Diagnostic Classification of Autism using Resting-State fMRI Data and Conditional Random Forest.

Abstract

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder that is associated with atypical connectivity within and between brain regions. In this study, we attempted to classify functional Magnetic Resonance Images (fMRI) of Typically Developing (TD) and ASD participants using conditional random forest and random forest. Restingstate fMRI images of TD and ASD participants (N=320 for training and N=80 for validation) were obtained from the Autism Imaging Data Exchange; ABIDE-I, ABIDE-II. Images were preprocessed using a standard pipeline. A Functional Connectivity (FC) matrix was calculated using 237 cortical, subcortical, and cerebellar Regions of Interest (ROIs). The dimensionality of the FC matrix was reduced using conditional random forests and at each dimension classification accuracy was tested using random forests. Results suggest that in the current dataset, the random forest is able to classify the TD and ASD with a peak accuracy of 65% using 143 features. Remarkably, the Cingulo-Opercular Task Control (COTC) region contributed the highest number of features linked to more accurate classification, and connectivity between COTC and the dorsal attention network distinguished ASD and TD participants.