Hearing & AI
BACKGROUND
Hearing plays a vital role in cognitive, linguistic, and social development across the human lifespan. When hearing impairment is left untreated, the resulting consequences can be lifelong and far-reaching — including delayed language acquisition in early childhood, reduced academic and employment opportunities, social withdrawal, and significantly increased risk of dementia in later life. More than 1.5 billion people worldwide are affected by hearing loss, including 34 million children, and these numbers continue to rise.
For those with severe to profound deficits, cochlear implants (CIs) are a powerful and transformative intervention. However, their effectiveness relies heavily on the accuracy of the fitting process that determines the maximum comfortable loudness level (MCL) for each individual user.
Concepts
This project addresses this gap through a clinically validated and explainable AI framework that fully automates cochlear implant customisation by detecting the electrically evoked stapedius reflex (eSR) — a robust neurobiological biomarker strongly correlated with loudness perception.
The eSR provides an objective indicator of the MCL without requiring behavioural participation, making it suitable for patients of any age or cognitive ability.
Challenges
Current CI fitting protocols are manual, time-consuming, and reliant on behavioural responses, making them unsuitable for patients who cannot provide reliable feedback, especially infants.
As CI surgeries are now commonly performed within the first year of life— and implantation between 6 and 12 months offers the best speech and language outcomes — the need for objective, scalable, age-independent fitting methods has become critical.
Publications
Desoyer C, Proksch J, Berger P, Baumgartner C. Machine and deep learning-based detection of the stapedius reflex for automatic cochlear implant fitting. SIPAIM'24, Antigua, Guatemala, 2024,1-4.