Predicting autism in children at an early stage using eye tracking

Abstract

Autism spectrum disorder (ASD) refers to a collection of conditions characterized by challenges in areas such as social interactions, communication, and repetitive behavior. Children with autism spectrum disorders often experience difficulties in processing and responding to social cues, which can lead to deficits in social skills and nonverbal communication. Children with ASD have been observed to have problems in maintaining eye contact. The main aim of this study is to use the eye tracking scan path images as a biological indicator to identify children with autism. The dataset used in this study has 547 visualized scanpath images collected from 59 children. The aim of this study is to utilize these scanpath images and formulate an autism diagnosis technique with the help of machine learning algorithms. The proposed model extracts the trainable features from the images and it is fed to a logistic regression classifier and a multi-layer perceptron classifier (MLP). A comparative study between the performance of the proposed model and a custom convolutional neural network is also presented.