Deviant functional connectivity patterns in the EEG and their potential use for dyslexia screening

Abstract

Developmental dyslexia (DD) is a common learning disorder with potential neural origins. While EEG-based brain activation measures combined with machine learning have shown promise for DD screening, these approaches often lack validation on independent participants, which is a crucial step for practical application. This study developed an EEG-based screening approach and investigated the neural correlates of DD in Chinese children. EEG signals were recorded from 130 children (82 with DD, 48 typically developing) aged betweeen 7 and 11 during resting-state and working memory tasks. The EEG data were preprocessed into clean segments to compute functional connectivity (FC) matrices across four frequency bands (delta, theta, alpha, beta). The segments were split into two independent samples, ensuring independence at the participant level: Sample 1, used for training and five-fold cross validation of the convolutional neural networks, and Sample 2, used for cross-sample validation with the trained model. The beta-band FC index in the eyes-open condition achieved the highest within-sample classification accuracy (98%) and cross-sample accuracy (70%, p < .001). Discriminative FC patterns revealed that children with DD exhibited reduced temporal-parietal and central connectivity but increased frontal-central connectivity, likely reflecting compensatory mechanisms. Within the DD group, stronger FCs showed significant negative correlations with Chinese word reading accuracy and fluency. These findings suggest that EEG-based FC measures can effectively distinguish DD and reveal neural markers associated with impaired reading performance. This approach shows promise for noninvasive screening and deeper insight into the neural basis of DD, particularly in non-alphabetic language systems.