[Insight-users] Fast Marching Fine Vessel Segmentation

[Wu Zheying]

Hi all,

does anyone know how i can improve on my segmentation of vascular techniques? I am currently using the Fast Marching Image Segmentation algorithm from the ITK examples on my MRI Max Intensity Projection brain image. I have 5 seed positions at different parts of the vessel and i am giving it the following parameters:

sigmoid = 0.5
alpha = 0.5
beta = 3.5
timethreshold = 420
stopping value = 420

The problem is it segments well for contours that are wide apart, however for very thin vessel lines, it is almost impossible to segment it even when i have place a seed position at that location. I have noticed the gradient magnitude filter does a pretty good job of displaying the contour lines, however when it comes to the sigmoid filter, most of the fine thin contours are distorted/merged by some noise...I have tried to alter the curvature anisotropic filter as best as i can to smooth the image, still it is not good. How can i improve this? or is this the best that the Fast Marching method can do?

Hope someone can help me with this...

Zheying