A Bimodal Autism Spectrum Disorder Detection Using fMRI Images

Autism Spectrum Disorder (ASD) presents significant challenges in diagnosis despite its high prevalence. Recent studies have increasingly turned to utilize neuroimaging data to enhance diagnostic accuracy and clinical applicability. Particularly, functional Magnetic Resonance Imaging (fMRI) data are explored as input for the proposed machine learning based solutions. However, from fMRI, several brain atlases are provided to describe functional connections between the detected regions of interest. In this paper, we propose a new bimodal method based on deep learning for ASD diagnosis. For instance, we use two atlases of fMRI data: CC200 and AAL. An experimental study was carried out on the famous ABIDE (Autism Brain Imaging Data Exchange) dataset and the reported accuracy of 75.26% proved the efficiency of the proposed method.