IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO.

4, APRIL 2007 611

Hybrid Finite Element Method for Describing the

Electrical Response of Biological Cells to
Applied Fields
Wenjun Ying* and Craig S. Henriquez

Abstract—A novel hybrid finite element method (FEM) for mod- Of the numerical methods currently used to model field stim-
eling the response of passive and active biological membranes to ulation of biological cells, the finite difference method (FDM)
external stimuli is presented. The method is based on the differ- is the easiest to implement. Because the FDM uses Cartesian
ential equations that describe the conservation of electric flux and
membrane currents. By introducing the electric flux through the grids to approximate derivatives or equivalent circuit elements,
cell membrane as an additional variable, the algorithm decouples the geometry of interest is often represented as a piecewise rect-
the linear partial differential equation part from the nonlinear or- angular domain. The application of FDM to geometrically com-
dinary differential equation part that defines the membrane dy- plex domains is more challenging [13] and obtaining good ac-
namics of interest. This conveniently results in two subproblems: curacy requires very fine grids or the use of special techniques
a linear interface problem and a nonlinear initial value problem.
The linear interface problem is solved with a hybrid FEM. The ini- [14], [15]. In contrast, the finite element method (FEM) [11],
tial value problem is integrated by a standard ordinary differen- [12], [16] or finite volume method (FVM) [17], [18] is a more
tial equation solver such as the Euler and Runge-Kutta methods. flexible approach for modeling field simulation of arbitrarily
During time integration, these two subproblems are solved alter- shaped cells. The FEM partitions the intracellular and extra-
natively. The algorithm can be used to model the interaction of cellular spaces into simple elements, such as triangles in two
stimuli with multiple cells of almost arbitrary geometries and com-
plex ion-channel gating at the plasma membrane. Numerical ex- space dimensions (2-D) and tetrahedrons in three space dimen-
periments are presented demonstrating the uses of the method for sions (3-D). With elements aligned with the membrane of cells,
modeling field stimulation and action potential propagation. the FEM usually yields higher order accuracy than the FDM.
Index Terms—Field stimulation, hybrid finite element method, When the cellular media are assumed to have piecewise homo-
interface problem, Laplace equation, transmembrane potential. geneous and isotropic conductivities, the electric potential equa-
tion simply reduces to the Laplace’s equation. In this case, the
integral-based boundary element method (BEM) is applicable
I. INTRODUCTION [19]–[21]. The BEM only discretizes the cell membrane and,
thus, requires fewer grid nodes compared to FDM, FEM, and

F IELD stimulation of biological cells has a wide range of

applications, including gene transfection [1], [2], elec-
trochemotherapy of tumors [3] and cardiac defibrillation [4].
FVM. As a result, BEM can be very efficient and accurate if it
is carefully designed and implemented.
The FDM, FEM, and BEM all approximate the partial dif-
Much of the understanding on how cells respond to external ferential equations (PDE) that describe the continuity and con-
fields is based on theory derived for a single isolated cell. servation of electric flux and membrane currents. The accuracy
Analytical expressions for an idealized, passive membrane and ultimate solution times of the methods depend, in part, on
response has been derived for single cells that are spherical, the fineness of the grid. In some implementations of the methods
prolate or oblate spheroidal [5]–[10]. to study field stimulation, the membrane and the gap junctions
In contrast to a single cell, the response of multiple cells in that electrically connect cells are assumed to have finite thick-
suspension or in tissue to an electric field is less well under- ness [22], requiring fine grids in the membrane regions and
stood. The induced transmembrane potential inside a multiple smooth transitions of element sizes away from these regions.
cell system depends on not only cell density but also on the The tractability of the methods is also affected when the non-
arrangement of cells and their positions [11], [12] in the field. linear ion-channel gating and complex membrane dynamics are
Because of the confounding effects of cell size, position, prop- included. In these cases, the PDE-based numerical methods in-
erties and packing, the electric potential distribution of multiple volve the solution of a boundary value problem with nonlinear
cells can only be studied using numerical methods and computer and time-dependent boundary conditions at each time instance.
simulations. To overcome the difficulties of standard methods, some
circuit-based numerical methods, such as the equivalent circuit
Manuscript received May 15, 2006; revised September 3, 2006. This work methods (ECM) [23], [24] and the transport lattice method
was supported in part by the National Institutes of Health (NIH) under Grant (TLM) [13], [25]–[27], have been recently developed. The
R01HL76767. Asterisk indicates corresponding author.
*W. Ying is with the Department of Biomedical Engineering, Duke Univer- ECM introduced by Fear and Stuchly [23] was initially used to
sity, Durham, NC 27708-0281 USA (e-mail: [email protected]). model the response of cells connected via gap junctions. The
C. S. Henriquez is with the Departments of Biomedical Engineering and ECM represents the cytoplasm and medium by resistances,
Computer Science, Duke University, Durham, NC 27708-0281 USA.
Digital Object Identifier 10.1109/TBME.2006.889172
which are connected in parallel by membrane impedance,

0018-9294/$25.00 © 2007 IEEE

612 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO. 4, APRIL 2007

and the membrane is represented by a parallel resistance-ca-

pacitance unit. Similar to the FDM, the ECM [24] also uses
Cartesian grids but with two levels of resolution: a fine grid is
built around the membranes while a coarse grid occupies the
major area far away from the membranes.
The TLM is based on spatially distributed transport networks,
which are solved by Kirchhoff’s laws to determine the potential
distribution. It is essentially a spatial representation of the cell
membrane and medium using the ECM, and is equivalent to the
FEM in that they both solve complicated PDEs. Taking advan-
tage of circuit solving software tools, the TLM offers a wide
range of flexibility and modularity for representing highly non-
linear transport processes such as ion channels, ion pumps and
membrane electroporation and heat transport within tissue with
perfusion [13], [26].
In this paper, a modification of the classical PDE-based nu- Fig. 1. A uniform cylindrical cell fiber: (a) transverse cross section of the cylin-
merical methods for field stimulation of biological cells is pre- drical fiber in a rectangular domain; (b) longitudinal cross section of a cylin-
sented. First, the membrane is assumed to be an interface with drical fiber surrounded by a layer of fluids; (c) closeup of a matching triangula-
tion of the intracellular and extracellular spaces; (d) closeup of a nonmatching
zero thickness. Second, the electric flux through the cell mem- triangulation of the intracellular and extracellular spaces.
brane is introduced as an additional variable into the PDEs. As
will be shown, the introduction of the flux variable decouples
the time-dependent boundary conditions from the PDEs and, during the stimulation period and at the positions where the
thus, makes it easier to include nonlinear membrane dynamics. stimuli are applied. In the case that both the intracellular and
It also eliminates the need for matching grids in the extracel- extracellular spaces have homogeneous and isotropic conduc-
lular space and intracellular space at the interface. The introduc- tivities, each of (2.1) simply reduces to the Poisson or Laplace’s
tion of the flux variable decomposes the original problem into equation.
two much simpler subproblems: a linear time-independent in- Let , be the unit outward normals to the boundaries
terface problem and a nonlinear space-independent initial value and of the intracellular and extracellular spaces, respec-
problem. In this approach, the linear interface problem is solved tively. Let be the unit outward normal pointing outside of a
with a hybrid FEM. The initial value problem is integrated by a cell on the membrane . It is obvious that
standard ordinary differential equation (ODE) solver such as the
Euler and Runge–Kutta methods. These two subproblems are
solved alternatively during time integration. In this sense, the al-
gorithm is highly modular and avoids the need to directly solve a
boundary value problems with nonlinear boundary conditions. As usual, it is assumed that the computational domain is
Simulation results are presented showing that the method can bounded and no-flux boundary conditions are applied on the ex-
be used to model cells of arbitrary geometries with complex terior boundary [see Fig. 1(a)], i.e.,
ion-channel gating.
(2.2)
II. INITIAL BOUNDARY VALUE PROBLEM
As shown in Fig. 1(a), the cell membrane is assumed to The normal component of the intracellular and extracel-
be an interface that separates the computational domain into lular current density at the membrane surface is continuous
an intracellular space and an extracellular space . The ef- and equal to the membrane current (units:
fective electrical conductivity coefficients (units: ) of ), i.e.,
the intracellular and extracellular spaces are and , respec-
tively. The time-dependent intracellular and extracellular poten-
(2.3)
tials (units: mV) are given by and ,
respectively. Here, and denotes the temporal and spatial The membrane current has two components: a capacitive
variables. (displacement) current and a resistive (conductive) current. Both
By the conservation of electric flux, the electric potentials the capacitive and the resistive currents depend on the trans-
and are governed by the second-order elliptic equations membrane potential , which is the difference of the intra-
cellular and extracellular potentials across the membrane, i.e.,
(2.1a)
(2.1b)
(2.4)
Here, and are the current (source)
density (units: ) due to the voltage stimuli applied in Let be the membrane capacitance per unit area (units:
the intracellular and extracellular spaces. They are nonzero only ) and be a set of state variables including ion con-
YING AND HENRIQUEZ: HYBRID FEM FOR DESCRIBING THE ELECTRICAL RESPONSE OF BIOLOGICAL CELLS TO APPLIED FIELDS 613

centrations, which define the physiological state of the cellular of the flux through the membrane . The interface flux condition
structures. Let be the membrane resistive/ionic (2.3) is naturally incorporated into these two identities.
current. The membrane current is then explicitly given by Note that the intracellular and extracellular potentials are cou-
pled by the transmembrane potential through the membrane,
which imposes an essential interface condition. To work with
(2.5) the FEM, the interface condition (2.4) has to be weakly en-
forced. Multiplied by a test function and
The state variables are typically governed by a system of or- integrated over the membrane , the interface condition (2.4)
dinary differential equations: can be expressed as

(2.6)

Here, is a set of model-dependent functions.

Note that (2.4) and the first identity in (2.3) make up two for any test function .
interface conditions for both the potentials and the flux. Thus, Furthermore, let be the product
a linear interface problem is defined by (2.1) subject to the space. Let , , ,
homogeneous Neumann boundary condition (2.2) and the and . The interface problem can be equiva-
interface conditions. Given transmembrane potential and lently and concisely written in the following variational form:
provided that the cell membrane is smooth enough, the find and , such that
interface problem has a unique solution , up to an
additive constant. (3.1a)
Moreover, by the potential-current relationship (2.3), the
membrane current can be computed in terms of the solution
to the interface problem. Thus, the membrane cur- (3.1b)
rent is uniquely determined by the transmembrane potential
. In this sense, the membrane current can be thought of for any test functions and . Here,
as a linear function of the transmembrane potential , i.e., denotes the jump of the test function across the mem-
. Equation (2.5) can then be treated as an ODE brane . The flux unknown is introduced as a variable as well
for the transmembrane potential in the following form: as the potential variables and , which makes the varia-
tional problem (3.1) hybrid. In the literature, the flux variable
(2.7) is called the Lagrange multiplier and the space is called the
Given appropriate initial values for the transmembrane poten- Lagrange multiplier space [29].
tial and the state variables , the ODEs (2.6), (2.7) has a The hybrid variational formulation (3.1) makes up a saddle
unique solution. The overall initial boundary value problem for point problem. It has a unique solution up to a constant
the electric potential distribution in a cellular structure is, thus, [30]. The solution equals the electric flux across the membrane
well-posed. , i.e.,

III. HYBRID VARIATIONAL FORMULATION

(3.2)
Let and be the Sobolev spaces consisting
of up to first-order derivatives square-integrable functions on which balances the membrane current in (2.7).
the bounded domains and , respectively. Let be
the trace space of the Sobolev space or on the IV. FINITE ELEMENT DISCRETIZATION
membrane . Let be the dual space of the trace space Let and be regular triangulations of the intracellular
[28]. and extracellular spaces with mesh parameters and , re-
Multiplying each of the Poisson equations (2.1) by a corre- spectively. Let and be the corresponding finite ele-
sponding test function or , using ment spaces associated with the partitions. Denote by
the technique of integration by parts, and imposing the no-flux the finite dimensional subspace of the unconstrained
boundary condition (2.2) on the exterior boundary yield the product space . Here, represents the mesh
following integral identities: parameter of the global triangulation . It is im-
portant to note that the triangulation does not need to be con-
formal. The two partitions, and , for the intracellular and
extracellular spaces, respectively, may be nonmatching on the
membrane [see Fig. 1(d)]. In this case, the membrane can in-
herit its partition from either or . The corresponding
FEM is called the mortar FEM [31]–[33]. Let be the trace
for any test functions and . Here, space of one of the finite element spaces and on the
denotes the normal component membrane . Let be the dual space of the trace space .
614 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO. 4, APRIL 2007

Corresponding to its continuous counterpart, the dual space

is called the discrete Lagrange multiplier space.
The discrete version of the variational problem (3.1) reads as
follows: find and , such that

(4.1a)

(4.1b) E
Fig. 2. A single circular cell in an external electric field. The electric field
is established by a pair of external electrodes and is assumed uniform in the
vicinity of the cell. (a) The space is separated into two parts by the membrane
0: the intracellular space and the extracellular space . Assume that the
for any test functions and . In general, espe- n
cell has diameter d and the unit normal points from the intracellular to the
cially when the intracellular and extracellular grids and extracellular space. (b) The marked points are evenly spaced around the circle.
are not matching along the membrane , the finite dimensional
spaces and have to satisfy a stability condition in order
for the hybrid variational problem to be uniquely solvable (up
VI. TEST PROBLEM
to a constant) [30]. For the situations where the grids match on
the membrane [see Fig. 1(c)], the stability condition is always
satisfied by the spaces and introduced above and so is Analytical solutions are possible for certain geometries and
the unique solvability of the discrete variational problem guar- conditions. These solutions can be used to validate the numer-
anteed as long as the membrane is smooth enough [31]. ical schemes. As in Krassowska and Neu [35], an idealized
In matrix-vector notation, the final linear system corre- model of a single cell in an external electric field , shown in
sponding to the discrete variational problem has the saddle Fig. 2, is considered. The cell is assumed to have a diameter
point form that is small compared with the extracellular region and is lo-
cated far away from the electrodes such that the electric field is
approximately uniform in the vicinity of the cell. The extracel-
(4.2) lular potential far away from the cell corresponds to the uniform
electric field
Here, the block 2 2 coefficient matrix is symmetric but indefi-
nite even though the stiffness component at the top-left corner
is nonnegative definite. In the literature, a large amount of work (6.1)
has been devoted to the problem of solving such linear systems
[34]. For this particular problem, due to the biorthogonality be-
tween the nodal basis functions of the finite element trace space Assuming that the membrane is passive, such that the
and the discrete Lagrange multiplier space , the flux membrane current depends on the
vector can be locally eliminated from the saddle point for- transmembrane potential only, the problem corresponds
mulation, resulting in a positive definite algebraic system [33]. to a cylindrical cell in a transverse electric field. The solution
The linear system is then solved by a standard iterative solver,
to the problem can be obtained by separation of variables and
such as the conjugate gradient iteration method.
expressed in cylindrical coordinates as functions of radius ,
angle and time as follows:

V. ALGORITHM
(6.2a)
Assuming that the transmembrane potential and the state (6.2b)
variables have been initialized at , the algorithm for inte-
grating the electric potential equations (2.1)–(2.7) is formulated (6.2c)
as the following two-step procedure.
• Step 1: With the old transmembrane potential at Here
, the electric flux , which approximates the
membrane current , is calculated by solving the
linear system (4.2).
• Step 2: With the old transmembrane potential , old state
variables at and the membrane current obtained
in the previous step, the ODEs (2.6) and (2.7) are integrated
by a time step , yielding new values of and at
. and
Repeat these two steps above until the final computational time
is reached.
YING AND HENRIQUEZ: HYBRID FEM FOR DESCRIBING THE ELECTRICAL RESPONSE OF BIOLOGICAL CELLS TO APPLIED FIELDS 615

The potentials at steady-state are given by

VII. COMPUTATIONAL CONSIDERATIONS

The algorithms for the Hybrid FEM and Delaunay trian- Fig. 3. Stimulation across a cylindrical cell with passive membrane. (a) Steady-
gulation were implemented in custom codes written in C++. state iso-potential contours around the cell with matching grids as shown in
Fig. 1(c) (duration the entire simulation, the relative error of the potentials, rel-
Simulations were all performed on a dual Xeon 3.6 GHz ative to the maximum transmembrane potential at steady-state, is bounded by
computer. For the test problem of a circular cell with active 2.71%); (b) steady-state iso-potential contours around the cell with nonmatching
membrane in a uniform field, the domain comprises approxi- grids as shown in Fig. 1(d) (during the entire simulation, the relative error of the
potentials, relative to the maximum transmembrane potential at steady-state, is
mately 4250 nodes. When the membrane was described with bounded by 2.94%).
the Hodgkin–Huxley model [36], the simulation required a time
step of to resolve the dynamics of the very fast
early phase and the CPU time per time step was approximately around cell) and the locations (indicated by small squares) of
13.8 ms. Thus, for a fixed time step of , a sim- the peak positive (depolarized) portion (facing the cathode) and
ulation of 10.0 ms of activity required approximately 2.5 hrs. peak negative (hyperpolarized) region (facing the anode). The
The average CPU time per time step scales roughly linearly solutions for the matching and nonmatching grids are the same,
with the number of nodes. with less than 0.25% error at steady-state.
Fig. 4 shows the variation of the intracellular, extracellular
VIII. NUMERICAL RESULTS and transmembrane potential as a function of time at 16 evenly
The hybrid FEM was applied to several problems of an active spaced points on the circular membrane [see Fig. 2(b)]. The
or passive cell in an electric field. In all experiments, continuous computed solutions are compared with the analytic solutions
piece-wise linear finite elements are used to discretize the hybrid given by (6.2). As is shown, the computed solutions and analyt-
variational problem (3.1). The time integration for the ODEs ical solutions match very well for all time steps. During the en-
(2.6) and (2.7) simply follows the forward Euler method. The in- tire simulation, the relative error of the potentials, relative to the
tracellular and extracellular conductivities are , maximum transmembrane potential at steady-state, is bounded
, respectively. The membrane capacitance is by 2.71%. Simulations using the nonmatching grids at the inter-
. For the passive cells, the membrane resis- face produced nearly identical results. Tables I and II show the
tance is . For the active cells, the ionic currents relative error bounds at the 16 points with both matching and
were represented by the Hodgkin–Huxley model [36]. nonmatching grids at six different times.
In the first experiment, a passive circular cell with diameter One advantage of the hybrid finite element scheme is that it
is placed between two planar electrodes, with the can be used to model cases in which the membrane has nonlinear
anode on the left and the cathode on the right (see Fig. 2). The properties. Another simulation was performed in which the pas-
distance between the electrodes, 0.01 cm, is large compared to sive membrane was replaced with one whose dynamics were
the cell diameter, so the analytical expression (6.2) can be used represented by the Hodgkin–Huxley model [36]. The external
for validation. The electric field , generated by electric field was increased to to facilitate stimula-
the planar electrodes, is kept on for the entire duration of the tion since the active membrane model has much larger threshold
simulation. action potential than the passive one. The intracellular, extracel-
Fig. 1(c) and (d) shows the triangular grids used in the sim- lular and transmembrane potentials at nine points around the cell
ulation. In one case, the grid points on extracellular side of [see Fig. 2(b)] are plotted in Fig. 5 from 0 to 10 ms. The results
the membrane match with the grid points on the intracellular are consistent with those given by Krassowska and Neu [35], in
side of the membrane. The grid on the membrane has a spa- which the stimulation process has two distinct phases. In the first
tial length of approximately . In the other case, phase (Fig. 5), the intracellular, extracellular and membrane po-
the grid points on the two sides of the membrane do not match tential all show some variation within the first microsecond and
and the membrane inherits the partition from the intracellular then reach a steady-state. The results also show that after the
space side, which has a coarser triangulation [see Fig. 1(d)]. initial polarization, all points in the intracellular domain follow
This means the mesh parameter on the membrane in the non- the same time course, whereas the extracellular potential is con-
matching case is , twice that of the matching stant in time at a given location.
case. In both cases, however, the forward Euler Method with a The response of a passive and active fiber to a re-orientation
time step is used to integrate the ODE (2.7). of the electric field along the long axis of the cylinder was also
Fig. 3(a) and (b) shows the steady-state potentials and the modeled using the hybrid FEM. For simplicity, the domain was
isocontours at time , respectively. Fig. 3(a) and (b) modeled as a 2-D rectangular cross section of a cylindrical fiber
also shows the variation of transmembrane potential around [see Fig. 1(b)] with dimensions 0.2 cm by 15 , or 0.4 cm by
the perimeter of the cell (as an exaggerated greyscale strip 15 , bounded by a thin layer of fluid with thickness of 5 .
616 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO. 4, APRIL 2007

Fig. 4. Electric potentials at 16 evenly spaced points around the cylindrical circle (passive membrane) from the simulation with matching grids [see Fig. 1(c)].
+
The exact solutions are plotted with lines. The numeric data are marked by points. The plus marker “ ” denotes transmembrane potential, the cross marker “ ” 2
3
denotes extracellular potential and the asterisk marker “ ” denotes intracellular potential. Duration the entire simulation period, the relative error of the potentials
at each point, relative to the maximum transmembrane potential at steady-state, is bounded by 2.71%.

TABLE I TABLE II
RELATIVE ERRORS OF POTENTIALS FROM THE RELATIVE ERRORS OF POTENTIALS FROM THE
SIMULATION WITH MATCHING GRIDS SIMULATION WITH NONMATCHING GRIDS

cathode. These results are consistent with those of Weidmann

Fig. 6 shows the steady-state distribution of the intracellular, ex- [37].
tracellular and membrane potentials along the two fibers. For Again, the hybrid FEM can also be used to investigate active
each of the fibers, there is a linear change in extracellular poten- and propagated responses. Fig. 7 shows the results from the
tial and a variation of intracellular potential at two ends corre- application of both a local (current) stimulus (at one end)
sponding to a hyperpolarization of transmembrane potential at and a field (voltage) stimulus along a uniform cell fiber (of
the anode and a depolarization of transmembrane potential at the length 0.2 cm) with an active membrane described by the
YING AND HENRIQUEZ: HYBRID FEM FOR DESCRIBING THE ELECTRICAL RESPONSE OF BIOLOGICAL CELLS TO APPLIED FIELDS 617

Fig. 5. Electric potentials at nine points around the upper semi-circle of the cell shown in Fig. 2(b). The membrane dynamics follows the Hodgkin–Huxley model.
(a) From 0 to 0.001 ms and (b) from 0 to 10 ms, the nine lines from top to bottom correspond to the nodes from 0 to 8 in Fig. 2(b). The transmembrane potential has
maximum value at node 0, minimum value at node 8 and a value of 0 at node 4; (c) from 0 to 0.001 ms; (d) from 0 to 10 ms, the intracellular potential is spatially
uniform at steady-state. (e) From 0 to 0.001 ms and (f) from 0 to 10 ms, the nine lines from bottom to top correspond to the nodes from 0 to 8 in Fig. 2(b). The
extracellular potential has maximum value at node 8, minimum value at node 0 and a value of 0 at node 4, and is constant in time at steady-state.

Hodgkin–Huxley model. In the first case, a corner at the far end into the hyperpolarized zone. Fig. 7(b) shows the intracellular,
of the fiber is grounded; the local current stimulus with strength extracellular and transmembrane potentials along the fiber,
0.2 is applied at the near end for 2 s. As a result, demonstrating that the method can be used to study complex
an electrical wave is initiated at the stimulus site as shown in interactions of the electric field with active and propagated
Fig. 7(a). In the second case, the field stimulation with strength responses in realistic geometries.
0.25 V/cm is applied for the entire duration. Because of the Finally, the method was used to consider the response of mul-
short fiber length, both the intracellular and extracellular poten- tiple cells in a uniform electric field. Here, the cells are expected
tials show variations from classical core conductor predictions to locally perturb the field. Fig. 8 shows closeups of the grids
[38], [39], in which the potentials are simply scaled versions of (nonmatching) and complex potential distributions for a cluster
the transmembrane potential (with opposite sign) [40]. In the of four circular and elliptical cells. The perturbations of the field
case of field stimulation Fig. 7(b), the membrane depolarizes act to modify the location of the maximum and minimum trans-
at one end of the fiber and hyperpolarizes at the other end. The membrane potentials on the cells (square markers) such that they
depolarized end initiates an action potential that propagates are not symmetric about the middle axis of the cell as is the
618 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO. 4, APRIL 2007

Fig. 6. Steady-state potential distribution due to field stimulation along uniform cell fibers with passive membrane. The dotted lines mean the extracellular po-
tential, the dashed lines denote the intracellular potential and the solid lines represent the transmembrane potential. (a) The uniform cell fiber has length 0.2 cm.
(b) The uniform cell fiber has length 0.4 cm.

Fig. 7. Electric potentials sampled at 11 evenly spaced points along the 0.2 cm uniform cell fiber, which include the endpoints. The membrane dynamics follows
the Hodgkin–Huxley model. (a) A current stimulus with strength 0.2 mA=cm is applied at the near end of the fiber for 2 ms. (b) A field stimulus with strength
0.25 V/cm is applied for the entire duration of simulation.
YING AND HENRIQUEZ: HYBRID FEM FOR DESCRIBING THE ELECTRICAL RESPONSE OF BIOLOGICAL CELLS TO APPLIED FIELDS 619

By treating the membrane current as a parameter, a range of

ODE solvers can be used depending on the desired accuracy. In
the present implementation, only explicit methods such as the
forward Euler method and the classic fourth-order Runge–Kutta
method are used. Explicit methods have stability restrictions
that effect the range of timesteps for the time-dependent prob-
lems. To overcome the restriction on timestep size, implicit
methods such as the backward Euler, implicit Runge–Kutta
methods and backward differentiation formula (BDF) could
also be used. A future paper will describe the implementation
of the method with adaptive time stepping.
The time spent in solving the space-independent nonlinear
ODE part scales linearly with the number of nodes on the mem-
brane. The efficiency of the overall algorithm strongly depends
on the linear system solver used for the linear PDE part. In the
literature, there are optimal algorithms, such as the multigrid
method, which could solve the linear system (4.2) in a work
linearly proportional to the number of nodes. In a word, the im-
Fig. 8. Stimulation of multiple cells under an external electric field (along the plementation of the method may be very efficient if an optimal
horizontal direction): (a) close-up of nonmatching triangulations of the intra-
cellular and extracellular spaces. (b) Steady-state iso-potential contours around linear system solver is selected for the PDE part.
the cells from the simulation on the nonmatching grids that respectively result While the efficiency of the current implementation of the hy-
from uniform refinement of the ones shown in (a) by connecting the midpoints brid FEM can be improved, it can nevertheless be applied to
of edges. The square markers denote the local maximum and minimum points
of the transmembrane potential, which is defined on the coarse grid side. a wide range of problems in biology involving cell-to-cell in-
teractions. The approach is attractive for studying the affect of
case with a single cell. These results clearly show that single cell the extracellular space on current flow during propagation of
theory cannot be used to predict the response of multiple cells. action potentials, stimulation and field mediated drug delivery.
While the results presented were all in two-dimensions corre-
sponding to the response of infinite cylinders in three-dimen-
IX. DISCUSSION sions, the response of more realistic (e.g. spherical) cells to an
electric field would require a full three-dimensional representa-
The method proposed in this work provides a powerful and tion. Fortunately, the method can also be extended straightfor-
flexible tool for simulating the response of biological cells to an wardly to three dimension by using tetrahedral elements instead
external electric field. In each time step, the algorithm decou- of triangular elements for the linear interface problem. This ex-
ples the linear partial differential equations from the nonlinear tension will make it suitable to study intramural stimulation and
ordinary differential equations. It not only avoids the need to impulse propagation in cells of arbitrary shape.
solve a boundary value problem with nonlinear boundary con-
ditions but also makes the solver highly modularized. The ap-
proach makes it straightforward to incorporate various models ACKNOWLEDGMENT
of membrane ion transport or pore formation [41]–[48]. By sim- The authors would like to thank D. Rose for the useful
ilarly introducing the electric flux through gap junctions as an discussions.
independent variable, the method also can be used to efficiently
model the response of assemblies of biological cells connected
by gap junctions. REFERENCES
As demonstrated, the hybrid FEM can be used to model the
[1] L. M. Mir, M. F. Bureau, J. Gehl, R. Rangara, D. Rouy, J. M. Caillaud,
response in complex domains of multiple cells with arbitrary ge- P. Delaere, D. Branellec, B. Schwartz, and D. Scherman, “High-effi-
ometries (see Fig. 8). The introduction of the electric flux vari- ciency gene transfer into skeletal muscle mediated by electric pulses,”
able in (3.1) avoids the need to discretize the thin membrane and Proc. Nat. Acad. Sci. USA, vol. 96, no. 8, pp. 4262–4267, 1999.
[2] E. Neumann, S. Kakorin, and K. Toensing, “Fundamentals of electro-
provides a more accurate approximation of the membrane cur- porative delivery of drugs and genes,” Bioelectrochem. Bioenerg., vol.
rent than methods that compute the derivatives of the potential 48, no. 1, pp. 3–16, 1999.
variables. A particular strength of the method is that the intracel- [3] G. Sersa, T. Cufer, M. Cemazar, M. Rebersek, and R. Zvonimir, “Elec-
trochemotherapy with bleomycin in the treatment of hypernephroma
lular and extracellular spaces can be independently partitioned metastasis: case repeat and literature review,” Tumori, vol. 86, no. 2,
such that the grids do not need to match on the membrane. In pp. 163–165, 2000.
some simulations, this will yield some savings in computer re- [4] T. Ashihara, T. Yao, T. Namba, M. Ito, T. Ikeda, A. Kawase, S. Toda,
T. Suzuki, M. Inagaki, M. Sugimachi, M. Kinoshita, and K. Nakazawa,
sources. In the experiments reported in the work, the intracel- “Electroporation in a model of cardiac defibrillation,” J. Cardiovasc.
lular potential inside a cell is typically smoother and has smaller Electrophysiol, vol. 12, no. 12, pp. 1393–1403, 2001.
variation. Thus, it is more appropriate to use larger elements in- [5] H. Fricke, “The electric permittivity of a dilute suspension of mem-
brane-covered, ellipsoids,” J. Appl. Phys., vol. 24, pp. 644–646, 1953.
side a cell than outside. However, the hybrid FEM also allows [6] H. P. Schwan, “Electrical properties of tissue and cell suspensions,”
larger elements on the outside of a cell if necessary. Adv. Biol. Med. Phys., vol. 5, pp. 147–209, 1957.
620 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO. 4, APRIL 2007

[7] T. Kotnik and D. Miklavcic, “Analytical description of transmembrane [32] B. I. Wohlmuth, Discretization Methods and Iterative Solvers Based on
voltage induced by electric fields on spheroidal cells,” Biophys. J., vol. Domain Decomposition, ser. Lecture Notes in Computational Science
79, no. 2, pp. 670–679, 2000. and Engineering. Berlin, Germany: Springer-Verlag, 2001, vol. 17.
[8] J. Gimsa and D. Wachner, “Analytical description of the transmem- [33] B. P. Lamichhane and B. I. Wohlmuth, “Mortar finite elements for in-
brane voltage induced on arbitrarily oriented ellipsoidal and cylindrical terface problems,” Computing, vol. 72, pp. 333–348, 2004.
cells,” Biophys. J., vol. 81, no. 4, pp. 1888–1896, 2001. [34] M. Benzi, G. H. Golub, and J. Liesen, “Numerical solution of saddle
[9] B. Valic, M. Golzio, M. Pavlin, A. Schatz, C. Faurie, B. Gabriel, J. point problems,” Acta Numerica, pp. 1–137, 2005.
Teissie, M. P. Rols, and D. Miklavcic, “Effect of electric field induced [35] W. Krassowska and J. C. Neu, “Response of a single cell to an external
transmembrane potential on spheroidal cells: theory and experiment,” electric field,” Biophys. J., vol. 66, pp. 1768–1776, 1994.
Eur. Biophys. J., vol. 32, no. 6, pp. 519–528, 2003. [36] A. L. Hodgkin and A. F. Huxley, “A quantitative description of mem-
[10] D. C. Lee and W. M. Grill, “Polarization of a spherical cell in a nonuni- brane current and its application to conduction and excitation in nerve,”
form extracellular electric field,” Ann. Biomed. Eng., vol. 33, no. 5, pp. J. Physiol., vol. 117, pp. 500–544, 1952.
603–615, 2005. [37] S. Weidmann, “Electrical constants of trabecular muscle from mam-
[11] R. Susil, D. Semrov, and D. Miklavcic, “Electric field induced trans- malian heart,” J. Physiol., vol. 210, no. 4, pp. 1041–1054, 1970.
membrane potential depends on cell density and organization,” Electro. [38] J. Clark and R. Plonsey, “A mathematical evaluation of the core con-
Magnetobiol., vol. 17, no. 3, pp. 391–399, 1998. ductor model,” Biophys. J., vol. 6, no. 1, pp. 95–112, 1966.
[12] M. Pavlin, N. Pavselj, and D. Miklavcic, “Dependence of induced [39] R. C. Barr, R. Plonsey, and C. R. Johnson, “Membrane current from
transmembrane potential on cell density, arrangement and cell position transmembrane potentials in complex core-conductor models,” IEEE
inside a cell system,” IEEE Trans. Biomed. Eng., vol. 49, no. 6, pp. Trans. Biomed. Eng., vol. 50, no. 4, pp. 405–411, Apr. 2003.
605–612, Jun. 2002. [40] R. Plonsey and R. C. Barr, Bioeleciricity: a Quantitative Approach.
[13] D. A. Stewart, T. R. Gowrishankar, K. C. Smith, and J. C. Weaver, New York: Kluwer Academic/Plenum, 2000.
“Cylindrical cell membranes in uniform applied electric fields: valida- [41] E. Neumann, A. E. Sowers, and C. A, Jordan, Electroporation and
tion of a transport lattice method,” IEEE Trans. Biomed. Eng., vol. 52, Electrofusion in Cell Biology. Norwell, MA: Kluwer Academic.,
no. 10, pp. 1643–1653, Oct. 2005. 1989.
[14] C. S. Peskin, “Numerical analysis of blood flow in the heart,” J. [42] J. C. Weaver, “Electroporation; a general phenomenon for manipu-
Comput. Phys., vol. 25, pp. 220–252, 1977. lating cells and tissues,” Cell. Biochem., vol. 51, no. 4, pp. 426–435,
[15] R. J. LeVeque and Z. Li, “The immersed interface method for elliptic 1993.
equations with discontinuous coefficients and singular sources,” SIAM [43] K. A. DeBruin and W. Krassowska, “Electroporation and shock-in-
J. Numer. Anal., vol. 31, no. 4, pp. 1019–1044, 1994. duced transmembrane potential in a cardiac fiber during defibrillation
[16] D. Semrov and D. Miklavcic, “Numerical modeling for in vivo elec- strength shocks,” Ann. Biomed. Eng., vol. 26, no. 4, pp. 584–596, 1998.
troporation,” in Electrochemotherapy, Electrogenetherapy and Trans- [44] ——, “Modeling electroporation in a single cell: I. effects of field
dermal Drug Delivery: Electrically Mediated Delivery of Molecules to strength and rest potential,” Biophys. J., vol. 77, no. 3, pp. 1213–1224,
Cells, M. J. Jaroszeski, R. Heller, and R. Gilbert, Eds. Totowa, NJ: 1999.
Humana, 1999, pp. 63–81. [45] ——, “Modeling electroporation in a single cell: II. Effects of ionic
[17] R. C. Penland, D. M. Harrild, and C. S. Henriquez, “Modeling impulse concentration,” Biophys. J., vol. 77, pp. 1225–1233, 1999.
propagation and extracellular potential distributions in anisotropic car- [46] J. C. Weaver, “Electroporation of cells and tissues,” IEEE Trans.
diac tissue using a finite volume element discretization,” Comput. Vi- Plasma Sci., vol. 28, no. 1, pp. 24–33, Feb. 2000.
sual. Sci., vol. 4, pp. 215–226, 2002. [47] K. C. Smith, J. C. Neu, and W. Krassowska, “Model of creation and
[18] S. Zozor, O. Blanc, V. Jacquemet, N. Virag, J. M. Vesin, E. Pruvot, L. evolution of stable electropores for dna delivery,” Biophys. J., vol. 86,
Kappenberger, and C. S. Henriquez, “A numerical scheme for mod- no. 5, pp. 2813–2826, 2004.
eling wavefront propagation on a monolayer of arbitrary geometry,” [48] T. R. Gowrishankar, A. T. Esser, Z. Vasilkoski, K. C. Smith, and J. C.
IEEE Trans. Biomed. Eng., vol. 50, no. 44, pp. 412–420, Apr. 2003. Weaver, “Microdosimetry for conventional and supra-electroporation
[19] L. J. Leon and F. A. Roberge, “A new cable model formulation based in cells with organelles,” Biochem. Biophys. Res. Commun., vol. 341,
on Green’s theorem,” Ann. Biomed. Eng., vol. 18, no. 1, pp. 1–17, 1990. pp. 1266–1276, 2006.
[20] ——, “A model study of extracellular stimulation of cardiac cells,”
IEEE Trans. Biomed, Eng., vol. 40, , no. 12, pp. 1307–1319, Dec. 1993. Wenjun Ying was born in 1976 in Zhejiang, China.
[21] K. R. Foster and A. E. Sowers, “Dielectrophoretic forces and potentials He received the B.S. degree and the M.S. degrees in
induced on pairs of cells in an electric field,” Biophys. J., vol. 69, no. applied mathematics from Tsinghua University, Bei-
3, pp. 777–784, 1995. jing, China, in 1997 and 2000, respectively. He re-
[22] E. C. Fear and M. A. Stuchly, “Modeling assemblies of biological cells ceived the Ph.D. degree in mathematics from Duke
exposed to electric fields,” IEEE Trans. Biomed. Eng., vol. 45, no. 10, University, Durham, NC, in May, 2005. His Ph.D. de-
pp. 1259–1271, Oct. 1998. gree dissertation is on the application of a multilevel
[23] ——, “A novel equivalent circuit model for gap-connected cells,” Phys. adaptive method to modeling electrical wave propa-
Med. Biol., vol. 43, no. 6, pp. 1439–1448, 1998. gation in the heart.
[24] A. Ramos, A. Raizer, and J. L. Marques, “A new computational ap- He currently works as a Research Associate at the
proach for electrical analysis of biological tissues,” Bioelectrochem- Computational Electrophysiology Laboratory in the
istry, vol. 59, no. 1–2, pp. 73–84, 2003. Department of Biomedical Engineering, Duke University. His research interests
[25] T. R. Gowrishankar and J. C. Weaver, “An approach to electrical mod- include computational electrophysiology, scientific computing and numerical
eling of single and multiple cells,” Proc. Nat. Acad. Sci. USA, vol. 100, analysis.
no. 6, pp. 3203–3208, 2003.
[26] D. A. Stewart, T. R. Gowrishankar, and J. C. Weaver, “Transport lattice
approach to describing cell electroporation: use of a local asymptotic
model,” IEEE Trans. Plasma Sci., vol. 32, no. 4, pp. 1696–1708, Aug. Craig S. Henriquez received the B.S.E. degree in
2004. biomedical and electrical engineering from Duke
[27] ——, “Three dimensional transport lattice model for describing action University, Durham, NC, in 1981. After teaching
potentials in axons stimulated by external electrodes,” Bioelectrochem-
high school for two years, he returned to Duke and
istry, 2006, (Jan 26, Epub ahead of print). received the Ph.D. degree in biomedical engineering
[28] R. A. Adams, Ed., Sobolev Spaces. New York: Academic, 1975.
in 1988.
[29] I. Babuska, “The finite element method with lagrangian multipliers,” He became a Research Assistant Professor in
Numer. Math., vol. 20, pp. 179–192, 1973.
1989 and an Assistant Professor of Biomedical
[30] F. Brezzi and M. Fortin, Mixed and Hybrid Finite Element Methods. Engineering in 1991 at Duke University. He is
New York: Springer, 1991. currently the Jeffrey N. Vinik Professor of Biomed-
[31] J. Huang and J. Zhou, “A mortar element method for elliptic prob-
ical Engineering and Computer Science at Duke
lems with discontinuous coefficients,” IMA J. Numer. Anal., vol. 22, University. He is also the Co-Director of Duke’s Center for Neuroengineering.
pp. 549–576, 2002.
His research interests include cardiac and neural electrophysiology, large-scale
computer modeling, and neural analysis.