Anatomically based Bayesian decoding of the cortical response to intracortical microstimulation

Sensory prostheses must encode complex surrogate sensory signals as patterns of microstimulation, and yet no established framework exists for quantifying the information delivered to the brain. Here we develop a Bayesian decoder based on the underlying anatomy in cortex and evaluate it on population neural activity recorded from the vibrissa region of rodent primary somatosensory cortex with voltage sensitive dye imaging in response to intracortical microstimulation (ICMS). The anatomically based decoder accurately classified stimulus location from the multi-electrode array, indicating a degree of similarity between the whisker evoked neural response and that generated by intracortical microstimulation. However, when the decoder was modified to discriminate between whisker and electrical stimuli, it did so with high performance. Ultimately, the results presented here establish the beginnings of a generalized decoding framework for evaluating stimulus encoding models designed to deliver surrogate sensory signals to the brain.