Morphological bias of the MNI152 brain

Abstract

The MNI152 template is widely treated as a representative average brain in neuroimaging, computational modeling, and neuromodulation research, yet its fidelity to true population morphology has not been systematically evaluated. In this study, we compared the MNI152 template to 430 individual MRI scans from a publicly available dataset spanning Asian, Black, and White participants. We additionally generated an alternative template using ANTsPy to assess whether a modern diffeomorphic approach yields a more anatomically representative average. We conducted affine registrations and deformation-based morphometry to detect and quantify gross morphological differences as well as local voxel-level deformations. Across all racial groups, the MNI152 template required consistent global contraction, and its Jacobian fields revealed spatially heterogeneous deformations, indicating systematic mismatches in both size and shape. In contrast, the ANTsPy template showed mean scaling and deformation values near 1.0, reflecting closer correspondence to real human anatomy. These findings demonstrate that the MNI152 template does not accurately represent the morphology of the studied population and that linear registration alone cannot correct its inherent biases. Reliance on MNI152 may therefore introduce unintended distortions in applications requiring anatomically realistic head models. More robust, unbiased template-generation pipelines, and potentially demographic-specific templates, may be necessary for improved accuracy.