Systematic bias in surface area asymmetry measurements from automatic cortical parcellations.

Anatomical asymmetry is a hallmark of the human brain and may reflect hemispheric differences in its functional organization. Widely used software like FreeSurfer can automate neuroanatomical measurements and facilitate studies of hemispheric asymmetry. However, patterns of surface area lateralization measured using FreeSurfer are curiously consistent across diverse samples. Here, we demonstrate systematic biases in these measurements obtained from the default processing pipeline. We compared surface area asymmetry measured from reconstructions of original brains vs. the same scans after flipping their left-right orientation. The default pipeline returned implausible asymmetry patterns between the original and flipped brains: Many structures were always left- or right-lateralized. Notably, these biases occur prominently in key speech and language regions. In contrast, manual labeling and curvature-based parcellations of key structures both yielded the expected reversals of left/right lateralization in flipped brains. We determined that these biases result from discrepancies in how regional labels are defined between the cortical parcellation atlases' left and right hemispheres. These biases are carried into individual parcellations because the parcellation algorithm prioritizes vertex correspondence to the template over individual neuroanatomical variation, meaning such biases could exist in any asymmetric atlas-based parcellation. We further demonstrate several straightforward, bias-free approaches to measuring surface area asymmetry, including using symmetric registration templates and parcellation atlases, vertex-wise analyses, and within-subject curvature-based parcellations. These results highlight theoretical concerns about using only atlas-based parcellations to make inferences about population-level brain asymmetry and underscore the need for validating bias-free neuroanatomical measurements, particularly to better examine how structural lateralization underlies functional lateralization.