Enhanced Signal Detection by Adaptive Decorrelation of Interspike Intervals

摘要

Spike trains with negative interspike interval (ISI) correlations, in whichlong/short ISIs are more likely followed by short/long ISIs, are commonin many neurons. They can be described by stochastic models with aspike-triggered adaptation variable. We analyze a phenomenon in thesemodels where such statistically dependent ISI sequences arise in tandemwith quasi-statistically independent and identically distributed (quasi-IID) adaptation variable sequences. The sequences of adaptation statesand resulting ISIs are linked by a nonlinear decorrelating transformation.We establish general conditions on a family of stochastic spikingmodels that guarantee this quasi-IID property and establish bounds onthe resulting baseline ISI correlations. Inputs that elicit weak firing ratechanges in samples with many spikes are known to be more detectiblewhen negative ISI correlations are present because they reduce spikecount variance; this defines a variance-reduced firing rate coding benchmark.We performed a Fisher information analysis on these adaptingmodels exhibiting ISI correlations to show that a spike pattern code basedon the quasi-IID property achieves the upper bound of detection performance,surpassing rate codes with the same mean rate—including thevariance-reduced rate code benchmark—by 20% to 30%. The informationloss in rate codes arises because the benefits of reduced spike countvariance cannot compensate for the lower firing rate gain due to adaptation.Since adaptation states have similar dynamics to synaptic responses,the quasi-IID decorrelation transformation of the spike train is plausiblyimplemented by downstream neurons through matched postsynaptic kinetics.This provides an explanation for observed coding performance in sensory systems that cannot be accounted for by rate coding, for example,at the detection threshold where rate changes can be insignificant.