Inferring Neural Connectivity via Measured Delay in Directed Information Estimates

Directed information between two neural signals has previously been used to successfully determine if there is a synapse connecting one neuron to the other. However, there are situations in which the directed information method leads to false positives for detecting connections due to the influence of a third neuron. We propose a method that accounts for these cases using the delay-profile of the causally-conditioned entropy rate to find a measured delay range and its intersection with a connectionbased delay range. Through simulations in NEURON, we show how this method outperforms connectivity inference based on directed information alone.