A Machine Learning-Based Analysis of Game Data for Attention Deficit Hyperactivity Disorder Assessment

Area of Science:

• Neuroscience
• Computer Science
• Psychiatry

Background:

• Attention deficit hyperactivity disorder (ADHD) affects 9.5% of the U.S. population, presenting lifelong challenges.
• Current ADHD diagnosis relies on subjective evaluations by parents and teachers, often missing nuanced cognitive deficits.
• Accurate diagnosis by psychiatrists is frequently inaccessible, particularly for children with less severe symptoms.

Purpose of the Study:

• To develop an engaging and objective tool to aid medical providers in diagnosing ADHD.
• To leverage interactive video game technology for cognitive assessment.

Main Methods:

• Developed
• Groundskeeper,
• an interactive video game on Sifteo Cubes, designed to exercise skills impacted by ADHD.
• Collected gameplay data from 52 patients (with and without ADHD) and transformed it into ADHD-indicative features.
• Applied machine learning algorithms to develop diagnostic models, validated against expert psychiatric assessments.

Main Results:

• Predictive algorithms demonstrated high accuracy in diagnosing ADHD subtypes and related disorders.
• F-measure scores for diagnostic accuracy: ADHD, inattentive type (78%), ADHD, combined type (75%), anxiety disorders (71%), and depressive disorders (76%).

Conclusions:

• The
• Groundskeeper
• game represents a novel and promising approach for ADHD screening.
• This technology offers a potential solution for more accessible and objective ADHD assessment.