ADHD diagnosis in children using Common spatial pattern and Nonlinear Analysis of Filter banked EEG

Attention Deficit Hyperactivity Disorder (ADHD) is currently the most common neuropsychiatric disorder of childhood, which is marked by abnormalities in hyperactivity, inattention, and impulsivity stages. Early diagnosis of ADHD may effectively reduce the symptoms and side effects of this disorder. Recording and analysis of Electroencephalogram (EEG) is one of the proper functional methods to detect such disorders, which merits independency from subjects variability. This study considers classification of ADHD and healthy subjects by using EEG signals and temporal and frequency features such as common spatial pattern (CSP), selective nonlinear attributes, and through usage of filter banks and time windowing approaches. In order to evaluate our proposed method, we applied it on a benchmark dataset, namely NBML, which consists of 328 epochs of recorded EEG signals from 61 subjects in two classes of ADHD and healthy. The proposed method achieved the discrimination accuracy of 83.33% of the two classes by using the proposed KNN classifier, which proved superior to previously reported results on the same dataset.