Abstract
Abstract
Diffusion tractography is the prominent in-vivo technique to study and investigate white-matter pathways in the human brain. While tractography is a powerful method, recent work suggests that different tractography methods can produce strikingly different representations of the same white-matter pathway. This multitude of differing options and diverging pipelines makes tractography related group-effects difficult to generalize, as it is currently unclear whether group-level inferences made using one tractography pipeline can be expected to hold when a different, equally defensible pipeline is applied to the same data. Here, we test the generalizability of sex-related changes on tractography-derived features by analyzing the exact same datasets with two equally reasonable pipelines which differ in model fitting, tractography reconstruction, and microstructure and volumetric analysis. We found that despite differences in analysis, the resulting patterns and biological interpretations of sex effects rarely disagreed across methods. Microstructural sex-related effects between methods were remarkably consistent between protocols, only displaying one statistically significant disagreement out of 343 comparisons (0.29%). However, discrepancies were more common among normalized pathway-volume effects, producing 8 out of the 9 total statistically significant disagreements between methods. Moreover, we found that the two tractography methods are differentially sensitive to tractography-derived features, as bundles derived from targeted tractography were much more sensitive to volumetric effects than tractogram-based tractography, potentially explaining the volumetric discrepancy between methods. This study indicates that reasonable methodological choices are unlikely to lead two investigators to fundamentally opposing conclusions about sex differences in white-matter, and that the robustness of tractography findings is similar to established fields of science. More broadly, this study presents an optimistic outlook on the future of tractography, and provides an initial empirical comparison of cross-pipeline robustness.