Head Movement Analysis in Differentiating Children with ASD from TD under Various Sensory Stimuli Conditions: A Machine Learning Approach

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by behavioral symptoms in sensory processing, social interaction, and motor skills. The present study aimed to conceptually validate whether participants’ head movements could differentiate children with ASD from their typically developing peers under various sensory stimuli conditions. Four sensory stimulation conditions were included: visual, auditory, tactile, and olfactory. Data were collected from 40 individuals who were exposed to four different sensory stimuli, resulting in the construction of datasets comprising a total of 160 head movement data entries. Machine learning algorithms were utilized for classification purposes. Results revealed that under olfactory stimulation conditions, the decision tree model achieved the highest accuracy (90.0%) employing the rotation range in Z axis (ZRR) and the angular rotation per minute in Z axis (ZARPM) features. For visual conditions, both the decision tree with 2 features and the neural network with 3 features exhibited optimal performance (87.5%). Overall, classifiers demonstrated robust performance across sensory conditions, indicating potential for early ASD diagnosis.