Segmenting the skull base

Hello Slicer Community

Background: I am a relatively new 3D slicer user. I am a neurosurgery resident with no programming experience. I am interested in using Slicer to segment neuroimaging for the purposes of 3D printing and education.

My question is, what is the best way you have found to segment the skull base from CT scans? I have tried a few methods but due to the extremely thin bone and generally poor image resolution in this region of the body, my models do not appear anatomically accurate. When using the threshold tool, if the threshold is too high, there are many holes where there should be bone (such as the orbital roof or posterior clinoids). However if the threshold is too low the small foramen disappear or there is “noise” that doesn’t resolve well with smoothing tools.

I am hopeful that an extension such as Total Segmenter could be created for the skull base and was curious if anyone is currently doing this work or at least knows of a better way to segment skull bases with improved anatomic accuracy.

I don’t think I’ve seen anything specific, but I would think that with some effort you could train MONAI Label to do that task. It may take several dozen example segmentations to use for training. If you find a public dataset maybe multiple people could contribute to a project.