Increased Temporal Variability of Gait in ASD: A Motion Capture and Machine Learning Analysis

Area of Science:

• Neuroscience
• Biomechanical Engineering
• Developmental Psychology

Background:

• Motor deficits, particularly atypical gait, are frequently observed in individuals with autism spectrum disorder (ASD).
• The underlying mechanisms and precise characteristics of gait differences in ASD remain incompletely understood.
• Gait timing presents a measurable and accessible metric for exploring motor differences.

Purpose of the Study:

• To investigate if temporal gait parameters alone can characterize autistic gait.
• To determine if temporal gait analysis, enhanced by machine learning, can serve as a classifier between individuals with ASD and typically developing (TD) individuals.
• To explore the potential of gait timing analysis as a diagnostic tool for ASD.

Main Methods:

• High-resolution temporal analysis of gait was conducted on two groups of male participants: high-functioning ASD (N=16) and TD (N=16), aged 7–35 years.
• Data were collected using a VICON® 3D motion analysis system.
• Machine learning models, including random forest, were applied to temporal gait variability data for classification.

Main Results:

• The ASD group exhibited significantly increased temporal variability across all tested gait parameters compared to the TD group (p < 0.001).
• Machine learning analysis demonstrated that temporal gait variability effectively classified participants into ASD and TD groups.
• Random forest emerged as the best-performing model among twelve tested algorithms.

Conclusions:

• Temporal gait analysis reveals significant differences in gait timing variability between individuals with ASD and TD individuals.
• Machine learning algorithms can effectively utilize gait timing variability for group classification.
• This approach holds promise as a potential future diagnostic aid for ASD.