Identification of autism spectrum disorder based on functional near-infrared spectroscopy using dynamic multi-attribute spatio-temporal graph neural network

Zhengqi Fan, Ziheng Gao, Lingyu Xu, Jie Yu, Jun Li

Abstract: Highlights•A new GNN model for ASD identification achieves a classification accuracy of 96.8%.•The dynamic brain functional connectivity network is modeled.•Extract correlation features from multiple fNIRS attributes separately and fuse them.•The inferior frontal gyrus (IFG) is strongly correlated with ASD children.•ASD children have weaker resting-state long-range FC in bilateral IFG and TL.