Comparison Of Diagnostic Techniques Of ASD In Children Using EEG By Applying Feature Extraction Techniques Like AR Covariance And AR Yule Walker Methods Using PRNN And NARXNN

Abstract

Investigators from Karpagam Academy of Higher Education studied about the feasibility and accuracy EEG(Electroencephalogram) signals on effectively diagnosing Autism Spectrum Disorders (ASD) in children. 4 children with already diagnosed ASD and six controls aged 6 to 12 years were enrolled in this study. EEG was done on these children and the desired features of EEG were extracted using AR(Auto-regressive) Covariance technique and AR Yule Walker Method. The acquired signals were classified using two Artificial neural networks named PRNN (Pattern recognition neural network) and NARXNN(Nonlinear Autoregressive with External Input Neural Network). Mean Recognition performance of PRNN using AR Modified covariance and AR Yule walker both accounted to 91% accuracy in diagnosing ASD. Whereas accuracy of NARXNN using AR Modified covariance and AR Yule walker techniques both scored 95% diagnosis accuracy. The results of this study provide a promising future of early diagnosis of ASD when the latter technique is used for the same purpose..