Shape Constraint Incorporated Geometric Flows for Blood Vessels Segmentation

摘要

Flux maximizing geometric flows have been widely used to segment elongated structures such as blood vessels. Only using the magnitude and direction of an appropriate vector field to segment blood vessels often results in leakages at areas where the image information is ambiguous. To overcome this problem, we combine image statistics and shape information to derive a geometric flow to segment tubular structures and penalize leakages. To accelerate the segmentation speed, a two-stage segmentation framework is presented. Results on cases demonstrate it is the mostly accuracy and efficiency of the approach.