A SHAPE INDUCED ANISOTROPIC FLOWFOR VOLUMETRIC VASCULAR SEGMENTATION IN MRA

Evolutionary schemes based on the level set theory are effective tools for medical image segmentation. In this paper, a shape prior is introduced which can be useful for vessel segmentation and can produce elongated structures. For a hypothetical evolving implicit surface, using the gradient vectors of its signed distance transform (SDT) a shape measure is introduced that is maximized whenever the local surface resembles a cylinder. Using this shape prior, a new functional is defined and the optimization is obtained by applying Frechet derivative. We show that this yields an anisotropic expansion term that propagates the surface in the tangential direction of vessel. This prior is then combined with edge information to produce a complete level set scheme for vessel segmentation. We have applied our method to five real MRA data sets and comparison has been made with a state-of-the-art vessel segmentation method. Presented results indicate that using this method a significant improvement is achievable and the method can be an effective tool to extract vessels in MRA intracranial images