Authors: Sarah Verhulst1, Sarineh Keshishzadeh1, Viacheslav Vasilkov1, Markus Garrett2
1Hearing Technology @ WAVES, Dept of Information Technology, Ghent University
2Medizinische Physik and Cluster of Excellence “Hearing4all”, Oldenburg University
Background: With the discovery of cochlear synaptopathy, a form of sensorineural hearing loss (SNHL) which targets the synapses between the inner-hair-cells and the auditory nerve, the search for non-invasive diagnostic markers for human diagnosis has begun. At the same time, hearing restoration strategies should consider compensating for the perceptual consequences of synaptopathy. A factor which hitherto has limited the development of both diagnostics and treatment, is that patients likely have unknown mixtures of synaptopathy and OHC damage. These aspects have different functional (and nonlinear) properties and hence affect sound coding as well as the generators of auditory-evoked-potentials (AEP) differently. To disentangle SNHL sources and provide individual treatment of frequency-specific SNHL, biophysical models of the auditory periphery can be used. These models can simulate the how different SNHL aspects, in isolation and combination, affect AEP-based hearing screening markers and sound coding in the impaired ear.
Methods: In this work, we show how a transmission-line cochlear model with a variety of auditory-nerve fibers and brainstem neurons can be used to develop AEP-based stimuli which are sensitive to synaptopathy, even in listeners with co-existing outer-hair-cell damage. At the same time, we will present how a number of experimental AEP-based markers can be used to derive parameters for individualized models of the auditory periphery.
Results: Our combined experimental and modeling study on AEPs shows that the stimuli and analysis methods for EFRs can be optimized to emphasize individual differences in the degree of (age-related) synaptopathy. At the same time, we used a combined model and experimental approach to investigate which markers derived from a set of individual ABR/EFR recordings (e.g. ABR wave amplitudes, latencies, growth slopes, EFR strength), were best able to predict the individual SNHL profile. Here, we conclude that a combination of the optimized EFR with the ABR wave-I latency-level slope was best able to predict an individual SNHL profile.
Conclusion: The SNHL parameters of computational auditory models can be individualized on the basis of AEP markers of hearing and can afterwards be used as a basis for numerical or machine-learning methods in the design of signal processing which compensates for individual synaptopathy/OHC damage profiles.
Work supported by the European Research Council (ERC-StG-678120, RobSpear)