A machine learning model diagnoses ADHD via the pupil

Abstract

Attention Deficit/Hyperactive Disorder (ADHD) is diagnosed in children based on reports from parents, teachers, and close contacts. Although considered a neurodevelopmental disorder, there are no neurological or biological tests that are commonly used for diagnosis. However, the causative neurological circuits affect physiological markers in a measurable way. Differences in pupil behaviour in ADHD were identified in 2018 in a dataset in which the pupil response to a working memory task was recorded from 22 control, and 28 ADHD-diagnosed children. With our trained neural network, by assessing a child over a number of trials it was possible predict whether a child has ADHD with moderate accura..