Detecting autism in children through drawing characteristics using the visual-motor integration test

Area of Science:

• Neurodevelopmental Disorders
• Developmental Psychology
• Machine Learning Applications

Background:

• Autism Spectrum Disorder (ASD) diagnosis relies on behavioral observation.
• Early detection of ASD is crucial for effective intervention.
• Visual-motor integration skills are often affected in individuals with ASD.

Purpose of the Study:

• To introduce a novel classification method for distinguishing children with autism from typically developing children.
• To identify specific drawing patterns indicative of ASD.
• To explore the potential of a cross-cultural tool for early autism detection.

Main Methods:

• Recruited 50 school-age children (6-12 years) in Taiwan.
• Collected drawing data from a visual-motor integration test.
• Trained deep learning classification models using ensemble learning.

Main Results:

• Achieved a classification accuracy of 0.934.
• Identified five key drawing patterns differentiating children with and without ASD.
• Children with ASD showed difficulties with patterns involving circles and spatial relationships.

Conclusions:

• The novel drawing-based classification method shows high accuracy in identifying ASD.
• Specific visual-motor deficits, particularly with circles and spatial relationships, are linked to ASD in drawings.
• This method offers a potential cross-cultural tool for early autism detection and intervention.