Development and Feasibility of a Yoga-Based Motor Assessment Tool (Y-MAT) with AI-Enabled Scoring for Children with ADHD
V. P, B. Holla, H. Bhargav, E. Sharma, R. Naidu LIntroduction
Motor dysfunction in ADHD is increasingly recognized as a core manifestation of neurodevelopmental disruption rather than an incidental problem. Yet there is no consensus gold-standard motor assessment, and most tools emphasize broad milestones rather than neuromuscular dysregulation.
Objectives
Aim: Develop a Yoga-Based Motor Assessment Tool (Y-MAT) for children with ADHD.
Primary objectives: Develop and validate the tool; pilot test feasibility.
Secondary objective: Explore artificial intelligence and video-analytics for digitizing motor assessment.
Methods
Seven motor domains were identified through review of literature and mapped to Yoga-tasks through expert consensus (n = 6). The Y-MAT was administered to 11 children with ADHD and 10 controls by a trained psychiatrist, video-recorded, and scored by 2 blinded human raters. AI-based scoring was also implemented using Google’s MediaPipe and YOLOv8 for pose recognition and movement analysis. All children were also assessed using the ADHD Rating Scale (ADHD-RS) and the Vineland Social Maturity Scale (VSMS) to evaluate symptom severity and social functioning. Feasibility, Inter-rater reliability (Cohen’s kappa) and Concurrent Validity was measured.
Results
The Y-MAT demonstrated high acceptability, engagement and a minimal dropout rate.The ADHD scores show a moderate negative correlation with both ratings, with slightly stronger associations for AI-based scoring, indicating that greater ADHD severity was associated with poorer neuromuscular regulation. Inter-rater reliability ranged from fair to moderate agreement, with Impulse Control of Vrikshasana (p = 0.0001) and Motor Sequencing of Surya Namaskar (p = 0.0032) exhibiting the strongest and statistically significant agreement. AI-based scores moderately correlated with human ratings.
Selected Yoga-tasks mapped to their corresponding motor domains Table 1. Long description.