Large-scale electrophysiology and deep learning reveal distorted neural signal dynamics after hearing loss

Abstract
Data availability
Article and author information
Metrics

Abstract

Listeners with hearing loss often struggle to understand speech in noise, even with a hearing aid. To better understand the auditory processing deficits that underlie this problem, we made large-scale brain recordings from gerbils, a common animal model for human hearing, while presenting a large database of speech and noise sounds. We first used manifold learning to identify the neural subspace in which speech is encoded and found that it is low-dimensional and that the dynamics within it are profoundly distorted by hearing loss. We then trained a deep neural network (DNN) to replicate the neural coding of speech with and without hearing loss and analyzed the underlying network dynamics. We found that hearing loss primarily impacts spectral processing, creating nonlinear distortions in cross-frequency interactions that result in a hypersensitivity to background noise that persists even after amplification with a hearing aid. Our results identify a new focus for efforts to design improved hearing aids and demonstrate the power of DNNs as a tool for the study of central brain structures.

Data availability

The metadata, ABR recordings, and a subset of the IC recordings analyzed in this study are available on figshare (DOI:10.6084/m9.figshare.845654). We have made only a subset of the IC recordings available because they are also being used for commercial purposes. These purposes (to develop improved assistive listening technologies) are distinct from the purpose for which the recordings are used in this manuscript (to better understand the fundamentals of hearing loss). Researchers seeking access to the full set of neural recordings for research purposes should contact the corresponding author via e-mail to set up a material transfer agreement. The custom code used for training the deep neural network models for this study is available at github.com/nicklesica/dnn.

The following data sets were generated

1. Nicholas Lesica
(2023) Data from Sabesan et al., 2023
10.6084/m9.figshare.845654.

https://figshare.com/articles/dataset/Data_from_Sabesan_et_al_2023/22651816

Article and author information

Author details

Shievanie Sabesan

Ear Institute, University College London, London, United Kingdom

Competing interests
No competing interests declared.
Andreas Fragner

Perceptual Technologies Ltd, London, United Kingdom

Competing interests
No competing interests declared.
Ciaran Bench

Ear Institute, University College London, London, United Kingdom

Competing interests
No competing interests declared.
Fotios Drakopoulos

Ear Institute, University College London, London, United Kingdom

Competing interests
No competing interests declared.
Nicholas A Lesica

Ear Institute, University College London, London, United Kingdom

For correspondence
n.lesica@ucl.ac.uk

Competing interests
Nicholas A Lesica, is a co-founder of Perceptual Technologies.

"This ORCID iD identifies the author of this article:" 0000-0001-5238-4462

Funding

Wellcome Trust (200942/Z/16/Z)

Shievanie Sabesan
Nicholas A Lesica

Engineering and Physical Sciences Research Council (EP/W004275/1)

Ciaran Bench
Fotios Drakopoulos
Nicholas A Lesica

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Ethics

Animal experimentation: All experimental protocols were approved by the UK Home Office (PPL P56840C21). Every effort was made to minimize suffering.

Copyright

This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.