Questions of overshading of Grown from seeds option to build up bronchus model

Support
,
1

Hi everyone here, after I checked the video https://www.youtube.com/watch?v=9iiOBmaP8bA of growing from seeds, I immediately started up my trial on my CT case, however, even though I zoomed in a lot on the small bronchi, the airway build always takes part from the lung, and the smaller bronchi can’t be built nicely. How can I fix these problems?

2

It should be no problem segmenting up to 3-4 generations (level of branching) of bronchi, which should be sufficient for surgical planning. Up to how many generations are you trying to segment?Can you post a few screenshots?

One thing to pay attention when you correct seeds in low-contrast areas that you always need to paint on both sides of the boundary (when you paint a bit of “airway” inside a bronchus then always paint a bit of “parenchyma” outside the bronchus to prevent leaking).

I would also recommend to check out fully automatic “AI” segmentation tools (LungCTAnalyzer, MONAIAuto3DSeg, TotalSegmentator, DensityLungSegmentation), as you may find that they offer good enough segmentation for your use case.

3

Thank you for ur answer, that is a nice advice on painting parenchyma around the small bronchus, I struggled with the problem of leaking so much before, as I was following only the Lung, airway, and others for seeds growing. And I will try the AI segmentation tool today.

4

By the way, May I kindly ask if there is any possible downloading link for the AI tool LungCTSegmentor as I am also watching youtube videos about that but nothing available in the extensions

5
6

Hello

Few questions

  1. What is your image spacing? and have you tried resampling?
  2. Have you tried Slicer/SlicerAirwaySegmentation: CLI module for airway segmentation starting from chest CT images
  3. What about New extension: MONAI Auto3DSeg - raise the bar in AI medical image segmentation - Announcements - 3D Slicer Community
7

Hi, my previous image spacing is 1.5mm, which is a bit too thick. I also tried the airway segmentation model, but the result is weird as it doesn’t only contain the airway but also the whole lung.

1 Like
8

1.5 is Alright but with lung I wouldn’t expect fine details. If you use MONAI, the LUNGS task uses a 1.5 mm resampling but if you resample your image beforehand you will be able to extract more details but might require smoothing.

image
image1284×640 54.8 KB