Detecting ADHD children based on EEG signals using Graph Signal Processing techniques

Attention Deficit Hyperactivity Disorder (ADHD) is a neurological disorder that is the most common childhood disorder. A significant lack of attention and concentration of the child is one of the apparent symptoms of this disorder. Accurate and early diagnosis of this disorder in preschool ages can help the control process and prevent the school problems caused by ADHD. Medical methods for preschool-age children can be problematic and slow down the control process. In these cases, Electroencephalogram (EEG) signals are useful and efficient tools, because of the non-invasiveness, being quite available, and having high temporal resolution. In this paper, we proposed a method to detect ADHD/Normal EEG signals recorded from children in an online and open access dataset. Our proposed method uses the Structural and Functional information of the EEG signals. Structural and functional based features were extracted using Graph Signal Processing (GSP) and Graph Learning (GL) techniques, respectively, which are the generalization of the Classic methods and can consider EEG signals as graph signals on the underlying graph of electrodes. We reached detection accuracies of 79.03% and 82.36% for using GSP and GL based features, respectively. But when we used the fusion of these feature sets, we got a high detection accuracy of 93.47% which shows these feature sets are complementary and consider thorough aspects of EEG signals.