Need advice on Segmentation of CT scanned fossil specimen (lungfish)

Support
1

Hi, folks!

I am not an expert in 3D Slicer, I need advice from the community. I have a TM-scanned fossil fish skull with inner structures (canals for nerves and vessels) of rather low contrast (the slice below is after Local Contrast Enhancement in ImageJ). I tried thresholding and Total segmentator, but without success. What might be the most promising method to segment the inner canals?

My PC: Windows 10; Intel(R) Core™ i5-10400F CPU 2.9GHz; RAM 32,0 GB

Rhinodipterus-iter200183
Rhinodipterus-iter2001831024×740 63.7 KB

2

What is TM-scanning? I am not sure what that imaging modality is.

I can’t recognize any structures in this cross-section, but you should give the NNinteractive prompt-based segmentation extension a try. You simply iterate over by giving negative (does not belong) and positive (belongs) prompts (points, curves, ROIs etc) and guide the AI.

3

Sorry for wrong terminology)) I ment CT - computed tomography. The ‘structures’ are those that are shown by dotted outlines (some of them). They represent the canals within the matrix, as seen on the rendered volume.

Thank you very much for your recomendation.

Screenshot
Screenshot903×865 298 KB

Rhino
Rhino1024×740 73 KB

4

I think manually segmenting those manually would be quite cumbersome and I don’t think any of the pre-trained modules (like totalSegmentator) would be of any help. I still think your best shot will be NNInteractive extension. it is manual, but the AI aspect of it should increase your throughput quite considerably.

5

Yes, I am just trying to download the pixi package and to install Python and pytorch.

Thank you once more!

6

You may want to give it a try on MorphoCloud: https://morphocloud.org

We have specific instructions on how to set it up on MC. Tutorials/MorphoCloud/NNInteractive.MD at main · SlicerMorph/Tutorials · GitHub