Eeg Based Diagnosis of Autism Spectrum Disorder Using Static and Dynamic Neural Networks

Electroencephalography (EEG) signals can be used to monitor the brain activities of all human beings. As a result, it can be used to detect abnormalities in the brain functioning. In this study, using Artificial Neural Network (ANN) EEG signals of children with Autism Spectrum Disorder (ASD) and non-ASD children were classified. Two neural network models namely Pattern Recognition Neural Network (Pattern Net) and Layered Recurrent Neural Network (LRN) were used. Auto regressive (AR) Burg and LRN combination were found to have the highest classification accuracy rate of 94.62%. Moreover, Bit Transfer Rate (BTR) of the signals were calculated for each network in order to evaluate the Human Machine Interface system performance. Maximum BTR of 6.08 bit/sec was achieved for AR Burg and LRN combination. The proposed method has obtained promising results. For the study, real-time dataset which was obtained from ASD children of various special schools in Coimbatore has been used.