Robot-Assisted ADHD Screening in Diagnostic Process

Abstract

This paper presents a novel approach to screening children with ADHD, one of the most prevalent childhood disorders. We have designed the robot-assisted, game-like test that directly reflects children’s behavior on measuring ADHD symptoms. Using the sensors in the robot system, most of the children’s behavior is automatically measured during the entire course of the test. We collected real data by carrying out tests to 326 children from 3rd to 4th grades in the field. A unified frame that classifies multiple categories of ADHD, ADHD-at-Risk and normal has been set up to investigate a wide spectrum of classifiers and their optimal hyper-parameters. The results of the data analysis show highly confident ADHD classification, up to 97% (F1 score). It could be a practical tool for clinicians and special teachers to use in the diagnosis of childhood ADHD.