Real-time Artificial Intelligence-based Speech Enhancement Methods for Hearing Aid Improvement

Institut Pasteur40 enrolled

Overview

Individuals with Auditory Neuropathy Spectrum Disorders (ANSD) represent 1-10% of adults with hearing loss. These individuals have little or no benefit from current hearing aids because ANSD is a continuum of hearing impairments due to synaptic or neural dysfunction in the peripheral and central parts of the auditory pathways, which impairs temporal information processing without necessarily affecting auditory sensitivity. There is a need to find ad-hoc denoising methods, based on the expert knowledge of audiologists, to improve the noise comprehension performance of these patients. Implemented denoising methods, based on artificial intelligence, will also greatly benefit more standard hearing loss cases.

Longitudinal study consisting of multiple visits with multiple examinations over a total duration of approximately 3 years per ANSD participant and 1 year per normal hearing participant. * normal hearing participant : year 1 : Inclusion visit, Phenotyping visit (Audiological tests), Follow-up Visit 1 (Noise Comprehension Tests) * ANSD participant : year 1 : Inclusion visit, Phenotyping visit (Audiological tests), Follow-up Visit 1 (Noise Comprehension Tests) year 2 : Follow-up Visit 2 (Noise Comprehension Tests) year 3 : Follow-up Visit 3 (Noise Comprehension Tests)

Study Type

INTERVENTIONAL

Allocation

Purpose

BASIC_SCIENCE

Masking

NONE

Enrollment

Conditions

Deafness Neurosensory

Eligibility

Sex: ALLMin age: 18 YearsHealthy volunteers:

Medical Language ↔ Plain English

Inclusion Criteria: For all participants: * Age greater than or equal to 18 years, * French mother tongue, * Have an average tonal hearing loss (calculated according to the BIAP method) \< 30 dB HL, on headphones, for each ear, * Be professionally active, not be pre-retired and not be unemployed * Be affiliated with a social security plan, For participants with TSNA: \- Have degraded comprehension in noise (threshold of Intelligibility in Noise \> 3 dB compared to the norm). For controls: \- Have normal comprehension in noise (Intelligibility in Noise threshold ≤ 3 dB from the norm). Exclusion Criteria: For all participants: * Have a conductive or mixed hearing loss, which is when the hearing loss affects both the outer and/or middle ear and the inner ear. * Have an asymmetric hearing loss, i.e. a difference in average hearing loss between the left and right ear greater than 30 dB, * Have sequelae of ear infections and/or a history of ENT disease that permanently affects hearing (vestibular schwannoma, Meniere's disease, sudden or fluctuating deafness, congenital hypoacusis) * Being under guardianship, * Being deprived of liberty by judicial or administrative decision, or being subject to legal protection.

Outcomes

Primary Outcomes

Comparison of percentage of words correctly repeated by participants (ANSD and controls)

Evaluate intelligibility, i.e. the percentage of words correctly repeated, for different speech and noise conditions. Tests will be repeated with and without application of speech enhancement methods to the sound.

Time frame: 3 years

Secondary Outcomes

Auditory Brain Response (ABRs)

Electrical activity of auditoral pathways, from the auditory nerve to the brain, in response to sound stimuli.

Time frame: 3 years

Percentage of words correctly repeated by participants with different angles of separation of speech and noises sources

Evaluate intelligibility and performance of source localisation for different angles of separation of speech and noises sources

Time frame: 3 years

Percentage of phonemes correctly repeated by participants (Verbo-Frequential Audiometry in the presence of calibrated noise: the DODELE AVfB test)

Evaluate speech perception in noise with phonemes as testing speech

Time frame: 3 years

Percentage of words correctly repeated by participants at different sound level of noise

Assess intelligibility in noise. Vocal tests will be repeated with and without tactile cues.

Time frame: 3 years

Myogenic vestibular evoked potentials (cervical VEMP or c-VEMP).

c-VEMP assesses the vestibular function through saccular and inferior vestibular nerve function performances. The collected parameters of c-VEMP testing include c-VEMP threshold, the latencies of the initial positivity (p1) and negativity (n1), and the p1-n1 inter-amplitude. We will collect these parameters and assess differences in controls vs. ANSDs.

Time frame: 3 years

Subjective evaluation of listening comfort using a visual analog scale

This self-assessment involves placing a cursor on a scale ranging from "very uncomfortable to very comfortable" using a scale ranging from 0 to 10.

Time frame: 3 years

Subjective evaluation of listening efficiency using a visual analog scale

This self-assessment involves placing a cursor on a scale ranging from "very inefficient to very efficient" using a scale ranging from 0 to 10.

Time frame: 3 years

Real-time Artificial Intelligence-based Speech Enhancement Methods for Hearing Aid Improvement

Overview

Conditions

Interventions

Eligibility

Locations (1)

Outcomes

Primary Outcomes

Secondary Outcomes

Central Contacts