In the images I usually analyze, it is common to get vascular structures that have more than one path between endpoints (vessels that originate at one point and converge afterwards). As I understand it, the centerline model extracted by VMTK only detects the shortest path between two endpoints, which causes the longer path (a different vascular segments) to be excluded from the centerline model. However, the network model does include the missing vessel, but I would like to have it in the centerline model as centerlines look much smoother and better than the network model. Any ideas on how to do that?
I attach a couple of images of an example case. The left vertebral artery is missing in the centerline model (first image). The Voronoi diagram inside the vessel and the network model look good, though (second image).