Functional connectivity magnetic resonance imaging classification of autism spectrum disorder using the multisite ABIDE dataset

Abstract

The goal of this paper is to apply machine learning algorithms to classify autism spectrum disorder (ASD) patients and typically developing (TD) participants using resting-state functional MRI (rs-fMRI) data from a large multisite data repository ABIDE(Autism Brain Imaging Data Exchange) and identify the important brain connectivity features. In this study, we implemented a data-driven approach to classify ASD patients and TD participants by using the rs-fcMRI features extracted from rs-fMRI data. We applied several classical machine learning classifiers such as support vector machines, logistic regression, and ridge. Our contribution has mainly three parts: (1) We used a cross-validation grid search method to find the optimal parameters for each classifier. By using the optimal parameters, the best accuracy we achieved is 71.98%, which is slightly higher than the present best accuracy 70% using deep learning and the same data from the multisite repository ABIDE. We also obtained satisfactory recall and precision results. (2) We implemented the same experiments for seven different brain atlas data in ABIDE, and we identified the most promising brain atlas is Craddock 400 (CC400). (3) We identified the top five highly correlated and anti-correlated region of interests (ROIs) from the brain atlas CC400 for the ASD group and TD group.