Personalizing Transient Noise Reduction Algorithm Settings for Cochlear Implant Users.

The analysis was based on 15 participants, after one participant left and two were excluded for very poor speech thresholds, so the result is small. The participants were better than average in quiet speech recognition, with consonant-vowel-consonant scores of at least 82% and a mean of 91%, so less successful cochlear implant users may not respond the same way. The experimenter knew which condition was being tested and manually scored responses, which could introduce bias. 5 of 15 participants could not understand the speech during the personalization task and chose settings by comfort or least annoyance instead of intelligibility. The algorithms and the automatic gain control were simulated on a computer, so the findings reflect that simulated chain rather than a normal processor setting. Subjective ratings were analysed in only nine participants, and no algorithm differed from no reduction on those ratings. The group-based multiband setting did not improve the primary speech-in-noise measure, so the benefit depended on each person's own settings. The study did not correct all multiple comparisons across several analyses, so some positive findings need cautious interpretation.

No author conflict-of-interest statement is supplied. The metadata lists non-U.S. government research support. The study used Advanced Bionics processors, and default multiband settings came from informal pilot experiments at Advanced Bionics’ European Research Center.

Do not read the 1.26 dB laboratory improvement as proof that cochlear implant users will understand everyday noisy environments better or prefer the personalized multiband algorithm. The study did not show subjective rating benefits, the clinical standard broadband algorithm did not improve speech recognition, and the personalized broadband algorithm worsened speech recognition.