How Does Deep Neural Network-Based Noise Reduction in Hearing Aids Impact Cochlear Implant Candidacy?

Area of Science:

• Audiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Adults with hearing loss often struggle with sentence recognition, impacting cochlear implant candidacy.
• Deep neural network (DNN)-based noise reduction is a novel approach to enhance hearing aid performance.

Purpose of the Study:

• To evaluate the impact of DNN-based noise reduction on speech perception in hearing-impaired adults.
• To assess how DNN technology influences cochlear implant candidacy criteria.

Main Methods:

• Speech perception was tested using AzBio sentences in quiet and noisy conditions (multi-talker babble) at varying signal-to-noise ratios (SNRs).
• Sentence recognition scores were compared across three hearing aid programs: calm, speech in noise, and DNN-based noise reduction.
• Bench analyses included phase inversion for SNR prediction and Hearing-Aid Speech Perception Index (HASPI v2) for intelligibility prediction.

Main Results:

• The DNN-based noise reduction program improved speech perception by 20-32% points compared to the standard program.
• Phase inversion predicted a 4-5 dB SNR improvement with DNN noise reduction.
• HASPI v2 slightly overestimated measured speech intelligibility.

Conclusions:

• DNN-based noise reduction enhances speech perception in noisy environments.
• This technology may hinder cochlear implant qualification for some patients with residual hearing.
• The adoption or necessity of cochlear implants could be altered by advanced noise reduction.