Title

What brain connectivity patterns from EEG tell us about hearing loss: A graph theoretic approach

Abstract

We investigated brain connectivity patterns from the electroencephalogram (EEG) to classify and understand hearing loss using a graph theoretic approach. In particular, we investigated global and nodal graph features of normal hearing (NH) and hearing impaired (HI) participants' brain networks via functional connectivity while they responded to clear and noise-degraded speech stimuli. We found that HI listeners had higher eccentricity, diameter and characteristic path length than the NH listeners for clear speech sounds, and larger radii for noisy speech signal. Moreover, we classified groups based on these graph theoretic features using support vector machine (SVM). Maximum classifier accuracy was 85.71 % for clear speech (Fl-score=86.00%) and 71.42% (Fl=67%) for degraded speech signals, respectively. Group classification based on nodal connectivity measures was more accurate in the left brain hemisphere, consistent with the leftward laterality of auditory-linguistic processing. Our data suggest HI listeners have more extended communication pathways and less efficient information exchange among brain regions than NH, establishing new biomarkers of hearing loss based on full-brain connectivity.

Publication Title

ICECE 2018 - 10th International Conference on Electrical and Computer Engineering

Share

COinS