Information theoretic optimization of cochlear implant electrode usage probabilities

Xiao Gao, David B. Grayden, Mark D. McDonnell

Abstract: Cochlear implants are neural prostheses that can restore hearing. Contemporary cochlear implant devices consist of up to 22 electrodes. An open question for future cochlear implants is whether new electrode designs that enable less current spread may provide improved hearing performance through more precise control of neural activation, more electrodes, or more precise positioning. Here we use a recently proposed information theoretic model for the electro-neural interface that enables estimates to be made of the optimal number of electrodes for different amounts of current spread. We apply information theoretic approaches for finding the channel capacity in the model to enable estimates of optimal electrode usage probabilities and positions. We also compare the performance in the model when auditory nerve fibers are assumed to be heterogenous, with a random distribution of firing thresholds and relative spreads, versus an assumption that they are all identical.