BundleMorphometry: Spatially Resolved White Matter Bundle Shape Analysis Using Nonlinear Streamline-based Registration

Diffusion MRI tractometry quantifies microstructural properties along white matter (WM) bundles but does not directly measure their geometry, potentially overlooking localized morphological alterations. Here, we introduce BundleMorphometry, a framework for quantifying local WM bundle shape using the analytical Jacobian of nonlinear streamline-based registration. Individual subject bundles are registered to a reference atlas bundle using BundleWarp, a streamline-based nonlinear registration method. We derive the analytical Jacobian of the resulting deformation field and compute its log-determinant (log det J) to quantify local expansion and compression at each streamline point. To enable spatially resolved group comparisons, these measurements are projected onto the two-dimensional grid, capturing variations along and across each bundle. We evaluated BundleMorphometry using two clinical datasets comprising patients with glioma and Parkinson's disease (PD). In the glioma dataset, BundleMorphometry revealed spatially extensive deformation patterns in lesion-affected WM bundles that were not captured by displacement magnitude alone. Additionally, in a group analysis of the PD dataset (32 patients and 32 healthy controls), we identified significant localized morphological differences in the corpus callosum forceps major, mid-body and right spinothalamic tract, with distinct patterns of expansion and compression. These findings demonstrate the potential of BundleMorphometry to advance conventional tractometry by enabling spatially resolved analysis of bundle morphology.