Methods and results
A supra-threshold filtered Poisson process with an intensity of 10 [s-1] was applied into an array of 50 stochastic Hodgkin-Huxley (HH) neuron models possessing stochastic sodium and potassium channels with a patch area of 100, 200, ..., and 700 [um2]. The stochastic ion channel gating was implemented by the channel-number tracking algorithm [2]. Each output spike train of neuron models was gathered and moving-averaged for calculating the rate of spike trains. Ten kinds of input realizations were applied repeatedly ten times to the array in order to estimate the total and noise entropies of the spike firing rate for calculating mutual information. Figures 1 and 2.
(left). A sample path of supra-threshold input signals (top column), raster plots of spike trains for 50 neuron models (middle column), and raster plots expanded temporally (bottom column) as a function of time at a patch area of 700 [um2]. The standard deviation of spike firing times (Jitter: JT) was estimated to be 0.331[ms]. (middle). A patch area of 400 [um2]. (JT=0.434[ms]). Figure 1 (right). A patch area of 100 [um2]. Spontaneous spike firings are observed, suggesting information lost (JT=0.491[ms]). Decreasing patch areas tended to increase randomness of spike firing times, and eventually to generate spontaneous spike firings.