Early Detection of Autism Spectrum Disorder Using Non-Invasive EEG

The Autistic Spectrum Disorder (ASD) is a neurodevelopmental disorder that has been increasingly diagnosed in children. Symptoms are commonly noticed in childhood and include impairments in communication and social interaction. Anticipating the diagnosis to before the onset of symptoms would allow different therapies to be started without compromising the child's development. Hence, several studies have searched for ASD biomarkers using non-invasive electroencephalography (EEG), a low-cost technique, using different strategies. In this scenario, this work compares different classifiers to automatically identify the ASD from the EEG records and analyses which features best describes the data set to assist in early diagnosis. The Support Vector Machine (SVM) using recursive feature elimination with cross validation (RFECV) and considering epoch of 60 seconds have shown high accuracy (at least 97.5%) to identify the ASD before the first year of life.