Wavelet Scattering Transform and Deep Learning Networks based Autism Spectrum Disorder Identification using EEG Signals

Autism Spectrum Disorder (ASD), a neurological abnormality that influences how an individual perceives and interacts with others, which leads to issues social interaction communication. In accordance the Centers for Disease Control Prevention, 1 in every 44 children USA is affected by ASD. The identification of ASD based on behavioural characteristics it generally takes long time from initial observation signs final diagnosis, due complexity diversity symptoms. application Electroencephalography (EEG) signals, recorded 14 healthy controls, as potential biomarker categorisation, was analysed this study. After pre-processing, second-order Wavelet Scattering Transform (WST) coefficients were extracted EEG signals Deep Learning (DL) detection networks (WST-ASDNets) used categorisation control subjects. Long Short Term Memory Network (LSTM) WST-ASDNet Convolution Neural (CNN) achieved accuracy 94% 92% respectively, subject identification. results demonstrate proposed WST-ASDNets can efficiently classify usage WST be categorisation.