Autism Spectrum Disorder Detection Using Multidomain Features of RS-FMRI Bold Time Series and Machine Learning

Autism spectrum disorder (ASD) is a complex neurodevelopmental condition characterized by atypical brain connectivity. Traditional diagnosis relies mainly on behavioral assessments, which are often subjective and time-consuming. In this study, we used resting-state functional magnetic resonance imaging (rs-fMRI) of ASD and typically developing (TD) individuals to build a diagnostic classification model. After preprocessing, the brain was divided into regions of interest (ROIs), and time series were extracted from each ROI. Statistical features from time and frequency domains, along with Pearson and coherence correlations, were computed. Our results showed 1,580 and 18 significant features from Pearson and coherence respectively. Significant features identified through statistical testing were used to train support vector machine (SVM) and random forest (RF) classifiers. The SVM achieved a higher 5-fold accuracy (98.2%) than RF (92.0%), indicating its suitability for high-dimensional rsfMRI features and supporting the potential of rs-fMRI-based ASD detection.