Automatic diagnosis of ADHD based on nonlinear analysis of actimetry registries

Attention-Deficit Hyperactivity Disorder (ADHD) is the most common mental health problem in childhood and adolescence. Its diagnosis is commonly performed in a subjective manner since current objective measurements are either expensive or time-consuming. However, subjective methods tend to overestimate the severity of the pathology. In this paper, we propose a novel methodology for automatic diagnosis of ADHD based on signal processing methods. The method is constructed in two stages: 1) An automatic activity/rest detection filter which allows for a separate analysis of both types of periods and 2) A feature extraction module based on nonlinear regularity quantification of either the global signal or the detected epochs. Results on real data show that the proposed methodology can discriminate between patients and controls with sensibility and specificity values approaching 80%.