Neuron Modeling Rattay-Greenberg-Resatz

1100_ch03.
fm Page 39 Wednesday, October 16, 2002 8:45 AM
chapter three
Neuron modeling
Frank Rattay, Robert J. Greenberg, and Susanne Resatz
Contents
3.1 Introduction
3.2 Models for the cell membrane
3.2.1 Axon models of the Hodgkin–Huxley type
3.2.2 Influence of temperature
3.2.3 Compartment models
3.2.4 Activating function
3.2.5 Activating function for long fibers
3.2.6 Membrane current fluctuations
3.3 The electrically stimulated retina
3.4 Concluding remarks
References
3.1 Introduction
A breakthrough in our understanding of the physics of neural signals, which
propagate as membrane voltage along a nerve fiber (axon), was achieved by
the ingenious work of Hodgkin and Huxley1 on the nonmyelinated squid
axon. This work helped explain the action potential or “spike” that conveys
the all-or-none response of neurons. To explore the complicated gating mech-
anism of the ion channels, the stimulating electrode was a long uninsulated
wire, thus every part of the neural membrane had to react in the same way;
that is, propagation of signals was prevented (Figure 3.1A). Refinements of
their method as well as the application of patch clamp techniques have
supplied us with models for different neural cell membranes. Reliable pre-
diction of membrane voltage V as a function of time is possible for arbitrary
stimulating currents Istimulus with proper membrane models.
©2003 CRC Press LLC

1100_Book Page 40 Tuesday, October 15, 2002 3:44 PM
A UNMYELINATED AXON
ELECTRODE
Hodgkin-Huxley B SPHERICAL CELL

membrane dynamics ELECTRODE
C
50mV
Ve Vi
5ms
Constant membrane conductance
Istimulus=100pA
Figure 3.1 Stimulation without signal propagation (space clamp). A stimulating elec-
trode in the form of an uninsulated wire inserted along an unmyelinated nerve fiber
(A) or current injection in a spherical cell (B) causes the same inside voltage, Vi, for
every part of the cell membrane. (C) In the first subthreshold phase, the voltage
across the cell membrane follows the simple constant membrane conductance model
(dashed line). Figure shows simulation of a 30-µm spherical cell sheltered by a
membrane with squid axon ion channel distribution (i.e., by solving the
Hodgkin–Huxley model with original data). Membrane voltage V is the difference
between internal and external voltage: V = Vi – Ve; Ve = 0.
The main equation for internal stimulation of the soma, or any other
compartment where current flow to other processes or neighbored compart-
ments is prevented, always has the same form: One part of the stimulating
current is used to load the capacity Cm (in Farads) of the cell membrane and
the other part passes through the ion channels; that is,
dV
I stimulus = Cm + I ion (3.1)
dt
The rate of membrane voltage change, dV/dt, follows as:
dV
dt
[ ]
= − I ion + I stimulus / Cm (3.2)
where the ion currents Iion are calculated from appropriate membrane mod-
els. Usually, the membrane models are formulated for 1 cm2 of cell membrane
and the currents in Eq. (3.1) become current densities.

A positive stimulating current applied at the inside of an axon or at any

other part of a neuron will cause an increase of V according to Eq. (3.1), if
the membrane has been in the resting state (Istimulus = 0 and dV/dt = 0) before.
In order to generate a spike, this positive stimulus current has to be strong
enough for the membrane voltage to reach a threshold voltage, which causes
many of the voltage-sensitive sodium channels to open. By sodium current
influx, the transmembrane potential increases to an action potential without
the need of further stimulating support. This means that as soon as the solid
line in Figure 3.1C is some few millivolts above the dashed line we can
switch off the stimulus without seeing any remarkable change in the shape
of the action potential.
In general the excitation process along neural structures is more com-
plicated than under space clamp conditions as shown in Figure 3.1. Current
influx across the cell membrane in one region influences the neighboring
sections and causes effects such as spike propagation (Figure 3.2). Besides
modeling the natural signaling, the analysis of compartment models helps
to explain the influences of applied electric or magnetic fields on represen-
tative target neurons. Typically, such a model neuron consists of functional
subunits with different electrical membrane characteristics: dendrite, cell
body (soma), initial segment, myelinated nerve fiber (axon), and nonmyeli-
nated terminal. Plenty of literature exists on stimulated fibers, but little has
been written about external stimulation of complete neurons.
In 1976, McNeal2 presented a compartment model for a myelinated nerve
fiber and its response to external point-source stimulation. He inspired many
authors to expand his model for functional electrical stimulation of the
peripheral nerve system, such as analysis of external fiber stimulation by
the activating function,3,4 unidirectional propagation of action potentials,5
stimulation of a nerve fiber within a6,7 selective axon stimulation,8,9,10 and
influence of fiber ending.11 Of specific interest is simulation of the threshold
and place of spike initiation generated with stimulating electrodes (e.g., by
the near field of a point source or dipole) by finite element calculation for a
specific implanted device or when the farfield influence from surface elec-
trodes is approximated by a constant field.12 Stronger electric stimuli or
application of alternating currents generate new effects in neural tissue. All
effects depend essentially on the electric properties of the neural cell mem-
brane and can be studied with compartment modeling. The ion channel
dynamics can be neglected during the first response of the resting cell, but
the complicated nonlinear membrane conductance becomes dominant in the
supra-threshold phase (Figure 3.1C). Consequently, the behavior can be ana-
lyzed with simple linear models or with more computational effort by sys-
tems of differential equations that describe the ion channel mechanisms in
every compartment individually.
Modeling the sub-threshold neural membrane with constant conduc-
tances allows the analysis of the first phase of the excitation process by the
activating function as a rough approach. Without inclusion of the compli-
cated ion channel dynamics, the activating function concept explains the

Figure 3.2 Simulation of the natural excitation of a human cochlear neuron by stim-
ulation with a 50-pA, 250-µsec current pulse injected at the peripheral end. This
bipolar cell is a non-typical neuron: (1) the dendrite is myelinated and often called
peripheral axon; (2) in contrast to animal cochlear neurons, the soma and the pre-
and post-somatic regions are unmyelinated; (3) a single synaptic input from the
auditory receptor cell (inner hair cell) generates a spike that propagates with a
remarkable delay across the current consuming somatic region. Note the decay of
the action potential in every internode that again is amplified in the next node of
Ranvier. Simulation uses internode with constant membrane conductance, a “warm”
Hodgkin–Huxley model (k = 12) in the active membranes of the soma, and a 10-fold
ion channel density in the peripheral terminal (node 0), all nodes, and pre- and post-
somatic compartments. The lines are vertically shifted according to their real location
of the rectified neuron. For details, see Rattay et al.35
basic mechanism of external stimulation, the essential differences between

anodic and cathodic threshold values, its dependence on the geometric sit-
uation, the mechanism of one side firing, and the blocking of spike propa-
gation by hyperpolarized regions, as well as several other phenomena.13,14
Beside nerve fiber analysis, the activating function, which represents the
direct influence of the electric field in every compartment, is useful for
magnetic stimulation15,16 and for direct stimulation of denervated muscle
fibers17 or, in generalized form, cardiac tissue18,19 and arbitrary neurons.20,12
In previous work we have shown how the shape of a neuron affects the
excitation characteristics and that several surprising phenomena may occur.

This is demonstrated, for example, by comparing the threshold current for

neurons with small and large soma. Whereas thick axons are easier to stimulate
than thin ones (known as the inverse recruitment order21), this relation does
not hold for the size of the soma: a 100-µs pulse from a point electrode 430
µm above the soma requires –2.3 mA to stimulate a 30-µm-diameter soma
neuron with a 2-µm axon but only –1.1 mA for a 10-µm soma. Increasing the
electrode distance to 1000 µm results in the same –2.2-mA threshold for both
cases; that is, the second surprise is that for the large soma the neuron excita-
tion threshold increases slightly when the electrode is moved within a specific
range toward the soma.12 The explanation is that the axon, which is generally
more excitable than the soma, loses more current to load the larger capacity
of the large soma; this effect is more pronounced when the place of spike
initiation within the axon is close to the soma.
Comparing a pyramidal cell with a cochlear neuron demonstrates the
variety in architecture and signal processing principles in the dendrite and
soma region. Some neuron types, such as the afferent bipolar cochlear neurons,
are effective transmission lines, where nearly all of the spikes initiated at the
synaptic contact with an inner hair cell arrive with some delay and small
temporal variation (jitter) in the terminal region (Figure 3.2). In contrast to the
cochlear neuron, a single synaptic input at the dendritic tree of a pyramidal
cell produces only minimal change in membrane voltage at the soma and at
the initial segment, and usually the collective effect of many synaptic activities
is necessary to influence spike initiation significantly. Typically, the pyramidal
cell response is dominated by the internal calcium concentration, which
depends on two types of voltage-dependent calcium channels: (1) low-voltage-
activated channels respond in the subthreshold range and may include gen-
eration of low-threshold spikes, and (2) high-voltage-activated channels in the
dendrites respond, for example, to backpropagating sodium spikes.22–24
3.2 Models for the cell membrane

The cell membrane is not a perfect insulator for ion transport. When one Na
channel opens for a short time, some thousand sodium ions may be driven
into the cell by the high outside concentration. The second driving element
is the voltage across the membrane, which requires a Na+ current according
to Ohms law. The conductance of a patch of cell membrane for a specific ion
type depends on the number of open channels and is defined by Ohm’s law
and a battery voltage according to the Nernst equation. A different internal
concentration, ci, and external ion concentration, ce, cause the membrane
voltage, Em, when one type of ions is involved:
RT c e
Em = ln (3.3)
zF ci
where the gas constant R = 8.31441 J/(mol.K); temperature, T, is in Kelvin;

z is valence and the Faraday constant represents the charge of one mol of

single-valenced ions F = 96485 c/mol. Note: At room temperature (T =

20°C = 293.15 K), the factor RT/F is about 25 mV.
In their experiments Hodgkin and Huxley1 found the sodium current
proportional to E – ENa. Approximating the Nernst equation [Eq. (3.3)] with
their data gives ENa = 45 mV = (25 mV)ln(ce/ci) and results in ce/ci =
exp(45/25) = 6.05; that is, the outside sodium concentration is six times the
inner one, and EK = –82 mV = (25 mV)ln(ce/ci) requires that the potassium
outside:inside concentration is 1:26.5.
The Goldman equation defines the steady state membrane voltage Em =
Erest when several ion types are involved. When K+, Na+, and Cl– ions are
considered it reads as:
RT PK [K ]e + PNa [Na]e + PCl [Cl]i

Em = ln (3.4)
F PK [K ]i + PNa [Na]i + PCl [Cl]e
where [K] is the potassium concentration and the suffixes i and e stand for
inside and external, respectively. PK, PNa, and PCl are permeabilities measured
in centimeters per second (cm/sec). Note that sodium and potassium are
anions, but chloride is cathodic, therefore, [Cl]i appears in the numerator in
contrast to the anionic concentrations. In the resting state, the membrane is
most permeable to potassium ions, and the resting membrane voltage of
about –70 mV (i.e., the inside is more negative compared to the extracellular
fluid) is close to the potassium Nernst potential.
The nonlinear conductance of the neural cell membrane depends on
different classes of ion channels25 (1 through 3, below) and on the activity
of ion pumps (4, below):
1. Voltage-dependent gating dominates excitation and neural signal

propagation in the axon; open/close kinetics depend on the voltage
across the cell membrane.
2. Calcium-dependent gating occurs mainly at the soma and dendrites;
the calcium concentration regulates the opening of the channels as
opening depends on intracellular calcium ion binding.
3. Transmitter gating and second messenger gating occurs at the pre-
and postsynaptic membrane.
4. Ion pumps are membrane molecules that consume energy pump ions
across the cell membrane in order to restore the high individual ion
concentration on one side of the cell (e.g., high external sodium and
high internal potassium concentration).
3.2.1 Axon models of the Hodgkin–Huxley type

The Hodgkin–Huxley model (HH model) was developed from a homoge-
neous nonmyelinated squid axon. It includes sodium, potassium, and leak-
age currents and has the form:

dV
dt
[ ]
= − g Na m 3 h(V − VNa ) − g K n 4 (V − VK ) − g L (V − VL ) + istimulus / c (3.5)
dm
dt
[
= −(α m + β m )m + α m k ] (3.6)
dh
dt
[
= −( α h + β h ) h + α h k ] (3.7)
dn
dt
[
= −(α n + β n )n + α n k ] (3.8)
k = 3 0.1T −0.63 (3.9)
where V is the reduced membrane voltage, resulting from internal, external,

and resting potential: V = Vi – Ve – Vrest; gNa, gK, and gL are the maximum
conductances for sodium, potassium, and leakage per cm2, respectively; m,
h, and n are probabilities (with values between 0 and 1) that reduce the
maximum conductances of sodium and potassium according to experimental
gating data; VNa, VK, and VL are the sodium, potassium, and leakage battery
voltages, respectively, according to the Nernst equation [Eq. (3.3)]; istimulus is
the stimulus current (µA/cm2); c is the membrane capacity per cm2; α and β
are voltage-dependent variables fitted from experimental data to quantify the
ion channel kinetics; k is a temperature coefficient that accelerates the gating
process for temperatures higher than the original experimental temperature
of 6.3°C; and temperature T is in °C. Although a temperature conversion is
possible, it is interesting to note that, unlike real tissue, the Hodgkin–Huxley
equations do not propagate action potentials above 31°C.14 Parameter values,
units, and further expressions of the HH model are listed in Table 3.1.
In 1964, Frankenhaeuser and Huxley26 developed a model for the myeli-
nated frog axon node (FH model) assuming HH-like gating mechanisms,
but they derived the ion current formulation from the Nernst–Planck equa-
tion and added a nonspecific current density, iP:
dV
= [−iNa − iK − iP − iL + istimulus ]/ c (3.10)
dt
EF 2 [Na]o − [Na]i exp(EF / RT )

iNa = PNa m 2 h (3.11)
RT 1 − exp(EF / RT )
EF 2 [K ]o − [K ]i exp(EF / RT )
iK = PK n 2 (3.12)

46
Table 3.1 Expressions and Constants for Axon Membrane Models
HH Model FH Model CRRSS Model SE Model SRB Model
αm 0.36(V − 22) 97 + 0.363V 1.87(V − 25.41) 4.6(V − 65.6)
2.5 − 0.1V
 22 − V   31 − V   25.41 − V   −V + 65.6 
exp(2.5 − 0.1V ) − 1 1 − exp  1 + exp  1 − exp  1 − exp 
 3   5.3   6.06   10.3 
βm 0.4(13 − V ) αm 3.97(21 − V ) 0.33(61.3 − V )
 V
4.exp −   V − 13   V − 23.8   V − 21   V − 61.3 
 18  1 − exp  exp  1 − exp  1 − exp 
 20   4.17   9.41   9.16 
αn 0.02(V − 35) 0.13(V − 35) 0.0517(V + 9.2)
0.1 − 0.01V
 35 − V   35 − V   −V − 9.2 
exp(1 − 0.1V ) − 1 1 − exp  1 − exp  1 − exp 
 10   10   1.1 
βn 0.05(10 − V ) 0.32(10 − V ) 0.092(8 − V )
 V
0.125.exp −   V − 10   V − 10   V − 8
 80  1 − exp  1 − exp  1 − exp 
 10   10   10.5 
αh 0.1(V + 10) βh 0.55(V + 27.74) 0.21(V + 27 )
Handbook of Neuroprosthetic Methods

 V − − −
0.07.exp −   V + 10   V − 5.5   V + 27.74   V + 27 
 20  1 − exp  exp  1 − exp  1 − exp 
 6   5   9.06   11 
βh 4.5 15.6 22.6 14.1
1
 45 − V   24 − V   56 − V   55.2 − V 
exp(3 − 0.1V ) + 1 1 + exp  1 + exp  1 + exp  1 + exp 
 10   10   12.5   13.4 
αp 0.006(V − 40) 0.0079(V − 71.5)
 40 − V   71.5 − V 
1 − exp  1 − exp 
 10   23.6 
βp 0.09(V + 25) 0.00478(V − 3.9)
− −
 V + 25   V − 3.9 
1 − exp  1 − exp 
 20   21.8 
Vrest (mV) –70 –70 –80 –78 –84

VNa (mV) 115 — 115 — —
VK (mV) –12 — — — 0
VL (mV) 10.6 0.026 –0.01 0 0
gNa (kΩ–1/cm–2) 120 — 1445 — —
gK,fast (kΩ–1/cm–2) 36 — — — 30
gK,slow (kΩ–1/cm–2) — — — — 60
gL (kΩ–1/cm–2) 0.3 30.3 128 86 60
c (kΩ–1/cm–2) 1 2 2.5 2.8 2.8
V(0) 0 0 0 0 0
m(0) 0.05 0.0005 0.003 0.0077 0.0382
n(0) 0.32 0.0268 — 0.0267 0.2563
h(0) 0.6 0.8249 0.75 0.76 0.6986
p(0) — 0.0049 — — 0.0049
T0 6.3°C 293.15 K = 20°C 37°C 310.15 K = 37°C 310.15 K = 37°C
Q10(αm) 3 1.8 3 2.2 2.2
Q10(βm) 3 1.7 3 2.2 2.2
Q10(αn) 3 3.2 3 3 3
Q10(βn) 3 2.8 3 3 3
Q10(αh) 3 2.8 3 2.9 2.9
Q10(βh) 3 2.9 3 2.9 2.9
PNa (cm/s) — 0.008 — 0.00328 0.00704
PK (cm/s) — 0.0012 — 0.000134 —
PP (cm/s) — 0.00054 — — —
[Na]o (mmol/l) — 114.5 — 154 154
[Na]i (mmol/l) — 13.7 — 8.71 30
[K]o (mmol/l) — 2.5 — 5.9 —
[K]i (mmol/l) — 120 — 155 —
Note: HH model = Hodgkin–Huxley model; FH model = Frankenhaeuser–Huxley; CRRSS model = Chiu, Ritchie, Rogert, Stagg, and Sweeney model;
SE model = Schwarz–Eikhof model; and SRB model = Schwarz, Reid, and Bostock model. Faraday constant F = 96485 C/mol; gas constant R =
8314.4 mJ/(mol.K).


iP = PP p 2 (3.13)
iL = g L (V − VL ) (3.14)
E = V + Vrest (3.15)
dm
= −(α m + β m )m + α m (3.16)
dt
dn
= −(α n + β n )n + α n (3.17)
dt
dh
= −( α h + β h ) h + α h (3.18)
dt
dp
= −( α p + β p ) p + α p (3.19)
dt
Note that membrane voltage is denoted as E in Eqs. (3.11) to (3.13) and

as V (reduced membrane voltage) in Eqs. (3.10) and (3.14). Temperature T
is measured in degrees K; for temperatures other than 20°C = 293.15 K, the
α and β values in Eqs. (3.16) to (3.19) must be modified. Parameter values
and further expressions of the FH model are listed in Table 3.1. Sodium
current plays the dominant role in the action potential of the mammalian
node of Ranvier, and in contrast to the axon model of squid and frog there
are almost no potassium currents.27–29
The CRRSS model, named after Chiu, Ritchie, Rogert, Stagg, and
Sweeney, describes a myelinated rabbit nerve node extrapolated from orginal
14°C data to 37°C:28,30
dV
dt
[ ]
= − g Na m 2 h(V − VNa ) − g L (V − VL ) + istimulus / c (3.20)
dm
dt
[
= −(α m + β m )m + α m k ] (3.21)
dh
dt
[
= −( α h + β h ) h + α h k ] (3.22)
k = 3 0.1T − 3.7 (3.23)

Note that the temperature factor k = 1 for T = 37°C. Parameter values and
further expressions of the CRRSS model are listed in Table 3.1.
Schwarz and Eikhof obtained a model of FH type from voltage clamp
experiments on rat nodes.29 From the original data, the Schwarz–Eikhof (SE)
model results by assuming a nodal area of 50 µm2.14,31
dV
= [−iNa − iK − iL + istimulus ]/ c (3.24)
dt

iNa = PNa m 3 h (3.25)
EF 2 [K ]o − [K ]i exp(EF / RT )
iK = PK n 2 (3.26)
iL = g L (V − VL ) (3.27)
E = V + Vrest (3.28)
dm
= −(α m + β m )m + α m (3.29)
dt
dn
= −(α n + β n )n + α n (3.30)
dt
dh
= −( α h + β h ) h + α h (3.31)
dt
Parameter values and additional expressions of the SE model are listed in

Table 3.1.
Schwarz et al.33 derived the SRB model from human nerve fibers at room
temperature. Single-action potentials are little effected by removing the fast
or slow potassium currents, but a slow K conductance was required to limit
the repetitive response of the model to prolonged stimulating currents. SRB
model follows:
dV
= [−iNa − iK , fast − iK ,slow − iL + istimulus ]/ c (3.32)
dt

iNa = PNa m 3 h (3.33)

iK , fast = g K n 4 (V − VK ) (3.34)
iK ,slow = g K ,slow p(V − VK ) (3.35)
iL = g L (V − VL ) (3.36)
E = V + Vrest (3.37)
dm
= −(α m + β m )m + α m (3.38)
dt
dn
= −(α n + β n )n + α n (3.39)
dt
dh
= −( α h + β h ) h + α h (3.40)
dt
dp
= −( α p + β p ) p + α p (3.41)
dt
Additional SRB model data for 37°C are listed in Table 3.1. (Axon mem-
brane models for biomedical applications are further discussed in References
14 and 31 through 36.)
Some authors neglect the weak K currents in the mammalian axon
totally because of their small amounts, but up to five types of K ion
channels are shown to have notable influences on axonal signaling, espe-
cially in the internode, an element often modeled as perfect insulator with
capacity C = 0. Curious effects such as spontaneous switching between
high and low threshold states in motor axons are suggested to be a
consequence of K+ loading within small internodal spaces.37 Phenomena
such as different excitability fluctuations after action potentials in sensory
and motor neurons require far more sophisticated modeling. This means
that for many applications the modeler’s work is not finished by selecting
the SRB model (as it is based on human data) or by fitting its parameters
according to the shape and propagation velocity of a measured action
potential.
The dendrite and especially the soma region is more difficult to simulate
because a variety of ion channel types are involved.37 Some membrane mod-
els are available,39–45 but even in these models we cannot rely on a precise
individual measurement of all the ion components that vary in channel
density and open/close kinetics. Repetitive firing (bursting) may occur as
response to stimulation or in pacemaker neurons without any input.

1100_ch03.fm Page 51 Wednesday, October 16, 2002 8:47 AM
3.2.2 Influence of temperature

Most of the membrane model data were gathered at low temperatures, but
for biomedical applications they have to be extrapolated to physiological
temperatures. Usually, a specific constant, Q10, is introduced to account for
the acceleration in membrane-gating dynamics when the temperature is
increased by 10°C. Hodgkin and Huxley used Q10 = 3 as a common coeffi-
cient for all gating variables m, n, and h [Eqs. (3.6)–(3.8)]:
( T −T0 )/10
k = Q10 (3.42)
Raising the temperature causes shortening of the action potential and an

increase of the spike propagation velocity in the unmyelinated squid axon.
The spike duration in Figure 3.1 with original HH data (6.3°C) is about ten
times longer than that of the cochlea neuron in Figure 3.2, which also was
simulated with HH kinetics. The temporal membrane behavior in cat46 was
fitted with k = 12, corresponding to a temperature of 29°C (Eq. (3.42)); see
Motz and Rattay47 and Rattay,14 p. 165). The HH spike amplitude becomes
rather small for higher temperatures. An essentially reduced amplitude is
not able to preserve the excitation process along the fiber, and for tempera-
tures higher than 31 to 33°C action potentials will not propagate any more
in squid axons (heat block).14,48,49 Such amplitude reductions are not seen in
the membrane models of myelinated axons (FH, CRRSS, SE, and SRB mod-
els), and the heat block phenomenon is absent both in myelinated and
unmyelinated fibers of warm-blooded animals. In all axon models, spike
duration shortens considerably when temperature is increased — for exam-
ple, from 20°C (usual for data acquisition) to 37°C.14,32 A temperature factor,
k = 12, in the HH-model fits the temporal characteristics of action potentials
in unmyelinated axons of humans and warm-blooded animals at 37°C.14,35,50
The original FH model51 incorporates thermal molecular motion accord-
ing to the laws of gas dynamics [Eqs. (3.11)–(3.13)], but the authors did not
include the necessary Q10 values presented in Table 3.1 for the gating variables
m, n, h, and p. The gating process becomes essentially accelerated for high
temperatures and dominates spike duration.14,32,48,52 Also, the threshold cur-
rents are influenced by the temperature: The warmer membrane is easier to
excite (Figure 3.3). Therefore, the Q10 factors have to be involved in functional
electrical nerve stimulation modeling. Note that many published results
obtained with the original FH model must be corrected to be valid for 37°C.
3.2.3 Compartment models

Small pieces of a neuron can be treated as isopotential elements and a whole
neuron is represented by an electric network (Figure 3.4). A current injected
at the nth compartment has to cross the membrane as capacitive or ionic
current and leave to the left or right side as longitudinal current; that is,
application of Kirchhoff’s law is an extension of Eq. (3.1):

1000
Ielectrode /d
[µA/µm]
20°C
100
37°C
20°C
2*RHEOBASE
37°C
RHEOBASE
10
0.01 PULSE DURATION 1ms
CHRONAXIE 0.1
Figure 3.3 Threshold currents as a function of pulse duration (strength duration

relation). A myelinated fiber of diameter d and internodal length ∆x = 100d is stim-
ulated by a point source located at a distance ∆x above a node. Excitation is easier
for high temperatures (full lines) and cathodic pulse. The threshold current for long
pulses becomes time-independent and is called rheobase. The pulse duration belong-
ing to the doubled rheobase current is called chronaxie. Simulations use the FH model;
internode membrane is assumed to be a perfect insulator, with c = 0.
DENDRITE
SOMA
INITIAL SEGMENT
NODE UNMYELINATED
INTERNODE AXON
Ve,0
Ve,1 Ve,2
Cm
Gm OUTSIDE
MEMBRANE
R/2 R/2 INSIDE
Vi,0 Vi,1 Vi,2
Figure 3.4 Scheme of a neuron with subunits and part of the simplified equivalent
electrical network (batteries resulting from different ion concentrations on both sides
of the membrane are not shown).

d(Vi ,n − Ve ,n ) Vi ,n − Vi ,n−1 Vi ,n − Vi ,n+1

I injected ,n = Cn + I ion ,n + + (3.43)
dt Rn / 2 + Rn−1 / 2 Rn / 2 + Rn+1 / 2
where n indicates the nth compartment, C is its membrane capacity, and R/2
is the internal resistance between the center and the border of the compart-
ment. Introducing the reduced membrane voltage V = Vi – Ve – Vrest leads
to the following system of differential equations for calculating the time
courses of Vn in every compartment:
Vn − 1 − Vn Vn + 1 − Vn
dVn
dt
[
= − Iion , n + +
Rn − 1 / 2 + Rn / 2 Rn + 1 / 2 + Rn / 2
+ ...
(3.44)
Ve , n − 1 − Ve , n Ve , n + 1 − Ve , n
+ +
Rn − 1 / 2 + Rn / 2 Rn + 1 / 2 + Rn / 2
+ ... + Iinjected , n ]/C n
The dots in Eq. (3.44) stand for terms that have to be added in cases of
more than two neighbor elements (e.g., at the cell body [soma] or at branch-
ing regions). For the first and last compartments, Eq. (3.44) has a reduced
form. The membrane surface, An, of every compartment has to be calculated
to find Cn = Ancn (cn is the specific membrane capacitance) and Iion = Aniion.
The ionic membrane current density, iion, is computed with an appropriate
membrane model; for cylinder elements (d, diameter; ∆x, length), we obtain
An = dnπ∆xn, R / 2 = 2ρi ∆xn / (dn2 π) , where the internal resistivity ρi is often
assumed as ρi = 0.1 kΩcm.
In a spherical cell body with several processes (as in Figure 3.4), the
internal resistances to the neighbor compartments depend on the compart-
ment diameters; that is, Rsoma→dendrite1/2 < Rsoma→axon/2 if ddendrite1 > daxon. In more
detail, Asoma = 4rsoma2π − Σ (2rsomaπhj), with subscript j indicating the jth process)
such that hj = rsoma – zj, where zj = rsoma 2 − (dprocess , j / 2) . The somatic resis-
2
tance to the border of the jth process is:
Rsoma , j ρ  rsoma + zj 
= i ln  (3.45)
2 2rπ  rsoma − zj 
The electric and geometric properties change from compartment to compart-

ment and the modeler has to decide about the degree of complexity that
should be involved. For example, the surface area of a dendrite compartment
and the soma can be simulated by a cylinder and a sphere or an additional
factor is included to represent the enlargement by spines or windings. Halter
and Clark53 and Ritchie54 presented detailed ion current data for the intern-
ode, whereas many authors neglect internodal membrane currents. A com-
promise is to simulate the internode as a single compartment with both
membrane conductance Gm and membrane capacity C proportional to 1/N,
where N is the number of myelin layers.12

Synaptic activation of a neuron can be simulated by current injection at

the soma or at dendrite compartments (Figure 3.2). In such cases, all external
potentials, Ve, are assumed to be 0 in Eq. (3.44). On the other hand, neuro-
prosthetic devices generate neural activities by application of electric fields.
Instead of current injection, the terms including the external potentials Ve
become the stimulating elements in Eq. (3.44).
3.2.4 Activating function

The driving term of the external potential on compartment n (Eq. (3.44)) is
called the activating function fn:
− Ve , n Ve , n + 1 − Ve , n
fn = [RV e, n − 1
n − 1 / 2 + Rn / 2
+
Rn + 1 / 2 + Rn / 2
+ ... ]/C
n (3.46)
The physical dimension of fn is V/sec or mV/msec. If the neuron is in the

resting state before a stimulating current impulse is applied, the activating
function represents the rate of membrane voltage change in every compart-
ment that is activated by the extracellular field. That is, fn is the slope of
membrane voltage Vn at the beginning of the stimulus. Regions with positive
activating function are candidates for spike initiation, whereas negative acti-
vating function values cause hyperpolarization.
The temporal pattern generated in a target neuron by a single active
electrode of a cochlear implant is analyzed in the following with the acti-
vating function concept. The electrode is assumed as an ideal point source
in an infinite homogeneous extracellular medium, which results in the
simple relation:
ρ e .I electrode
Ve = (3.47)
4 πr
where Ve is measured at a point with distance r from the point source and
ρe is the extracellular resistivity.
The insert of Figure 3.5A shows the values of Ve and f along a cochlear
neuron when stimulated with 1 mA from a monopolar electrode. The alter-
nating sequence of positive and negative activating function values indicates
first spike initiation at peripheral node P2 (Figure 3.5B) and a second spike
initiation region in the central axon for stronger positive monophasic stimuli
(Figure 3.5C). Cathodic stimulation also generates two regions of spike ini-
tiation — first, excitation in P1 (Figure 3.5D), then excitation in P4 (Figure
3.5E). The activating function predicts easier excitation with cathodic cur-
rents with values of –3890 V/sec vs. 2860 V/sec in P1 and P2, respectively.
Note, however, that f quantifies the very first response, which is flattened
by the capacity (compare dashed line in Figure 3.1) and is further influenced
by the neighbor elements.

Figure 3.5 Human auditory nerve response evoked by a stimulating impulse from a
cochlear implant. (A) Position of a cochlear neuron relative to a spherical electrode
(same geometry as in Figure 3.2; unmyelinated terminal P0, nodes P1–P5 of periph-
eral myelinated thin process, unmyelinated region around the soma, and a thicker
central axon). Extracellular resistivity ρe = 0.3 kΩcm, and calculation of potential from
a 1-mA point source [Eq. (3.47)] results in 1 V at r = 0.24 mm, which is the electrode
radius. The values of the activating function f shown in insert (A) are proportional
to the slopes of the neural response curves at stimulus onset in (B) and (C). The
largest f value of P2 predicts the place of spike initiation in (B). All slopes of membrane
voltage in the central nodes are positive and strengthen the stimulus current to 1.2
mA, causing additional spike initiation at peripheral node C2 (C). The activating
function value of P0 is also positive, but the much stronger negative activity in P1
compensates the excitation process quickly; insert (C) shows the part of the mem-
brane voltage of the P0 compartment during the stimulus pulse as marked by arrows,
with 5× magnification. In part (D), –800 µA per 100-µsec pulse initiates a spike at P1
which is in accordance with the activating function concept. In part (E), the smaller
intensity of f at P4 allows spike initiation at –1200 µA. The P4 spike develops after
the P1, but nevertheless the P4 spike activates the central axon. Note the different
arrival times at the measuring electrode C5.
3.2.5 Activating function for long fibers

In 1976, McNeal introduced the first efficient model for computing peripheral
nerve fiber responses in functional electrical nerve stimulation. Many
authors followed his simplification of an ideal internode membrane, where
both membrane conductance and membrane capacity are assumed to be 0.
In this case, the main equation, Eq. (3.44), reduces to:4,12
d∆x  Vn − 1 − 2Vn + Vn + 1 Ve , n − 1 − 2Ve , n + Ve , n + 1 

dVn
dt
[
= −iion , n + 
4ρi L  ∆x 2
+
∆x 2


]/ c (3.48)

with constant fiber diameter d, node-to-node distance ∆x, node length L, and
axoplasma resistivity ρi; both membrane capacity c and nodal ion current iion
are per cm2. The activating function:
d∆x Ve , n − 1 − 2Ve , n + Ve , n + 1
fn = (3.49)
4ρi Lc ∆x 2
becomes proportional to the second difference of the extracellular potential

along the fiber.
Equations (3.48) and (3.49) model the unmyelinated fiber when the
length of the active membrane is equal to the compartment length: L = ∆x.
With ∆x → 0, Eq. (3.49) reads as:
d ∂ Ve
2
f= (3.50)
4ρi c ∂x 2
Equations (3.49) and (3.50) verify the inverse recruitment order; that is,
thick fibers are easier to stimulate3,21 (Figure 3.5) by the linear relation
between d and f.
For a monopolar point source in an infinite homogeneous medium, the
activating function is easy to evaluate. For example, for a straight axon (Eq.
(3.47); Figure 3.6A,B), we obtain:
ρ e .I el
((x − x ) + z )
2 −0.5
Ve = 2
(3.51)
4π el el
where el stands for electrode and z measures the center-to-center distance

between a small spherical electrode and the axon, and (see Eq. (3.50)):
(
dρ e .I el
) (2(x − x ) − z )
−2.5
(x − xel ) + zel2
2 2
f= 2
(3.52)
16ρi cπ el el
Equation (3.50) implies the relation between the curvature of the Ve graph
and the regions of excitation in a target fiber; solving f(x) = 0 gives the points
of contraflexure of Ve that separate the regions of depolarization (f > 0) from
the hyperpolarized ones (f < 0). In our example, this results in the vertical
dashed lines at x = xel ± 0.707zel. In the excited (f > 0) regions, the Ve graph
is always above its tangents (Figure 3.6B; regions I and III for anodic stim-
ulation). Strong curvature of Ve means large f values; therefore, the central
part of region II becomes extremely hyperpolarized at anodic stimulation
(Figure 3.6C). In relation to the shape of the activating function, the threshold
for cathodic stimulation is about four times weaker (Figure 3.6E and F) which
is in accordance with experimental findings.55

ELECTRODE 2
A Z
r = (x-xelectrode)2+zelectrode
X 70∞
AXIS OF AXON
B 500mV
r eIelectrode
Ve= 2 2
1mm 4p (x-xelectrode) +zelectrode
Ielectrode= 6mA
C
Ielectrode= 1.5mA
D
E Ielectrode= -1.5mA
f
F
Ve
REGION I II III
Figure 3.6 Stimulation of an unmyelinated axon with a monopolar electrode. During

the 100-µsec stimulus pulse, the activating function defines three regions of responses,
separated by vertical dashed lines. (B)–(D) Vertically shifted snapshots of membrane
voltages in 100-µsec steps. Thick lines demonstrate the similarity between the acti-
vating function and the membrane voltage at the end of the 100-µsec stimulus signal.
Simulation is of HH data at 29°C, ρe = 0.3 kΩ, electrode fiber distance z = 1 mm, and
fiber diameter 10 µm.
Deviations between the thick lines for f and for Ve at the end of the 100-
µsec pulse (just above threshold) are caused by (1) an inner-axonal current
component resulting from membrane currents at the other compartments,
(2) exponential loading process of the membrane capacity (in the activating
function concept, the dashed line in Figure 3.1 is approximated by the tan-
gent at stimulus onset), and (3) the beginning of the voltage-sensitive ion
channel-gating processes. Compared to the prediction by f, there is a small
enlargement of the hyperpolarized region II seen at the zero-crossings of the
thick line in Figure 3.6C. This is mostly caused by the inner-axonal current,
where the strong hyperpolarized central part has a negative influence at the

borderlines to regions I and III. For the same reason the points of spike
initiation (Figure 3.6C) are lateral to the maximum points of f (Figure 3.6B).
With some computational effort, the linear theory can be improved by
considering the influences of the inner-axonal current and the exponential
loading process of the membrane capacity.56 Especially for close electrode
fiber distances, the subthreshold steady-state membrane voltage will con-
siderably deviate from the first response.57,58 But, the activating function
concept demonstrates the instructive relation between a simple formula and
observed phenomena. For example, Eq. (3.49) indicates that a straight part
of an axon in a constant field region is not stimulated directly; however, the
bending part in a constant-field region will be de- or hyperpolarized,
depending on the orientation of the axon and field. Such effects are studied
in spinal root stimulation (see Figure 3.3 of Reference 59).
Note that the equations of this section require constant data along the
fiber. At a fiber ending, widening or branching the treatment is according
to the description of Eq. (3.44). The fiber-end influence may be modeled
by a boundary field driving term in addition to the activating function.11
In some applications (heart and skeletal muscle, nerve bundle), many
excitable fibers are closely packed and the effects of current redistribution
are handled by the bidomain model. The tissue is not assumed to be a
discrete structure but rather two coupled, continuous domains: one for the
intracellular space and the other for the interstitial space. This space-
averaged model is described with a pair of coupled partial differential
equations that have to be solved simultaneously for the intracellular and
interstitial potentials.7,18,19,60 Despite the complex situation, several stimu-
lation phenomena of the heart muscle are already predicted by the activat-
ing function concept from single-fiber analysis.60
3.2.6 Membrane current fluctuations

Single-fiber recordings show variations in the firing delays even when stim-
ulated with current pulses of constant intensity.62,63 The main component of
the stochastic arrival times (jitter) is explained by ion channel current fluc-
tuations. Models of the Hodgkin–Huxley type define the gating variables m,
n, and h as probabilities, but no variance is included. The addition of a simple
noise term in single compartment modeling can fit several of measured
cochlear nerve data.47 Other phenomena — for example, a bimodal distri-
bution of spike latencies, as discussed later (Figure 3.7B) — require full
compartment analysis with a stochastic term in every active compartment.35
Including membrane current fluctuations is significant in the following
applications: (1) modeling temporal firing patterns (e.g., in the cochlear
nerve), (2) detailed analysis of fiber recruitment (for example, constant stim-
ulation conditions generate three types of axons within a nerve bundle: one
group fires at every stimulus, the second fires stochastically, and the third
shows solely subthreshold responses), (3) the transition from the myelinated
to the unmyelinated branching axon is less secure in regard to signal transfer,

Figure 3.7 Simulation of ion current fluctuations in the cochlear neuron membrane.
(A) Membrane voltage in the resting (left) and activated (right) neuron in a human
and cat. Stimulation by current injection at the lateral end is close to natural spike
initiation by synaptic activation. As a consequence of the RC circuit properties of
the network, the voltage fluctuations become low pass filtered and neighboring
compartment behavior is correlated. The unmyelinated terminals are assumed to
have the same 1-µm diameter with same ion channel density as the peripheral
nodes, but a length relation of 10 to 1.5 µm results in 2.58 times stronger membrane
current fluctuation at the endings. The strong influence of the first compartment
to the network is still seen in the cat soma region (left) but it is lost in the human
case because of the longer peripheral axon (five nodes and a large unmyelinated
soma region). Voltage fluctuations in the soma region are generally smaller than
in the nodes. Loading the large capacity of the unmyelinated human soma requires
the spike about 400 µsec to cross this region, which is three times the value of the
myelinated cat case (right). (B) Responses of the fifth node in the central human
axon (compartment 25 marked as C5 in Figure 3.5) to anodic external stimulation
with 100 µsec pulses (10 runs for every stimulus intensity; situation shown in Figure
3.5). Weak stimuli fail sometimes (600 µA: 8 of 10) and have larger jitter, and their
arrival is more delayed, as they are generated in the periphery. At 900 µA, spikes
are also initiated in the central axon. Note that this experimentally known bimodal
response is a consequence of ion channel current fluctuations. For details, see Rattay
et al.35 and Rattay.70
and spikes can stochastically be lost50,64, (4) during the integration of neural
inputs of an integrate-and-fire neuron.
The elaborate models of single-channel current influences are recom-
mendable for detailed investigations.65–67 Combining this work with a com-
partment model will show most of the noisy behavior in nerve fibers as

reported by Verween and Derksen.66 Membrane noise increases with devia-

tion from the resting voltage and is greater for hyperpolarization than for
depolarization. To reduce the enormous computational effort of simulating
every channel activity we follow a suggestion of Rubinstein69 and assume
the noise current to be proportional to the square root of the number of
sodium channels.35,70 At every compartment, a Gaussian noise variable,
GAUSS (mean = 0, standard deviation = 1), is introduced that changes its
value every 2.5 µsec. The amplitude of the noise term is proportional to the
square root of the number of sodium channels involved. Noisy membrane
currents (in µA) are defined as:
I noise ,n = GAUSS ⋅ knoise An ⋅ g Na (3.53)
where An denotes the membrane area of compartment n (in cm2), and gNa
(mS cm–2) is the maximum sodium conductance. A standard factor knoise =
0.05 µA.mS–1/2 common to all compartments fits the observations of Verveen
and Derksen.66
Simulated voltage fluctuations are compared for cochlear neurons in cats
and humans. Both neurons are assumed to have the same peripheral (1 µm)
and central (2 µm) axon diameters, but in the cat the peripheral axon is
shorter (distance from soma to peripheral terminal in cats is 0.3 mm; in
humans, 2.3 mm), and the cat somatic region is insulated by myelin sheets.
Only the 10-µm-long peripheral terminal and the nodes are without myelin;
additionally, in humans, the soma region is unmyelinated (Figure 3.2). The
following effects are seen in Figure 3.7A:
1. Thick fibers show less membrane fluctuations. The peripheral com-

partments (1 and 5 in cats, 1 and 11 in humans) have smaller noise
amplitudes than the central axon (11 in cats, 19 in humans). The
phenomenon observed by Verveen and Derksen66 can be explained
in a simple model with corresponding states of homogeneous my-
elinated axons. During the excitation process, the nodal currents
are proportional to axon diameter, but the noisy current increases
with the square root of diameter only (nodal length is independent
of diameter).
2. Membrane voltage fluctuations in one compartment have two com-
ponents: the own membrane current and the noisy inner currents
from the neighboring compartments. Note the correlated sinusoidal-
like shapes of the three lower left cat curves. The cat myelinated soma
is assumed to have no sodium channels; therefore, the displayed
membrane voltage has lost the first component and becomes the
averaged neighboring compartment noise filtered by the soma ca-
pacity effect.
And, as shown in Figure 3.7B:

3. Threshold voltage has to be redefined (e.g., as 50% probability of

generating a spike); effectiveness increases from 600 µA (20%) to 700
µA (100%).
4. Weak stimuli have larger jitter because of the longer noise influence
during the subthreshold part of excitation. Jitter of peripherally
evoked spikes decreases from 600 to 900 µA; the same for the central
ones — from 900 to 1200 µA.
5. Spike initiation regions can be sensitive to stochastic components
(900 µA) .
3.3 The electrically stimulated retina

Photoreceptor loss due to retinal degenerative diseases such as age-related
macular degeneration and retinitis pigmentosa is a leading cause of blind-
ness. Although these patients are blind, they possess functioning bipolar and
ganglion cells that relay retinal input to the brain. Retina implants are in
development in order to evoke optical sensations in these patients. The
success of such an approach to provide useful vision depends on elucidating
the neuronal target of stimulation. A compartmental model for extracellular
electrical stimulation of the retinal ganglion cell (RGC) has been developed.71
In this model, a RGC is stimulated by extracellular electrical fields with active
channels and realistic cell morphology derived directly from neuronal trac-
ing of an amphibian retina cell.
The goal of the simulation was to clarify which part of the target cell is
most excitable. It was previously shown that patients blind from retinitis
pigmentosa resolve focal phosphenes when the ganglion cell side of the
retinal surface is stimulated. None of the patients reported the perception of
wedges; all reported spots of light or spots surrounded by dark rings. Where
is the initiation point of the signal that causes these optical sensations? Is
the threshold to propagate an action potential down the axon lowest for an
electrode over the axon, the soma, or somewhere over the dendritic arbor,
which may spread up to 500 µm in diameter and overlap the dendritic field
of other ganglion cells?72
To explore the relative thresholds of ganglion cell axons, somas, and
dendrites a mudpuppy (Necturus maculosus) RGC with a large dendritic field
and a long axon was chosen. This cell has been mapped in three dimensions.73
The entire traced cell, except the soma, is divided into cylindrical compart-
ments. The soma can be modeled as a compartmentalized sphere71 or as a
single spherical compartment as described above. The number of compart-
ments was increased until the thresholds did not change by more than 1%.
Finally, more than 9000 compartments ensure fine numerical solutions for
the RGC, as shown in Figure 3.8.
An extracellular field from an ideal monopolar point or disk electrode
in a homogeneous medium were applied to these compartments. An elec-
trical network as shown in Figure 3.4 represents the whole retinal ganglion
cell. Several constants were specified based on whole-cell recording data

E
B
F
D
100µm
C
Figure 3.8 A topographical view of the target cell shown with its axon projecting
upwards. The soma of the retinal ganglion cell is modeled as a 24-µm sphere. This
size was chosen to approximate the diameter of the actual traced mudpuppy soma.
The electrode is modeled as an ideal point source in an infinite homogeneous medium.
For calculation of Ve, see Eq. (3.47); ρe = 1/60 Ωcm, the resistivity of normal (0.9%)
saline, which is similar to the resistivity of the vitreous humor.76 The height of the
electrode was held fixed at 30 µm above the center of the soma, a distance comparable
to that of the retinal ganglion cells to the surface of the retina in our region of interest.
The electrode positions are marked by the letters A through G. The linear distances
from the soma (in µm) are: (A) over the axon ~503, (B) over the axon ~130, (C) opposite
the axon ~334, (D) perpendicular to the axon ~160, (E) directly above the soma, (F)
perpendicular to the axon ~302, and (G) opposite the axon ~121. The vertical distance
(not shown) from the electrode to the center of the nearest compartment is (in µm):
(A) 33.0, (B) 30.0, (C) 34.0, (D) 30.0, (E) 30.0, (F) 29.5, and (G) 38.0.
at 22°C, including membrane capacitance (c = 1 µF/cm2) and cytoplasmic

resistance (0.11 kΩcm).75 These values are assumed to be uniform through-
out the cell.
For comparison, the electric conductance of the cell membrane was
evaluated with a linear passive model, a Hodgkin–Huxley model with pas-
sive dendrites (HH),1 and a model composed of all active compartments
with five nonlinear ion channels (FCM).76 At the start of all simulations, the
membrane potential is initialized to a resting potential of –70 mV. The linear
passive model reduces each patch of membrane to a simple parallel RC
circuit with a leak. The leak conductance was modeled as a battery at –70
mV in series with a conductance of 0.02 mS/cm2.

The retinal ganglion cell in many species (including humans) has a

nonmyelinated axon inside the eye. Some simulations were performed with
the classic description for unmyelinated axons, the Hodgkin–Huxley
model. The Hodgkin–Huxley channel kinetics are applied to the soma and
the axon, with the constants gNa = 120 mS/cm2, ENa = 50 mV, gK = 36
mS/cm2, EK = –77 mV, gl = 0.3 mS/cm2, El = –54.3 mV. The dendrites are
passive in this model.
The FCM model includes five nonlinear ion channels and one linear
leakage channel.44,45,76 The kinetic model was constructed on the basis of
voltage clamp data in retinal ganglion cells of tiger salamanders and rats.77,78
The model consists of INa and ICa; the delayed rectifier, IK; IK,A; and the Ca-
activated K current, IK,Ca. IK,Ca was modeled to respond to Ca influx, and a
variable-rate, Ca-sequestering mechanism was implemented to remove cyto-
plasmic calcium. ICa and IK,Ca do play an important role in regulating firing
frequency. The basic mathematical structure for voltage-gating is based on
the Hodgkin–Huxley model:
dE
Cm = − g Na m 3 h(E − ENa ) − g Ca c 3 (E − ECa ) − g K n 4 (E − EK )
dt (3.54)
− g A a hA (E − EK ) − g K ,Ca (E − EK ) − g l (E − El ) + I stimulus
3
where
g Na = 50 mS cm 2 , ENa = 35 mV , g Ca = 2.2 mS cm 2 , g K = 12 mS cm 2 ,
EK = −75 mV , g A = 36 mS cm 2 , g l = 0.005 mS cm 2 , El = −60 to − 65 mV.
The gating of IK,Ca is modeled as
g K ,Ca = g K ,Ca
([Ca ] Ca )
2+
i
2+ 2
diss
(3.55)
1 + ([Ca ] Ca )
2+
i
2+ 2
diss
where g K ,Ca = 0.05 mS/cm2 and the Ca2+-dissociation constant, Ca2+diss, is

10–3 mmole/l. [Ca2+]i is allowed to vary in response to ICa. The inward flowing
Ca2+ ions are assumed to be distributed uniformly throughout the cell; the
free [Ca2+] above a residual level — Ca2+res = 10–4 mmole/l — are actively
removed from the cell or otherwise sequestered with a time-constant (τ =
1.5 msec). Thus, [Ca2+]i follows the equation:
d[Ca 2+ ]i − sI Ca [Ca 2+ ]i − [Ca 2+ ]res

= − (3.56)
dt 2 νF τ

where F is the Faraday constant, s/ν is the ratio of surface to volume of the
compartment being studied, and the factor of 2 on ν is the valency. For a
spherical soma with r = 12 µm, Eq. (3.56) becomes:
d[Ca 2+ ]i
= −0.000013 I Ca − 0.666667 ([Ca 2+ ]i − 0.0001) (3.57)
dt
ECa is modeled as variable according to the already mentioned Nernst equa-

tion:
RT [Ca 2+ ]e
ECa = ln (3.58)
2 F [Ca 2+ ]i
where [Ca2+]e is the external calcium ion concentration = 1.8 mmole/l.

Similar to the Hodgkin–Huxley model, the rate constants for m, h, c, n,
a, and hA solve the first-order kinetic equation (Table 3.2):
dx
= −( α x + β x ) x + α x (3.59)
dt
The multi-compartment model includes representations for dendritic trees,

soma, axon hillock, a thin axonal segment, and an axon distal to the thin
segment. On these parts, the five ion channels were distributed with varying
densities, simulated by varying the value of gmax (mS/cm2) for each channel
(Table 3.3). 44,45,76 Using the multicompartmental simulation package
NEURON79, the action potential threshold values were computed for the
electrode positions as marked in Figure 3.8.71
Table 3.2 Rate Constants for Voltage-Gated Ion Channels

α β
−0.6(E + 30)
αm = β m = 20e − ( E + 55 ) 18
e −0.1( E + 30 ) − 1
6
α h = 0.4 e − ( E + 50 ) 20 βh =
e −0.1( E + 20 ) + 1
−0.3(E + 13)
αc = β c = 10e − ( E + 38 ) 18
e −0.1( E +13 ) − 1
−0.02(E + 40)
αn = β n = 0.4 e − ( E + 50 ) 80
e −0.1( E + 40 ) − 1
−0.006(E + 90)
αA = β A = 0.1e − ( E + 30 ) 10
e −0.1( E + 90 ) − 1
0.6
α hA = 0.04 e − ( E + 70 ) 20 β hA =
e −0.1( E + 40 ) + 1

Table 3.3 FCM Model Channel Densities at Soma, Dendrite, and Axon
Soma Axon Axon Axon
gmax (mS/cm2) Dendrite (0–3%)a (3–9%)a (9–100%)a
gNa 70 40 150 100 50
gCa 1.5 3.6 1.5 0 0
gK 18 12 18 12 15
gKa 54 36 54 0 0
gK,Ca 0.065 0.065 0.065 0 0
a Percent total axon length from soma; total axon length approximately 1 mm.
A practical retinal prosthesis will perhaps have disk electrodes, because

flat electrodes are easy to produce. To replace the point electrode source by
a field from an equipotential metal disk in a semi-infinite medium, the
external potential, Ve, under the electrode is calculated by:
2V0 2a
V (r , z) = ⋅ arcsin (3.60)
π (r − a) + z + (r + a)2 + z 2
2 2
where (r, z) is the radial and axial distance from the center of the disk in
cylindrical coordinates z ≠ 0.80 V0 is the potential, and α is the radius of the
disk. In simulation, both 50- and 100-µm-diameter disks required the same
cathodic threshold voltage. In human trials, the current required to reach
threshold also did not vary when a monopolar stimulating electrode was
changed from 50-µm-diameter to 100-µm.71,81
3.4 Concluding remarks

Calcium currents and the inside calcium concentration often have significant
influence on the temporal spiking pattern of electrically stimulated neurons,
because calcium is an important intracellular signaling molecule with rapid
effects on many processes. A rather simple approach demonstrates the agree-
ment between the FCM model and the firing behavior of retinal ganglion
cells;44,45,76 a similar model exists for pyramidal cells.42 Detailed modeling
includes the variation of calcium concentration with distance from the cell
membrane, buffer-elements and pumps.82
Selective stimulation of neural tissue is a great challenge for biomedical
engineering. A typical example in bladder control is activation of the detru-
sor muscle without activation of the urethral sphincter and afferent fibers
when stimulating sacral roots with tripolar cuff electrodes. Analysis with
the activating function is of help for finding the polarized and hyperpo-
larized regions but nonlinear membrane current modeling has to be
included to quantify cathodic and anodic block or anodic break phenomena
in selected target fibers and to determine the stimulus pulse parameters
for an operating window.14,83,84

Table 3.4 Cathodic 100-µsec Current Thresholds for

Point-Source Stimulationa
Passive HH FCM
Electrode Model Model Model
A 0.900 1.58 1.73
B 1.00 1.79 1.97
C 0.954 75.1 55.2
D 0.699 14.4 2.72
E 1.00 1.00 1.00
F 0.812 65.1 2.80
G 1.05 9.58 2.63
a Normalized for position over the soma and calculated with the
three membrane models. The absolute threshold current at the
soma (electrode position E) for the passive model was –32.9 µA;
for the Hodgkin–Huxley model, –43 µA; and for the FCM (five
nonlinear ion channels) model, –71 µA. The threshold response
for the passive model is defined as a 15-mV depolarization.13
Stimulation of neurons in a network is often difficult to explore because

of the complex neural geometry. Every type of the involved neurons has
to be analyzed in regard to the response and its sensitivity. Note that
neurotransmitter release can be initiated or influenced by the applied field
without the generation of a spike.12 This may be of specific importance at
the retina where several types of neurons never generate action potentials
but their degree of neurotransmitter release heavily depends on transmem-
brane voltage.
The extracellular potential along a target cell is input data for the exci-
tation process. Some case studies are based on the combination of ideal point
sources in infinite homogeneous media or idealized surface electrodes,71,80,85
but most clinical applications require more precise calculation of the field.
Finite element software such as ANSYS is commonly used70,86,87 but so, too,
are finite differences6 and the boundary element method.88
Simulation tools such as NEURON (http://www.neuron.yale.edu/)79
and GENESIS (http://www.genesis-sim.org/GENESIS/)89 are recommended
for solving neural compartment models. Compartment modeling (i.e., Eq.
(3.44)), together with a set of equations for the ion currents, results in a
system of ordinary differential equations that can be solved directly. This
procedure (methods of lines) will cause numerical problems for fine spatial
segmentation. Integration routines for stiff differential equations, implicit
integration methods, or solving with the Crank Nicholson method are of
help to reduce the computational effort.90
References
1. Hodgkin, A.L. and Huxley, A.F., A quantitative description of membrane
current and its application to conduction and excitation in nerve, J. Physiol.,
117, 500–544, 1952.

2. McNeal, D.R., Analysis of a model for excitation of myelinated nerve, IEEE

Trans. Biomed. Eng., BME-23, 329–337, 1976.
3. Rattay, F., Analysis of models for external stimulation of axons, IEEE Trans.
Biomed. Eng., 33, 974–977, 1986.
4. Rattay, F., Analysis of models for extracellular fiber stimulation, IEEE Trans.
Biomed. Eng., 36, 676–682, 1989.
5. Sweeney, J.D. and Mortimer, J.T., An asymmetric two electrode cuff for gen-
eration of unidirectionally propagated action potentials, IEEE Trans. Biomed.
Eng., BME-33, 541–549, 1986.
6. Veltink, P.H., van Alste, J.A., and Boom, H.B.K., Simulation of intrafascicular
and extraneural nerve stimulation, IEEE Trans. Biomed. Eng., BME-35, 69–75,
1988.
7. Altman, K.W. and Plonsey, R., Point source nerve bundle stimulation: effects
of fiber diameter and depth on simulated excitation, IEEE Trans. Biomed. Eng.,
37, 688–698, 1990.
8. Sweeney, J.D., Ksienski, D.A., and Mortimer, J.T., A nerve cuff technique for
selective excitation of peripheral nerve trunk regions, IEEE Trans. Biomed.
Eng., BME-37, 706–715, 1990.
9. Veraart, C., Grill, W.M., and Mortimer, J.T., Selective control of muscle acti-
vation with a multipolar nerve cuff electrode, IEEE-Trans. Biomed. Eng., 40,
640-653, 1990.
10. Tyler, D.J. and Durand, D.M., Intrafascicular electrical stimulation for selec-
tively activating axons, IEEE Eng. Med. Biol. 13, 575-583, 1994.
11. Nagarajan, S.S., Durand, D., and Warman, E.N., Effects of induced electric
fields on finite neuronal structures: a simulation study, IEEE Trans. Biomed.
Eng., BME-40, 1175–1188, 1993.
12. Rattay, F., The basic mechanism for the electrical stimulation of the nervous
system, Neuroscience, 89, 335–346, 1999.
13. Coburn, B., Neural modeling in electrical stimulation review, Crit. Rev. Biomed.
Eng., 17, 133–178, 1989.
14. Rattay, F., Electrical Nerve Stimulation: Theory, Experiments and Applications,
Springer-Verlag, New York, 1990.
15. Basser, P.J., Wijesinghe, R.S., and Roth, B.J., The activating function for mag-
netic stimulation derived from a 3-dimensional volume conductor model,
IEEE Trans. Biomed. Eng., 39, 1207–1210, 1992.
16. Garnham, C.W., Barker, A.T., and Freeston, I.L., Measurement of the activating
function of magnetic stimulation using combined electrical and magnetic
stimuli, J. Med. Eng. Technol., 19, 57–61, 1995.
17. Reichel, M., Mayr, W., and Rattay, F., Computer simulation of field distribu-
tion and excitation of denervated muscle fibers caused by surface electrodes,
Artif. Organs, 23, 453–456, 1999.
18. Sobie, E.A., Susil R.C., and Tung, L., A generalized activating function for
predicting virtual electrodes in cardiac tissue, Biophys. J., 73, 1410–1423, 1997.
19. Efimov, I.R., Aguel, F., Cheng, Y., Wollenzier, B., and Trayanova, N., Virtual
electrode polarization in the far field: implications for external defibrillation,
Am. J. Physiol. Heart. Circ. Physiol., 279, H1055–H1070, 2000.
20. Rattay, F., Analysis of the electrical excitation of CNS neurons, IEEE Trans.
Biomed. Eng., 45, 766–772, 1998.
21. Blair, E.A. and Erlanger, J., A comparison of the characteristics of axons through
their individual electrical responses, Am. J. Physiol., 106, 524–564, 1933.

22. Huguenard, J.R., Low-threshold calcium currrents in central nervous system

neurons, Annu. Rev. Physiol., 58, 329–348, 1996.
23. Schiller, J., Helmchen, F., and Sakmann, B., Spatial profile of dendritic calcium
transients evoked by action potentials in rat neocortical pyramidal neurones,
J. Physiol., 487, 583–600, 1995.
24. Mainen, Z.F. and Sejnowski, T.J., Modeling active dendritic processes in py-
ramidal neurons, in Methods in Neuronal Modeling: From Ions to Networks, 2nd
ed., Koch, C. and Segev, I., Eds., MIT Press, Cambridge, MA, 1998, pp.
171–209.
25. Destexhe, A, Mainen, Z.F., and Sejnowski, T.J., Synthesis of models for excit-
able membranes, synaptic transmission, and neuromodulation using a com-
mon kinetic framework, J. Comput. Neurosci., 1, 195–230, 1994.
26. Frankenhaeuser, B., Sodium permeability in toad nerve and in squid nerve,
J. Physiol., 152, 159–166, 1960.
27. Horáckova, M., Nonner, W., and Stämpfli, R., Action potentials and voltage
clamp currents of single rat Ranvier nodes, Proc. Int. Union Physiol. Sci., 7,
198, 1968.
28. Chiu, S.Y., Ritchie, J.M., Rogart, R.B., and Stagg, D., A quantitative description
of membrane currents in rabbit myelinated nerve, J. Physiol., 313, 149–166, 1979.
29. Schwarz, J.R. and Eikhof, G., Na currents and action potentials in rat myeli-
nated nerve fibres at 20 and 37°C, Pflügers Arch., 409, 569–577, 1987.
30. Sweeney, J.D., Mortimer, J.T., and Durand, D., Modeling of mammalian my-
elinated nerve for functional neuromuscular electrostimulation, in IEEE 9th
Annu. Conf. Eng. Med. Biol. Soc., Boston, MA, 1987, pp. 1577–1578.
31. Rattay, F., Simulation of artificial neural reactions produced with electric
fields, Simulation Practice Theory, 1, 137–152, 1993.
32. Rattay, F. and Aberham, M., Modeling axon membranes for functional elec-
trical stimulation, IEEE Trans. Biomed. Eng., BME 40, 1201–1209, 1993.
33. Schwarz, J.R., Reid, G., and Bostock, H., Action potentials and membrane
currents in the human node of Ranvier, Eur. J. Physiol., 430, 283–292, 1995.
34. Wesselink, W.A., Holsheimer, J., and Boom, H.B., A model of the electrical
behaviour of myelinated sensory nerve fibres based on human data, Med.
Biol. Eng. Comput., 37, 228–235, 1999.
35. Rattay, F., Lutter, P., and Felix, H., A model of the electrically excited human
cochlear neuron. I. Contribution of neural substructures to the generation and
propagation of spikes, Hear. Res., 153, 43–63, 2001.
36. Burke, D., Kiernan, M.C., and Bostock, H., Excitability of human axons, Clin.
Neurophysiol., 112, 1575–1585, 2001.
37. Baker, M.D., Axonal flip-flops and oscillators, Trends Neurosci., 23, 514–519,
2000.
38. Magee, J.C., Voltage-gated ion channels in dendrites, in Dendrites, Stuart, G.,
Spruston, N., and Häusser, M., Eds., Oxford University Press, London, 1999,
pp. 139–160.
39. Belluzzi, O. and Sacchi, O., A five conductance model of the action potential
in the rat sympathetic neurone, Progr. Biophys. Molec. Biol., 55, 1–30, 1991.
40. Winslow, R.L. and Knapp, A.G., Dynamic models of the retinal horizontal
cell network, Prog. Biophys. Mol. Biol., 56, 107–133, 1991.
41. McCormick, D.A. and Huguenard, J.R., A model of the electrophysiological
properties of thalamocortical relay neurons, J. Neurophysiol., 68, 1384–1400,
1992.

42. Traub, R.D., Jefferys, J.G., Miles, R., Whittington, M.A., and Toth, K., A branch-
ing dendritic model of a rodent CA3 pyramidal neurone, J. Physiol., 481, 79–95,
1994.
43. DeSchutter, E. and Smolen, P., Calcium dynamics in large neuronal models,
in Methods in Neuronal Modeling: From Ions to Networks, 2nd ed., Koch, C. and
Segev, I., Eds., MIT Press, Cambridge, MA, 1999, pp. 211–250.
44. Fohlmeister, J.F. and Miller, R.F., Impulse encoding mechanisms of ganglion
cells in the tiger salamander retina, J. Neurophysiol., 78, 1935–1947, 1997.
45. Fohlmeister, J.F. and Miller, R.F., Mechanisms by which cell geometry controls
repetitive impulse firing in retinal ganglion cells, J. Neurophysiol., 78,
1948–1964, 1997.
46. Hartmann, R., Topp, G., and Klinke, R., Discharge patterns of cat primary
auditory fibers with electrical stimulation of the cochlea, Hearing Res., 13,
47–62, 1984.
47. Motz, H. and Rattay, F., A study of the application of the Hodgkin–Huxley
and the Frankenhaeuser–Huxley model for electrostimulation of the acoustic
nerve, Neuroscience, 18, 699–712, 1986.
48. Hodgkin, A.L. and Katz, B., The effect of temperature on the electrical activity
of the giant axon of the squid, J. Physiol., 109, 240–249, 1949.
49. Huxley, A.F., Ion movements during nerve activity, Ann. N.Y. Acad. Sci., 81,
221–246, 1959.
50. Rattay, F., Propagation and distribution of neural signals: a modeling study
of axonal transport, Physics Alive, 3, 60–66, 1995.
51. Frankenhaeuser, B. and Huxley, A.L., The action potential in the myelinated
nerve fibre of Xenopus laevis as computed on the basis of voltage clamp data,
J. Physiol., 171, 302–315, 1964.
52. Frankenhaeuser, B. and Moore, L.E., The effect of temperature on the sodium
and potassium permeability changes in myelinated nerve fibers of Xenopus
laevis, J. Physiol., 169, 431–437, 1963.
53. Halter, J.A. and Clark, J.W., A distributed-parameter model of the myelinated
nerve fiber, J. Theor. Biol., 148, 345–382, 1991.
54. Ritchie, J.M., Physiology of axons, in The Axon: Structure, Function and Patho-
pysiology, Waxman, S.G., Kocsis, J.D., and Stys, P.K., Eds., Oxford University
Press, Oxford, 1995, pp. 68–96.
55. Ranck, J.B., Jr., Which elements are excited in electrical stimulation of mam-
malian central nervous system: a review, Brain Res., 98, 417–440, 1975.
56. Warman, E.N., Grill, W.M., and Durand, D., Modeling the effects of electric
fields on nerve fibers: determination of excitation thresholds, IEEE Trans.
Biomed. Eng., 39, 1244–1254, 1992.
57. Plonsey, R. and Barr, R.C., Electric field stimulation of excitable tissue, IEEE
Trans. Biomed. Eng., 42, 329–336, 1995.
58. Zierhofer, C.M., Analysis of a linear model for electrical stimulation of axons:
critical remarks on the “activating function concept,” IEEE Trans. Biomed. Eng.,
48, 173–184, 2001.
59. Rattay, F., Minassian, K., and Dimitrijevic, M. R., Epidural electrical stimula-
tion of posterior structures of the human lumbosacral cord: 2. Quantitative
analysis by computer modeling, Spinal Cord, 38, 473–489, 2000.
60. Henriquez, C.S., Simulating the electrical behavior of cardiac tissue using the
bidomain model, Crit. Rev. Biomed. Eng., 21, 1–77, 1993.

61. Knisley, S.B., Trayanova, N., and Aguel, F., Roles of electric field and fiber
structure in cardiac electric stimulation, Biophys. J., 77, 1404–1417, 1999.
62. Van den Honert, C. and Stypulkowski, P.H., Temporal response patterns of
single auditory nerve fibers elicited by periodic electrical stimuli, Hear. Res.,
29, 207–222, 1987.
63. Javel, E. and Shepherd, R.K., Electrical stimulation of the auditory nerve: II.
Effect of stimulus waveshape on single fibre response properties, Hear. Res.,
130, 171–188, 1999.
64. Horikawa, Y., Simulation study on effects of channel noise on differential
conduction at an axon branch, Biophys. J., 65, 680–686, 1993.
65. Holden, A.V., Models of the Stochastic Activity of Neurons, Springer-Verlag,
Berlin, 1976.
66. Sigworth, F.J., The variance of sodium current fluctuations at the node of
Ranvier, J. Physiol., 307, 97–129, 1980.
67. Verveen, A.A. and Derksen, H.E., Fluctuation phenomena in nerve mem-
brane, Proc. IEEE, 56, 906–916, 1968.
68. DeFelice, L.J., Introduction to Membrane Noise, Plenum Press, New York, 1981.
69. Rubinstein, J.T., Threshold fluctuations in an N sodium channel model of the
node of Ranvier, Biophys. J., 68, 779–785, 1995.
70. Rattay, F., Basics of hearing theory and noise in cochlear implants, Chaos,
Solitons Fractals, 11, 1875–1884, 2000.
71. Greenberg, R.J., Velte, T.J., Humayun, M.S., Scarlatis, G.N., de Juan, E., Jr., A
computational model of electrical stimulation of the retinal ganglion cell, IEEE
Trans. Biomed. Eng., 46, 505–514, 1999.
72. Toris, C.B., Eiesland, J.L., and Miller, R.F., Morphology of ganglion cells in
the neotenous tiger salamander retina, J. Comp. Neurol., 352, 535–559, 1995.
73. Velte, T.J. and Miller, R.F., Dendritic integration in ganglion cells of the mud-
puppy retina, Visual. Neurosci., 12, 165–175, 1995.
74. Coleman, P.A. and Miller, R.F., Measurement of passive membrane parame-
ters with whole-cell recording from neurons in the intact amphibian retina,
J. Neurophysiol., 61, 218–230, 1989.
75. Fohlmeister, J.F., Coleman, P.A., and Miller, R.F., Modeling the repetitive firing
of retinal ganglion cells, Brain Res., 510, 343–345, 1990.
76. Geddes, L.A. and Baker, L.E., The specific resistance of biological material-A
compendium of data for the biomedical engineer and physiologist, Med. Biol.
Eng., 5, 271–293, 1967.
77. Lukasiewicz, P. and Werblin, F., A slowly inactivating potassium current
trunkates spike activity in ganglion cells of the tiger salamander retina, J.
Neurosci., 8, 4470–4481, 1988.
78. Lipton, S.A. and Tauck, D.L., Voltage-dependent conductances of solitary
ganglion cells dissociated from rat retina, J. Physiol., 385, 361–391, 1987.
79. Hines, M.L. and Carnevale, N.T., NEURON: a tool for neuroscientists, Neu-
roscientist, 7, 123–135, 2001.
80. Wiley, J.D. and Webster, J.G., Analysis and control of the current distribution
under circular dispersive electrodes, IEEE Trans. Biomed. Eng., BME-29,
381–385, 1982.
81. Humayun, M.S., de Juan, E., Dagnelie, G., Greenberg, R.J., Propst, R., and
Phillips, D.H., Visual perception elicited by electrical stimulation of retina in
blind humans, Arch. Ophthalmol., 114, 40–46, 1996.

82. DeSchutter, E. and Bower, J.M., An active membrane model of the cerebellar
Purkinje cell: I. Simulation of current clamps in slice, J. Neurophysiol., 71,
375–400, 1994.
83. Rijkhoff, N.J., Holsheimer, J., Koldewijn, E.L., Struijk, J.J., Van Kerrebroeck,
P.E., Debruyne, F.M., and Wijkstra, H., Selective stimulation of sacral nerve
roots for bladder control: a study by computer modeling, IEEE Trans. Biomed.
Eng., 41, 413–424, 1994.
84. Bugbee, M., Donaldson, N.N., Lickel, A., Rijkhoff, N.J., and Taylor, J., An
implant for chronic selective stimulation of nerves, Med. Eng. Phys., 23, 29–36,
2001.
85. Rattay, F., Modeling the excitation of fibers under surface electrodes, IEEE-
Trans. Biomed. Eng., BME-35, 199–202, 1988.
86. Coburn, B. and Sin, W.K., A theoretical study of epidural electrical stimulation
of the spinal cord. Part I: Finite element analysis of stimulus fields, IEEE Trans.
Biomed. Eng., 32, 971–977, 1985.
87. Rattay, F., Leao, R.N., and Felix, H., A model of the electrically excited human
cochlear neuron. II. Influence of the three-dimensional cochlear structure on
neural excitability, Hear. Res., 153, 64–79, 2001.
88. Frijns, J.H., de Snoo, S.L., and Schoonhoven, R., Improving the accuracy of
the boundary element method by the use of second-order interpolation func-
tions, IEEE Trans. Biomed. Eng., 47, 1336–1346, 2000.
89. Bower, J. and Beeman, D., The Book of GENESIS, 2nd ed., TELOS, New York,
1997.
90. Mascagni, M.V. and Sherman, A., Numerical methods for neural modeling,
in Methods in Neuronal Modeling: From Ions to Networks, 2nd ed., Koch, C. and
Segev, I., Eds., MIT Press, Cambridge, MA, 1999, pp. 569–606.

Neuron Modeling Rattay-Greenberg-Resatz

Uploaded by

Copyright:

Available Formats

Neuron Modeling Rattay-Greenberg-Resatz

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Neuron Modeling Rattay-Greenberg-Resatz

Uploaded by

Copyright:

Available Formats

1100_ch03.

fm Page 39 Wednesday, October 16, 2002 8:45 AM

©2003 CRC Press LLC

Hodgkin-Huxley B SPHERICAL CELL

Constant membrane conductance

The rate of membrane voltage change, dV/dt, follows as:

©2003 CRC Press LLC

A positive stimulating current applied at the inside of an axon or at any

©2003 CRC Press LLC

basic mechanism of external stimulation, the essential differences between

©2003 CRC Press LLC

This is demonstrated, for example, by comparing the threshold current for

3.2 Models for the cell membrane

where the gas constant R = 8.31441 J/(mol.K); temperature, T, is in Kelvin;

©2003 CRC Press LLC

single-valenced ions F = 96485 c/mol. Note: At room temperature (T =

RT PK [K ]e + PNa [Na]e + PCl [Cl]i

1. Voltage-dependent gating dominates excitation and neural signal

3.2.1 Axon models of the Hodgkin–Huxley type

©2003 CRC Press LLC

k = 3 0.1T −0.63 (3.9)

where V is the reduced membrane voltage, resulting from internal, external,

EF 2 [Na]o − [Na]i exp(EF / RT )

©2003 CRC Press LLC

Handbook of Neuroprosthetic Methods

©2003 CRC Press LLC

©2003 CRC Press LLC

EF 2 [Na]o − [Na]i exp(EF / RT )

Note that membrane voltage is denoted as E in Eqs. (3.11) to (3.13) and

k = 3 0.1T − 3.7 (3.23)

©2003 CRC Press LLC

EF 2 [Na]o − [Na]i exp(EF / RT )

Parameter values and additional expressions of the SE model are listed in

EF 2 [Na]o − [Na]i exp(EF / RT )

©2003 CRC Press LLC

iK ,slow = g K ,slow p(V − VK ) (3.35)

©2003 CRC Press LLC

3.2.2 Inﬂuence of temperature

Raising the temperature causes shortening of the action potential and an

3.2.3 Compartment models

©2003 CRC Press LLC

Figure 3.3 Threshold currents as a function of pulse duration (strength duration

©2003 CRC Press LLC

d(Vi ,n − Ve ,n ) Vi ,n − Vi ,n−1 Vi ,n − Vi ,n+1

tance to the border of the jth process is:

The electric and geometric properties change from compartment to compart-

©2003 CRC Press LLC

Synaptic activation of a neuron can be simulated by current injection at

3.2.4 Activating function

The physical dimension of fn is V/sec or mV/msec. If the neuron is in the

©2003 CRC Press LLC

3.2.5 Activating function for long ﬁbers

d∆x  Vn − 1 − 2Vn + Vn + 1 Ve , n − 1 − 2Ve , n + Ve , n + 1 

©2003 CRC Press LLC

becomes proportional to the second difference of the extracellular potential

where el stands for electrode and z measures the center-to-center distance

©2003 CRC Press LLC

Figure 3.6 Stimulation of an unmyelinated axon with a monopolar electrode. During

©2003 CRC Press LLC