Within-network functional connectivity reduced for auditory & attention networks in the presence of hearing loss

Authors: Ivan Abraham *1; Amber Leaver 2; Brad Sutton 1; Yuliy Baryshnikov 1; Jonathan Peelle 3; Fatima Husain 1

1 University of Illinois, Urbana-Champaign
2 Northwestern University
3 Northeastern University

The collection of audiometry data is routine for the diagnosis and treatment of hearing disorders. However, audiometry data accompanied by detailed structural and functional magnetic resonance images (fMRI) of the brain is an infrequent occurrence. The Hearing Health Institute (HHI) at the University of Illinois has put together one of the largest such datasets using data harmonized from multiple sites (University of Illinois, Washington University at St. Louis, Wilford Hall Ambulatory Surgical Center & Northwestern University). The multi-site raw dataset was processed & analyzed the same way using standardized fMRI preprocessing and analysis routines (fMRIPrep & xcpEngine). A subset of the data having complete audiometry data was chosen (where N=168 with 75 controls and 108 participants identifying as male). Resting-state functional connectivity analysis based on a parcellation of four hundred regions-of-interest (ROIs) organized into seventeen networks (Schaefer400x17) showed that within-network functional connectivity is reduced with age and hearing loss for ROIs within the auditory (AU), dorsal attention (DA) & ventral attention (VA) networks of the brain. This reduction was also found to be well-correlated with hearing loss as evidenced by higher thresholds for pure tone averages (PTA) at 4 kHz (p-values: 1.64E-4 (AU), 1.3E-9 (VA) and 1.1E-5 (DA)) and to a lesser degree at 8 kHz (p-values: 7E-4 (AU), 2.2E-9 (VA) and 2.7E-5 DA)). Structural MRI analysis (freesurfer segmentation output) also revealed that cortical gray matter volume normalized by the total intracranial volume is inversely correlated with age & connectivity. The results point to the auditory and extra-auditory impact of hearing loss and age. They also underscore the feasibility of combining datasets across different studies and sites post hoc and the novel findings emerging from such combined large datasets.