Anatomical MRI Abnormalities in Autism?
Objectives: To identify anatomical abnormalities evident in MRI scans of individuals with ASD aged 6-35 years in comparison with IQ and gender matched individuals, and to assess their potential clinical significance.
Methods: We built an automated analysis pipeline that co-registered all MRI scans to a single anatomical template and parcellated each subject's brain into over 180 anatomical ROIs using a probablistic atlas. This was performed using Freesurfer image analysis suite and custom written software. Importantly, scans of all subjects were handled in an identical, automated manner so as to avoid potential biases inherent in manually performed analyses.
Results: Comprehensive univariate analyses using anatomical measures from over180 brain areas, revealed only a few significant differences between ASD and control brains, and these were mostly limited to ASD individuals with severe symptoms. Significant between-group differences were on the order of 2-4%, while between-subject differences within each group reached 80-100%. Small differences across groups were, therefore, greatly overshadowed by considerable within-group variability. In agreement with the univariate analyses, several multivariate classification analyses yielded marginal decoding accuracies of ASD and control individuals.
Conclusions: These sobering findings challenge previous reports of anatomical abnormalities in ASD individuals older than 6 years of age, and suggest that anatomical measures derived from MRI scans of these individuals are likely to be of low clinical and scientific significance for ASD research.