Social cognition and functional brain network in autism spectrum disorder: Insights from EEG graph-theoretic measures

Abstract

Prior social cognition studies are not proficient to revealing an adequate/accurate processing of whole-brain and thus, deteriorated ASD classification. The present paper purpose is two-fold-(i) to explore the topological configurations of whole-brain functional network (local and global) using novel graph-theory;(ii) to find neural markers that can predict cognition in context of perception task and improve ASD classification. In this direction, we derived weighted Visibility Graph (VG) networks from brain EEGsignals (recorded under resting/experimental task-state) of ASD(28;14.8 ± 3.4) and Typically Developing TDs(28;13.6 ± 2.4). The neural correlates are quantified using complex graph-based measures which revealed higher intra-connectivity (in frontal, temporal, and occipital over right hemisphere) and lower inter-connectivity of the regions during task, thus suggesting re-organization of whole-brain network under cognition in ASD. The poor network integration and high segregation reveal attenuated efforts in processing data in controlled way and further representing dense cognitive network with short functional reach in ASD. To classify ASD with an optimal set of neural correlates, different state-of-art classification models are constructed. The support Vector Machine (SVM) model demonstrated that a combined effect of two metrics-Average Weighted Degree (AWD) and Mutual Information (MI) has detection accuracy 92.34 %. Additionally, the variation of behavioral specificities and experiment-based neural metrics computed using correlation is found revealing ASD core traits (such as social cognition & communication), thus ensuring clinical value of proposed metrics in ASD. In sum, the present paper provides an improved understanding of whole brain reorganization with social cognition that may detect ASD and its pathological underpinnings in the prodromal stage.