Machine Learning Based Framework for Classification of Children with ADHD and Healthy Controls

Abstract

Electrophysiological (EEG) signals provide good temporal resolution and can be effectively used to assess and diagnose children with Attention Deficit Hyperactivity Disorder (ADHD). This study aims to develop a machine learning model to classify children with ADHD and Healthy Controls. In this study, EEG signals captured under cognitive tasks were obtained from an open-access database of 60 children with ADHD and 60 Healthy Controls children of similar age. The regional contributions towards attaining higher accuracy are identified and further tested using three classifiers: AdaBoost, Random Forest and Support Vector Machine. The EEG data from 19 channels is taken as input features in individual and combinatorial sets to classifiers. Evaluating all the classifiers' overall performance, the highest accuracy of 84% is obtained with the AdaBoost classifier when all the Right Hemisphere channels are taken into consideration. The higher sensitivity of 96% indicates a better true positive detection rate of the model created with the Right Hemisphere features. This study highlights the intrinsic physiological contrast prevalent in brain activity of ADHD and healthy children, which can be effectively utilized for diagnostic purposes.