Autism spectrum disorder (ASD) and Developmental Coordination Disorder (DCD) are distinct clinical groups with overlapping motor features. We attempted to (1) differentiate children with ASD from those with DCD, and from those typically developing (TD) (ages 8-17; 18 ASD, 16 DCD, 20 TD) using a 5-min coloring game on a smart tablet and (2) identify neural correlates of these differences. We utilized standardized behavioral motor assessments (e.g. fine motor, gross motor, and balance skills) and video recordings of a smart tablet task to capture any visible motor, behavioral, posture, or engagement differences. We employed machine learning analytics of motor kinematics during a 5-min coloring game on a smart tablet. Imaging data was captured using functional magnetic resonance imaging (fMRI) during action production tasks. While subject-rated motor assessments could not differentiate the two clinical groups, machine learning computational analysis provided good predictive discrimination: between TD and ASD (76% accuracy), TD and DCD (78% accuracy), and ASD and DCD (71% accuracy). Two kinematic markers which strongly drove categorization were significantly correlated with cerebellar activity. Findings demonstrate unique neuromotor patterns between ASD and DCD relate to cerebellar function and present a promising route for computational techniques in early identification. These are promising preliminary results that warrant replication with larger samples.
Keywords: Autism spectrum disorders; Developmental coordination disorder; Early detection; Machine learning; Smart tablet; fMRI.
© 2023. The Author(s).