Modeling speech intelligibility using inferior colliculus responses

Hearing loss diminishes speech intelligibility. While hearing aids restore some audibility by compensating for loss of outer hair cell (OHC) gain, many users struggle to understand speech in noisy environments. This persistent difficulty may arise because conventional hearing aids primarily address loss of OHC function, overlooking deficits at more central auditory sites. The Neurogram Similarity Index Measure (NSIM) has been proposed to relate auditory nerve responses to speech intelligibility [Hines and Harte, 2012; Mamun 2015; Wirtzfeld 2017; Cheema and Puria, 2025]. In this study, we simulate inferior colliculus (IC) responses using the Mitchell and Carney (2024) model and adapt the NSIM to relate neural representations of speech at the IC level for speech intelligibility. Our findings demonstrate that IC-based NSIM can accurately simulate psychometrically measured speech intelligibility curves, providing novel insights into objective measures of speech restoration that focus on central auditory processing. This enables development of IC-based loss functions. In the future, this work can be extended to devise new hearing aid processing strategies that minimize deficits resulting from cochlear neural degeneration that propagate into higher centers of auditory brain. [Work supported by the NIDCD of the NIH through R01-DC07910 Grant, as well as a MEE SharkTank award.]