Topography High Sensitivity of Discriminative Fibers in ASD Children with Parcellation-Free Machine Learning Method

Motivation: Abnormalities in white matter fibers are associated with the onset of Autism Spectrum Disorder(ASD) though early diagnosis of ASD is difficult. Goal(s): This study aims to establish a Parcellation-Free model for ASD with Diffusion Tensor Imaging that can accurately classify and identify abnormal white matter fibers. Approach: The Parcellation-Free model preserve fiber-level white matter information of the ASD children without brain region parcellation and averaging, and input it into the Vision Transformer classifier, then identify the white matter fibers contribute to the classification. Results: The classification model achieved an sensitivity of 90.4%, and output fibers that contribute to classification, confirming previous research. Impact: We aim to establish a reliable classification model for early diagnosis for ASD, which can output white matter fibers that contribute to classification at the fiber level,which may be meaningful for exploring the pathogenesis and potential intervention targets of ASD.