Autistic spectrum traits detection and early screening: A machine learning based eye movement study.

Area of Science:

• Neurodevelopmental Disorders
• Machine Learning Applications
• Ophthalmology and Vision Science

Background:

• Autism spectrum disorders (ASD) present significant challenges in social interaction, communication, and concentration, impacting daily life.
• Traditional ASD screening methods are often time-consuming, straining limited public health resources.

Purpose of the Study:

• To develop and validate a novel, objective, and rapid technique for early detection of autistic traits.
• To leverage eye-movement data combined with machine learning for improved ASD screening.

Main Methods:

• Raw eye-movement data was transformed into relevant features for analysis.
• Eight machine learning classification models were trained and tested, with a focus on an ensemble model.
• Participants viewed a virtual reality (VR) scene to capture eye-movement patterns.

Main Results:

• The ensemble model achieved an accuracy of 0.73, precision of 0.68, recall of 0.81, and an F1-score of 0.74 in preliminary experiments (n=107).
• The model demonstrated a recall of 0.77 in a test experiment with participants diagnosed with ASD (n=22).
• Continuous observation of figures and animals in the VR scene effectively distinguished individuals with varying degrees of autistic traits.

Conclusions:

• Eye-movement data serves as an effective tool for distinguishing autistic traits.
• This approach offers a faster and more objective method for early ASD screening.
• Early detection through this technique can facilitate timely intervention and symptom management.