Correlation of speech recognition by cochlear implant users with metrics based on thresholds and loudness levels

Abstract

Intersubject variability in speech recognition is a persistent characteristic of cochlear implant users, despite significant improvements in average performance. This study examines simple psychophysical measures, maximum comfortable loudness levels (C-levels) and dynamic ranges (DRs) that may serve for diagnosis of conditions that affect speech recognition. Eighteen subjects who had used Nucleus CI24 cochlear implants for a minimum of 1 year participated in the study. T and C levels were measured using the method of adjustment. Consonant, vowel, and sentence recognition were tested using the subjects everyday speech processors (SPEAK or ACE). Across-site variation (ASV) was quantified for each subject using the variance of the T and C levels across all tested stimulation sites. Mean DRs (C levels minus T levels) for all tested sites were calculated for each subject. The results showed a moderate but significant negative correlation between speech-recognition performance and ASV of the T and C levels measured with both monopolar and bipolar electrode configurations. Mean DRs also correlated with speech performance. These results support the hypothesis that the large across-site variation in detection thresholds and small dynamic range reflect conditions that are detrimental to speech-recognition with cochlear implants. [Work supported by NIDCD Grant DC03808.]