Complex network modelling of EEG band coupling in dyslexia: an exploratory analysis of auditory processing and diagnosis

TL;DR

This study uses complex network analysis of EEG phase-amplitude coupling to identify differences in auditory processing between dyslexic and control subjects, revealing loss of small-world topology in dyslexia and enabling classification.

Contribution

It introduces a novel graph-theoretic approach to analyze EEG coupling data for dyslexia diagnosis, highlighting network topology differences and applying SVM for classification.

Findings

Dyslexics show loss of small-world network topology.

Graph metrics can distinguish dyslexic from control subjects.

EEG coupling patterns correlate with auditory processing differences.

Abstract

Complex network analysis has an increasing relevance in the study of neurological disorders, enhancing the knowledge of brain's structural and functional organization. Network structure and efficiency reveal different brain states along with different ways of processing the information. This work is structured around the exploratory analysis of the brain processes involved in low-level auditory processing. A complex network analysis was performed on the basis of brain coupling obtained from Electroencephalography (EEG) data, while different auditory stimuli were presented to the subjects. This coupling is inferred from the Phase-Amplitude coupling (PAC) from different EEG electrodes to explore differences between controls and dyslexic subjects. Coupling data allows the construction of a graph, and then, graph theory is used to study the characteristics of the complex networks throughout time for controls and dyslexics. This results in a set of metrics including clustering coefficient, path length and small-worldness. From this, different characteristics linked to the temporal evolution of networks and coupling are pointed out for dyslexics. Our study revealed patterns related to Dyslexia as losing the small-world topology. Finally, these graph-based features are used to classify between controls and dyslexic subjects by means of a Support Vector Machine (SVM).