Automated Identification and Classification of Autism Spectrum Disorder using Behavioural and Visual Patterns in Children

Autism Spectrum Disorder is a challenging developmental disorder that can lead to serious social, communication, and behavioral difficulties. The primary goal is to resolve the problems these kids are having and to help them realize that this illness does not define or restrict who you can be or place you in a superior or inferior position to others. We propose a system to detect ASD in children using two approaches viz., behavioral traits and facial images. In the behavioral trait approach, eight individual traits that have been found to be useful in diagnosing ASD patients based on controls in behavioral science are added to the ten behavioral features (AQ-10-Child) that have been approved by the National Institute of Health Research (NHS) in the behavioral trait approach. In the facial recognition approach, the images of the children are used and pass them into the pretrained CNN EfficientNet model which detects the presence of autism. Finally, we integrate these machine learning models into a fully functional website which will act as a link between the users and the admin team.