Applying machine learning to identify autistic adults using imitation: An exploratory study

Area of Science:

• Neuroscience
• Biomedical Engineering
• Developmental Psychology

Background:

• Autism spectrum condition (ASC) diagnosis relies on behavioral symptoms, with motor aspects often overlooked.
• Quantitative kinematic analysis of autistic individuals' movement patterns is under-researched, hindering understanding of motor impairments and diagnostic potential.
• Current diagnostic methods lack objective, quantitative biomarkers for motor characteristics in ASC.

Purpose of the Study:

• To investigate the use of machine learning (ML) for identifying discriminative kinematic parameters and test conditions for classifying autism spectrum condition (ASC) and neurotypical controls.
• To explore the potential of data-driven methods in analyzing movement patterns for ASC identification.
• To assess if kinematic data can offer novel insights into motor differences in ASC.

Main Methods:

• Utilized data from 16 ASC participants and 14 controls imitating hand movements.
• Analyzed 40 kinematic parameters across eight imitation conditions using ML-based methods.
• Applied machine learning to identify significant kinematic parameters and optimal test conditions for classification.

Main Results:

• Identified two optimal imitation conditions and nine significant kinematic parameters that differentiate between ASC and controls.
• Demonstrated the feasibility of applying ML to high-dimensional kinematic data for classification.
• Showcased the potential of ML in analyzing complex movement data for identifying biomarkers.

Conclusions:

• Machine learning methods show promise for analyzing kinematic data in the context of autism spectrum condition (ASC).
• This study suggests the potential for developing kinematic biomarkers to aid in the diagnostic classification of ASC.
• Further research with larger sample sizes is warranted to validate these findings and explore clinical applications.