Abstract
Autism spectrum disorder (ASD) is associated with white matter microstructural abnormalities, particularly in the corpus callosum (CC). This study employed free water corrected diffusion magnetic resonance imaging (fwc-dMRI) to investigate CC subregion-specific microstructural alterations in preschool children with ASD, which mitigates partial volume effects from extracellular free water. Sixty-one ASD children (6.03 ± 1.08 years) and 62 typically developing (TD) controls (6.49 ± 1.45 years) were enrolled in this study. In the ASD group, the symptom severity was assessed by the Autism Behavior Checklist (ABC). Fwc-dMRI technique, a bi-tensor tractography method, was used to investigate the white matter microstructure, which models free water and brain tissues through isotropic and anisotropic tensors to eliminate the partial volume effects caused by extracellular free water. The CC was segmented into seven subregions automatically according to its alignment to the cortex by a robust machine learning approach based on an anatomically curated white matter atlas. Fwc-dMRI-derived metrics were extracted for each CC subregion. Then we compared diffusion metrics between the two groups, and the correlation between the fractional anisotropy tissue (FA(t)) and the scores of the ABC scale was analyzed in ASD. Significant group differences were localized to CC6 (temporal lobe projections), showing reduced FA(t) (t = -3.251, p < 0.01) and elevated radial diffusivity tissue (t = 3.632, p < 0.01), and CC1 (orbital lobe projections), exhibiting decreased free water (t = -3.068, p < 0.05). FA(t) in CC2-5 negatively correlated with ABC scores (r = -0.36 to -0.52, p < 0.01), linking frontoparietal connectivity to the symptom severity of ASD. Fwc-dMRI identified distinct microstructural disruptions in CC subregions, implicating dysmyelination in temporal pathways (CC6) and abnormal axonal development in frontal projections (CC1). These findings highlight fwc-dMRI's potential for early ASD diagnosis and intervention monitoring.