A sensitive estimator for crosscorrelograms

Abstract

The best established method for finding interactions between extracellularly recorded neurons is the crosscorrelation technique. The method is simple and useful, but it has some drawbacks. One of them is its limited sensitivity to weak interactions, which are common in the mammalian cerebral cortex. In the present paper a new method for the estimation of interaction strength is presented. This method is based on the intensity representation of point processes, and provides an optimal estimator for the intensity of the postsynaptic spike train. The estimator is complicated to use, but it can be approximated by a simple estimator, similar to ordinary measures of synaptic efficacy like the area under the crosscorrelogram peak. Simulation results, showing the advantage of the new estimator over the commonly used efficacy estimators and some measure of its robustness to deviations from model assumptions, are presented. Finally, application of the estimator to the analysis of simultaneous recordings of physiological single units is demonstrated.