Spike train entropy and information transmission for a mathematical model of a bursting neuron.

In this paper we present calculations of the entropy and information transmission associated with spike trains produced by the circle/circle bursting model neuron in response to filtered white-noise stimuli. For most computations, we treated the bursts as unitary objects and estimated the entropy from the time intervals between the first spikes in consecutive bursts. In one case, we considered the intervals associated with all the spikes in the burst train or only the first and last spikes of each burst. We found that the entropy per burst was maximized when the stimulus was well matched to the neuron’s natural burst frequency and that the entropy/spike increased considerably when the duration of the burst was considered. Moreover, for a noisy stimulus for which the deterministic part of the stimulus was close to the natural burst frequency but most of the noise was at much higher frequencies, the bursting neuron transmitted considerably more information in bits/spike than a spiking model neuron constructed using similar ionic conductances.