Autism Spectrum Disorder Classification Based on ML and DL Techniques

Abstract

This study presents a novel approach to classifying Autism Spectrum Disorder (ASD) using eye-tracking data and advanced machine-learning techniques. The study employed a comprehensive dataset of eye movements from 28 children, 14 with ASD and 14 typically developing (TD) controls, collected while viewing 300 natural scene images. This methodology encompassed traditional machine learning algorithms (Logistic Regression, Random Forest, and Gradient Boosting) and deep learning models (Convolutional Neural Network and a hybrid CNN-LSTM architecture). The hybrid CNN-LSTM model achieved the highest accuracy of 99.89%, outperforming other approaches and comparable studies in literature. Notably, even our simpler Logistic Regression model attained 99.37% accuracy, demonstrating the robust discriminative power of eye-tracking data for ASD classification. Our results suggest that integrating spatial and temporal patterns in eye movements significantly enhances ASD detection accuracy. This study contributes to the growing body of research on automated, data-driven approaches to ASD diagnosis, potentially facilitating earlier and more accurate identification of ASD in clinical settings.