Using convex optimization to compute channel capacity in a channel model of cochlear implant stimulation

The goal of cochlear implants is to restore hearing by replacing the function of missing inner hair cells using an array of electrodes to directly stimulate auditory nerves. Previous studies have already established an information theoretic modeling framework for the electrode-neural interface, which enables prediction of the optimal number of electrodes by calculating mutual information between channel input (choice of electrodes) and channel output (defined as a function of the active nerve fibers in response to an electrode choice). In order to estimate whether the positions or usage probabilities of electrodes could impact on the performance of cochlear implants from an information theoretic perspective, a convex optimization method is adapted to compute channel capacity and capacity-achieving input distribution of the cochlear implant model in this paper. We relate the capacity-achieving input distribution to the optimal locations and usage probabilities of electrodes. The result shows that the optimized distribution of electrodes increases the maximum mutual information and predicts the best performance of cochlear implants.