Advances in Computational Modeling of Cochlear Implant Physiology and Perception

Models of cochlear implant physiology and perception have historically utilized deterministic descriptions of auditory-nerve (AN) responses to electrical stimulation, which ignore stochastic activity present in the response. Physiological models of AN responses have been developed that do incorporate stochastic activity [8][13][14][27][38][39], but the consequences of stochastic activity for the perception of cochlear implant stimulation have not been investigated until recently [3]. Such an investigation is prompted by inaccuracies in predicting cochlear implant perception by deterministic models. For example, studies of single-fiber responses, where only an arbitrary deterministic measure of threshold is recorded, do not accurately predict perceptual threshold versus phase duration (strength-duration) curves for sinusoidal stimulation [24] or for pulsatile stimulation [25][26]. Furthermore, strength-duration curves of cochlear implant users are not well predicted by deterministic Hodgkin–Huxley type models [25] [30]. However, the complexity of previous stochastic physiological models has made the computation of responses for large numbers of fibers both laborious and time-consuming. Furthermore, the parameters of these models are often not easily matched to the fiber characteristics of the auditory nerve in humans or other mammals. This has prompted us to develop a simpler and more computationally efficient model of electrical stimulation of the auditory nerve [1][2][4] which is capable of direct and rapid prediction of perceptual data [3].