Automated Delimitation and Classification of Autistic Disorder Using EEG Signal
Asit Kumar Subudhi, Monalisa Mohanty, Santanu Kumar Sahoo, Saumendra Kumar Mohanty, Bibhuprasad Mohanty
Abstract
Autism Spectrum Disorder (ASD) is a disturbing neuro-developmental disorder affecting the behavioural ability and communication skills of an individual. Early detection of such prevalent abnormality through EEG pattern analysis with the help of computer-aided diagnostic tools; provide significant relief to the health care professionals for better management and treatment of such patients. This work has been proposed as an investigation of the sensory responsiveness of the children affected by ASD. This ASD is analyzed by extracting the non-linear features of the EEG signal. A total of 73 EEG signals are collected from different patients out of which 41 are affected with ASD and 32 with normal neural activity. These signals are then pre-processed and filtered using a low pass filter and then separated using Independent Component Analysis (ICA) into additive subcomponents. Significant features are extracted, which can be further investigated to identify event-related potential and Non-biological Artifacts. These features are then classified using Support Vector Machine (SVM) with an accuracy of 90.41%. As a result, appropriate sensory profiles can be obtained that may help proper diagnosis and treatment at an early stage.