Hypergraph representation of multilayer brain network enhances autism spectrum disorder detection

We present a hypergraph-based framework for analyzing functional brain networks in children with autism spectrum disorder (ASD) using resting-state electroencephalography data. Moving beyond conventional multilayer network approaches, our method captures previously undetectable higher-order connectivity patterns through a two-stage analysis: (1) constructing multilayer networks via recurrence quantification analysis to model within- and cross-frequency interactions and (2) transforming these networks into hypergraphs to better represent complex neural relationships. Our results identify distinctive connectivity signatures in ASD, particularly in bilateral frontal regions, with hypergraph representations revealing patterns obscured in traditional analyses. Most significantly, hypergraph-derived features achieved 81% classification accuracy (F1-score) using support vector machines, outperforming 57% achieved with multilayer network features. These findings demonstrate how hypergraphs can provide more stable and informative biomarkers for ASD, offering both a powerful analytical framework for studying neurodevelopmental disorders and a promising pathway toward more objective diagnostic tools. The improvement in classification performance underscores the clinical potential of this approach.