This week Kitware announced a new partner library for ITK and VTK that brings tubular segmentation and registration support to the community at large.
A guiding premise of TubeTK is that by focusing on 1D and 2D manifolds we can devise methods that are insensitive to the modality, noise, contrast, and scale of the images being analyzed and to the arrangement and deformations of the objects in them. In particular, we propose that TubeTK’s manifold methods offer improved performance for many applications, compared to methods involving the analysis of independent geometric measures (e.g., edges and corners) or requiring complete shape models.
Tubular segmentation is big-business in the biomedical space, isolated out blood vessels and other organs within larger structures. Surely they’ll be the first big users of this library, but eventually it could come to a huge variety of fields.