Statistics of inverse interspike intervals: The instantaneous firing rate revisited.

The rate coding hypothesis is the oldest and still one of the most accepted and investigated scenarios in neuronal activity analyses. However, the actual neuronal firing rate, while informally understood, can be mathematically defined in several different ways. These definitions yield distinct results; even their average values may differ dramatically for the simplest neuronal models. Such an inconsistency, together with the importance of "firing rate," motivates us to revisit the classical concept of the instantaneous firing rate. We confirm that different notions of firing rate can in fact be compatible, at least in terms of their averages, by carefully discerning the time instant at which the neuronal activity is observed. Two general cases are distinguished: either the inspection time is synchronised with a reference time or with the neuronal spiking. The statistical properties of the instantaneous firing rate, including parameter estimation, are analyzed, and compatibility with the intuitively understood concept is demonstrated.

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.

Presynaptic Spontaneous Activity Enhances the Accuracy of Latency Coding.

The time to the first spike after stimulus onset typically varies with the stimulation intensity. Experimental evidence suggests that neural systems use such response latency to encode information about the stimulus. We investigate the decoding accuracy of the latency code in relation to the level of noise in the form of presynaptic spontaneous activity. Paradoxically, the optimal performance is achieved at a nonzero level of noise and suprathreshold stimulus intensities. We argue that this phenomenon results from the influence of the spontaneous activity on the stabilization of the membrane potential in the absence of stimulation. The reported decoding accuracy improvement represents a novel manifestation of the noise-aided signal enhancement.

The Gamma renewal process as an output of the diffusion leaky integrate-and-fire neuronal model.

Statistical properties of spike trains as well as other neurophysiological data suggest a number of mathematical models of neurons. These models range from entirely descriptive ones to those deduced from the properties of the real neurons. One of them, the diffusion leaky integrate-and-fire neuronal model, which is based on the Ornstein-Uhlenbeck (OU) stochastic process that is restricted by an absorbing barrier, can describe a wide range of neuronal activity in terms of its parameters. These parameters are readily associated with known physiological mechanisms. The other model is descriptive, Gamma renewal process, and its parameters only reflect the observed experimental data or assumed theoretical properties. Both of these commonly used models are related here. We show under which conditions the Gamma model is an output from the diffusion OU model. In some cases, we can see that the Gamma distribution is unrealistic to be achieved for the employed parameters of the OU process.

Coding accuracy is not fully determined by the neuronal model.

It is automatically assumed that the accuracy with which a stimulus can be decoded is entirely determined by the properties of the neuronal system. We challenge this perspective by showing that the identification of pure tone intensities in an auditory nerve fiber depends on both the stochastic response model and the arbitrarily chosen stimulus units. We expose an apparently paradoxical situation in which it is impossible to decide whether loud or quiet tones are encoded more precisely. Our conclusion reaches beyond the topic of auditory neuroscience, however, as we show that the choice of stimulus scale is an integral part of the neural coding problem and not just a matter of convenience.

Stein's neuronal model with pooled renewal input.

The input of Stein's model of a single neuron is usually described by using a Poisson process, which is assumed to represent the behaviour of spikes pooled from a large number of presynaptic spike trains. However, such a description of the input is not always appropriate as the variability cannot be separated from the intensity. Therefore, we create and study Stein's model with a more general input, a sum of equilibrium renewal processes. The mean and variance of the membrane potential are derived for this model. Using these formulas and numerical simulations, the model is analyzed to study the influence of the input variability on the properties of the membrane potential and the output spike trains. The generalized Stein's model is compared with the original Stein's model with Poissonian input using the relative difference of variances of membrane potential at steady state and the integral square error of output interspike intervals. Both of the criteria show large differences between the models for input with high variability.

Estimating latency from inhibitory input.

Stimulus response latency is the time period between the presentation of a stimulus and the occurrence of a change in the neural firing evoked by the stimulation. The response latency has been explored and estimation methods proposed mostly for excitatory stimuli, which means that the neuron reacts to the stimulus by an increase in the firing rate. We focus on the estimation of the response latency in the case of inhibitory stimuli. Models used in this paper represent two different descriptions of response latency. We consider either the latency to be constant across trials or to be a random variable. In the case of random latency, special attention is given to models with selective interaction. The aim is to propose methods for estimation of the latency or the parameters of its distribution. Parameters are estimated by four different methods: method of moments, maximum-likelihood method, a method comparing an empirical and a theoretical cumulative distribution function and a method based on the Laplace transform of a probability density function. All four methods are applied on simulated data and compared.

Metabolic cost of neuronal information in an empirical stimulus-response model.

The limits on maximum information that can be transferred by single neurons may help us to understand how sensory and other information is being processed in the brain. According to the efficient-coding hypothesis (Barlow, Sensory Comunication, MIT press, Cambridge, 1961), neurons are adapted to the statistical properties of the signals to which they are exposed. In this paper we employ methods of information theory to calculate, both exactly (numerically) and approximately, the ultimate limits on reliable information transmission for an empirical neuronal model. We couple information transfer with the metabolic cost of neuronal activity and determine the optimal information-to-metabolic cost ratios. We find that the optimal input distribution is discrete with only six points of support, both with and without a metabolic constraint. However, we also find that many different input distributions achieve mutual information close to capacity, which implies that the precise structure of the capacity-achieving input is of lesser importance than the value of capacity.

On the estimate of the rate constant in the homogeneous dissolution model.

The aim is to determine how well the rate parameter of the homogeneous model of dissolution can be estimated in dependency on the chosen times to measure the empirical data. The approach is based on the theory of Fisher information. We show that if the probability distribution of the measurement errors is known, the data should be collected at a single time instant or its close proximity in order to obtain the best estimate. This is in sharp contrast with commonly used experimental protocols. Further, from the properties of the Fisher information we deduce how suitable is the model of measurement error and we show that asymmetric distribution of data close to the time origin is unavoidable.

A simple neuronal model with intrinsic saturation of the firing frequency.

We present a comparison of the intrinsic saturation of firing frequency in four simple neural models: leaky integrate-and-fire model, leaky integrate-and-fire model with reversal potentials, two-point leaky integrate-and-fire model, and a two-point leaky integrate-and-fire model with reversal potentials. "Two-point" means that the equivalent circuit has two nodes (dendritic and somatic) instead of one (somatic only). The results suggest that the reversal potential increases the slope of the "firing rate vs input" curve due to a smaller effective membrane time constant, but does not necessarily induce saturation of the firing rate. The two-point model without the reversal potential does not limit the voltage or the firing rate. In contrast to the previous models, the two-point model with the reversal potential limits the asymptotic voltage and the firing rate, which is the main result of this paper. The case of excitatory inputs is considered first and followed by the case of both excitatory and inhibitory inputs.

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.

Estimation of time-dependent input from neuronal membrane potential.

The set of firing rates of the presynaptic excitatory and inhibitory neurons constitutes the input signal to the postsynaptic neuron. Estimation of the time-varying input rates from intracellularly recorded membrane potential is investigated here. For that purpose, the membrane potential dynamics must be specified. We consider the Ornstein-Uhlenbeck stochastic process, one of the most common single-neuron models, with time-dependent mean and variance. Assuming the slow variation of these two moments, it is possible to formulate the estimation problem by using a state-space model. We develop an algorithm that estimates the paths of the mean and variance of the input current by using the empirical Bayes approach. Then the input firing rates are directly available from the moments. The proposed method is applied to three simulated data examples: constant signal, sinusoidally modulated signal, and constant signal with a jump. For the constant signal, the estimation performance of the method is comparable to that of the traditionally applied maximum likelihood method. Further, the proposed method accurately estimates both continuous and discontinuous time-variable signals. In the case of the signal with a jump, which does not satisfy the assumption of slow variability, the robustness of the method is verified. It can be concluded that the method provides reliable estimates of the total input firing rates, which are not experimentally measurable.

First-spike latency in the presence of spontaneous activity.

A new statistical method for the estimation of the response latency is proposed. When spontaneous discharge is present, the first spike after the stimulus application may be caused by either the stimulus itself, or it may appear due to the prevailing spontaneous activity. Therefore, an appropriate method to deduce the response latency from the time to the first spike after the stimulus is needed. We develop a nonparametric estimator of the response latency based on repeated stimulations. A simulation study is provided to show how the estimator behaves with an increasing number of observations and for different rates of spontaneous and evoked spikes. Our nonparametric approach requires very few assumptions. For comparison, we also consider a parametric model. The proposed probabilistic model can be used for both single and parallel neuronal spike trains. In the case of simultaneously recorded spike trains in several neurons, the estimators of joint distribution and correlations of response latencies are also introduced. Real data from inferior colliculus auditory neurons obtained from a multielectrode probe are studied to demonstrate the statistical estimators of response latencies and their correlations in space.

Fano Factor: A Potentially Useful Information.

The Fano factor, defined as the variance-to-mean ratio of spike counts in a time window, is often used to measure the variability of neuronal spike trains. However, despite its transparent definition, careless use of the Fano factor can easily lead to distorted or even wrong results. One of the problems is the unclear dependence of the Fano factor on the spiking rate, which is often neglected or handled insufficiently. In this paper we aim to explore this problem in more detail and to study the possible solution, which is to evaluate the Fano factor in the operational time. We use equilibrium renewal and Markov renewal processes as spike train models to describe the method in detail, and we provide an illustration on experimental data.

Competitive and noncompetitive odorant interactions in the early neural coding of odorant mixtures.

Most olfactory receptor neurons (ORNs) express a single type of olfactory receptor that is differentially sensitive to a wide variety of odorant molecules. The diversity of possible odorant-receptor interactions raises challenging problems for the coding of complex mixtures of many odorants, which make up the vast majority of real world odors. Pure competition, the simplest kind of interaction, arises when two or more agonists can bind to the main receptor site, which triggers receptor activation, although only one can be bound at a time. Noncompetitive effects may result from various mechanisms, including agonist binding to another site, which modifies the receptor properties at the main binding site. Here, we investigated the electrophysiological responses of rat ORNs in vivo to odorant agonists and their binary mixtures and interpreted them in the framework of a quantitative model of competitive interaction between odorants. We found that this model accounts for all concentration-response curves obtained with single odorants and for about half of those obtained with binary mixtures. In the other half, the shifts of curves along the concentration axis and the changes of maximal responses with respect to model predictions, indicate that noncompetitive interactions occur and can modulate olfactory receptors. We conclude that, because of their high frequency, the noncompetitive interactions play a major role in the neural coding of natural odorant mixtures. This finding implies that the CNS activity caused by mixtures will not be easily analyzed into components, and that mixture responses will be difficult to generalize across concentration.

Efficient olfactory coding in the pheromone receptor neuron of a moth.

The concept of coding efficiency holds that sensory neurons are adapted, through both evolutionary and developmental processes, to the statistical characteristics of their natural stimulus. Encouraged by the successful invocation of this principle to predict how neurons encode natural auditory and visual stimuli, we attempted its application to olfactory neurons. The pheromone receptor neuron of the male moth Antheraea polyphemus, for which quantitative properties of both the natural stimulus and the reception processes are available, was selected. We predicted several characteristics that the pheromone plume should possess under the hypothesis that the receptors perform optimally, i.e., transfer as much information on the stimulus per unit time as possible. Our results demonstrate that the statistical characteristics of the predicted stimulus, e.g., the probability distribution function of the stimulus concentration, the spectral density function of the stimulation course, and the intermittency, are in good agreement with those measured experimentally in the field. These results should stimulate further quantitative studies on the evolutionary adaptation of olfactory nervous systems to odorant plumes and on the plume characteristics that are most informative for the 'sniffer'. Both aspects are relevant to the design of olfactory sensors for odour-tracking robots.

On the estimation of refractory period.

The refractory period concept is exploited in many neurophysiological studies as a part of developed methods. Computations connected to these methods are influenced by the estimated value of refractory period. In this article, parametric and nonparametric refractory period estimation methods (minimum, maximum likelihood estimate, minimum risk estimate, etc.) are compared for three neuronal models. The results acquired from simulated data illustrate both accuracy and bias of different techniques. Experimental data are also investigated to show the limitations of the methods discussed.

Classification of stimuli based on stimulus-response curves and their variability.

Neuronal responses evoked in sensory neurons by static stimuli of various intensities are usually characterized by their input-output transfer function, i.e. by plotting the firing frequency (or any other measurable neuron response) versus the corresponding stimulus intensity. The aim of the present article is to determine the stimulus intensities which can be considered as "the most important" from two different points of view: transferring as much information as possible and coding the intensity as precisely as possible. These two problems are very different because, for example, an informative signal may be difficult to identify. We show that the role of noise is crucial in both problems. To obtain the range of stimuli which are the best identified, we propose to use measures based on Fisher information as known from the theory of statistical inference. To classify the most important stimuli from the point of view of information transfer, we suggest methods based on information theory. We show that both the most identifiable signal and the most informative signal are not unique. To study this, a generic model of input-output transfer function is analyzed under the influence of several different types of noise. Finally, the methods are illustrated on a model and data pertaining to olfactory sensory neurons.

A review of the methods for signal estimation in stochastic diffusion leaky integrate-and-fire neuronal models.

Parameters in diffusion neuronal models are divided into two groups; intrinsic and input parameters. Intrinsic parameters are related to the properties of the neuronal membrane and are assumed to be known throughout the paper. Input parameters characterize processes generated outside the neuron and methods for their estimation are reviewed here. Two examples of the diffusion neuronal model, which are based on the integrate-and-fire concept, are investigated--the Ornstein--Uhlenbeck model as the most common one and the Feller model as an illustration of state-dependent behavior in modeling the neuronal input. Two types of experimental data are assumed-intracellular describing the membrane trajectories and extracellular resulting in knowledge of the interspike intervals. The literature on estimation from the trajectories of the diffusion process is extensive and thus the stress in this review is set on the inference made from the interspike intervals.

Errors in estimation of the input signal for integrate-and-fire neuronal models.

Estimation of the input parameters of stochastic (leaky) integrate-and-fire neuronal models is studied. It is shown that the presence of a firing threshold brings a systematic error to the estimation procedure. Analytical formulas for the bias are given for two models, the randomized random walk and the perfect integrator. For the third model considered, the leaky integrate-and-fire model, the study is performed by using Monte Carlo simulated trajectories. The bias is compared with other errors appearing during the estimation, and it is documented that the effect of the bias has to be taken into account in experimental studies.