Multi-atlas Hypergraph Fusion Based on Brain Regions Overlap Amount for Diagnosis of ASD

Recently, the accurate classification of autism spectrum disorder (ASD) has attracted a lot of attention in human brain analysis. However, the existing diagnosis method only focus on single atlas or relationship among brain regions, neglecting the high-order and spatial relationship among the brain regions in different atlases. To tackle this weakness, in this paper, we propose a multi-atlas hypergraph fusion method based on brain regions (BRs) overlap amount to diagnosis ASD/TC. We calculate the correlations of BRs for each subject and formally introduce the measure of multi-atlas overlap (MAO), brain regions overlap amount (BRsOA). Then, we fuse multiple hypergraphs calculated from multi-atlas by MAO and generate a multi-atlas hypergraph (MAHG). In the final, the MAHG is used to perform the classification task. The experimental results on Autism Brain Imaging Data Exchange (ABIDE) verify that our proposed method outperforms classical methods in ASD/TC classification.