Model-Based Inference of Electrode Distance and Neuronal Density from Measured Detection Thresholds in Cochlear Implant Listeners.

Purpose: Cochlear implants (CI) are a highly successful neural prosthesis that can restore hearing in individuals with sensorineural hearing loss. However, the extent of hearing restoration varies widely. Two major factors likely contribute to poor performance: (1) the distances between electrodes and surviving spiral ganglion neurons and (2) the density of those neurons. Reprogramming the CI at a poor electrode-neuron interface, using focused tripolar stimulation or remapping the electrodes, would benefit from understanding the cause of the poor interface.

Methods: We used a cochlear model with simplified geometry and neuronal composition to investigate how the interface affects stimulation thresholds. We then inverted the model to infer electrode distance and neuronal density from monopolar and tripolar threshold values obtained behaviorally. We validated this inverted model for known scenarios of electrode distance and neuronal density. Finally, we assessed the model using data from 18 CI users whose electrode distances were measured from CT imaging.

Results: The inverted model accurately inferred electrode distance and neuronal density for known scenarios. It also reliably reproduced behavioral monopolar and tripolar threshold profiles for CI users, with mean prediction errors within 1 dB for 17/18 subjects. Fits of electrode distance were more variable; accuracy depended on the assumed value of temporal bone resistivity. Twelve subjects had minimum distance error (0.31 mm) using low resistivity (70 Ω-cm) while the others had better fits (0.30 mm) with higher resistivity (250 Ω-cm).

Conclusion: This inverted model shows promise as a simple, practical tool to better assess and understand the electrode-neuron interface.