Liver Vessel Segmentation

Liver vessels generated from computed tomography are usually pretty small, which poses big challenges to satisfactory vessel segmentation, including 1) scarcity of high-quality and large-volume vessel masks, 2) difficult to capture vessel-specific features and 3) heavily imbalanced distribution of vessels and liver tissues. To help the improvement of vessel segmentation, we build a high-quality and large-volume dataset having both liver and vessel masks available. Comprehensive experiments conducted on the dataset as well as existing well known datasets show that the averaged dice coefficient score is lifted by at least 12.8%. We believe that this dataset can be of great help for this area.