Vessel segmentation toolbox for susceptibility imaging: Region-growing algorithm and deep learning

Motivation: χ-separation generates paramagnetic and diamagnetic susceptibility maps, potentially reflecting iron and myelin distribution. However, large vessels introduce erroneous values, hindering accurate quantification of these components. Goal(s): Our goal is to develop a user-friendly, GUI-based vessel segmentation toolbox that generates a high-quality vessel mask to exclude vessels in χ-separation maps. Approach: Two vessel segmentation methods are developed: a region-growing algorithm-based method with handy hyperparameter tuning options and a deep learning-based method with no hyperparameter tuning. Results: The toolbox produces a high-quality vessel mask for enhanced χ-separation analysis. Impact: The vessel segmentation toolbox generates a high-quality vessel mask through a user-friendly GUI, supporting the reliability of analysis of χ-separation results by effectively excluding vessel artifacts for improved quantification of iron and myelin.