A Hearing-Model-Based Active-Learning Test for the Determination of Dead Regions

Area of Science:

• Audiology
• Hearing Science
• Biomedical Engineering

Background:

• Cochlear dead regions significantly impact hearing, characterized by impaired inner hair cell function.
• Estimating the edge frequency (f_e) is crucial for understanding dead region extent and guiding hearing aid fitting.
• Current methods for f_e estimation can be time-consuming or less precise.

Purpose of the Study:

• To develop and validate a Bayesian active-learning procedure for accurate and efficient estimation of cochlear edge frequency (f_e).
• To compare the performance of the new method against established techniques like "Fast PTCs" and extensive measurements.
• To assess the clinical utility of the method for hearing aid fitting.

Main Methods:

• A Bayesian active-learning algorithm was employed to estimate hearing model parameters, including f_e and outer hair cell loss.
• The procedure adaptively selects masker frequency and level to maximize information gain about model parameters.
• The method utilizes simple yes-no judgments, allowing for estimation of psychometric function slope and response reliability.

Main Results:

• The Bayesian active-learning procedure yielded accurate f_e estimates comparable to "Fast PTCs" and more extensive measurements.
• Reliable f_e estimates were achieved in an average of 33 trials, requiring only 5 to 8 minutes.
• The method demonstrated comparable efficiency to "Fast PTCs" while offering improved response reliability assessment.

Conclusions:

• The developed Bayesian active-learning method provides a fast, accurate, and reliable approach for estimating cochlear edge frequency (f_e).
• Its efficiency and ability to assess response reliability make it suitable for clinical audiology, particularly for hearing aid fitting.
• This technique offers a valuable advancement in characterizing cochlear dead regions and personalizing hearing rehabilitation.