Models of Neural Activation for Patient-Specific Cochlear Implant Programming

Abstract

The cochlear implant (CI) is a neural prosthetic designed to restore the sensation of hearing through direct stimulation of auditory nerve fibers (ANFs) by an array of electrodes inserted into the cochlea. Although many patients achieve success with the device, a significant number receive little to no benefit. For this population, post-operative adjustments to the parameters that control how stimuli are delivered to the electrode array are the primary means for improving their outcomes. However, this process can be time-consuming, requiring multiple clinical visits over several months, causing strains on both patients and clinical resources. Additionally, programming is often performed without specific knowledge of the electrode array placement, which can further complicate programming if the true placement strays significantly from the expected location. For these reasons, it is advantageous for all to develop strategies that hasten the identification of optimal settings. In this dissertation, we present several such methods based on computational electroanatomical models (EAMs) of the inner ear with a focus on the incorporation of patient-specific features. The first and second aims of this work describe the design of algorithmic and visualization techniques for selecting the active electrode set from a distance-based model of the electric field. We then explore a more comprehensive, highly-detailed volume conduction model, including its use for the development of novel programming strategies that better reflect the individuality of each patient. This includes the third aim, which investigated the influence of neural model parameters on simulations of neural behavior, and the fourth aim, which explores the creation of a computationally-efficient patient-specific EAMs. The final aim is a model-based speech processing strategies which dynamically selects the electrode firing sequence based on simulated neural responses. These methods have shown an increased benefit to the clinical programming process compared to existing methods and expand the potential clinical applications of these computational models.