PP410 - Effectiveness of a New Microphone Array with an Acoustic-Enhanced Deep Learning Model

Abstract:
Current hearing aids offer minimal help in real-world noisy settings. A novel hearing device by AudioFocus addresses this issue by using a "direct-to-reverberation energy ratio" cue. Subjects with normal hearing and mild to moderate SNHL were recruited. QuickSIN, Acceptable Noise Level and Listening Effort data were collected. Results revealed a statistically significantly lower listening effort in subjects with AudioFocus but no statistically significant differences in Quick SIN and ANL. These findings suggest that while the speech in noise benefits of AudioFocus remain unclear, it may help in reducing patients' effort in adverse listening environments, and potentially increase device adaptation rate.

Summary:
Rational/Purpose
Acoustically challenging environments can make individuals with hearing loss feel excluded and inadequate, which can lead to social isolation and depression. These conditions can significantly speed up cognitive decline and increase the risk of developing dementia and Alzheimer’s disease. However, only 20% of people aged 55-74 with hearing loss use hearing aids, with the most common reasons for underuse being "limited benefit" and "difficulty in noisy or background noise situations." Current hearing aids offer minimal help in real-world noisy settings, highlighting the urgent need for improvements in hearing aid performance in complex acoustic environments. A novel hearing device by AudioFocus addresses this issue by using an underexplored cue known as the "direct-to-reverberation energy ratio". The device leverages this cue with a deep learning AI model to amplify important nearby voices while suppressing distant voices that patients often perceive as noise. This study aimed to evaluate the clinical performance of the AudioFocus device, particularly in comparison to other high-end hearing aids that also use AI.

Methods
Subjects with normal hearing and mild to moderate sensorineural hearing loss (SNHL) were recruited to compare their speech perception in noise across different devices, as well as their subjective impressions of sound quality. Each participant was tested in a sound booth with one speaker in front of them and one speaker behind them. Participants' speech in noise understanding was assessed with the QuickSIN test at 50 dB SPL. Three conditions: No device, AI assisted traditional premium level hearing aid (Big 5 AI), and novel mic array device with deep learning model (AudioFocus) were assessed using a single QuickSIN list. The Acceptable Noise Level test was used to measure participants’ tolerance to background noise using the different hearing devices: Subjects reported the level of background increased to that were “put up with for a long time” while listening to the targeted acoustic signal. Lastly, participants answered a series of questionnaires regarding their listening effort with the different hearing aids, where likelihood scales between 1 and 7 were used, with 1 being the least effort. One-Way ANOVAs were carried out to examine the effect of testing conditions on QuickSIN and ANL, while a pair-wised t-test was used on Listening Effort between device conditions.

Results, Conclusions and Implications for Practice
Results from this study revealed a statistically significantly lower listening effort in subjects with AudioFocus compared to Big 5 AI (t(11) = 0.99, p = 0.0023). However, no statistically significant differences were found in the subjects’ sound field QuickSIN scores (F(2, 15) = 3.32, p = 0.06), nor in their ANL outcomes (F (2, 15) = 0.99, p = 0.39), between the two devices and the no device condition. These findings suggest that while the speech in noise benefits of the novel mic array device remain unclear, the advances in this hardware and software technology may help in reducing patients' effort in adverse listening environments, thus enhancing their hearing experiences and potentially increasing the likelihood of device adaptation in the future.

Brief Summary of Clinical Takeaways:
An overview of the effectiveness of a novel hearing device using an Acoustic-Enhanced Deep Learning Model.