Location, location, location- choice of Voxel-Based Morphometry processing pipeline drives variability in the location of neuroanatomical brain markers

Abstract

Fundamental and clinical neuroscience has benefited from the development of automated computational analyses of Magnetic Resonance Imaging (MRI) data, such as Voxel-based Morphometry (VBM). VBM determines regional gray matter variations with high spatial resolution and results are commonly interpreted in a regional-specific manner, for instance with respect to which specific brain regions differ in volume between women and men. In excess of 600 papers using VBM are now published every year and a number of different automated VBM processing pipelines are frequently used in analyses although it remains to be fully and systematically assessed whether they come up with the same answers. Here we have therefore examined variability between four commonly used VBM pipelines in two large brain structural datasets. Spatial similarity, reproducibility and reliability of the processed gray matter brain maps was generally low between pipelines. Examination of sex-differences and age-related changes in gray matter volumes revealed considerable differences between the pipelines in terms of the specific regions identified as well as meta-analytic characterization of their function. In contrast, applying machine learning-based multivariate analyses allowed an accurate prediction of sex or age based on the gray matter maps across pipelines, although prediction accuracy differed strongly between them. Together the findings suggest that the choice of pipeline alone leads to considerable variability in brain structural analyses which poses a serious challenge for reproducibility as well as interpretation.