A computational study to model the effect of electrode-to-auditory nerve fiber distance on spectral resolution in cochlear implant

摘要

Spectral ripple discrimination (SRD) has been widely used to evaluate the spectral resolution in cochlear implant (CI) recipients based on its strong correlation with speech perception performance. However, despite its usefulness for predicting speech perception outcomes, SRD performance exhibits large across-subject variabilities even among subjects implanted with the same CIs and sound processors. The potential factors of this observation include current spread, nerve survival, and CI mapping. Previous studies have found that the spectral resolution reduces with increasing distance of the stimulation electrode from the auditory nerve fibers (ANFs), attributable to increasing current spread. However, it remains unclear whether the spread of excitation is the only cause of the observation, or whether other factors such as temporal interaction also contribute to it. In this study, we used a computational model to investigate channel interaction upon non-simultaneous stimulation with respect to the electrode–ANF distance, and evaluated the SRD performance for five electrode–ANF distances. The SRD performance was determined based on the similarity between two neurograms in response to standard and inverted stimuli and used to evaluate the spectral resolution in the computational model. The spread of excitation was observed to increase with increasing electrode–ANF distance, consistent with previous findings. Additionally, the preceding pulses delivered from neighboring channels induced a channel interaction that either inhibited or facilitated the neural responses to subsequent pulses depending on the electrode–ANF distance. The SRD performance was also found to decrease with increasing electrode–ANF distance. The findings of this study suggest that variation of the neural responses (inhibition or facilitation) with the electrode–ANF distance in CI users may cause spectral smearing, and hence poor spectral resolution. A computational model such as that used in this study is a useful tool for understanding the neural factors related to CI outcomes, such as cannot be accomplished by behavioral studies alone.