Brics Before brains there was no color or sound in the universe, nor was there any flavor or

aroma and probably little sense and no feeling or emotion.

-Roger W. Sperry

Exploring Biological Neuron Models

By D Mishra, A Yadav, S Ray and P K Kalra

1. Introduction other via action potentials, these models
The human brain is one of the most complex basically represent neuronal behavior in
objects in the universe. Many attempts have terms of membrane potential and action
been made to investigate and model the potential. Many Neuron Models have been
functionalities of the brain. We still do not proposed by researchers, some most popular
know exactly how brain learns. The research in models are listed below:
this filed is aimed to know about the brain with Ø Hodgkin-Huxley
more and more intricacy. This ongoing Ø Integrate-and-fire
research will make it possible to develop more Ø Integrate-and-fire with adapt
realistic models of human brain. These models Ø Integrate-and-fire-or-burst
are important tools for characterizing what Ø Resonate-and-fire
nervous systems do, determining how they Ø Quadratic integrate-and-fire
function, and understanding why they operate Ø FitzHugh-Nagumo
in particular ways. In this article, we are Ø Morris-Lecar
focusing on the modeling aspects of the basic Ø Wilson
unit of the brain, i.e. Neuron. Ø Izhikevich
2. The Neuron Ø Hindmarsh-Rose
Neurons form the basis of the brain. These neuron models represent some or all
Therefore, a sufficient in-depth knowledge of the characteristics of the responses of real
about neurons is necessary for the study of the neurons. In [6], Izhikevich has discussed these
brain. A typical human brain consists of characteristics and corresponding neuron
approximately 100 billion neurons, each models in detail. In the next section, we
neuron having at least 10,000 connections with describe the dynamics for some of these
other neurons. Electrochemical signals flow in neuron models. Responses of neurons are
neurons, originating at the dendrites or cell summarized in Table 1.
body in response to stimulation from other 3.1 Hodgkin-Huxley Model
neurons and propagating along axon branches
which terminate on the dendrites or cell bodies In 1952, Hodgkin and Huxley introduced this
of perhaps thousands of other neurons. The neuron model [5]. This model is important
connections between the ends of axons and the not only because their parameters are
dendrites or cell bodies of other neurons are biophysically meaningful and measurable,
specialized structures called synapses [1, 3, 7]. but also because they allow us to investigate
questions related to synaptic integration,
3. Neuron Model dendritic cable filtering, effects of dendritic
Various mathematical models for biological morphology, the interplay between ionic
neurons have been developed to represent currents, and other issues related to single cell
there biological activities. As it is generally dynamics. It is described by the following
believed that neurons communicate with each four nonlinear differential equations:

13
 dv 1
C  I Na  I K  I L  I EXT ……. (1) 0.9
m
dt h
n
0.8

Gating Variables: m, h & n

dm ……. (2) 0.7
m  m  m
dt 0.6

dn 0.5
n  n  n ……. (3)
dt 0.4

dh 0.3
h  h  h ……. (4) 0.2
dt
0.1

where, 0
0 20 40 60 80 100
I Na  m hGNa VNa  v 
3
……. (5) Time in milliseconds

Figure 2: Response of Hodgkin-Huxley

I K  n 4GK VK  v  ……. (6) model: Gating variables

The activity of a nerve cell is found by solving

I L  GL VL  v  ……. (7) the set of differential equations (1) – (4)
numerically. Figure 1 and Figure 2 show the
response of this model. Simulink model for the
In this representation of the model, symbols
Hodgkin-Huxley equations are shown in
have the following meanings:
Figure 3. Figure 4 shows the circuit diagram of
n = Membrane potential this model.
IEXT = External Current 3.2 Fitzhugh-Nagumo Model
GNa = Maximum sodium conductance
GK = Maximum potassium conductance The analysis of the Hodgkin-Huxley equations
is extremely difficult because of the
GL = Leakage Conductance
nonlinearities and the large number of
VNa = Sodium reversal voltage
variables. Mathematical analysis would be
VK = Potassium reversal voltage helpful if it is performed on simpler equations
VL = Leakage voltage whose solutions shared the qualitative
INa= Sodium Current properties of those of the Hodgkin-Huxley
IK = Potassium Current equations. Analysis of such simpler systems
IL = Leakage Current may lead to the discovery of new phenomena,
m = Sodium activation gating variable which may be searched for in the original
h = Sodium inactivation gating variable system and also in experimental preparations.
n = Potassium activation gating variable Such a simplified system of equations has its
40 origin in the works of Fitzhugh and
Nagumo[2] and is known as the Fitzhugh-
20
Nagumo equations. In the Hodgkin-Huxley
Membrane voltage in mV

0
system the variables v (voltage) and m (sodium
activation) have similar (mostly fast) time
-20 courses and the variables n (potassium
activation) and h (sodium inactivation) have
-40
similar time courses. Heuristically speaking, in
-60
the Fitzhugh-Nagumo system, v and m are
regarded as mimicked by a single variable v(t),
-80
0 20 40 60 80 100
which we call the voltage, and n and h are
mimicked by a single variable w(t), which is
Time in milliseconds
called as the recovery variable. The Fitzhugh-
Figure 1: Response of Hodgkin-Huxley
model: Membrane potential
Nagumo (FHN) equations in their general

14
form are given by the equations (8) and (9). The system of FHN equations is an example
of autonomous system, since the derivatives do
dv v3 not explicitly depend on time. We made a
 v w I ……. (8)
dt 3 Simulink model for these equations. It is shown
dw in Figure 5. Figure 6 shows the response of this
  v  a  bw ……. (9) model.
dt

Figure 3: Simulink model for Hodgkin -Huxley equations

Table 1: Responses of Neurons (See [6] for details.)

Type of Response Description

Tonic Spiking While the input is on, the neuron continues to fire a train of spikes.
Neuron fires only a single spike at the onset of the input
Phasic Spiking
and remains quiescent afterwards.
Tonic Bursting Neuron fires periodic bursts of spikes when stimulated.
Phasic Bursting Similarly to the phasic spikers, some neurons are phasic bursters.
Mixed Model(Burs- Neurons fire a phasic burst at the onset of stimulation and
-ting Then Spiking) then switch to the tonic spiking mode.
Spike Frequency Neuron fires tonic spikes with decreasing frequency.
Adaptation
Class 1 Excitability The ability to fire low-frequency spikes when the input is weak.
Class 2 Excitability Neurons are either quiescent or fire a train of spikes with a
certain relatively large frequency
Spike Latency Neurons fire spikes with a delay that depends on the strength of the input signal.

15
Subthreshold Oscillations Neurons act as band-pass filters.
Frequency Preference Neurons having oscillatory potentials can respond
and Resonance selectively to the inputs having frequency content
similar to the frequency of subthreshold oscillations.
Integration and Coincidence Neurons without oscillatory potentials act as integrators.
Detection They prefer high-frequency input; the higher the
frequency the more likely they fire.
When a neuron receives and then is released from an
Rebound Spike inhibitory input, it may fire a post-inhibitory (rebound)
spike.
Rebound Burst When a neuron receives and then is released from an
inhibitory input, it may fire a post-inhibitory (rebound)
burst.
Threshold Variability Threshold is not constant.
Some neurons can exhibit two stable modes of
Bistability of Resting and operation: resting and tonic spiking (or even bursting).
Spiking States An excitatory or inhibitory pulse can switch between the
modes.
After firing a spike, the membrane potential of a neuron
Depolarizing After may exhibit a prolonged after-hyperpolarization (called
Potentials AHP) or a prolonged depolarized afterpotential (called
DAP)
-
Neurons are extremely sensitive to brief coincident
Accommodation inputs, but may not fire in response to a strong but
slowly increasing input.
A bizarre feature of many thalamo-cortical neurons is
that they are quiescent when there is no input, but fire
Inhibition-Induced Spiking when hyperpolarized by an inhibitory input or an
injected current.
Instead of spiking, a thalamocortical neuron can fire
Inhibition-Induced Bursting tonic bursts of spikes in response to a prolonged
hyperpolarization.

3.3 Morris-Lecar Model

Morris and Lecar suggested a simple two
dimensional model to describe oscillations in
barnacle giant muscle fiber. Because it has
biophysically meaningful and measurable
parameters, the model became quite popular in
computational neuroscience community. It
consists of a membrane potential equation
with instantaneous activation of Ca-current
and an additional equation describing slower
Figure 4: Circuit diagram for
Hodgkin-Huxley equations activation of K-current. Following equations

16
describe this model [8]: It is assumed that the calcium current is always
dv at equilibrium, with its activation curve given by
C = - ( I Ca + I K + I m ) + I …. (10)
dt  v  1  …. (15)
m v   0.5 1  tanh 
dw w¥ (v ) - w …. (11)  15 
=
dt t w (v ) The potassium activation variable follows the
standard first order equation with steady-state
where v is the absolute membrane potential (in activation
mV), C is the membrane capacitance, w is the  v 1
activation variable for potassium, and ICa, IK w v   0.5 1  tanh  …. (16)
and IL are ionic currents which are given by  30 
and time constant
5
I Ca = GCa m¥ (v )(v - VCa ) .….. (12)  w v   …. (17)
v
cosh
60
I K = GK w (v - VK ) …... (13) The other parameters are GCa = 1.1, GK = 2.0,
Gm = 0.5, VCa = 100, EK = -70, and Vrest = -50 (all
2
I m = Gm (v - Vrest ) conductances are in mS/cm and potentials in
….. (14)
mV).

5: Simulink model for Fitzhugh-Nagumo equations Figure 8: Simulink model for Morris-Lecar equations

Figure 6: Response of Fitzhugh-Nagumo model: Figure 9: Response of Morris -Lecar model:

Activation variable (v) Absolute membrane potential (v

Figure 7: Response of Fitzhugh-Nagumo model: Figure 10: Response of Morris-Lecar Model:

Refractoriness (w) Potassiumactivation variable (w)

17
3.4 Wilson-Cowan Model
This model is represented in terms of two
nonlinear differential equations that represent
the interaction between excitatory and
inhibitory populations of neurons [9].
dV
= -V + f (rV + aV - bR )…. (18)
dt
dR
= - R + f (r R + cV - dR ) …. (19)
dt
where a, b, c, d, rV and rR are parameters. The
function f is assumed to be sigmoid.
1
f u   …. (20) Figure 13: Simulink model for four-dimensional
1  eu Wilson Model

We have made a Simulink model for four 3.5 Integrate-and-Fire Model

dimensional Wilson-Cowan model which is
shown in Figure 13. This model is the simplest model but it
captures almost all the important properties of
the cortical neuron. The basic circuit of this
model consists of a capacitor C in parallel with
a resistor R driven by a current I. Figure 14
depicts this circuit.

Figure 14: Circuit diagram of an Integrate -and-Fire

neuron model

The driving current can be split into two

components, I = IR + IC. IR is the resistive
current which passes through the linear
Figure 11 Simulink Model for two dimensional resistor, R and IC charges the capacitor C. Thus
Wilson-Cowan Model
v dv
I= +C …. (21)
R dt
where, v is the membrane potential. A spike
occurs when v reaches a threshold VTH. After
the occurrence of a spike, next spike cannot
occur during refractory period, TREF. Figure 15
on page 7 depicts the response of this model.

3.6 Izhikevich Spiking Model

Figure 12: Response of two-dimensional This model is represented in terms of the

Wilson-Cowan model

18 5
Following differential equations and reset
mechanism [6]. dv
 u  F (v )  I  w …. (25)
dt
dv du
 0.04v 2  5v  140  u  I …. (22)  G (v )  u …. (26)
dt dt
du …. (23) dw 1
 a (bv  u )  H (v)  w …. (27)
dt dt 
 v  c …. (24)
if v  30 mV , then  where F, G, and H are some functions. Figure
u  u  d 16 shows the response of this model.
All of the responses in appendix are obtained
using this model.

Figure 16: Response of the Hindmarsh -Rose model

4. Conclusion
We have presented different dynamical models
Figure 15: Response of the Integrate-and-Fire model of biological neuron or nerve cell. We have
plotted their response obtained by solving their
3.7 Hindmarsh-Rose Model dynamical equations numerically. We have also
presented the Simulink models of these
The Hindmarsh-Rose model of thalamic neuron models. A brief discussion of different
neuron can be written in the general form [4,6]: types of neuronal responses has been made.

References

[1] Dyan, P., & Abbott, L.F. (1999). Theoretical neuroscience: computational and mathematical modeling of neural
systems, The MIT press, Massachusetts.
[2] FitzHugh, R. (1969). Mathemati cal models for excitation and propagation in nerve. Biological Engineering. H.P.
Schawn (Ed.), New York: McGraw -Hill.
[3] Gerstner & Kistler (2002). Spiking Neuron Models. Single Neurons, Populations, Plasticity. Cambridge
University Press, Cambridge.
[4] Hindmarsh, J. L., and R. M. Rose. (1984). A model of neuronal bursting using three coupled first order
differential equations. Proc. R. Soc. Lond. Biol. 221, 87 -102
[5] Hodgkin, A., & Huxley, A. (1952). A quantitative description of membrane current and its application to
conduction and excitation in nerve. J.Phisiol. (Lond.). 117, 500 -544.
[6] Izhikevich, E. M. ( 2004). Which Model to Use for Cortical Spiking Neurons ?, IEEE Transaction on neural
networks, 15, 1063-1070.
[7] Koch, C. (1999). Biophysics of Computation: Information Processing in Single Neurons. New York: Oxford
University Press.
[8] Morris, C., & Lecar, H. (1981). Voltage oscillations in the barnacle giant muscle fiber, Biophysics J., 35, 193-
213.
[9] Wilson, H., R. (1999). Simplified dynamics of human and mammalian neocortical neurons.Journal of theoretical
biology, 200, 375-388.

19
 Appendix: Different Types of Neuronal (E) Mixed mode
Responses

Membrane Voltage in mV
Following figures show different types of
neuronal responses as discussed by
Izhikevich [7]. (A) tonic spiking
20
Membrane Voltage in mV

-20
Time in milliseconds
-40

(F) spike freq. adapt

-60
20

Membrane Voltage in mV
-80
0
0 10 20 30 40 50 60 70 80 90 100
Time in milliseconds
-20

B) phasic spiking
-40
20
-60
Membrane Voltage in mV

0
-80

-20
0 10 20 30 40 50 60 70 80

Time in milliseconds
-40

-60

(G) Class 1 excitable

-80

0 20 40 60 80 100 120 140 160 180 200

Membrane Voltage in mV

Time in milliseconds
(C) tonic bursting

20
Membrane Voltage in mV

-20

-40

Time in milliseconds
-60

-80

0 20 40 60 80 100 120 140 160 180 200 220 (H) Class 2 excitable
Time in milliseconds
20
(D) phasic bursting
Membrane Voltage in mV

20
0
Membrane Voltage in mV

0 -20

-20 -40

-40
-60

-60
-80

-80 0 50 100 150 200 250 300

Time in milliseconds
0 20 40 60 80 100 120 140 160 180 200
Time in milliseconds

20 5
 (I)(I)spike
spikelatency
latency (M) rebound spike

20 20
mV
ltage ininmV

Membrane Voltage in mV
0 0
Voltage

-20 -20
Membranee Vo

-40 -40
Membran

-60 -60

-80 -80

0 10 20 30 40 50 60 70 80 90 100 0 20 40 60 80 100 120 140 160 180 200

Time in
Time inmilliseconds
milliseconds Time in milliseconds

(J) subthreshold osc. (N) rebound burst

20 20
Membrane Voltage in mV

Membrane Voltage in mV
0 0

-20 -20

-40 -40

-60 -60

-80 -80

0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200

Time in milliseconds Time in milliseconds

(K) resonator O) thresh. variability

20 20
Membrane Voltage in mV

Membrane Voltage in mV

0 0

-20 -20

-40 -40

-60 -60

-80 -80

0 50 100 150 200 250 300 350 400 0 10 20 30 40 50 60 70 80 90 100

Time in milliseconds Time in milliseconds

(L) integrator (P) bistability

20 20
Membrane Voltage in mV
Membrane Voltage in mV

0 0

-20 -20

-40 -40

-60 -60

-80 -80

0 10 20 30 40 50 60 70 80 90 100 0 50 100 150 200 250 300

Time in milliseconds Time in milliseconds

21
 (Q) DAP
(S) inh. induced sp
20
20
Membrane Voltage in mV

Membrane Voltage in mV
0
0

-20 -20

-40 -40

-60 -60

-80
-80
0 50 100 150 200 250 300 350
0 5 10 15 20 25 30 35 40 45 50
Time in milliseconds
Time in milliseconds

(R) accomodation (T) inh. induced brst.

20 20
Membrane Voltage in mV

Membrane Voltage in mV
0 0

-20 -20

-40
-40

-60
-60

-80
-80
0 50 100 150 200 250 300 350
0 50 100 150 200 250 300 350 400

Time in milliseconds Time in milliseconds

About the author: Deepak Mishra is pursuing his Ph. D. in the Department of Electrical
Engineering in Indian Institute of Technology Kanpur, India. He obtained Masters of
Technology in Instrumentation from Devi Ahilya University, Indore in 2003. His major
field of study is Neural Networks and Computational Neuroscience. (Home page:
http://home.iitk.ac.in/~dkmishra)

About the author: Abhishek Yadav is working as Assistant Professor in the Department of
Electrical Engineering, College of Technology, G. B. Pant University of Agriculture and
Technology, Pantnagar, Uttaranchal, India. He obtained Masters of Technology in
Electrical Engineering from Indian Institute of Technology Kanpur, India in 2005. His
major field of study is Computational Neuroscience.

About the author: Sudipto Ray is working as Faculty Associate in the Indian Institute of
Information Technology, Allahabad, India. He obtained Masters of Technology in
Electrical Engineering from Indian Institute of Technology Kanpur, India in 2005. His
major field of study is Computational Neuroscience.

About the author: Professor Prem Kumar Kalra obtained his PhD in Electrical
Engineering from University of Manitiba, Canada. He is currently Professor and Head of
Electrical Engineering Department at IIT Kanpur, India. His research interests are Power
systems, Expert Systems applications, HVDC transmission, Fuzzy logic and Neural
Networks application and KARMAA (Knowledge Acquisition, Retention, Management,
Assimilation & Application).

22 5