Auxiliary diagnostic method for children with autism spectrum disorder based on virtual reality and eye-tracking technology

In the behavioral analysis of children with Autism Spectrum Disorder (ASD), virtual reality (VR)-based eye-tracking technology offers a precise method for assessing social and cognitive characteristics. It overcomes the limitations of traditional diagnostic methods, such as clinician subjectivity and experience bias. VR also addresses ASD-related challenges like attention instability and emotional variability during social interactions. This paper combines eye-tracking with VR environments to analyze gaze patterns in children with ASD. It proposes a new diagnostic framework to improve objectivity and accuracy.The gaze estimation model integrates head and eye movement data to predict gaze direction. It enhances precision using binocular fusion and employs multi-scale convolutional kernels to extract hierarchical eye movement features. The model simplifies network connections to retain essential information. A lightweight Transformer architecture models long-range temporal dependencies in eye movements. A Bayesian decision model is used to classify fixations, saccades, and smooth pursuit.To test the model, an emotion recognition task was designed in a WebVR environment. Gaze data from children with ASD were collected, key features were extracted, and abnormal patterns were identified for diagnostic support. The experimental results showed an 85.88% accuracy rate. This confirms the effectiveness of combining VR and eye-tracking technology in ASD diagnosis, advancing intelligent medical tools, and reducing reliance on subjective clinical judgment.