Modelling the user-specific human cochlea

Abstract

Cochlear implants have successfully been used to enable hundreds of thousands of profoundly deaf people to regain some perception of hearing. Hearing performance does, however, vary greatly among individual implant users. In order to gain a better understanding of the underlying factors that cause these inter-user performance differences, insight into the functioning of individual implant users’ hearing systems is required. Some of the parameters unique to an implanted user’s hearing system may be measured non-invasively using psychoacoustics or the measurement of electrically evoked compound action potentials. While these methods provide information of the macro response of a user’s hearing system to stimulation, individual parameters, for example the individual neurons that are excited cannot be measured. Some individual parameters are difficult or even impossible to measure in a living human as it is not technically feasible or invasive surgery is required. When obtaining measurements inside the inner ear of a living human is not an option, an alternative that mimics the human hearing system is required from which measurements can be predicted: models.

This study describes the development of a method to construct an electrical computational three-dimensional finite element model of the implanted cochlea of a specific living individual. This method is presented as a tool for researchers to probe the cochleae of specific implanted users non-invasively. Data from a low resolution computer tomography scan is used to construct a geometric representation of the bony outer cochlear structures and augmented with histologic data to construct the smaller inner cochlear structures. A detailed skull geometry with brain and scalp volumes that includes the user’s return electrode is also constructed. The user’s electrode array is modelled in its intra-cochlear location and stimulation is simulated using finite element modelling. The cochleae of five individual ears were modelled and intra-cochlear and neural node potentials were predicted along with neural excitation patterns.

Having models that can predict user-specific outcomes, predictions that include the variability between implanted ears are obtained. This allowed the comparison of modelled data to common trends found in literature and enabled the investigation of questions frequently asked by modellers. These include the effect that bone resistivity, head volume shape, return electrode implementation and return electrode position have on modelled results. These findings were incorporated and contributed to higher detailed models being produced than are currently described in literature.

The models were then practically applied in two areas. The first was in the quantification of potential decay in the cochlea where a simple model is derived to predict decay at the neuron level based on the location of an electrode. The second was in the translation of the model into the clinical domain where the mismatch between the perceived pitch and mapped frequencies of specific implanted individuals were predicted. Along with these predictions it was found that neural excitation and intra-cochlear potential spread are highly dependent on individual cochlear morphology. This warrants the inclusion of user-specific morphology in volume conduction models of the implanted cochlea where user-specific outcomes are predicted.