Machine learning approach for classification of ADHD adults

Abstract

Machine learning techniques that combine multiple classifiers are introduced for classifying adult attention deficit hyperactivity disorder (ADHD) subtypes based on power spectra of EEG measurements. The analyzed sample includes 117 adults (67 ADHD, 50 controls). The measurements are taken for four different conditions: two resting conditions (eyes open and eyes closed) and two neuropsychological tasks (visual continuous performance test and emotional continuous performance test). We divide the sample into four data sets, one for each condition. Each data set is used for training of four different support vector machine classifiers, while the output of classifiers is combined using logical expression derived from the Karnaugh map. The results show that this approach improves the discrimination between ADHD and control groups, as well as between ADHD subtypes.