Model-based design of optimal neurostimulation in the NHP hippocampus for enhancing behavioral task performance

A general model-based methodology is presented for the optimal design of stimulation patterns in the CA1 region of the hippocampus of non-human primates (NHP) that seeks to enhance performance in a Delayed-Match-to-Sample task. The methodology follows a hierarchical Volterra-type modeling approach that expresses the probability of a correct behavioral outcome in terms of multi-convolutional modules involving "Triggering Likelihood Functions" (TLFs), which represent the interactions among multiple neuronal spikes as they impact the outcome. This TLF-based model is estimated from experimental spike-train data recorded in the CA1 region of the hippocampus with multi-electrode arrays. The model can be used to compute the likelihood of a correct outcome for any given set of spike-train data of CA1 multi-neuron activity. This enables the design of the optimal stimulation pattern through a computational search procedure under proper constraints on mean firing rates. We present results of the TLF-based model obtained from experimental NHP data and initial experimental validation of the designed optimal stimulation pattern.