Coefficient of variation vs. mean interspike interval curves: What do they tell us about the brain?

Altered Primary Motor Cortex Neuronal Activity in a Rat Model of Harmaline-Induced Tremor During Thalamic Deep Brain Stimulation

Although deep brain stimulation (DBS) is a clinically effective surgical treatment for essential tremor (ET), and its neurophysiological mechanisms are not fully understood. As the motor thalamus is the most popular DBS target for ET, and it is known that the thalamic nucleus plays a key role in relaying information about the external environment to the cerebral cortex, it is important to investigate mechanisms of thalamic DBS in the context of the cerebello-thalamo-cortical neuronal network. To examine this, we measured single-unit neuronal activities in the resting state in M1 during VL thalamic DBS in harmaline-induced tremor rats and analyzed neuronal activity patterns in the thalamo-cortical circuit. Four activity patterns - including oscillatory burst, oscillatory non-burst, irregular burst, and irregular non-burst - were identified by harmaline administration; and those firing patterns were differentially affected by VL thalamic DBS, which seems to drive pathologic cortical signals to signals in normal status. As specific neuronal firing patterns like oscillation or burst are considered important for information processing, our results suggest that VL thalamic DBS may modify pathophysiologic relay information rather than simply inhibit the information transmission.

The Jacobi diffusion process as a neuronal model

The Jacobi process is a stochastic diffusion characterized by a linear drift and a special form of multiplicative noise which keeps the process confined between two boundaries. One example of such a process can be obtained as the diffusion limit of the Stein's model of membrane depolarization which includes both excitatory and inhibitory reversal potentials. The reversal potentials create the two boundaries between which the process is confined. Solving the first-passage-time problem for the Jacobi process, we found closed-form expressions for mean, variance, and third moment that are easy to implement numerically. The first two moments are used here to determine the role played by the parameters of the neuronal model; namely, the effect of multiplicative noise on the output of the Jacobi neuronal model with input-dependent parameters is examined in detail and compared with the properties of the generic Jacobi diffusion. It appears that the dependence of the model parameters on the rate of inhibition turns out to be of primary importance to observe a change in the slope of the response curves. This dependence also affects the variability of the output as reflected by the coefficient of variation. It often takes values larger than one, and it is not always a monotonic function in dependency on the rate of excitation.

A novel approach to probabilistic characterisation of neural firing patterns

BACKGROUND

The advances in extracellular neural recording techniques result in big data volumes that necessitate fast, reliable, and automatic identification of statistically similar units. This study proposes a single framework yielding a compact set of probabilistic descriptors that characterise the firing patterns of a single unit.

NEW METHOD

Probabilistic features are estimated from an inter-spike-interval time series, without assumptions about the firing distribution or the stationarity. The first level of proposed firing patterns decomposition divides the inter-spike intervals into bursting, moderate and idle firing modes, yielding a coarse feature set. The second level identifies the successive bursting spikes, or the spiking acceleration/deceleration in the moderate firing mode, yielding a refined feature set. The features are estimated from simulated data and from experimental recordings from the lateral prefrontal cortex in awake, behaving rhesus monkeys.

RESULTS

An efficient and stable partitioning of neural units is provided by the ensemble evidence accumulation clustering. The possibility of selecting the number of clusters and choosing among coarse and refined feature sets provides an opportunity to explore and compare different data partitions.

CONCLUSIONS

The estimation of features, if applied to a single unit, can serve as a tool for the firing analysis, observing either overall spiking activity or the periods of interest in trial-to-trial recordings. If applied to massively parallel recordings, it additionally serves as an input to the clustering procedure, with the potential to compare the functional properties of various brain structures and to link the types of neural cells to the particular behavioural states.

Conductance-based refractory density approach: comparison with experimental data and generalization to lognormal distribution of input current

The conductance-based refractory density (CBRD) approach is an efficient tool for modeling interacting neuronal populations. The model describes the firing activity of a statistical ensemble of uncoupled Hodgkin-Huxley-like neurons, each receiving individual Gaussian noise and a common time-varying deterministic input. However, the approach requires experimental validation and extension to cases of distributed input signals (or input weights) among different neurons of such an ensemble. Here the CBRD model is verified by comparing with experimental data and then generalized for a lognormal (LN) distribution of the input weights. The model with equal weights is shown to reproduce efficiently the post-spike time histograms and the membrane voltage of experimental multiple trial response of single neurons to a step-wise current injection. The responses reveal a more rapid reaction of the firing-rate than voltage. Slow adaptive potassium channels strongly affected the shape of the responses. Next, a computationally efficient CBRD model is derived for a population with the LN input weight distribution and is compared with the original model with equal input weights. The analysis shows that the LN distribution: (1) provides a faster response, (2) eliminates oscillations, (3) leads to higher sensitivity to weak stimuli, and (4) increases the coefficient of variation of interspike intervals. In addition, a simplified firing-rate type model is tested, showing improved precision in the case of a LN distribution of weights. The CBRD approach is recommended for complex, biophysically detailed simulations of interacting neuronal populations, while the modified firing-rate type model is recommended for computationally reduced simulations.

Note on the coefficient of variations of neuronal spike trains

It is known that many neurons in the brain show spike trains with a coefficient of variation (CV) of the interspike times of approximately 1, thus resembling the properties of Poisson spike trains. Computational studies have been able to reproduce this phenomenon. However, the underlying models were too complex to be examined analytically. In this paper, we offer a simple model that shows the same effect but is accessible to an analytic treatment. The model is a random walk model with a reflecting barrier; we give explicit formulas for the CV in the regime of excess inhibition. We also analyze the effect of probabilistic synapses in our model and show that it resembles previous findings that were obtained by simulation.

Variable Effects of Acoustic Trauma on Behavioral and Neural Correlates of Tinnitus In Individual Animals

The etiology of tinnitus is known to be diverse in the human population. An appropriate animal model of tinnitus should incorporate this pathological diversity. Previous studies evaluating the effect of acoustic over exposure (AOE) have found that animals typically display increased spontaneous firing rates and bursting activity of auditory neurons, which often has been linked to behavioral evidence of tinnitus. However, only a subset of studies directly associated these neural correlates to individual animals. Furthermore, the vast majority of tinnitus studies were conducted on anesthetized animals. The goal of this study was to test for a possible relationship between tinnitus, hearing loss, hyperactivity and bursting activity in the auditory system of individual unanesthetized animals following AOE. Sixteen mice were unilaterally exposed to 116 dB SPL narrowband noise (centered at 12.5 kHz) for 1 h under ketamine/xylazine anesthesia. Gap-induced prepulse inhibition of the acoustic startle reflex (GPIAS) was used to assess behavioral evidence of tinnitus whereas hearing performance was evaluated by measurements of auditory brainstem response (ABR) thresholds and prepulse inhibition PPI audiometry. Following behavioral assessments, single neuron firing activity was recorded from the inferior colliculus (IC) of four awake animals and compared to recordings from four unexposed controls. We found that AOE increased spontaneous activity in all mice tested, independently of tinnitus behavior or severity of threshold shifts. Bursting activity did not increase in two animals identified as tinnitus positive (T+), but did so in a tinnitus negative (T−) animal with severe hearing loss (SHL). Hyperactivity does not appear to be a reliable biomarker of tinnitus. Our data suggest that multidisciplinary assessments on individual animals following AOE could offer a powerful experimental tool to investigate mechanisms of tinnitus.

Distinguishing the Causes of Firing with the Membrane Potential Slope

In this letter, we aim to measure the relative contribution of coincidence detection and temporal integration to the firing of spikes of a simple neuron model. To this end, we develop a method to infer the degree of synchrony in an ensemble of neurons whose firing drives a single postsynaptic cell. This is accomplished by studying the effects of synchronous inputs on the membrane potential slope of the neuron and estimating the degree of response-relevant input synchrony, which determines the neuron's operational mode. The measure is calculated using the normalized slope of the membrane potential prior to the spikes fired by a neuron, and we demonstrate that it is able to distinguish between the two operational modes. By applying this measure to the membrane potential time course of a leaky integrate-and-fire neuron with the partial somatic reset mechanism, which has been shown to be the most likely candidate to reflect the mechanism used in the brain for reproducing the highly irregular firing at high rates, we show that the partial reset model operates as a temporal integrator of incoming excitatory postsynaptic potentials and that coincidence detection is not necessary for producing such high irregular firing.

Learning optimisation by high firing irregularity

In a network of leaky integrate-and-fire (LIF) neurons, we investigate the functional role of irregular spiking at high rates. Irregular spiking is produced by either employing the partial somatic reset mechanism on every LIF neuron of the network or by using temporally correlated inputs. In both the benchmark problem of XOR (exclusive-OR) and in a general-sum game, it is shown that irrespective of the mechanism that is used to produce it, high firing irregularity enhances the learning capability of the spiking neural network trained with reward-modulated spike-timing-dependent plasticity. These results suggest that the brain may be utilising high firing irregularity for the purposes of learning optimisation.

Firing variability is higher than deduced from the empirical coefficient of variation

A convenient and often used summary measure to quantify the firing variability in neurons is the coefficient of variation (CV), defined as the standard deviation divided by the mean. It is therefore important to find an estimator that gives reliable results from experimental data, that is, the estimator should be unbiased and have low estimation variance. When the CV is evaluated in the standard way (empirical standard deviation of interspike intervals divided by their average), then the estimator is biased, underestimating the true CV, especially if the distribution of the interspike intervals is positively skewed. Moreover, the estimator has a large variance for commonly used distributions. The aim of this letter is to quantify the bias and propose alternative estimation methods. If the distribution is assumed known or can be determined from data, parametric estimators are proposed, which not only remove the bias but also decrease the estimation errors. If no distribution is assumed and the data are very positively skewed, we propose to correct the standard estimator. When defining the corrected estimator, we simply use that it is more stable to work on the log scale for positively skewed distributions. The estimators are evaluated through simulations and applied to experimental data from olfactory receptor neurons in rats.

Does high firing irregularity enhance learning?

In this note, we demonstrate that the high firing irregularity produced by the leaky integrate-and-fire neuron with the partial somatic reset mechanism, which has been shown to be the most likely candidate to reflect the mechanism used in the brain for reproducing the highly irregular cortical neuron firing at high rates (Bugmann, Christodoulou, & Taylor, 1997; Christodoulou & Bugmann, 2001), enhances learning. More specifically, it enhances reward-modulated spike-timing-dependent plasticity with eligibility trace when used in spiking neural networks, as shown by the results when tested in the simple benchmark problem of XOR, as well as in a complex multiagent setting task.

Comparison of local measures of spike time irregularity and relating variability to firing rate in motor cortical neurons

Spike time irregularity can be measured by the coefficient of variation. However, it overestimates the irregularity in the case of pronounced firing rate changes. Several alternative measures that are local in time and therefore relatively rate-independent were proposed. Here we compared four such measures: CV(2), LV, IR and SI. First, we asked which measure is the most efficient for time-resolved analyses of experimental data. Analytical results show that CV(2) has the less variable estimates. Second, we derived useful properties of CV(2) for gamma processes. Third, we applied CV(2) on recordings from the motor cortex of a monkey performing a delayed motor task to characterize the irregularity, that can be modulated or not, and decoupled or not from firing rate. Neurons with a CV(2)-rate decoupling have a rather constant CV(2) and discharge mainly irregularly. Neurons with a CV(2)-rate coupling can modulate their CV(2) and explore a larger range of CV(2) values.

Receptor-Agonist Interactions in Service-Theoretic Perspective, Effects of Molecular Timing on the Shape of Dose-Response Curves

Service-theoretic concepts and methods, widely used in other fields (e.g., telecommunication and operations research), are useful also in a biochemical setting because the treatment of biocatalysts (enzymes, receptors) as servers and their ligands as customers, based on the established formal methods of service or queuing theory, may lead to insights and results unobtainable by conventional, mass-action-law-based theories. In this article, we apply the service-theoretic approach to receptor-agonist systems and show how by changing the stochastic time pattern of "operationally relevant" point events (e.g., instants of agonist arrival, instants of post-climax agonist departure) a great variety of dose-response curves may be generated, even in very simple reaction schemes, which, according to mass action kinetics, invariably lead to hyperbolic r(A) curves (r and A stand for response and agonist concentration, respectively). The molecular timing inherent to a hyperbolic response system is not optimal: for instance, at the agonist concentration A(50), half of the agonist molecules are rejected ("lost") because of unfortunate timing of the arrival events. The fraction of lost arrivers can be diminished considerably if the arrivals are better timed: "sub-Poisson" arrivals improve the timing and, thus, convert hyperbolic r(A) curves into "lifted" nonhyperbolic ones. Conversely, "super-Poisson" arrivals make the non-optimal timing in hyperbolic response systems even worse and, thus, convert hyperbolic r(A) curves into "depressed" nonhyperbolic ones. Furthermore, under special timing conditions, nonhyperbolic r(A) curves can be generated, which are partly lifted, partly depressed relative to the reference hyperbola, and which resemble in shape well-known nonhyperbolic forms of enzyme and receptor kinetics (negatively cooperative, positively cooperative, and sigmoidal kinetics). In addition unusual (undulatory and sawtooth-like) r(A) curves can be generated solely by changing the temporal pattern of arrival and service completion instants. Virtually any shape of dose-response curves may be obtained by allowing for probability distributions whose characteristic shape varies with their mean; we call such distributions "variomorphic" and apply them to the arrival process of agonist molecules.

Impact of synaptic noise and conductance state on spontaneous cortical firing

Cortical neurons in-vivo operate in a continuum of overall conductance states, depending on the average level of background synaptic input throughout the dendritic tree. We compare how variability, or fluctuations, in this input affects the statistics of the resulting 'spontaneous' or 'background' firing activity, between two extremes of the mean input corresponding to a low-conductance (LC) and a high-conductance (HC) state. In the HC state, we show that both firing rate and regularity increase with increasing variability. In the LC state, firing rate also increases with input variability, but in contrast to the HC state, firing regularity first decreases and then increases with an increase in the variability. At high levels of input variability, firing regularity in both states converge to similar values.

Reproducing bursting interspike interval statistics of the gustatory cortex

Cortical neurons in vivo generate highly irregular spike sequences. Recently, it was experimentally found that the local variation of interspike intervals, LV, is nearly constant for every spike sequence for the same neurons. On the contrary, the coefficient of variation, CV, varies over different spike sequences. Here, we first show that these characteristic features are also applicable in bursting spike sequences that are obtained from the rat gustatory cortex. Next, we show that the conventional leaky integrate-and-fire model does not fully account for reproducing these statistical features in data of real bursting spike sequences. We resolve this difficulty by proposing an alternative neuron model which is a reduction of the bursting neuron model involving the persistent sodium current. Our study implies that (1) the characteristic features of CV and LV are the results of the endogenous bursting and (2) the bursting behavior in the gustatory cortex is caused mainly by the persistent sodium current.

Analytical integrate-and-fire neuron models with conductance-based dynamics for event-driven simulation strategies

Event-driven simulation strategies were proposed recently to simulate integrate-and-fire (IF) type neuronal models. These strategies can lead to computationally efficient algorithms for simulating large-scale networks of neurons; most important, such approaches are more precise than traditional clock-driven numerical integration approaches because the timing of spikes is treated exactly. The drawback of such event-driven methods is that in order to be efficient, the membrane equations must be solvable analytically, or at least provide simple analytic approximations for the state variables describing the system. This requirement prevents, in general, the use of conductance-based synaptic interactions within the framework of event-driven simulations and, thus, the investigation of network paradigms where synaptic conductances are important. We propose here a number of extensions of the classical leaky IF neuron model involving approximations of the membrane equation with conductance-based synaptic current, which lead to simple analytic expressions for the membrane state, and therefore can be used in the event-driven framework. These conductance-based IF (gIF) models are compared to commonly used models, such as the leaky IF model or biophysical models in which conductances are explicitly integrated. All models are compared with respect to various spiking response properties in the presence of synaptic activity, such as the spontaneous discharge statistics, the temporal precision in resolving synaptic inputs, and gain modulation under in vivo-like synaptic bombardment. Being based on the passive membrane equation with fixed-threshold spike generation, the proposed gIF models are situated in between leaky IF and biophysical models but are much closer to the latter with respect to their dynamic behavior and response characteristics, while still being nearly as computationally efficient as simple IF neuron models. gIF models should therefore provide a useful tool for efficient and precise simulation of large-scale neuronal networks with realistic, conductance-based synaptic interactions.

Sensitive dependence of the coefficient of variation of interspike intervals on the lower boundary of membrane potential for the leaky integrate-and-fire neuron model

After the report of Softky and Koch [Softky, W.R., Koch, C., 1993. The highly irregular firing of cortical cells is inconsistent with temporal integration of random EPSPs. J. Neurosci. 13, 334-350], leaky integrate-and-fire models have been investigated to explain high coefficient of variation (CV) of interspike intervals (ISIs) at high firing rates observed in the cortex. The purpose of this paper is to study the effect of the position of a lower boundary of membrane potential on the possible value of CV of ISIs based on the diffusional leaky integrate-and-fire models with and without reversal potentials. Our result shows that the irregularity of ISIs for the diffusional leaky integrate-and-fire neuron significantly changes by imposing a lower boundary of membrane potential, which suggests the importance of the position of the lower boundary as well as that of the firing threshold when we study the statistical properties of leaky integrate-and-fire neuron models. It is worth pointing out that the mean-CV plot of ISIs for the diffusional leaky integrate-and-fire neuron with reversal potentials shows a close similarity to the experimental result obtained in Softky and Koch [Softky, W.R., Koch, C., 1993. The highly irregular firing of cortical cells is inconsistent with temporal integration of random EPSPs. J. Neurosci. 13, 334-350].

Measurement of time-dependent changes in the irregularity of neural spiking

Irregularity of firing in spike trains has been associated with coding processes and information transfer or alternatively treated as noise. Previous studies of irregularity have mainly used the coefficient of variation (CV) of the interspike interval distribution. Proper estimation of CV requires a constant underlying firing rate, a condition that most experimental situations do not fulfill either within or across trials. Here we introduce a novel irregularity metric based on the ratio of adjacent intervals in the spike train. The new metric is not affected by firing rate and is very localized in time so that it can be used to examine the time course of irregularity relative to an alignment marker. We characterized properties of the new metric with simulated spike trains of known characteristics and then applied it to data recorded from 108 single neurons in the motor cortex of two monkeys during performance of a precision grip task. Fifty-six cells were antidromically identified as pyramidal tract neurons (PTNs). Sixty-one cells (30 PTNs) exhibited significant temporal modulation of their irregularity during task performance with the contralateral hand. The irregularity modulations generally differed in sign and latency from the modulations of firing rate. High irregularity tended to occur during the task phases requiring the most detailed control of movement, whereas neural firing became more regular during the steady hold phase. Such irregularity modulation could have important consequences for the response of downstream neurons and may provide insight into the nature of the cortical code.

Similarity of interspike interval distributions and information gain in a stationary neuronal firing

The Kullback-Leibler (KL) information distance is proposed for judging similarity between two different interspike interval (ISI) distributions. The method is applied by a comparison of four common ISI descriptors with an exponential model which is characterized by the highest entropy. Under the condition of equal mean ISI values, the KL distance corresponds to information gain coming from the state described by the exponential distribution to the state described by the chosen ISI model. It has been shown that information can be transmitted changing neither the spike rate nor coefficient of variation (CV). Furthermore the KL distance offer an indication of the exponentiality of the chosen ISI descriptor (or data): the distance is zero if, and only if, the ISIs are distributed exponentially. Finally an application on experimental data coming from the olfactory sensory neurons of rats is shown.

What is signal and what is noise in the brain?

The response of a cortical neuron to a stimulus can show a very large variability when repeatedly stimulated by exactly the same stimulus. This has been quantified in terms of inter-spike-interval (ISI) statistics by several researchers (e.g., [Softky, W., Koch, C., 1993. The highly irregular firing of cortical cells is inconsistent with temporal integration of random EPSPs. J. Neurosci. 13(1), 334-350.]). The common view is that this variability reflects noisy information processing based on redundant representation in large neuron populations. This view has been challenged by the idea that the apparent noise inherent in brain activity that is not strictly related or temporally coupled to the experiment could be functionally significant. In this work we examine the ISI statistics and discuss these views in a recently published model of interacting cortical areas [Knoblauch, A., Palm, G., 2002. Scene segmentation by spike synchronization in reciprocally connected visual areas. I. Local effects of cortical feedback. Biol. Cybernet. 87(3), 151-167.]. From the results of further single neuron simulations we can isolate temporally modulated synaptic input as a main contributor for high ISI variability in our model and possibly in real neurons. In contrast to alternative mechanisms, our model suggests a function of the temporal modulations for short-term binding and segmentation of figures from background. Moreover, we show that temporally modulated inputs lead to ISI statistics which fit better to the neurophysiological data than alternative mechanisms.

Synchronized firing of FitzHugh-Nagumo neurons by noise

We investigate the influence of noise on synchronization between the spiking activities of neurons with external impulsive forces. We first analyze the dependence of the synchronized firing on the amplitude and the angular frequency of the impulsive force in the noise-free system. Three cases (regular spiking, traveling wave, and chaotic spiking) with low synchronized firing are chosen to study effects due to noise. In each case we find that small noise can be a promoter of synchronization phenomena in neural activities, by choosing an appropriate noise intensity acting on some of the neurons.

The discharge variability of neocortical neurons during high-conductance states

In vivo recordings have shown that the discharge of cortical neurons is often highly variable and can have statistics similar to a Poisson process with a coefficient of variation around unity. To investigate the determinants of this high variability, we analyzed the spontaneous discharge of Hodgkin-Huxley type models of cortical neurons, in which in vivo-like synaptic background activity was modeled by random release events at excitatory and inhibitory synapses. By using compartmental models with active dendrites, or single compartment models with fluctuating conductances and fluctuating currents, we found that a high discharge variability was always paralleled with a high-conductance state, while some active and passive cellular properties had only a minor impact. Furthermore, a balance between excitation and inhibition was not a necessary condition for high discharge variability. We conclude that the fluctuating high-conductance state caused by the ongoing activity in the cortical network in vivo may be viewed as a natural determinant of the highly variable discharges of these neurons.

A spiking neuron model: applications and learning

This paper presents a biologically inspired, hardware-realisable spiking neuron model, which we call the Temporal Noisy-Leaky Integrator (TNLI). The dynamic applications of the model as well as its applications in Computational Neuroscience are demonstrated and a learning algorithm based on postsynaptic delays is proposed. The TNLI incorporates temporal dynamics at the neuron level by modelling both the temporal summation of dendritic postsynaptic currents which have controlled delay and duration and the decay of the somatic potential due to its membrane leak. Moreover, the TNLI models the stochastic neurotransmitter release by real neuron synapses (with probabilistic RAMs at each input) and the firing times including the refractory period and action potential repolarisation. The temporal features of the TNLI make it suitable for use in dynamic time-dependent tasks like its application as a motion and velocity detector system presented in this paper. This is done by modelling the experimental velocity selectivity curve of the motion sensitive H1 neuron of the visual system of the fly. This application of the TNLI indicates its potential applications in artificial vision systems for robots. It is also demonstrated that Hebbian-based learning can be applied in the TNLI for postsynaptic delay training based on coincidence detection, in such a way that an arbitrary temporal pattern can be detected and recognised. The paper also demonstrates that the TNLI can be used to control the firing variability through inhibition; with 80% inhibition to concurrent excitation, firing at high rates is nearly consistent with a Poisson-type firing variability observed in cortical neurons. It is also shown with the TNLI, that the gain of the neuron (slope of its transfer function) can be controlled by the balance between inhibition and excitation, the gain being a decreasing function of the proportion of inhibitory inputs. Finally, in the case of perfect balance between inhibition and excitation, i.e. where the average input current is zero, the neuron can still fire as a result of membrane potential fluctuations. The firing rate is then determined by the average input firing rate. Overall this work illustrates how a hardware-realisable neuron model can capitalise on the unique computational capabilities of biological neurons.