Machine Learning Model to Predict Autism Spectrum Disorder Using Eye Gaze Tracking

Autism Spectrum Disorder (ASD) is caused by a group of complex neurological disorders which inhibit the development of a person’s social, behavioral and verbal skills. It is of paramount importance for early diagnosis of this disorder as studies have shown that starting the ASD treatment in the early stages of a patient’s life can provide fruitful results; treatment plans are better able to work when the child is still at his developing age. However, the current treatment is long and requires professionals, so many children are not diagnosed with ASD in their early childhood. In this paper, we propose a robust framework for detecting Autism Spectrum Disorder (ASD) in patients using their tracked eye-gaze as features. Our approach consists of using supervised learning models to predict children with ASD. The result shows the ability to classify patients with reasonable accuracy. This approach can assist doctors because it only requires five minutes of video with a camera rather than several hours of behavioral and cognitive tests conducted by a licensed professional.