A Study on Improvement of Renal Artery Segmentation Using Hybrid Method

摘要

For the purpose of giving better aid by computers for kidney surgery, more accurate vessel segmentation method is pressing needed. However, since the blood vessels in the kidneys have a low contrast, segmentation still remain a challenge. The blood vessel segmentation methods based on the Hessian-matrix (HM) filter have really a great performance, but over-segmentations are serious. We propose a novel hybrid method that combines graph-cut algorithm with template model tracking algorithm to segment the blood vessels. We use the graph-cut in collaboration with tubular filter to get the rough segmentations and use the model tracking to find the missed smaller vessels. Experiments on CT volumes showed that the proposed method have a great performance on small blood vessel segmentations with more than 70% overlapping.