EEG-based multi-feature fusion assessment for autism

Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental disorder which affects the developmental trajectory in several behavioral domains, including impairments of social communication, cognitive and language abilities. In this paper, multi-feature fusion method based on EEG signal is used to extract as many as possible features including power spectrum analysis, bicoherence, entropy and coherence methods, then we use minimum redundancy maximum correlation (mRMR) algorithm to choose the features, which are applied to input to three classifiers to obtain accuracy classification results. We try to find some key biomarkers of ASD by examining the accuracy of classifier, using different models which use the combination of multiplex features. The results show when nine features are selected by SVM-linear classifier, the accuracy is up to 91.38%. This method might provide objective basis for clinical diagnosis of autism.