Cognitive Assessment of Autism Spectrum Disorder Using an EEG-based Social Interaction Platform

Previous studies have suggested a strong correlation between social interaction skills and the social environment, as well as personal experiences. Neuroimaging studies have further demonstrated the link between social interaction and brain oscillations. In this study, we developed a game-based social training platform that recorded electroencephalography (EEG) signals to provide neurophysiological indicators for cognitive assessment in autistic teenagers. Both typically developed (TD) participants and individuals with autism spectrum disorder (ASD) were recruited to participate in the social training game, with 12-channel EEG signals recorded during the sessions. The EEG signals underwent preprocessing, and the event-related potentials (ERPs) and event-related spectral perturbations (ERSPs) were analyzed and compared between the two groups. Six machine learning frameworks were employed for computer-aided assessment of ASD. The results revealed that the support-vector machine with a polynomial kernel function achieved the highest balanced accuracy of 91.67% when both ERPs and ERSPs were utilized as features for evaluating autism. These findings suggest the potential use of EEG features incorporated into a game-based social training interface for the cognitive assessment of ASD.