User:Nir.nossenson/sandbox

= Biological Neuron Model =

Model Scope and Aims
A biological neuron model (also known as spiking neuron model) is a mathematical description of the properties of certain cells in the nervous system that generate sharp electrical potentials, roughly one millisecond in duration, as shown in Fig. 1. The amplitude, and the exact shape of the action potential can vary according to the exact experimental technique used for acquiring the signal. Measurement techniques that penetrate the cell membrane (as in Fig. 2) yield spike amplitudes of about 100 mV peak-to-peak, whereas extra-cellular techniques (see Fig.3 for illustration), result-in lower spike amplitudes, as low as few tens of micro-Volts, often with both sharp positive and negative peaks. It is worth noting that not all the cells that are classified as neurons by morphological or cell staining criteria actually produce the type of spikes that define the scope of the spiking neuron models. For example, cochlear hair cells, retinal receptor cells, and retinal bipolar cells do not spike. Furthermore, many cells in the nervous system are not classified as neurons by neither criteria but rather as glia. Ultimately, biological neuron models aim to explain the mechanisms underlying the operation of the nervous system for the purpose of restoring lost control capabilities such as perception (detection), motor movement decision making and execution, and continuous limb control (estimation). In that sense, biological neural models differ from artificial neuron models that do not presume to predict the outcomes of experiments involving the biological neural tissue (although artificial neuron models are also concerned with execution of detection and estimation tasks). An important aspect of biological neuron models is therefore experimental validation and the use of physical units to describe the experimental procedure associated with the model predictions.

Neuron models can be divided into two categories according to the physical units of the interface of the model. Each category could be further divided according to the abstraction\detail level:
 * 1)  Electrical inputs  / Membrane Voltage Models - These models produce a prediction for membrane output voltage as function of electrical stimulation at the input stage (either voltage or current). The various models in this category differ in the exact functional relationship between the input current and the output voltage, and in the detail level. Some models in this category are black box models and distinguish only between two measured voltage levels: the presence of a spike (also known as ”action potential”) or a quiescent state, whereas other models are more detailed and account for sub-cellular processes.
 * 2)  Natural or  Pharmacological Neuron Models - These models were inspired from experiments involving either natural or pharmacological stimulation, experiments which exhibit a clear stochastic behavior. The output of these models is therefore the probability of a spike event as function of the input stimulus which can be natural external stimulus or pharmacological, depending on the model. Typically, this output probability is normalized (divided by) a time constant, and the resulting normalized probability is called the "Firing Rate" and has units of Hertz. The models in within this category differ in the functional relationship connecting the input stimulus to the output probability where models that are categorized as  Markov models are simpler and yield more tractable results.  Multi unit measurement.png

Although it is not unusual in science and engineering to have several descriptive models for different abstraction/detail levels, the number of different, sometimes contradicting, biological neuron models is exceptionally high. This situation is partly the result of the many different experimental settings, and the difficulty to distinguish between measurements effects, interactions of many cells (network effects) to the intrinsic properties of a neuron. To accelerate the convergence to a unified theory, we list several models in each category, and where applicable, also references to experiments that support each model.

Electrical Input / Electrical Output Neuron Models
The models in this category connect between neuron membrane currents at the input stage, to membrane voltage at the output stage. The most extensive experimental inquiry in this category of models was made by Hodgkin–Huxley in the early the 1950's using an experimental setup that punctured the cell membrane and allowed to force membrane to a certain voltage/current. It is important to note that in modern application which rely on  an electrical neural interface, the stimulation is extra-cellular to avoid tissue damage and cell death due to membrane puncturing. Hence, it is not clear to what extent the electrical neuron models hold for extra-cellular stimulation (see e.g. ).

Integrate-and-fire
One of the earliest models of a neuron was first investigated in 1907 by Louis Lapicque. A neuron is represented in time by


 * $$I(t)=C_\mathrm{m} \frac{d V_\mathrm{m}(t)}{d t}$$

which is just the time derivative of the law of capacitance, $Q = CV$. When an input current is applied, the membrane voltage increases with time until it reaches a constant threshold $V_{th}$, at which point a delta function spike occurs and the voltage is reset to its resting potential, after which the model continues to run. The firing frequency of the model thus increases linearly without bound as input current increases.

The model can be made more accurate by introducing a refractory period $t_{ref}$ that limits the firing frequency of a neuron by preventing it from firing during that period. Through some calculus involving a Fourier transform, the firing frequency as a function of a constant input current thus looks like


 * $$\,\! f(I)= \frac{I} {C_\mathrm{m} V_\mathrm{th} + t_\mathrm{ref} I}$$.

A remaining shortcoming of this model is that it implements no time-dependent memory. If the model receives a below-threshold signal at some time, it will retain that voltage boost forever until it fires again. This characteristic is clearly not in line with observed neuronal behavior.

=== Hodgkin–Huxley  ===

The Hodgkin–Huxley model connects between ion currents crossing the neuron cell membrane to the membrane voltage. The model is based on experiments that allowed to force membrane voltage using an intra-cellular pipette. This model is based on the concept of membrane ion channels and relies on data from the squid giant axon. In terms of recognition by the scientific community, this model is a very successful as Hodgkin–Huxley won the Nobel Prize for their work.

We note as before our voltage-current relationship, this time generalized to include multiple voltage-dependent currents:


 * $$C_\mathrm{m} \frac{d V(t)}{d t} = -\sum_i I_i (t, V)$$.

Each current is given by Ohm's Law as


 * $$I(t,V) = g(t,V)\cdot(V-V_\mathrm{eq})$$

where $g(t,V)$ is the conductance, or inverse resistance, which can be expanded in terms of its constant average $ḡ$ and the activation and inactivation fractions $m$ and $h$, respectively, that determine how many ions can flow through available membrane channels. This expansion is given by


 * $$g(t,V)=\bar{g}\cdot m(t,V)^p \cdot h(t,V)^q$$

and our fractions follow the first-order kinetics


 * $$\frac{d m(t,V)}{d t} = \frac{m_\infty(V)-m(t,V)}{\tau_\mathrm{m} (V)} = \alpha_\mathrm{m} (V)\cdot(1-m) - \beta_\mathrm{m} (V)\cdot m$$

with similar dynamics for $h$, where we can use either $τ$ and $m_{∞}$ or $α$ and $β$ to define our gate fractions.

With such a form, all that remains is to individually investigate each current one wants to include. Typically, these include inward Ca2+ and Na+ input currents and several varieties of K+ outward currents, including a "leak" current.

The end result can be at the small end 20 parameters which one must estimate or measure for an accurate model, and for complex systems of neurons not easily tractable by computer. Careful simplifications of the Hodgkin–Huxley model are therefore needed.

==== Experimental Evidence Supporting the Model by Hodgkin and Huxley  : ====

Leaky integrate-and-fire
In the leaky integrate-and-fire model, the memory problem is solved by adding a "leak" term to the membrane potential, reflecting the diffusion of ions that occurs through the membrane when some equilibrium is not reached in the cell. The model looks like


 * $$I(t)-\frac{V_\mathrm{m} (t)}{R_\mathrm{m}} = C_\mathrm{m} \frac{d V_\mathrm{m} (t)}{d t}$$

where $R_{m}$ is the membrane resistance, as we find it is not a perfect insulator as assumed previously. This forces the input current to exceed some threshold $I_{th} = V_{th} / R_{m}$ in order to cause the cell to fire, else it will simply leak out any change in potential. The firing frequency thus looks like


 * $$f(I) =

\begin{cases} 0, & I \le I_\mathrm{th} \\ {[} t_\mathrm{ref}-R_\mathrm{m} C_\mathrm{m} \log(1-\tfrac{V_\mathrm{th}}{I R_\mathrm{m}}) {]}^{-1}, & I > I_\mathrm{th} \end{cases} $$

which converges for large input currents to the previous leak-free model with refractory period.

Exponential integrate-and-fire
In the Exponential Integrate-and-Fire, spike generation is exponential, following the equation:


 * $$ \frac{dX}{dt} = \Delta_T \exp \left( \frac{X - X_T} {\Delta_T} \right) $$.

where $$X$$ is the membrane potential, $$X_T$$ is the membrane potential threshold, and $$\Delta_T$$ is the sharpness of action potential initiation, usually around 1 mV for cortical pyramidal neurons. Once the membrane potential crosses $$X_T$$, it diverges to infinity in finite time.

FitzHugh–Nagumo
Sweeping simplifications to Hodgkin–Huxley were introduced by FitzHugh and Nagumo in 1961 and 1962. Seeking to describe "regenerative self-excitation" by a nonlinear positive-feedback membrane voltage and recovery by a linear negative-feedback gate voltage, they developed the model described by


 * $$\begin{array}{rcl}

\dfrac{d V}{d t} &=& V-V^3 - w + I_\mathrm{ext} \\ \\ \tau \dfrac{d w}{d t} &=& V-a-b w \end{array}$$

where we again have a membrane-like voltage and input current with a slower general gate voltage $w$ and experimentally-determined parameters $a = -0.7, b = 0.8, τ = 1/0.08$. Although not clearly derivable from biology, the model allows for a simplified, immediately available dynamic, without being a trivial simplification.

Morris–Lecar
In 1981 Morris and Lecar combined Hodgkin–Huxley and FitzHugh–Nagumo into a voltage-gated calcium channel model with a delayed-rectifier potassium channel, represented by


 * $$\begin{array}{rcl}

C\dfrac{d V}{d t} &=& -I_\mathrm{ion}(V,w) + I \\ \\ \dfrac{d w}{d t} &=& \phi \cdot \dfrac{w_{\infty} - w}{\tau_{w}} \end{array}$$

where $$I_\mathrm{ion}(V,w) = \bar{g}_\mathrm{Ca} m_{\infty}\cdot(V-V_\mathrm{Ca}) + \bar{g}_\mathrm{K} w\cdot(V-V_\mathrm{K}) + \bar{g}_\mathrm{L}\cdot(V-V_\mathrm{L})$$.

Hindmarsh–Rose
Building upon the FitzHugh–Nagumo model, Hindmarsh and Rose proposed in 1984 a model of neuronal activity described by three coupled first order differential equations:


 * $$\begin{array}{rcl}

\dfrac{d x}{d t} &=& y+3x^2-x^3-z+I \\ \\ \dfrac{d y}{d t} &=& 1-5x^2-y \\ \\ \dfrac{d z}{d t} &=& r\cdot (4(x + \tfrac{8}{5})-z) \end{array}$$

with $r^{2} = x^{2} + y^{2} + z^{2}$, and $r ≈ 10^{−2}$ so that the $z$ variable only changes very slowly. This extra mathematical complexity allows a great variety of dynamic behaviors for the membrane potential, described by the $x$ variable of the model, which include chaotic dynamics. This makes the Hindmarsh–Rose neuron model very useful, because being still simple, allows a good qualitative description of the many different patterns of the action potential observed in experiments.

Cable theory
Cable theory describes the dendritic arbor as a cylindrical structure undergoing a regular pattern of bifurcation, like branches in a tree. For a single cylinder or an entire tree, the input conductance at the base (where the tree meets the cell body, or any such boundary) is defined as


 * $$G_{in} = \frac{G_\infty \tanh(L) + G_L}{1+(G_L / G_\infty )\tanh(L)}$$,

where $L$ is the electrotonic length of the cylinder which depends on its length, diameter, and resistance. A simple recursive algorithm scales linearly with the number of branches and can be used to calculate the effective conductance of the tree. This is given by


 * $$\,\! G_D = G_m A_D \tanh(L_D) / L_D$$

where $A_{D} = πld$ is the total surface area of the tree of total length $l$, and $L_{D}$ is its total electrotonic length. For an entire neuron in which the cell body conductance is $G_{S}$ and the membrane conductance per unit area is $G_{md} = G_{m} / A$, we find the total neuron conductance $G_{N}$ for $n$ dendrite trees by adding up all tree and soma conductances, given by


 * $$G_N = G_S + \sum_{j=1}^n A_{D_j} F_{dga_j}$$,

where we can find the general correction factor $F_{dga}$ experimentally by noting $G_{D} = G_{md}A_{D}F_{dga}$.

Compartmental models
The cable model makes a number of simplifications to give closed analytic results, namely that the dendritic arbor must branch in diminishing pairs in a fixed pattern. A compartmental model allows for any desired tree topology with arbitrary branches and lengths, but makes simplifications in the interactions between branches to compensate. Thus, the two models give complementary results, neither of which is necessarily more accurate.

Each individual piece, or compartment, of a dendrite is modeled by a straight cylinder of arbitrary length $l$ and diameter $d$ which connects with fixed resistance to any number of branching cylinders. We define the conductance ratio of the $i$th cylinder as $B_{i} = G_{i} / G_{∞}$, where $$G_\infty=\tfrac{\pi d^{3/2}}{2\sqrt{R_i R_m}}$$ and $R_{i}$ is the resistance between the current compartment and the next. We obtain a series of equations for conductance ratios in and out of a compartment by making corrections to the normal dynamic $B_{out,i} = B_{in,i+1}$, as


 * $$B_{\mathrm{out},i} = \frac{B_{\mathrm{in},i+1}(d_{i+1}/d_i)^{3/2} }{ \sqrt{R_{\mathrm{m},i+1}/R_{\mathrm{m},i}} }$$
 * $$B_{\mathrm{in},i} = \frac{ B_{\mathrm{out},i} + \tanh X_i }{ 1+B_{\mathrm{out},i}\tanh X_i }$$
 * $$B_\mathrm{out,par} = \frac{B_\mathrm{in,dau1} (d_\mathrm{dau1}/d_\mathrm{par})^{3/2}} {\sqrt{R_\mathrm{m,dau1}/R_\mathrm{m,par}}} + \frac{B_\mathrm{in,dau2} (d_\mathrm{dau2}/d_\mathrm{par})^{3/2}} {\sqrt{R_\mathrm{m,dau2}/R_\mathrm{m,par}}} + \ldots$$

where the last equation deals with parents and daughters at branches, and $$X_i = \tfrac{l_i \sqrt{4R_i}}{\sqrt{d_i R_m}}$$. We can iterate these equations through the tree until we get the point where the dendrites connect to the cell body (soma), where the conductance ratio is $B_{in,stem}$. Then our total neuron conductance is given by


 * $$G_N = \frac{A_\mathrm{soma}}{R_\mathrm{m,soma}} + \sum_j B_{\mathrm{in,stem},j} G_{\infty,j}$$.

An example of a compartmental model of a neuron, with an algorithm to reduce the number of compartments (increase the computational speed) and yet retain the salient electrical characteristics, can be found in.

Natural Stimulus Neuron Models
The models in this category were derived following experiments involving natural stimulation which exhibit a clear stochastic behavior. Consequently, the following models generate a probabilistic relationship between the input stimulus to spike occurrences.

=== The Non-Homogeneous Poisson Process Model (Siebert 1965, Seibert 1970) === Siebret l suggested to model neuron spike firing pattern using the Non-Homogeneous Poisson Process model, following experiments involving the auditory system. According to Siebert, the probability of a spiking event at the time interval $$[t, t+\Delta_t]$$ is proportional to a non negative function $$g[s(t)]$$, where $$s(t)$$ is the raw stimulus.:

$$P_{spike}(t\in[t',t'+\Delta_t])=\Delta_t \cdot g[s(t)]$$

Siebert considered several functions as $$g[s(t)]$$, including $$g[s(t)] \propto s^2(t)$$ for low stimulus intensities.

The main advantage of Siebert's model is its simplicity. The shortcomings of the model is its inability to reflect properly the following phenomena: These shortcoming are addressed by the two state Markov Model.
 * The edge emphasizing property of the neuron in response to a stimulus pulse.
 * The saturation of the firing rate.
 * The values of inter-spike-interval-histogram at short intervals values (close to zero).

=== The Two State Markov Model   (Nossenson & Messer, 2010, 2012, 2015) ===

The spiking neuron model by Nossenson & Messer produces the probability of the neuron to fire a spike as a function of either an external or pharmacological stimulus. The model consists of a cascade of a receptor layer model and a spiking neuron model, as shown in Fig 5. The connection between the external stimulus to the spiking probability is made in two steps: First, a receptor cell model translates the raw external stimulus to neurotransmitter concentration, then, a spiking neuron model connects between neurotransmitter concentration to the firing rate (spiking probability). Thus, the spiking neuron model by itself depends on neurotransmitter concentration at the input stage. An important feature of this model is the prediction for neurons firing rate pattern which captures, using a low number of free parameters, the characteristic edge emphasized response of neurons to a stimulus pulse, as shown in Fig. 5. The firing rate is identified both as a normalized probability for neural spike firing, and as a quantity proportional to the current of neurotransmitters released by the cell. The expression for the firing rate takes the following form:

$$R_{fire}(t)=\frac{P_{spike}(t;\Delta_t)}{\Delta_t}=[y(t)+R_0] \cdot P_0(t)$$

where,
 * P0 is the probability of the neuron to be "armed" and ready to fire. It is given by the following differential equation:

$$\dot{P}_0=-[y(t)+R_0+R_1] \cdot P_0(t) +R_1$$

P0 could be generally calculated recursively using Euler method, but in the case of a pulse of stimulus it yields a simple closed form expression.
 * y(t) is the input of the model and is interpreted as the neurotransmitter concentration on the cell surrounding (in most cases glutamate) . For an external stimulus it can be estimated through the receptor layer model:

$$y(t) \simeq  g_{gain} \cdot s^2(t)$$, with $$s^2(t)$$ being stimulus power (given in Watt or other energy per time unit).
 * R0 corresponds to the intrinsic spontaneous firing rate of the neuron.
 * R1 is the recovery rate of the neuron from refractory state.

Other predictions by this model include:

1) The averaged Evoked Response Potential (ERP) due to population of many neurons in unfiltered measurements resembles the firing rate..

2) The voltage variance of activity due to multiple neuron activity resembles the firing rate (also known as Multi-Unit-Activity power or MUA)..

3) The inter-spike-interval probability distribution takes the form a gamma-distribution like function.

Non-Markovian Models
Below is a list of Non Markovian neuron models.
 * Johnson, and Swami
 * Berry and Meister
 * Kass and Ventura

Pharmacological Stimulus Neural Models
The models in this category produce predictions for experiments involving pharmacological stimulation.

Synaptic transmission (Koch & Segev, 1999)
The response of a neuron to individual neurotransmitters can be modeled as an extension of the classical Hodgkin–Huxley model with both standard and nonstandard kinetic currents. Four neurotransmitters primarily have influence in the CNS. AMPA/kainate receptors are fast excitatory mediators while NMDA receptors mediate considerably slower currents. Fast inhibitory currents go through GABAA receptors, while GABAB receptors mediate by secondary G-protein-activated potassium channels. This range of mediation produces the following current dynamics:


 * $$I_\mathrm{AMPA}(t,V) = \bar{g}_\mathrm{AMPA} \cdot [O] \cdot (V(t)-E_\mathrm{AMPA})$$
 * $$I_\mathrm{NMDA}(t,V) = \bar{g}_\mathrm{NMDA} \cdot B(V) \cdot [O] \cdot (V(t)-E_\mathrm{NMDA})$$
 * $$I_\mathrm{GABA_A}(t,V) = \bar{g}_\mathrm{GABA_A} \cdot ([O_1]+[O_2]) \cdot (V(t)-E_\mathrm{Cl})$$
 * $$I_\mathrm{GABA_B}(t,V) = \bar{g}_\mathrm{GABA_B} \cdot \tfrac{[G]^n}{[G]^n+K_\mathrm{d}} \cdot (V(t)-E_\mathrm{K})$$

where $ḡ$ is the maximal conductance (around 1S) and $E$ is the equilibrium potential of the given ion or transmitter (AMDA, NMDA, Cl, or K), while $[O]$ describes the fraction of receptors that are open. For NMDA, there is a significant effect of magnesium block that depends sigmoidally on the concentration of intracellular magnesium by $B(V)$. For GABAB, $[G]$ is the concentration of the G-protein, and $K_{d}$ describes the dissociation of G in binding to the potassium gates.

The dynamics of this more complicated model have been well-studied experimentally and produce important results in terms of very quick synaptic potentiation and depression, that is, fast, short-term learning.

=== The Two State Markov Model  (Nossenson & Messer, 2010, 2012, 2015) === The model by Nossenson and Messer (2010) translates neurotransmitter concentration at the input stage to probability of releasing neurotransmitter at the output stage. For a more detailed description of this model see this Section.

Applications
The question of neural modeling is at the heart of the following projects:

Electrical Retinal Prosthesis
Further reading on this subject:

Neurotransmitter based Retinal Prosthesis
Further reading on this subject

Artificial Limb Control and Sensation
Further reading on this subject see:

Conjecture 1: Relation between Artificial and Biological neuron models
The most basic model of a neuron consists of an input with some synaptic weight vector and an activation function or transfer function inside the neuron determining output. This is the basic structure used in artificial neurons, which in a neural network often looks like

$$y_i = \phi\left( \sum_j w_{ij} x_j \right)$$

where $y_{i}$ is the output of the $i$ th neuron, $x_{j}$ is the $j$th input neuron signal, $w_{ij}$ is the synaptic weight (or strength of connection) between the neurons $i$ and $j$, and $φ$ is the activation function. While this model has seen success in machine-learning applications, it is a poor model for real (biological) neurons, because it lacks the time-dependence that real neurons exhibit. Some of the earliest biological models took this form until kinetic models such as the Hodgkin–Huxley model became dominant.

In the case of modelling a biological neuron, physical analogues are used in place of abstractions such as "weight" and "transfer function". A neuron is filled and surrounded with water containing ions, which carry electric charge. The neuron is bound by an insulating cell membrane and can maintain a concentration of charged ions on either side that determines a capacitance $C_{m}$. The firing of a neuron involves the movement of ions into the cell that occurs when neurotransmitters cause ion channels on the cell membrane to open. We describe this by a physical time-dependent current $I(t)$. With this comes a change in voltage, or the electrical potential energy difference between the cell and its surroundings, which is observed to sometimes result in a voltage spike called an action potential which travels the length of the cell and triggers the release of further neurotransmitters. The voltage, then, is the quantity of interest and is given by $V_{m}(t)$.

=== Conjecture 2: Loops of spiking neurons for decision making ===

General Comments regarding the Modern Perspective of Scientific and Engineering Models
The models above are still idealizations. Corrections must be made for the increased membrane surface area given by numerous dendritic spines, temperatures significantly hotter than room-temperature experimental data, and nonuniformity in the cell's internal structure. Certain observed effects do not fit into some of these models. For instance, the temperature cycling (with minimal net temperature increase) of the cell membrane during action potential propagation not compatible with models which rely on modeling the membrane as a resistance which must dissipate energy when current flows through it. The transient thickening of the cell membrane during action potential propagation is also not predicted by these models, nor is the changing capacitance and voltage spike that results from this thickening incorporated into these models. The action of some anesthetics such as inert gases is problematic for these models as well. New models, such as the soliton model attempt to explain these phenomena, but are less developed than older models and have yet to be widely applied. Also improbable possibility of modelling of local chronobiology mechanisms.

Modern views regarding of the role of the scientific model suggest that ”All models are wrong but some are useful” (Box and Draper, 1987, Gribbin, 2009; Paninski et al., 2009).