Stochastic resonance in the mutual information between input and output spike trains of noisy central neurons

Within the framework of information theory, we investigate the phenomenon of stochastic resonance in a realistic stochastic model of central neurons. The essential point is that, for central neurons, the input is not a continuous signal but a train of spikes which has, in general, a non-periodic distribution. The behavior of an individual neuron is modeled by a combination of an Ornstein-Uhlenbeck diffusion process and a jump process corresponding to a discrete input spike train. The input spike train is assumed to be given by a Poisson distribution. If the parameters are such that the input spike train provokes a membrane potential that hovers under the threshold, the noise given by the diffusive part of the neuron model induces a maximum in the mutual information between the input and output spike trains. Thus, the stochastic resonance effect results in a strategy for efficient information transmission of subthreshold signals. Copyright (C) 1998 Elsevier Science B.V.