Vessel segmentation for visualization of MRA with blood pool contrast agent

Abstract

We present a method for the segmentation of vessel structures in 3D magnetic resonance angiography (MRA) images with bloodpool contrast agent, allowing artery–vein separation for occluding vessel removal from MIP visualization. The method first uses a front propagation algorithm to select a path along the vessel of interest. Two controlling speed functions are considered, a multi–scale vessel filter, and an approach based on a cylinder shape model. The cylinder based method uses orientation information which is propagated with the front and iteratively updated as the surface expands. Once a vessel of interest is selected, orientation and radius parameters are used to construct a deformable model of the vessel, which is then adapted to the image borders to refine the segmentation of the selected vessel. The results of a comparison with manual segmentations are presented. The extracted centre lines are compared with those from the manual segmentations, showing a mean deviation of 2.55mm for the multi–scale filter, and 1.06mm for the cylinder model, compared to voxel dimensions of 0.93mm. The mean deviation of the final segmentation from the surface of the manual segmentation was 0.59mm.