LungCTAnalyzer extension for lung CT segmentation and analysis for COVID-19 assessment

[rbumm (Rudolf Bumm)]

I can not yet reproduce your black preview in 3D, mine (in Slicer 4.13 and LCTA 2.47) is colored.

Anyway, if you enabled the 3D preview, you should “Hide preview in 3D” afterwards to obtain only the final output segmentation in the 3D view.

[Aaron_Tsai (Hsiu-Ming Tsai)]

Hi Rudolf,

I would like to test the package on rodents’ lung tissue analysis.
When I move to step 2, that is the step I would need to place a few seeds.
However, the image is kind of resample to very large voxels makes the images blurred.
Dose this package only good for human CT images?
Thank you very much,

Regards,
Aaron

[rbumm (Rudolf Bumm)]

Hi,
The package has been developed for human CT and the Lung CT Segmenter module does not yet work reliably with Micro CT. However, you could go to “Segment Editor”, and create a “lung segmentation” with “right lung” and “left lung” by using the “Grow From Seeds” effect. As soon as you have good lung masks you could then run the “Lung CT Analyzer” module and use the “lung segmentation” with the lung masks as input segmentation.
Regards
Rudolf

[lassoan (Andras Lasso)]

I think the only value hardcoded for human clinical CT is the preview image resolution of 2mm. This is probably too large for microCT.

You can easily modify this value by editing this line in LungCTSegmenter.py file:

rbumm/SlicerLungCTAnalyzer/blob/0ebbfee16245c44c5f2b3e848f93eb3e045ee6dc/LungCTSegmenter/LungCTSegmenter.py#L924

Alternatively, you can modify the spacing value of the volume in Volumes module to be 10x larger. This does not change the image in any way, it just makes all length measurements 10x larger.

@rbumm we could consider allowing users to edit this value in some advanced parameters section in the GUI.

[rbumm (Rudolf Bumm)]

@lassoan never thought you could set volume spacing from the Volumes extension directly.
So now I can confirm that setting a rodent microCT´s spacing from 0.133mm 10 x larger enables the use of Lung CT Segmenter

with the following lung thresholds:

The other option does not work for me yet. Testing …

[rbumm (Rudolf Bumm)]

@lassoan

parameters = {"outputPixelSpacing": "2.0,2.0,2.0", "InputVolume": self.inputVolume, "interpolationType": "linear", "OutputVolume": self.resampledVolume}

To my understanding setting

914. parameters = {"outputPixelSpacing": "0.,0.,0.", "InputVolume": self.inputVolume, "interpolationType": "linear", "OutputVolume": self.resampledVolume}

should leave the spacing untouched in resampling, 2D views are not blurred, but the lung preview looks like this:

… and “Apply” hangs Slicer.

Any ideas?

[lassoan (Andras Lasso)]

0 spacing is invalid, do not set that value. If you don’t resample then computation time will be too long for interactive use. If you do not want to expose spacing as a parameter to the users (because it is too technical) then we could add a “lung size” (diameter of the lungs along a chosen anatomical axis) and compute the spacing relative to that. It is important that this spacing is not related to the input image resolution but the level of detail that we want for the segmentation preview.

[rbumm (Rudolf Bumm)]

Setting spacing directly to 1 mm (original 0.188mm) and then using 2.0, 2.0, 2.0 works well. What would be your suggestion for a corresponding ouputPixelSpacing setting to achieve a similar result by leaving the input spacing at 0.188mm?

[lassoan (Andras Lasso)]

If you had to increase the spacing of the original image by approximately 5x (from 0.188mm to 1mm) then it means that you could instead decrease the spacing of the preview image by approximately 5x to have about the same level of detail. So, instead of 2.0mm spacing, use 0.4mm spacing for the preview image.

[rbumm (Rudolf Bumm)]

Yes, but it does not work for some reason. I will open an issue and we can discuss it there.