I need to segment the 3D vessels of the lung in HR-CT data to analyze ILD disease.
Best method for tree vessel detection after Identification and segmentation lung detection, because, I have not enough slices.
The blood vessels have high density, in contrast to the rest of the lung (well... tumor and atelectasis also have high density).
So, you can get away with simple thresholding (-400 HU?) if you want to proceed with voxels, or use marching cubes if you want to proceed with triangles.
Commonly used algorithms are using center line detection methods and reorientation along that center line in order to initialize segmentation algorithms, such as
E.Ukwatta et al. "Joint Segmentation of 3D Femoral Lumen and Outer Wall Surfaces from MR Images" MICCAI 2013. 534-541.
I've seen something nice at CVPR this year, that could work well too: Zhu et al. "Graph-Based Optimization with Tubularity Markov Tree for 3D Vessel Segmentation" CVPR 2013.
Other than that you can always apply a Frangi vesselness filter and threshold the probabilistic result with an easy matlab script. The filter can be found on Matlab central. Hope that helps! Good luck!
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages