Comprehensive Summary
The study presented by Deng et. al utilizes electroencephalography (EEG) signals from children viewing cartoons alongside machine learning to differentiate children with autism spectrum disorder (ASD) and typical developing children. ASD diagnosis is typically difficult and is usually diagnosed after critical developmental windows. The prospective study used a sample of 70 children (23 with ASD and 47 typically developing children) and collected EEG data while children watched cartoon stimuli designed to elicit semantic processing. The researchers found that with an accuracy of 76.9% for the validation model, the machine learning model could distinguish children with ASD from developing peers. The discriminative EEG patterns were linked to semantic processing differences as children with ASD showed different brain responses with the cartoon stimuli, which is consistent with altered language integration and meaning-making. The researchers emphasize how semantic-level neural processing can serve as a biomarker for ASD.
Outcomes and Implications
The researchers suggest the idea of objective diagnostic aids that complement behavior assessment for ASD, becoming a noninvasive biomarker. While a proof-of-concept idea, the researchers further that a larger cohort of children could contribute to a multimodal diagnostic toolkit for diagnosis. Eventually, these methods could also help with tracking treatment response or development trajectories.Lin Deng, Meng-Jie Lu, Le-tong Yang, Yue Zhang, Hang-yu Tan, Miao Cao, Fei Li