You are on page 1of 2

Harvard University, Neurobiology 101hfm.

Fundamentals in Computational Neuroscience Fall term 2014/15

Assignment 5 – Integrate-and-fire neurons


Alexander Mathis, Ashesh Dhawale
September 30, 2014, due date October 7, 2014

1 A simple neuron model: the integrate-and-fire neuron

= 0

Figure 1: Equivalent circuit of a passive neuron. Cm represents the total membrane capacitance, Rm the total
membrane resistance, EL is the reversal potential of the leakage conductance. The voltage within the
cell is Vin , is referenced to the potential outside the cell, Vout = 0 mV. When the stimulation current
delivered by the electrode Ie is 0, Vin = EL .

The time dependent change in voltage of a passive, single compartment neuron model is given by the following differential
equation

dV
τm = EL − V + Rm Ie (1)
dt
where V is the membrane potential, EL the reversal potential of the leakage current, Rm is the membrane resistance,
τm = Rm Cm the time constant of the cell (membrane capacitance = Cm ), and Ie an external current applied through an
electrode.
For a constant stimulation current I0 , this equation can be solved to yield the membrane potential as a function of time.
 
t
V (t) = EL + Rm I0 + (V (0) − EL − Rm I0 ) exp − (2)
τm

where V (0) is the voltage at time t = 0.

This passive circuit is the building block of one of the simplest neuron models: the integrate-and-fire neuron. In order to
mimic spiking behavior one simply adds the assertion that an action potential occurs, whenever the membrane potential
V crosses a threshold Vthresh . Immediately after firing a spike, the voltage is reset to a fixed value Vreset (that should of
course satisfy Vreset < Vthresh ).

1.1 Model an integrate-and-fire neuron using the Euler method


We are typically interested in modeling changes in membrane potential under conditions where the stimulation current
varies in a time-dependent manner. Therefore, instead of analytically obtaining the solution to V (t), as in Eq. (2), we would
rather solve Eq. (1) numerically. The Euler method is a simple technique to numerically solve differential equations. For a
general equation

dV
= f (V, t), (3)
dt
the solution can be approximated using the forward Euler method with discretization ∆t by:

V (t + ∆t) = V (t) + ∆t(f (V, t)). (4)

Page 1of 2
Apply the Euler method to numerically solve Eq.(1) and calculate V (t) for a time-period of 500 ms. Include the reset
condition outlined above to model an integrate-and-fire neuron in M ATLAB.
Use Rm = 10 MΩ, Vthresh = −54 mV, Vreset = −80 mV, EL = −70 mV, τm = 10 ms, Ie ≈ 10 nA. Use a time step of
∆t = 0.01 ms and the starting condition V (0) = −70 mV.

1.2 Firing rate of the integrate-and-fire neuron


Analytically derive the minimum electrode current Imin required for this integrate-and-fire neuron to fire at least one spike,
i.e. at time t → ∞. For simplicity, assume that Vreset = EL .

Use the integrate-and-fire neuron model in MATLAB to determine the relationship between the firing rate and input
current (this relationship is often called the F-I curve) for a reasonable range of current values. What happens if you
increase the current to very large values? What parameter would you change to successfully run the simulation even for
large current inputs?

1.3 Include a refractory period


To your simple integrate-and-fire neuron, add an absolute refractory period δabs during which you set V = Vreset , so that
the neuron cannot fire another spike during this interval. Assume δabs = 4 ms. Plot the F-I curve for this model and
compare it to the one you obtained in section 1.2.

How does the addition of a refractory period alter the F-I curve and what could be its physiological implication?

2 Voluntary extra work: Spike frequency adaptation


Many neurons respond to constant current stimulation with a reduction in spiking frequency over time. This phenomenon
is called spike frequency adaptation and can be modeled by including an additional adaptation current −Rm Ga (V − Ea )
through adaptation channel a that varies as a function of spiking and time. In this case, equation (1) becomes:

dV
τm = EL − V − Rm Ga (V − Ea ) + Rm Ie , (5)
dt
where Ea is the reversal potential of the adaptation current, Rm Ga is ratio of the adaptation channel conductance (conduc-
tance is the inverse of resistance: G = 1/R) to the membrane conductance. Let us assume that the conductance Ga decays
to zero with a time constant τa .

dGa
τa = −Ga . (6)
dt
Every time the neuron fires a spike, ∆Ga is added to the value of Ga , i.e.

Ga → Ga + ∆Ga (7)
Add an adaptation current to the integrate-and-fire neuron (without a refractory period) using equations (5), (6) and (7).
Assume Ea = −70 mV, τa = 100 ms, Rm ∆Ga = 0.5, and Rm Ga (0) = 0. Verify by simulating the model that the spike rate
indeed decreases over time in response to a constant stimulation current.

Estimate the spike frequency adapting neuron’s firing rate from its steady-state inter-spike interval, as a function of stim-
ulation current. How does this curve differ from the F-I curve obtained from the simple integrate-and-fire neuron?

Page 2of 2

You might also like