Segmentation of large nodule from CT

It turns out that part of the tumor is hollow:

With a little adaptation Iof threshold I (lower : -406) I get

which is probably reasonable.

Your reference was probably done with a fill-between-slices effect of the segment editor which requires more work.

Dear @rbumm,
thank you very much for your suggestion and updates.
I will perform another segmentation with the lower threshold of -406.
In your new case volume is around 13000 compared to 18000.
Is this a tolerable difference?
I will try to perform the same procedure for the other cases.
Concerning your observation “Your reference was probably done with a fill-between-slices effect of the segment editor which requires more work.”, is fill-between-slices effect a default option?

Fill between slices is on option in which you paint every volume slice of the tumor with an intensity modified paint tool and later use the fill between slices effect to create the actual segmentation. It is much more work and looks often ugly, like a layered cake.

If you create a tumor segmentation with local threshold and maybe add some pixes here and there that are missin or remove some that are representing vessels 13000 is the reference value. (net volume with holes)

Dear @rbumm,
thank you for support.
I will proceed with another case in order to evaluate if the differences still remains in the other cases of the same database. One more question:
-the results of segmentation in 3d seems to be in transparency. How to convert them as opaque models?

Good morning,
I’m testing other cases of the same dataset.

I’m having trouble defining the local threshold on the nodule for them. More specifically, after imposing the threshold range, and command left click (Mac), not only the region of interest but also the other regions the other regions are also selected. Can I have suggestions about it?


Could you provide one or two case numbers?

The number of the same dataset are SL0078 SL0080 SL0081

Where do I find these patients? Do you have a NBIA manifest for this series?

@rbumm I will share with you these cases soon. Thanks

(1) Create a ROI
(2) Place this ROI around the tumor
(3) use Local Threshold effect with Grow Cut enabled

(4) do some postprocessing (remove vessels)

Thank you very much @rbumm.
I will use your suggestions and update you.

Dear @rbumm,
sorry for delay to reply and thank for support.
Creating a ROI and performing the local threshold approach, I performed the segmentation.
The problem is that the ROI created disappeared after several slices. It means that the segmentation segmentation neglects the region from a certain slice onwards. Can you suggest how to fix it?
In addition, is it possible to create a “not cubed” ROI (a ROI with possibility to create curved ROI?)

@rbumm Can you please suggest me how to fix the last problems?

Unfortunately, there is no elliptical 3D ROI in Slicer.
I would recommend segmenting the right lung with Lung CT Segmenter and using it as a mask → Editable area “Inside right lung” for the Growcut operation.

Thanks a lot.
I will try. Concerning this problem “the ROI created disappeared after several slices” for the previous approach, can you please suggest how to fix it?

I would recommend to resample the volume 1,1,1 for a test. Create a new volume as output. Try to use the ROI with the new volume.

Thanks a lot.
Just the last question.
After performing Lung CT Segmenter, sometimes 3d models appear transparent and other times not.
What is the function that makes them transparent so you can see the module inside? Thanks again

Go “Segmentations”, select the structure you want, go “Advanced” and move the 3D slider.

1 Like