ABSTRACT
A new 3D segmentation method based on the level set technique is proposed. The main contribution is a robust evolutionary model that requires no fine tuning of parameters. A closed 3D surface propagates from an initial position toward the desired region boundaries through an iterative evolution of a specific 4D implicit function. The objective of this chapter is to extract the vascular tree from images of various modalities. Level sets are used to cover the complex topology of the vascular system. An accurate statistical approach is combined with the geometrical approach to give an accurate segmentation. The proposed statistical model is based on modelling the intensity histogram of the volume of the given data set using linear combination of Gaussians (LCG) with positive and negative components. The level set balloons are initialized inside the blood vessels with the largest cross-sectional areas. The surface evolves to cover the vascular tree. Experiments with simulated and real medical images (magnetic resonance angiography, and computed tomography) confirm the accuracy of our approach.
