Proceedings of the 2002 IEEE

international Conference on Robotics 8 Automation

Washington, DC May 2002

Needle Insertion Modelling and Simulation

S. P. DiMaio and S. E. Salcudean
Department of Electrical and Computer Engineering
University of British Columbia, Vancouver, Canada
[email protected], [email protected]

Abstmct-A methodology for estimating the force dis- alizable. In some cases, lookup tables determine n e e
tribution that occurs along a needle shaft during inser- dle force feedback versus depth for 1-DOF interaction
tion is described. To validate the approach, an exper-
imental system for measuring planar tissue deforma- [7,19]. Others employ elastic, viscous and viscoelas-
tion during needle insertions has been developed and tic models to approximate needle impedance along the
is presented. The planar motion of a soft tissue phan- axial direction [8,lo]. While perhaps effective for the
tom is measured during needle penetration and used in simulation of predominantly 1-DOF problems, these
conjunction with a two dimensional linear elastostatic
material model, discretised using the Finite Element approaches may not be suitable for problems involv-
Method, to derive contact force information that is not ing more complex soft tissue anatomy, needle place
directly measurable. This approach provides a method ment optimisation, trajectory planning and automatic
for quantifying the needle forces and soft tissue defor-
mations that occur during general needle trajectories control, where more detailed verifiable knowledge of
in multiple dimensions. A novel simulation algorithm the biomechanical interaction between surgical needles
that is based upon these results is also described. and soft tissues is required.
In prior work, needle insertion forces have been d e
I. INTRODUCTION termined for gelatine [20], ex vivo porcine and human
samples [8]. In each case, only the resultant force act-
One of the most common procedures employed in
ing at the proximal end of the needle was measured,
modern clinical practice is the subcutaneous insertion
while in fact penetration forces are distributed along
of needles and catheters. Such procedures range in the entire length of the needle axis, resulting from
complexity from superficial needle pricks to the biopsy
physical phenomena such as cutting/fracture, sliding,
of deep-seated tumors, and involve the subcutaneous
friction, stick-slip friction, tissue deformation, tissue
insertion of long, slender surgical tools and needles
displacement and peeling [8]. The needle driving forces
into soft, inhomogeneous tissue, usually without vi-
measured previously are the integration of this force
sual feedback from below the skin’s surface. Physicians distribution along the needle shaft.
and surgeons often rely only upon kinesthetic feedback
Consider the insertion of a needle into a volume
from the tool, correlated with their own mental 3-D vi-
of soft material, as depicted in Figure 1. In general,
sualisation of anatomic structures.
Complications arising from the complexity of
such interventions have been studied in biopsy [l],
brachytherapy [2] and particularly in anaesthesia [3-61,
where it is found that such complications are due, in
large part, to poor technique and needle placement [3].
Virtual-reality-based training systems for catheter
insertion [7], epidural lumbar puncture [8-lo], spine
biopsy [ll], breast biopsy [l] , neurosurgical probe in-
sertion [12], interstitial brachytherapy [13], prostate Fig. 1. Needle insertion into soft media.
needle biopsy [14], etc. constitute part of the present
trend toward computer-based simulators for medical as the needle penetrates the tissue, forces are applied
and surgical training [15]. Even experienced physi- both axially and laterally and the tissue deforms, af-
cians are taking advantage of these developments by fecting the path of the needle and in turn the force
using simulators to plan and rehearse complex proce- distribution itself. The resultant proximal force does
dures, to design and evaluate new methods or equip- not provide sufficient information for the tissue defor-
ment [16],and to control complex medical robotic sys- mation to be determined. As it will be shown in this
tems [17,18]. The majority of the abovementioned paper, the distribution of forces along the needle plays
simulation and manipulation systems have been built an important part and should be determined in order
using largely phenomenological and heuristic models to gain insight into the more general biomechanics of
that have not been validated, and that are not gener- needle insertion procedures.

0-7803-7272-7/02/$17.00 0 2002 IEEE 2098

 This paper presents a new methodology that has
been developed to determine needle forces during soft
tissue puncture, and is organised as follows. In Sec-
tion 11, an experimental system for measuring planar
tissue deformations during probing and needle inser-
tion, is described. Soft tissue modelling and param-
eterization using a linear elastostatic model, outlined
in Section 111, is used for estimating the force distri-
bution that occurs along a needle shaft during inser-
tion, as is presented in Section IV. A numerical simula-
tion of a needle insertion, based upon estimated needle
force distributions, is discussed in Section V, particu-
larly because Of its for training, planning and Fig. 3. The complete experimental =tup: a robotic manipulator
guidance in needlebased procedures. Ck"-Ksions and with instrumented epidural needle mounted, a tissue phantom
discussion of future work axe provided in Section VI. and a CCD camera.

11. EXPERIMENTAL
SYSTEM TO MEASURE
PLANAR A . Motion and Force Sensing
TISSUEDEFORMATIONS A blunt probing tool and a needle are instrumented
Needle insertion and probing experiments proceed with a 6-DOF force/torque sensor (AT1 Nano-17 SI-
as shown in Figure 2. A 3-DOF, planar robotic ma- 12-0.12) with 12.5" and 0.0625Nmm force/torque
nipulator [21] is used to drive an end-effector (blunt resolution, and can be attached to the 3-DOF planar
probe or needle) into a thin rectangular block of soft manipulator, as shown in Figure 4. The manipulator
elastic material that acts as a tissue phantom. Both mechanism moves the end-effector along probing or
tool force and tissue phantom deformation axe sam- insertion trajectories.
pled during the manipulator trajectory, to be used in
model identification and verification. The equipment

pmbelpenetrate

track marker motion

Fig. 2. Needle insertion and probing - experimental procedure.

is shown in Figure 3, as it is configured during ex-

periments. The soft tissue phantom is derived from a
compound composed of polyvinyl chloride (PVC) and Fig. 4. (a) Instrumented probe, and (b) Instrumented 17 guage
a liquid plasticizer, a phthalate ester (Soft Plastic, M-F Tuohy epidural needle.
Manufacturing, TX). By varying the amount of plas-
ticizer added to the PVC, one can attain soft materi- Force-displacement characteristics of a small sam-
als with elasticities (Young's Modulus) ranging from ple of the tissue phantom were measured using the
below lOkPa (approximately that of breast tissue [l]) robotic manipulator, as shown in Figure 5. The soft
to well over 100kPa, a range that covers many of the plastic material exhibits non-linear elastic behaviour
body's soft elastic tissues [22]. The tissue phantom lies with an initial slope of approximately 200Nm-l, which
upon a horizontal platform and is fixed along the rear corresponds to a Young's Modulus of 39kPa for the
face, as shown. It is mounted directly below a colour given sample. Subsequent probing and insertion ex-
CCD camera and is marked with a grid of black, circu- periments attempt to operate with strains close to this
lar marks that can be tracked by the camera as they linear force-displacementregion.
move. A silicone lubricant allows the tissue sample
to slide on the platform, parallel to the camera im- B. Image-based Deformation Estimation
age plane and needle insertion axis, with very little Images from a single CCD camera are processed t o
friction. provide an estimate of the 2-D deformation of the top

2099
 sated by means of a parametric rectification function:

where (z,y) are marker coordinates in the image

frame, with respect to an origin lying on the imag-
ing axis, ~ ( 2 , g )is the marker's radius with re-
spect to the same origin, and (d,g') are the rec-
tified marker coordinates. Model parameters P =
{ko, kl, k2, ks, PI,p z } are determined in a calibra-
Fig. 5. Compressive loading of a lOmm x lOmm x 13mm tion step. The result is a mean error of less than one
material sample. The measured forcedisplacement relationship
is shown on the right. pixel over a set of 182 calibration points distributed
thoughout the image, which corresponds to a mean
error of approximately 0.3mm, or 0.3% of the tissue
surface of the tissue sample. The camera is mounted phantom's width. Such systematic error has little im-
as shown in Figure 3 and is interfaced to the frame pact on the measured relative displacement of markers.
grabber of an SGI 0 2 workstation. With the given The camera system and marker tracking algorithm
geometry and a camera resolution of 640 x 480 pix- measure the motion of the tissue sample's upper sur-
els, each pixel corresponds to approximately 0.3" of face; however, the sample has finite thickness so that
tissue phantom deformation/motion. Fkames are sam- needle insertions and probes occur some distance be-
pled at 25-3Ofps and are processed at the same rate by low this surface, as illustrated in Figure 7. The thinner
a marker tracking algorithm that tracks 161 individual
markers attached to the top surface of the sample.
Markers are differentiated on the basis of colour
contrast. An image-based tracking algorithm locates
tissue markers by identifying clusters of black pixels,
which are tracked from one image frame to the next.
In each frame, each marker is located within a neigh-
bourhood surrounding its last known location. Search
neighbourhoods are adaptively expanded if the marker
cannot be found within this region, a scheme that
Fig. 7. A profile view of the tissue phantom, illustrating defor-
can accommodate marker motion with speeds in ex- mation perpendicular to the imaging plane.
cess of 20" s-l. Examples of marker tracking fields
are given in Figure 6. Centre of mass computation the material, the smaller the effects of deformations
perpendicular to the image plane; however, a mini-
mum thickness is needed to maintain rigidity against
. - . ..... tissue buckling.
.. .,.........
*.
* * *
............
,

... ...... 111. MATERIALMODELLING

AND PARAMETER

....,..... .. .....*.-
. a . .
.."..
a..."
IDENTIFICATION
For needle force modelling and simulation, it is nec-
essary to characterise the relationship between the
forces applied to the tissue phantom, and the resulting
phantom deformation. Tissue deformation is complex
Fig. 6. Imagebased marker tracking during calibration (a) and and is still the subject of much research (e.g., [23-251
boundary probing (b). Square search neighbourhoods are shown and many others). In general, tissue modelling is com-
at each tracked marker. plex because of inhomogeous, non-linear, anisotropic
elastic and viscous behaviour. As a first approximai
of marker locations results in sub-pixel tracking preci- tion, this study focuses on homogeneous, linear elasto-
sion between 0.1 and 0.3mm, depending upon location static models that predict tissue deformations in two
within the image. dimensions. Such models are characterised by two pa-
Lense barrel distortion significantly affects marker rameters, namely Young's Modulus and the Poisson
position estimation (see Figure 6(a)) and is compen- Ratio [26].

2100
A . Linear Elastostatic Model Derivation in Two Di- the response due to needle penetration. The set of
mensions model parameters { E v} that minimises the squared
The elastic material is discretised as shown in F ~ ~ -error between measured and predicted node positions
ure 8. For a linear elastic model, the total strain energy is determined the minimisation:
min (g- K ( E ,~ ) --l f ) ~ .-( K
g ( E ,v)-'f)- , (4)
( E ,V I
01 -
Q
where g is the vector of measured model node dis-
0.08.
placements, f is the vector of forces applied to each
0.06.
node by the probe, and K is the system stiffness ma-
F
I
trix, which is a function of model parameters E and
>
OM- v. This problem was solved using the Matlab@ O p
timization Toolbox's non-linear least squares, and has
002. produced consistent results over a large range of initial
values.
0-
I
Tissue motion at several node positions is sampled
-006 -0.04 40.2 0
X Iml
0.m 0.04 006 during boundary probing and is used in conjunction
with measured probe force to estimate model param-
Fig. 8. The continuous domain R is divided into a finite number eters E and v. For small node displacements, mate-
of discrete elements Re by a mesh of nodes.
rial strains are small, corresponding to the near-linear
Estrainover a solid body R, as a function of stress u region of the plot shown in Figure 5. Squared error
and strain E , is given by: is minimised at E = 34kPa and v = 0.34, for max-
imum and mean node displacement errors of 1.6"
and 0.18" respectively. The estimate of Young's
modulus is slightly lower than that measured previ-
and is minimised at static equilibrium. Each element ously, by approximately 10%. For larger deformations,
e reaches its static equilibrium state when the first
variation of the energy functional dEe vanishes. After - measuredncdepobition
. o,34- ..
*
...... ....
x modelnodeposition m s I
discretising (1) using linear shape functions (see [27]), o,34

*.....
I .
the static equillibrium condition is expressed as:

6E" =
Le Aege& - -
f" = 0 , (2) .' .' .' .' .. .'
0.32- - 0.32-
* * . I . . "

where the A" matrix characterises the elastic b e - . .. . . . ..

s
o,3-. 8 * . 0.3.
I

*
*

*
*

*
" .

= s s

. .. .. . . -.
haviour of element Re and ge and f" are displacement >
* * = a =
o,20-. = 0.20. *
and force vectors for those mesh nFdes that constitute
, * . . i *
element e [27]. Over the entire set of nodes on body . I
0,this leads to a set of 2n linear equations: 0.28- I " = " '. 0.26- ' *

I,
. . . . a . % * * *

K(2nx2n)G = f (3)
0.24- a = = i . 0.24- * .x
-, y( .
B. Model Parameter Identification
In biomechanics studies, tissue model parameters
such as Young's Modulus ( E ) and the Poisson Ratio
(v) are typically determined by testing small homo-
geneous tissue samples using rheometers and similar
materials testing equipment 1221. Measurements of
elasticity and viscoelasticity are usually obtained by the linear elastic model is unable to predict global t i s
loading the tissue sample between two parallel plates, sue deformation accurately, since strains in some re-
in a fashion similar to that shown in Figure 5. In this gions are large, lying beyond the linear region shown
research, model parameter identification is achieved by previously. This effect is evident in Figure 9(b) - there
applying known forces to the tissue phantom's bound- are no global linear model parameters that can account
ary, while measuring the resulting node/marker dis- for the motion of all tissue nodes, particularly those in
placements. This -approach allows for the measure- regions of high strain. The maximum and mean node
ment of non-homogenous tissue samples, as well as displacement errors are 3.6" and 0.72".

2101
 Force estimation during needle insertion may pro-
ceed using model parameters derived in this manner,
which may also be applicable to non-homogeneous,
multi-phase tissue samples and tissue phantoms.

IV. NEEDLEINSERTION
FORCE
MODEL
Direct measurement of needle forces by an instru-
mentation technique is a challenging problem. In the 0.m
sketch shown in Figure 1, it is clear that for soft tis-
sues, needle forces result in material motion and d e
formation. Therefore, an indirect means of estimating
the applied forces is to measure the resulting defor-
mations. If the relationship between force and tissue
displacement is known, then the distribution of force
applied along the needle shaft can be computed. Fig. 10. Needle force estimation averaged over multiple needle
Clinical needle insertion rates vary between 0.4" insertions.
and lOmm per second [20].While future investigations
will focus on possible insertion rate dependence, the
- integrated force distribution
present work characterises needle insertions at a single
axial needle displacement rate of 1"s-l. 4-

In experiments, a 17 guage Tuohy epidural needle is

inserted into the tissue phantom while needle position,
insertion force and tissue phantom deformation are
sampled. It is assumed that each sample is a snapshot
of static tissue deformation, for which the local strains
axe low enough that we may consider linear elastic be-
haviour, with a Young's modulus of 34kPa and Pois-
son ratio of 0.34. Needle flexion and radial forces are
small and axe neglected. These assumptions allow the
relationship between 2-D tissue marker displacements Fig. 12. Comparison of the measured total needle insertion
and applied needle forces to be characterised using the force and the integrated needle force distribution.
linear elastostatic model described previously. Given
a set of node displacements, the linear elastic model
computes the force applied at each node (f = Ku (3), estimated at the model nodes closest to the needle axis,
see Section 111-A). Model forces are quze sensitive as well as those just off axis, were summed in the in-
to small errors in node position measurements; how- tegration of the needle force distribution.
ever, because of the random nature of the image-based
marker position estimation errors, and the controlled V. NEEDLEINSERTION
SIMULATION
nature of the experiments, it is possible to average sev- For needle insertion simulation, an FEM-based lin-
eral sets of node forces obtained from identical needle ear tissue model (Section 111-A) is combined with the
insertions. This improves the estimation result signif- experimentally determined force profiles shown in Fig-
icantly. Figure 10 shows the y-axis component of this ure 11. Needle shaft forces conforming to the derived
force (direction of insertion), for a needle displacement force distribution are applied to model mesh nodes
of approximately 70" (from boundary intercept). that lie in the path of the needle. Finite elements are
Model-based forces, estimated at nodes lying along progressively subdivided in order to concentrate node
the needle shaft, are graphed in Figure ll(a). The density around the needle axis, such that needle forces
corresponding axial needle force distribution is shown can be applied at a fine scale within the model. A typ-
in Figure Il(b). Because the back edge of the tis-
sue phantom is fixed to the platform at y = 0.102m,
negative reaction forces result along this edge. The
estimated node forces that lie along the needle shaft
(shown in Figure ll(a)) are summed and compared to
the insertion force measured by the force/torque sen-
sor. This result is shown in Figure 12. Tissue forces Fig. 13. Mesh subdivision.

2102
 10-.5f
. .
. . .. .. .. .. ..
. .
-0.5 -0.5 -0.5
0.02 0.04 0.06 0.08 0.1 0.020.04 0.06 0.08 0.1 0.02 0.04 0.06 0.08 0.1
0.5 : : : . :

.. ... . . ... ... ... ... ...

.. .... .. .. . . . . .
-0.5 -0.5
0.02 0.040.06 0.08 0.1 0.02 0.040.06 0.08 0.1 tissue boundw needle tip
0.5 : : : : :
. . . .
....
. . . . . . .
.. .. .. .. ..

-
-0.5 -0.5
0.02 0.04 0.06 0.08 0.1 0.020.04 0.06 0.08 0.1

-0.5 ’ . . . . .
0.02 0.04 0.06 0.08 0.1
-0.5
0.02 0.04 0.06 0.08 0.1
-0.5
0.51 . . . . .I

0 .... ...
. . . . .. .. .. .. . . . .
-0.5 -0.5 -0.5
0.02 0.04 0.06 0.08 0.1 0.02 0.04 0.06 0.08 0.1 0.02 0.04 0.06 0.08 0.1

Fig. 11. (a)Estimated needle shaft force distribution (at nodes), with needle location also shown. (b) Estimated force “density” along
the needle shaft. (c) This altered needle force distribution is purely artificial, but is used in Section V for comparison.

ical subdivision is shown in Figure 13. If repeated ad

infinitum, element subdivision will eventually result
in exact needle-node intercepts. This is impractical;
therefore, after a single level of element subdivision,
a mesh adaptation scheme places existing nodes onto
the needle axis, as shown in Figure 14.

(a) @)

Fig. 15. Needle insertion using (a) the estimated needle force
distribution, and (b) a distribution with a force concentration
at the tip.

In a second simulation, the simulated needle force

Fig. 14. During simulation, as the needle tip moves from one profile is adjusted by increasing the needle tip force
element into another, the intersection point between the needle significantly with respect to the shaft force (illustrated
axis and the common edge is determined (a). The node closest
to this intersection point (na in the figure) is projected onto the in Figure ll(c)), as one might expect would occur
needle axis, along the direction of the common edge, by moving if needle-tissue friction where far less significant than
the nominal position of the candidate node (shown in the top cutting force (total integrated needle force remains un-
row) such that it lies on the needle axis after deformation (b). changed). While such a force distribution may not ex-
ist in practice, it is used to illustrate the sensitivity of
Figure 15(a) shows a simulated needle insertion the tissue response to the shape of the needle force dis-
- node displacements are shown to closely match tribution. Figure 16(b) shows that the simulated de-
those actually measured after a needle penetration of formation using the new forces differs markedly from
70”. Maximum and mean node displacement er- the actual tissue deformation. The maximum and
rors of 1.4“ and 0.6mm, respectively, are observed mean node displacement errors become: 4.4” and
(shown in Figure 16(a)). 1.2” respectively.

2103
 tissue phantom is relatively uniform along the needle
0.34 shaft. A force peak, located immediately behind the
needle tip, rises approximately 30% above the friction
0.32
* a a a * n force, and may be attributable to tissue cutting. The
quality of the estimated force profile is limited by: (a)

- 0.3
....*- 0.3
tissue deformation in the plane perpendicular to the
image plane, as shown in Figure 7; and (b) linear elas-
E tostatic model constraints, including small strain and
t
0.28 0.28 linearity assumptions.
Simulated needle insertions that are based upon the
0.26
estimated needle forces are shown to reproduce results
0.26
similar to those observed experimentally. In addition,
a virtual “needle biopsy” is shown in order to illustrate
0.24 0.24 the potential of physically-based needle insertion simu-
-0.06 -0.04 -0.02 0 lations for planning and training purposes. Real-time
X lml
@) execution of this simulation, for interactive training,
Fig. 16. Comparison between measured and simulated needle
is challenging due to the large system of equations
insertions with: (a) estimated force distribution, and (b) altered involved, as well as frequent topological and bound-
force distribution (half of each model is plotted, since the defor- ary condition changes that occur as the needle moves
mation is symmetric). into the tissue. We have developed a fast, optimised
algorithm for interactive needle insertion, with force
feedback. The haptic simulation, described in [28],
In a final simulation example, a needle is inserted
achieves a sample rate of 500Hz for a 2-D virtual tis-
into the side of a tissue model that is rigidly- fixed
sue model and planar haptic interface, using a desktop
along one edge, in order to investigate the effect of
tissue deformation on needle placement accuracy: The PC.
needle axis is initially coincident with a “virtual%biopsy Future work will explore 3-D modelling techniques,
target”. The series of simulation samples shown in Fig- non-linear material models, the effects of material in-
ure 17 illustrates the needle’s path through the.tissue homogeneities and dynamics, as well as model-based
sample, and its failure to intercept the target (shown planning and control of needle insertion procedures.
as a black dot).
REFERENCES
VI. CONCLUSION F. S. Azar, D. N. Metaxas, and M. D. Schnall, “A Finite
Element Model of the Breast for Predicting Mechanical D e
This paper presents a novel approach for estimat- formations during Biopsy Procedures,” in Proceedings of
ing needle shaft forces and tissue behaviour using a the IEEE Workshop on Mathematical Methods in Biomed-
ical Image Analysis, pp. 38-45, 2000.
tissue measurement system and soft tissue deforma- S. Nath, Z.Chen, N. Yue, S . Dumpore, and R. Peschel,
tion models. An algorithm for simulating virtual nee- “Dosimetric effects of needle divergence in prostate seed
dle insertions based on these results is also presented. implant using lZ5Iand Io3Pd radioactive seeds,” in Medical
Physics, pp. 1058-1066, 2000.
Such needle mechanics measurements and simulations J. D. Andres, M. A. Fteina, and A. Lopez-Garcia, “Risks
are of interest for the development of physically-based of regional anaesthesia: Role of equipment - Needle design,
catheters,” in Proceedings of the l P h Annual European
virtual planning and training systems that are aimed Society of Regional Anaesthesia Congress, Sept. 1998.
at reducing the incidence of complications in clinical S. Datta, “Complications of regional analgesia and anaes-
practice. thesia: in Proceedings of the lPh Annual European Soci-
The approach is not limited to 1-DOF insertions into ety of Regional Anaesthesia Congress, Sept. 1998.
B. Fiducane, “Complications of brachial plexus anaesthe-
rigid or static environments, as is the case in prior sia,” in Proceedings of the 1 Ph Annual European Society
work in this area. In addition, the method presented of Regional Anaesthesia Congress, Sept. 1998.
in this paper relies on the established Finite Element D. L. Brown, “What do we need t o advance regional a n a e s
thesia?,” in Proceedings of the Annual European Society of
Method and parameters identified during experiments. Regional Anaesthesia Congress, 1999.
While developed for 2D tissue phantoms, the method E. Gobbetti, M. Tuveri, G. Zanetti, and A. Zorcolo,
can easily be generalised to 3D, as long as a method “Catheter insertion simulation with co-registered direct
volume rendering and haptic feedback,” in Medicine Meets
for measuring deformation is available (e.g., MRI, U1- Virtual Reality 2000, pp. 96-98, Jan. 2000.
trasound). P.N. Brett, T. J. Parker, A. J. Harrison, T. A. Thomas,
The needle force distribution that is estimated from and A. Carr, “Simulation of resistance forces acting on sur-
gical needles,” in Inst. of Mech. Eng. Part H, J o u n a l of
experimental measurements, and used for simulation, Engineehng in Medicine, vol. 211, pp. 335-347, 1997.
indicates that axial friction between the needle and the P. Gorman, T. Krummel, R. Webster, M. Smith, and

2104
 01

008

006

-
P OM E OM
>
002 0 02

4 . / ,

-008 -004
,

-002
,

0
,

002
,

OM
,
OW
,
008
,
01
-0021,
-008 -001
,

-002
,

0
,

002
,
O M
,
008
,

008
,
01
4 0 2 [ ,

-008 -004
,

-002
,
0
,

002
,

OM
,

008
,

008
,
01
x [ml X Iml X Iml

Fig. 17. Simulated needle intercept of a small target embedded within elastic tissue.

D. Hutchens, “A Prototype Haptic Lumbar Puncture Sim- in ZEEE Transactions on Visualization and Computer
ulator,” in Med. Meets Virtual Reality, pp. 106-109, 2000. Graphics, vol. 5, pp. 62-73, 1999.
T. Dang, T. M. Annaswamy, and M. A. Srinivasan, “Devel- 1251 A. Hagemann, K. Rohr, H. S . Stiehl, U. Spetzger, and
opment and Evaluation of an Epidural Injection Simulator J. M. Gilsbach, “Nonrigid Matching of Tomographic Im-
with Force Feedback for Medical Training,“ in Proceedings ages Based on a Biomechanical Model of the Human Head,”
of Medicine Meets Virtual Reality, pp. 97-102, 2001. in Med. Imaging - Image Processing, pp. 583-592, 1999.
D.-S. Kwon, J.-U. Kyung, S. M. Kwon, J. B. Ra, H. W. [261 Y . C. Fung, A First Course in Continuum Mechanics.
Park, H. S . Kang, J. Zeng, and K. R. Cleary, “Realistic Prentice Hall, 1977.
Force Reflection in a Spine Biopsy Simulator,” in ZEEE [27] C. Cuvelier, A. Segal, and A. A. van Steenhoven, Finite
Znt. Conf. on Rob. and Auto., pp. 1358-1363, 2001. Element Methods and Navier-Stokes Equations. D. Reidel
K. B. Shimoga and P. K. Khosla, “Visual and Force feed- Publishing Company, 1986.
back to Aid Neurosurgical Probe Insertion,” in lGh Znt. PI S . P. DiMaio and S. E. Salcudean, “Simulated Interactive
Conf. of the ZEEE Engineering in Medicine and Biology Needle Insertion,” in Proc. of the loth Symp. on Haptic Zn-
Society., vol. 2, pp. 1051-1052, 1994. terfaces for Virtual Env. and Telwp. Systems, ZEEE Vir-
S . Miller, C. Jeffrey, J. Bews, and W. Kinsner, “Advances tual Reality, Mar. 2002.
in the Virtual Reality Interstitial Brachytherapy System,”
in Proc. of the Canadian Conf. on Elec. and Comp. Eng.,
pp. 34S354, May 1999.
J. Zeng, C. Kaplan, J. Bauer, J. Xuan, I. A. Sesterhenn,
J. H. Lynch, M. T. Freedman, and S. K. Mun, “Optimizing
prostate needle biopsy through 3-D simulation,” in Pro-
ceedings of SPIE Medical Imaging, Feb. 1998.
D. Sorid and S. K. Moore, “The Virtual Surgeon,” in ZEEE
Spectrum Magazine, pp. 26-31, July 2000.
L. M. Auer, A. Radetzky, C. Wimmer, G . Kleinszig,
F. Schroecker, D. P. Auer, H. Delingette, B. Davies, and
D. P. Pretschner, “Visualization for Planning and Simula-
tion of Minimally Invasive Neurosurgical Procedures,” in
Medical Image Computing and Computer-Assisted Znter-
vention, pp. 1199-1209, 1999.
A. Zivanovic and B. L. Davies, “A Robotic System for
Blood Sampling,” in IEEE Transactions on Information
Technology in Biomedicine, vol. 4, pp. 8-14, Mar. 2000.
K. Khodabandehloo, P. N. Brett, and R. 0. Buckingham,
“Special-Purpose Actuators and Architectures for Surgery
Robots,” in Computer-Integrated Surgery: Technology and
Clinical Applications, The MIT Press, 1996.
L. Hiemenz, D. J. McDonald, D. Stredney, and D. Ses-
sanna, “A Physiologically Valid Simulator for Training Res-
idents to Perform an Epidural Block,” in Proc. of the l P h
Southern Biomedical Engineering Conference, Mar. 1996.
L. Hiemenz, A. Litsky, and P. Schmalbrock, “Puncture Me-
chanics for the Insertion of an Epidural Needle,” in 2lSt
Annual Meeting of the American Soc. of Biomech., 1997.
M. R. Sirouspour, S. P. DiMaio, S. E. Salcudean, P. Abol-
maesumi, and C. Jones, “Haptic Interface Control - Design
Issues and Experiments with a Planar Device,” in Proc. of
the IEEE Int. Conf. on Robotics and Automation, 1999.
Y . C. Fung, Biomechanics - Mechanical Properies of Living
Tissues. Springer-Verlag, 1993.
D. L. James and D. K. Pai, “ArtDefo, Accurate Real Time
Deformable Objects,” in Computer Graphics - Proceedings
of SIGGRAPH ’99, 1999.
S . Cotin, H. Delingette, and N. Ayache, “Real-Time Elas-
tic Deformations of Soft Tissues for Surgery Simulation,”