Use combinatorial maps and topological invariants for medical image segmentation
Author: Guillaume Damiand (Univ. Lyons)
In this talk, I will show how to use combinatorial maps and topological invariants for medical image segmentation. First, I will introduce combinatorial maps, an efficient data structure allowing to describe quasi-manifold, and I will show how to define some efficient 2D and 3D image segmentation algorithms based on this data structure. Then I will show how we can use these combinatorial maps to compute incrementally some topological invariants, as Euler characteristic or Betti numbers. Then, these invariants can be used as topological criteria to guide the segmentation algorithms in order to improve the results of the image segmentation.