Autism Spectrum Disorder Detection

Abstract

Autism spectrum disorder (ASD) is a complex neuro developmental condition affecting social interaction and communication skills. Current diagnostic methods often rely on structural and resting-state functional magnetic resonance imaging (fMRI) with limited datasets, leading to high accuracy but limited generalizability. To address this, machine learning, pattern recognition, and other techniques have been used, achieving high accuracy but moderate generalization. This study introduces a novel approach to ASD detection using deep learning (DL), specifically a Convolutional Neural Network (CNN) classifier. By leveraging anatomical and functional connectivity indicators from fMRI data, our model aims to enhance the automated diagnosis of ASD. The proposed approach demonstrates significant improvement over existing methods, achieving an accuracy of approximately 85% in classifying autistic patients. Through the utilization of a ResNet model, this work showcases the potential of DL in advancing the accuracy and reliability of ASD diagnosis