CTA vessel segmentation: petrous ICA vs bone overlap problem

[sandcastle]

Hi all,

I’ve been working on segmenting the circle of Willis and proximal carotids from CTA data. This has been straightforward with 3DRA and TOF-MRA, but I’m running into difficulty with the petrous ICA being inseparable from adjacent bone on CTA.

The main issue seems to be that contrast-enhanced vessels and cortical bone have overlapping intensities, so:

• manual thresholding fails

• Grow from Seeds leaks into bone

• pre-processing with Swiss Skull Stripper has not helped

Does anyone have recommendations for separating vessels from bone in this region? For example:

• specific Slicer workflows or modules

• preprocessing techniques

• or alternative segmentation strategies beyond intensity-based methods

I’m relatively new to 3D Slicer, so any guidance on specific workflows or modules would be especially helpful. Thanks in advance!

[pieper (Steve Pieper (Isomics, Inc.))]

A lot depends on the quality of your CTA. I had some luck on exactly this problem using nnInteractive.

[chir.set (SET)]

You can use the Guided artery segmentation and the Guided vein segmentation modules in the SlicerVMTK extension. A sample result is below.

It’s indeed not straightforward inside a bone. Segmenting the carotids at the neck level is done in a breeze (except in very very tight stenosis). Segmenting the circle of Willis is even more tricky.

Since you are a beginner, I would advise to crop your volume to the target region of study to save time. Resampling to isotropic voxel size gives better results that are more good looking in addition.