New-generation Psychometric Tools for Characterizing Hearing efficiency and its deficits – NewPiTCH

The societal burden of hearing loss will worsen as life expectancy increases: virtually every one of us will experience it at some point
in their life. Current design of hearing aids is supported by audiometric tests that do not capture the full complexity of hearing loss,
leading to unsatisfactory performance of commercial aids for many potential users. We develop a new generation of auditory tools to
advance experimental and computational characterization of hearing deficits. These tools enable detailed dissection of the mechanisms underlying auditory representation, thus delivering a rich experimental blueprint that supports identification of specific model components impacted by hearing impairment. The tools are relevant to everyday tasks (e.g. speech comprehension) and are readily transferrable to the clinic, carrying the
potential to guide fine-tuning of hearing-aid design for translational results.

We identify three primary limitations with state-of-the-art knowledge in this area. First, hearing loss is a complex phenomenon that cannot be adequately understood only in terms of reduced audibility (the deficit primarily targeted by hearing aids): we must understand not only how humans detect auditory modulations, but also how they combine them in complex ways for interpreting articulated sounds like speech. Second, current approaches to the characterization of hearing efficiency often conflate two separate sources of impairment: poor tuning of the sensory mechanism versus excessive internal variability. Third, hearing loss presents large inter-individual differences, so that one-size-fits-all solutions are inevitably suboptimal. At present, we do not have access to personalized computational insight into hearing deficits, i.e. we cannot ascribe hearing deficits in a specific individual to a personalized description of how auditory computations are disrupted in that individual.

To address the above limitations, we set out 4 primary objectives: 1) develop a new generation of experimental tools (notably psychophysical reverse
correlation and internal noise estimation) that will deliver a fine-grained characterization of biological auditory processing, which we can exploit to impose highly constraining demands on putative computational models; 2) develop a new generation of computational models constrained by the above characterization, so that we are able to say, for a given hearing-impaired individual, where his/her deficits act at the level of a well-defined functional description of the auditory process; 3) make these experimental/computational tools relevant to everyday auditory tasks, by designing our protocols to span the entire processing hierarchy from mechanisms for detecting elementary auditory features to their role and organization for speech processing; 4) ensure that all tools can be feasibly transferred from the laboratory to the clinic by piloting new tools in both environments in a parallel fashion, so that we can optimize our protocols with immediate feedback to tailor them to the clinic.

Achieving the above-stated goals involves several challenges with associated risks. As insurance against those risks, we have put in place a series of mechanisms. First, we have structured our plan of action as an incremental progression from more controlled paradigms/protocols, for which we have clear evidence of successful delivery, to more articulated designs that bear direct relevance to natural auditory behaviour (e.g. speech comprehension). Second, we have developed a robust operational pipeline for interfacing with multiple clinical collaborators, so as to secure access to the hearing-impaired population. Third, we have assembled a team of experts with complementary skills at the highest level. Combined with the promising pilot data supporting this proposal, these elements make a strong case for the feasibility of the proposed research programme.