Assessing white matter pathway reproducibility from human whole-brain tractography clustering

Abstract

Diffusion MRI, together with tractography techniques, is a non-invasive tool to investigate the brains structural pathways (tracts). These tracts join together different regions of the brain and tract identification often involves the use of manual ROIs or automated techniques such as clustering. By studying these connections, current understanding of the connectome can be improved and changes due to disease in patient populations may be identified. We developed a tool to automatically identify all pathways in the human brain, including the short-range, U-shaped tracts, and map quantitative scalar metrics along the pathway trajectory for subsequent analysis. Pathways are identified via a spectral clustering technique on two datasets: Human Connectome Project (intersubject) and MyConnectome Project (intrasubject) and the reliability of the extracted tract and scalar values are evaluated. Average Euclidean distances and volumetric overlap were computed and indicated good spatial reliability. Intraclass correlations of the fractional anisotropy value mapped along the tract was calculated and exhibited good reproducibility within each dataset. Additionally, these evaluation metrics, together with the coefficient of variation of the mean streamline count is used to determine reliably identified U-shaped tracts across the datasets. The developed tract identification tool is an additional resource to studying the human connectome with increased confidence in the results. The identified reliable U-shaped tracts contributes to the identification of common structural connections across individuals and aids in advancing our understanding of the brains short-range pathways.