Title: SC-TauPath: A Structural Connectivity Attribution Framework for Mapping Tau Propagation Pathways in Alzheimer's Disease

Abstract:

A pivotal unresolved issue in Alzheimer’s disease (AD) research is determining how structural connections influence the spread of tau pathology. Current computational approaches often struggle with this, as they either depend too heavily on biophysical assumptions or fail to provide neurobiologically interpretable maps of propagation pathways. To address this, we introduce SC-TauPath, a novel attribution framework designed to map tau propagation routes using in vivo neuroimaging data.

The SC-TauPath method integrates a Network Diffusion Model (NDM)-enhanced multilayer perceptron with gradient $\times$ input attribution techniques. This combination allows for the scoring of each structural connectivity (SC) edge’s specific contribution to tau prediction. These attribution scores are subsequently converted into multi-scale pathway maps, identifying backbone edges, high-traffic routes, and hub regions of interest (ROIs). Notably, these maps align with established Braak staging anatomy.

We applied SC-TauPath to a cohort of 234 participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), utilizing paired diffusion tensor imaging (DTI) for SC and 18F-Flortaucipir PET for tau measurements. The framework demonstrated robust cross-validated performance in predicting tau levels. Furthermore, the resulting attribution-based pathway maps were consistent with known Braak staging patterns. These findings confirm that structural connectivity encodes spatially specific information regarding the regional distribution of tau in Alzheimer’s disease.

Source: arXiv Generated at: 2026-06-04 00:00:00 UTC