Authors: Ann-Kristin Seifer2, Nadja Schinkel-Bielefeld1, Hendrik Schröter2, Alberto N. Escalante-B.1
1WS Audiology 2Friedrich-Alexander Universität Erlangen-Nürnberg
Background: Hearing aids (HA) are configured to the wearer’s individual needs, which might vary greatly from user to user. Currently, it is common practice, that the initial HA gain settings are based on generic fitting formulas that link a user’s pure-tone hearing threshold to amplification characteristics. Subsequently, a time-consuming fine-tuning process follows, in which a hearing care professional (HCP) adjusts the HA settings to the user’s individual demands. An advanced, more personalized gain prescription procedure could support HCPs by reducing fine-tuning effort and facilitate over-the-counter HAs. We propose a machine learning based prediction for HA gain to minimize subsequent fine-tuning effort. The data-driven approach takes audiometic and personal variables into account, such as age, gender, and the user’s acoustical environment.
Method: A random forest regression model was trained on real-world HA fittings from the Connexx database (fitting software provided by Sivantos GmbH). Three months of data from Connexx version 188.8.131.524 were used. A data cleaning framework was implemented to extract a representative data set based on a list of machine learning and audiological criteria. These criteria include, for instance, using only ‘informative’ HCPs who perform fine-tuning for at least some patients. Furthermore, ‘informative’ HCPs are those who perform diagnostics beyond air conduction audiograms, use new technologies and special features. The resulting training data comprised 20,000 HA fittings and a 10-fold cross validation was used to train the random forest.
Results: The mean absolute error (MAE) of the random forest prediction is 3.88dB which is a 2.61dB (40%) improvement over the generic fitting formulas. The use of subject information in addition to audiogram (hearing loss) data improves the performance of the random forest, as the error decreases compared to a random forest trained solely on the user’s audiogram (4.77dB). The MAE of an additionally trained neural network is with 3.81dB marginally lower.
Conclusion: Data-driven fitting formula can reduce fine-tuning efforts compared to generic fitting formulas, the current state-of-the-art method. A key factor for the improved accuracy of the proposed machine learning method is the usage of additional personal variables.