Are sex differences in brain structure altered in autism? A machine learning study using multi-site structural MRI

Background Males are diagnosed with autism spectrum disorder (ASD) more frequently than females, but the neuroanatomical basis of this sex disparity remains unclear.Objective This study examined whether sex-related patterns of brain morphology differ between individuals with ASD and typically developing (TD) controls.Methods Multi-site structural MRI data were analyzed using a two-stage machine-learning framework. First, regional classifiers distinguished ASD from TD controls separately in males and females to compare diagnosis-related morphometric separability. Second, classifiers distinguished males from females within ASD and TD groups to compare sex-related morphometric separability. Voxel-wise analyses were used to localize gray matter differences.Results ASD-versus-TD classification was more accurate in males than in females, suggesting more consistent diagnosis-related structural differences in male participants. In contrast, male-versus-female classification was less accurate in ASD than in TD, consistent with altered structural sex differentiation in autism. Voxel-wise effects were strongest in the posterior cingulate cortex, precuneus, putamen, pallidum, and thalamus.Conclusion Sex-related morphometric patterns in ASD differ from those observed in typical development.