Computer Aided System for Autism Spectrum Disorder Using Deep Learning Methods

Abstract

The aim of this study is to apply machine learning algorithms to identify autism spectrum disorder (ASD) patients from a brain imaging dataset based only on brain activation patterns. ASD is a brain-based disorder normally characterized by repetitive and social behaviors, but in the present study, imaging data from a world-wide multisite database known as ABIDE (Autism Brain Imaging Data Exchange) is used for classification. A deep learning method that combines supervised and unsupervised machine learning methods has been employed to do the process. Input is based on the respective neural patterns of functional connectivity using resting state functional magnetic resonance imaging (rs-FMRI) present in the preprocessed ABIDE dataset, from which an associativity matrix is calculated between different regions of the brain, which shows an anti-correlation of brain function between anterior and posterior areas of the brain. Extracted features are then subjected to the pr e-training stage along with phenotypic information. Finally, the pre-trained weights are given as input to a convolutional neural network and classified as ASD or control type.