Central auditory processing of natural sound category following sensorineural hearing loss

Oregon Health & Science University, Portland OR

Investigators

Abstract

Project Summary Noise-induced hearing loss (NIHL) affects nearly one in four U.S. adults, making it the leading cause of acquired hearing loss, contributing to significant communication challenges and reduced quality of life. These challenges are particularly severe in complex listening conditions, such as when multiple sounds are present or in noisy backgrounds. These challenges often persist even with hearing aids. Currently, hearing-aid fitting is based on the audiogram, which measures hearing thresholds in quiet and does not fully address suprathreshold processing critical for real-world listening. Several neural mechanisms support complex-sound processing along the auditory pathway; how NIHL degrades these mechanisms is relatively well characterized in the auditory periphery, but not in the auditory cortex. Since the auditory cortex directly supports perception of complex sounds, it is critical to characterize how hearing loss affects cortical representations of complex sounds. A major impediment to understanding cortical processing in normal and impaired hearing has been the immense response diversity of auditory cortical neurons. We will address this limitation by using recent advances in large- scale electrophysiology and machine learning. In parallel, we will behaviorally assess categorization of sounds presented simultaneously, in noisy conditions – a task designed to capture real-world deficits due to NIHL. [K99 – normal hearing] Aim 1: We will perform large-scale single-unit recordings while presenting a massive natural sound corpus and use these data to identify the cortical manifold, which represents a low-dimensional representation of high-dimensional cortical activity. We will test whether manifold computations generalize across animals. Aim 2: Next, to reveal the neural basis of sound categorization, we will develop categorization models based on cortical single-unit data and validate these models using behavior. We will test whether neural models better match behavior than acoustic models not based on biology. [R00– NIHL] Aim 3: Using tools (cortical manifold and behavior) developed during K99, we will test whether frequency tuning broadens, inhibition reduces, and manifold degradation in noise correlates with behavioral deficits following NIHL. Aim 4: To directly test whether manifold-degradations underlie these perceptual deficits, we will design algorithms to restore the impaired cortical manifold to normal and compare their efficacy with traditional hearing-aid algorithms in improving neural representations and behavioral performance. These results will characterize, possibly for the first time, mechanisms by which NIHL impairs categorization of complex natural sound categories. Over the course of the project, the PI will be trained in cortical neurophysiology and modern machine-learning techniques, building on his existing expertise in peripheral neurophysiology, computation (modeling, signal processing), and translational neuroscience (noninvasive assays, hearing loss). This scientific and professional training will facilitate the PI's transition into an independent investigator position.