, mainly for the purpose
of early systemisation. On the other hand, uniform, quantitative, traceable procedures
for measurements in rehabilitation have not yet been introduced. One of the models
supposedly establishing a logical, understandable and thorough system in rehabilitation
was adopted by the World Health Organisation in 1980. Even though the model, known
as ICIDH (International Classification of Impairments, Disabilities and Handicaps) is
very detailed and subjective (rheumatologic diseases), it provides a solid framework for
understanding and treating neurological diseases and injuries. The most important
concept of the model is the assumption that every disease can be evaluated on four
levels, namely:
o pathology level,
o impairment level,
o disability level,
o handicap level.
In practice, the border between individual levels can be quite blurred. In general, a
distinction has to be made between measurements and scoring scales. Measurement
involves the use of a specific, traceable standard and the comparison of an actual value
of a certain quantity against this standard. Scales often rely on subjective observations
or rough measurements leading to a very few quantum descriptors. In rehabilitation,
measurements are much rarer than scoring scales. The simplest scoring of motor
364 M. Munih and T. Bajd / Rehabilitation Robotics

abilities and impairment of e.g. the upper extremity includes the squeezing of the
dynamometer for estimating muscle strength and measuring the range of movement of
individual joints. Self-scoring by patients is also present. The criteria of a good scoring
scale include:
o validity (the result actually refers to the aspects for which the test was carried
out),
o reliability (competent observer and time stability of results of such observer),
o sensitivity (detection, differentiation of sufficiently small changes that are still
relevant).
Standardised rehabilitation tests are for the most part tailored to a specific disease
and consider also the condition of other parts of the body and activity in general. Some
of these tests are Fugl-Meyr, Barthel Index and the widely-used Barthel Index, Katz
index, Nottingham Adl Index, Jebsen, FIM-test, Box and Block Test, Nine-Hole Peg
Test, Action Research Arm Test, Franchay Arm Test and others.
Unlike the scoring approach, haptic, robotic technology provides for the next stage
in rehabilitation, objective measurement during exercise. Quantities that vary with time
(position and orientation, velocities, accelerations, forces, torques) can often be
measured directly or derived. The presentation of these quantities for individual,
relevant points of the body offers an objective basis for evaluation, as in kinesiology,
and derived indexes can answer significant questions. Observation of measured data in
a frequency domain can result in new insights, for instance, in the case of Parkinson's
disease, the amplitude and frequency of shaking are determined by both place and time.
The measured quantities can be used as entries into various physical or
physiological models and through them new important parameters may be acquired, e.g.
active torques of the muscles of individual joints, passive torques in joints or even mass
and evaluation of moment of inertia of some body segments. In doing so, we have to be
aware of the fact that the more complex models usually comprise a higher number of
parameters, some roughly evaluated, carrying high measurement uncertainly, which in
the end may contribute to an even more uncertain final result.

7. Conclusion

Recent technological developments have made many things possible for the first time.
However, actual systems are only now being introduced into our lives in the fields of
computing, virtual reality visualisation, medicine in general and a narrow field of
rehabilitation. Previously most simple devices, accessories and physiotherapists’ hands
will in the future be complemented by computerized devices. These will help in
existing and, hopefully, also some new aspects of rehabilitation. Robotic approaches
undoubtedly allow for equally fast, perhaps even faster, but certainly more interesting
and entertaining rehabilitation methods. It has to be expected that at least some good
laboratory prototypes will develop into widely-available products. These products are
expected to have a suitable modular design, and developments in other areas, e.g.
improvement in speed of computer processing, improvement in video techniques,
software support, etc., will further contribute to the advancement in this area.
Unfortunately, the size of the rehabilitation robotic market will be considerably smaller
than the size of the market for computer games, which is why the volume of investment
will be smaller and the pace of development is expected to be slower.
 M. Munih and T. Bajd / Rehabilitation Robotics 365

Acknowledgements

The described research work is financed by the Slovenian Research Agency in the form
of a programme group Analysis and Synthesis of Movement in Man and Machine, and
by providing grants to young researchers.

References

[1] D. Khalili, M. Zomlefer, An intelligent robotic system for rehabilitation of joints and estimation of body
segment parameters, IEEE Trans Biomed Eng 35 (1988), 138-46
[2] B. Gjelsvik, E. Bassoe, Form und Function. Stuttgart, New York: Thieme Verlag, 2002.
[3] J. Howle, Neuro-Developmental Treatment Approach: Theoretical Foundations and Principles of
Practice, Laguna Beach, CA, NDTA, 2002
[4] B. Bobath, Observations on adult hemiplegia and suggestions for treatment, Physiotherapy 45 (1959),
279-289
[5] E. Panturin, The Bobath concept, Clin. Rehabil. 15 (2001), 111-113
[6] S. Raine, Defining the Bobath concept using the Delphi technique, Physiother. Res. Int. 11 (2006), 4-13
[7] F.A. Mussa-Ivaldi, N. Hogan, E. Bizzi, Neural, mechanical, and geometric factors subserving arm
posture in humans, J. Neurosci.. 5 (1985), 2732-2743
[8] R. Shadmehr, F.A. Mussa-Ivaldi, Adaptive representation of dynamics during learning of a motor task,
J. Neurosci. 14 (1994), 3208-3224
[9] T. Tsuji, P.G. Morasso, K. Goto, K. Ito, Human hand impedance characteristics during maintained
posture, Biol. Cybern..72 (1995), 475-485
[10] P. Lum, D. Reinkensmeyer, R. Mahoney, W.Z. Rymer, C. Burgar, Robotic devices for movement
therapy after stroke: current status and challenges to clinical acceptance, Top Stroke Rehabil 8 (2002),
40-53
[11] G. Fazekas, M. Horvath, A. Toth, A novel robot training system designed to supplement upper limb
physiotherapy of patients with spastic hemiparesis, Int. J. Rehabil. Res. 29 (2006), 251-254
[12] S. Hesse, G. Schulte-Tigges, M. Konrad, A. Bardeleben, C. Werner, Robot-assisted arm trainer for the
passive and active practice of bilateral forearm and wrist movements in hemiparetic subjects, Arch.
Phys. Med. Rehabil. 84 (2003) 915-920
[13] C.D. Takahashi, D.J. Reinkensmeyer, Hemiparetic stroke impairs anticipatory control of arm movement,
Exp. Brain Res. 149 (2003), 131-140
[14] J.L. Patton, F.A. Mussa-Ivaldi, Robot-assisted adaptive training: custom force fields for teaching
movement patterns, IEEE Trans. Biomed. Eng. 51 (2004), 636-646
[15] J.L. Patton, M.E. Stoykov, M. Kovic, F.A. Mussa-Ivaldi, Evaluation of robotic training forces that
either enhance or reduce error in chronic hemiparetic stroke survivors, Exp. Brain Res. 168 (2006),
368-383
[16] R.B. Salter et al., The effects of continuous compression on living articular cartilage, J. Bone Joint Surg.
42-A (1960)
[17] M.A. Ritter, V.S. Gandolf, K.S. Holston, Continuous passive motion versus physical therapy in total
knee arthroplasty, Clin. Orthop. Relat. Res. 244 (1989), 239-243
[18] H.I. Krebs, N. Hogan, B.T. Volpe, M.L. Aisen, L. Edelstein, C. Diels, Overview of clinical trials with
MIT-MANUS: A robot-aided neuro-rehabilitation facility, Technol. Health Care 7 (1999), 419–423
[19] D.J. Reinkensmeyer, L.E. Kahn, M. Averbuch, A. McKenna-Cole, B.D. Schmit, W.Z. Rymer,
Understanding and treating arm movement impairment after chronic brain injury: progress with the
ARM guide, J. Rehabil. Res. Dev., 37 (2000), 653–662
[20] C.G. Burgar, P.S. Lum, P.C. Shor, H.F. Machiel Van der Loos, Development of robots for rehabilitation
therapy: the Palo Alto VA/Stanford experience, J. Rehabil. Res. Dev., 37 (2000), 663–673
[21] S. Hesse, G. Schulte-Tigges, M. Konrad, A. Bardeleben, C. Werner, Robot-assisted arm trainer for the
passive and active practice of bilateral forearm and wrist movements in hemiparetic subjects, Arch.
Phys. Med. Rehabil. 84 (2003), 915–920
[22] S. Coote, E. Stokes, B. Murphy, W. Harwin, The effect of GENTLE/s robot-mediated therapy on upper
extremity dysfunction post stroke, Proc. 8th International Conference on Rehabilitation Robotics; 2003,
Daejeon, Korea, Human-Friendly Welfare Robot System Engineering Research Center, 59–61, 2003
[23] C. Fanin, P. Gallina, A. Rossi, U. Zanatta, S. Masiero, NeRe-Bot: A wire-based robot for
neurorehabilitation, Proc. 8th International Conference on Rehabilitation Robotics, Daejeon, Korea,
Human-Friendly Welfare Robot System Engineering Research Center, 23–26, 2003.
366 M. Munih and T. Bajd / Rehabilitation Robotics

[24] A. Toth, G. Fazekas, G. Arz, M. Jurak, M. Horvath, Passive robotic movement therapy of the spastic
hemiparetic arm with REHAROB: Report of the first clinical test and the follow-up system improvement,
Proc. 9th International Conference on Rehabilitation Robotics, Chicago, Illinois. Madison (WI),
Omnipress, 127–130, 2005.
[25] J. Dewald, M.D. Ellis, B.G. Holubar, T. Sukal T, A.M. Acosta, The robot application in the
rehabilitation of stroke patients, Neurol. Rehabil. 4 (2004), 7
[26] T. Nef, R. Riener, ARMin—Design of a novel arm rehabilitation robot, Proc. 9th International
Conference on Rehabilitation Robotics, Chicago, Illinois. Madison (WI), Omnipress, 57–60, 2005
[27] S.E. Fasoli, H.I. Krebs, J. Stein, W.R. Frontera, N. Hogan, Effects of robotic therapy on motor
impairment and recovery in chronic stroke, Arch. Phys. Med. Rehabil. 84 (2003), 477–482
[28] J. Stein, H.I. Krebs, W.R. Frontera, S.E. Fasoli, R. Hughes, N. Hogan, Comparison of two techniques of
robot-aided upper limb exercise training after stroke, Am. J. Phys. Med. Rehabil. 83 (2003), 720–728
[29] R.J. Sanchez, E. Wolbrecht, R. Smith, J. Liu, S. Rao, S. Cramer, T. Rahman, J.E. Bobrow, D.J.
Reinkensmeyer, A pneumatic robot for re-training arm movement after stroke: rationale and
mechanical design, Proc. 9th Int. Conf. Rehabilitation Robotics, Chicago, Illinois, Madison (WI),
Omnipress, 500–504, 2005
[30] R.F. Beer, J.D. Given, J.P. Dewald, Task-dependent weakness at the elbow in patients with hemiparesis,
Arch. Phys. Med. Rehabil. 80 (1999), 766–772
[31] R.F. Beer, J.P. Dewald, W.Z. Ryme, Deficits in the coordination of multijoint arm movements in
patients with hemiparesis: evidence for disturbed control of limb dynamics, Exp. Brain Res. 131 (2000),
305–319
[32] R.F. Beer, J.P. Dewald, M.L. Dawson, W.Z. Rymer, Targetdependent differences between free and
constrained arm movements in chronic hemiparesis, Exp. Brain Res. 156 (2004), 458–470
[33] J. Furusho, K. Koyanagi, Y. Imada, Y. Fujii, K. Nakanishi, K. Domen, K. Miyakoshi, U. Ryu, S.
Takenaka, A. Inoue, A 3-D rehabilitation system for upper limbs developed in a 5-year NEDO project
and its clinical testing, Proc. 9th Int. Conf. Rehabilitation Robotics, Chicago, Illinois. Madison (WI),
Omnipress, 53–56, 2005
[34] R. Colombo, F. Pisano, S. Micera, A. Mazzone, C. Delconte, M.C. Carrozza, P. Dario, G. Minuco,
Upper limb rehabilitation and evaluation of stroke patients using robot-aided techniques. Proc. 9th Int.
Conf. Rehabilitation Robotics, Chicago, Illinois, Madison (WI), Omnipress, 515–518, 2005
[35] L.E. Kahn, M. Averbuch, W.Z. Rymer, D.J. Reinkensmeyer, Comparison of robot-assisted reaching to
free reaching in promoting recovery from chronic stroke, In: M. Mokhtari (ed.), Integration of assistive
technology in the information age, Amsterdam (the Netherlands), IOS Press, 39–44, 2001
[36] G.B. Prange, M.J.A. Jannink, C.G.M. Groothuis-Oudshoorn, H.J. Hermens, M.J. Ijzerman, Systematic
review of the effect of robot-aided therapy on recovery of the hemiplegic arm after stroke, J.
Rehabilitation Research & Development, 43 (2006), 171–184
[37] S. Hesse, D. Uhlenbrock, A mechanized gait trainer for restoration of gait, J. Rehabilitation Research &
Development, 37 (2000), 701–708
[38] H. Schmidt, D. Sorowka, F. Piorko, N. Marhoul, R. Bernhardt, Entwicklung eines robotergestützten
Laufsimulators zur Gangrehabilitation (Development of a Robotic Walking Simulator for gait
rehabilitation), Biomed. Eng, 40 (2003), 281-286
[39] J.C. Perry, J. Rosen, Design of a 7-Degree-of-freedom upper limb powered exoskeleton, BioRob 2006,
The IEEE/RAS-EMBS Int. Conf. Biomedical Robotics and Biomechatronics, Pisa, Toscany, Italy, 805-
810, 2006
[40] H. Schmidt, D. Sorowka, F. Piorko, N. Marhoul, R. Bernhardt, Entwicklung eines robotergestützten
Laufsimulators zur Gangrehabilitation (Development of a robotic walking simulator for gait
rehabilitation), Biomedical Engineering 40 (2003), 281-286
[41] A. Frisoli, F. Rocchi, S. Marcheschi, A. Dettori, F. Salsedo, M. Bergamasco, A new force-feedback arm
exoskeleton for haptic interaction in Virtual Environments, First Joint Eurohaptics Conference and
Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems WHC, 2005.
[42] M. Casadio, V. Sanguineti, P.G. Morasso, V. Arrichiello, Braccio di Ferro: A new haptic workstation
for neuromotor rehabilitation, Technol. Health Care 14 (2006), 123–142
[43] K. Koyanagi, J. Furusho, U. Ryu, A. Inoue, Development of rehabilitation system for the upper limbs in
a NEDO project, Proc. 2003 IEEE Int. Conf. Robotics & Automation, Taipei, Taiwan (2003), 14-19
[44] J.C. Perry, J. Rosen, S. Burns, Upper-limb powered exoskeleton design, IEEE Transactions on
Mechatronics 12 (2007), 408-417
[45] U. Mali, M. Munih, HIFE-haptic interface for finger exercise, IEEE/ASME Transactions on
Mechatronics 11 (2006), 93-102
Basic Engineering for Medics and Biologists 367
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.

Subject Index
bending 13 kinetics 323
Bernoulli 45 laminar flow 45
bioimpedance 69 magnetic resonance imaging 302
biopotential 90 magnetic resonance spectroscopy 302
bipedal locomotion 323 measurement 353
brittle 13 mechanics 58
cardiogram 90 medical imaging 249
color Doppler 249 membrane potentials 90
computed tomography 274 moment 3
couple 3 muscle fatigue 343
diagnostic imaging 302 muscle recruitment 343
dielectric permittivity 69 myogram 90
differential equation 58 node 58
digital imaging 274 oculogram 90
Doppler flow measurement 249 orthotics 323
Doppler measurement 45 pressure gradient 45
ductile 13 prevention 81
effective orifice area and regurgitation 45
performance index 45 Reynolds number 45
electric conduction 69 rigid body dynamics 27
electrical injury 81 robot 353
electrical safety 81 shock 81
electrical stimulation parameters 343 static equilibrium 3
electrodes 343 stiffness 13
electrograms 90 stiffness matrix 58
encephalogram 90 stimulators 343
finite element analysis 58 strain 13
first aid 81 stress 13
force 3 tissue characterisation 69
free-body diagram 3 torsion 13
gait analysis 27 turbulent flow 45
haptic interface 353 ultrasound 249
heart valve 45 virtual reality 353
human movement 27 walking 323
internal force 3 X-ray 274
kinematics 323
This page intentionally left blank
Basic Engineering for Medics and Biologists 369
T.C. Lee and P.F. Niederer (Eds.)
IOS Press, 2010
© 2010 The authors and IOS Press. All rights reserved.

Author Index
Bajd, T. 343, 353 Mawson, S. 121
Black, N.D. 121, 140, 172 McCullagh, P.J. 121, 158
Campbell, V.A. 202 McGlade, K. 121
Craig, D. 172 Munih, M. 343, 353
Davies, R. 121, 172 Niederer, P.F. v, 249, 279, 302
Dendorfer, S. 231 Nugent, C.D. 158, 172
Donnelly, M.P. 158, 172 O’Brien, F.J. 187, 214
Dumont, K. 45 O’Connell, B. 202
Finlay, D.D. 158, 172 Partap, S. 187
Hallberg, J. 172 Plunkett, N. 214
Hammer, J. 231 Reilly, R.B. 90, 109
Hazenberg, J.G. 58 Schmid, J. 58
Jossinet, J. 69, 81 Sharma, P.K. 13
Koopman, H.F.J.M. 27 Verdonck, P. 45
Lee, T.C. v, 3, 58, 90, 109 Verkerke, G.J. 3, 58
Lenich, A. 231 Wang, H. 140, 158
Lyons, F. 187 Winder, J. 140
Matjačić, Z. 323 Zheng, H. 121, 140
This page intentionally left blank
This page intentionally left blank
This page intentionally left blank