REHEAR: Reverse Engineered Hearing Evaluation for Auditory-neuropathy Recognition

Description du projet

Un diagnostic précis d’un trouble auditif mal compris

Le spectre des neuropathies auditives touche jusqu’à 10 % des personnes atteintes d’une perte auditive. Contrairement à la perte auditive classique, les personnes touchées peuvent percevoir les sons mais ont du mal à comprendre la parole, car le nerf auditif transmet les signaux avec un décalage temporel. Les outils de diagnostic actuels ne permettent pas d’identifier les mécanismes neuronaux spécifiques en jeu, ce qui complique le choix du traitement. Soutenu par le programme Actions Marie Skłodowska-Curie, le projet REHEAR cherche à mettre au point le premier cadre de diagnostic computationnel pour la neuropathie auditive, en recourant à l’optimisation mathématique et à l’IA. L’objectif est de convertir les schémas de dysfonctionnement neuronal en tests de parole ciblés destinés à détecter les problèmes neuronaux sous-jacents. Cette approche sera validée auprès de personnes entendantes et de porteurs d’implants cochléaires, ouvrant ainsi la voie à un diagnostic de précision et à un traitement amélioré.

Objectif

"Auditory Neuropathy Spectrum Disorder (ANSD) affects up to 10% of individuals with hearing loss. Unlike typical cases, people with ANSD detect sounds but struggle with speech understanding due to disrupted auditory nerve timing. Current tests cannot identify specific neural mechanisms, making treatment unpredictable. ANSD patients often cycle through multiple failed treatments - hearing aids provide minimal benefit, and cochlear implant outcomes vary dramatically from excellent to poor. This creates prolonged periods of communication difficulty during critical developmental years for children and significant quality-of-life impacts for adults. Cochlear implants provide unique access to neural responses via electrical recordings (ECAPs), enabling mechanism-specific diagnosis. This project develops the first computational-to-clinical bridge for ANSD using reverse engineering: instead of measuring neural dysfunction directly, we embed ANSD-specific disruption patterns into speech materials, creating diagnostic tests that make neural problems ""hearable."" Two complementary frameworks—mathematical optimization and AI-based learning—convert neural dysfunction patterns into targeted speech tests while preserving disorder-specific diagnostic signatures. The methodology includes three work packages: developing computational frameworks, validating through behavioral experiments with normal-hearing listeners (n=40) and cochlear implant users (n=12-15 each) alongside neural validation, and optimizing clinical protocols. Based at University Hospital Zurich's Deep Hearing Lab under Prof. Goehring's supervision, this fellowship advances expertise in computational neuroscience and clinical translation, delivering the first validated framework for ANSD subtype diagnosis and precision hearing care for thousands of patients annually across Europe, with potential healthcare savings through reduced diagnostic workup times and improved treatment matching.

"