Questions about the Pulmonary Fissure Completeness workflow

[mauigna06]

Hi, thanks for the great work

I'm having trouble getting the fissure_phenotypes_workflow.py to finish. It always fails while executing ClassifyFissureParticles

Maybe the problem is my input data, I have read the appendix of the paper that published this code but maybe I'm missing something.

• Could you provide a test CT to input on the ComputeFeatureStrength and the parameters for the algorithm? I know I should use RidgeSurface but not sure about which method and its parameters (e.g. Frangi and alpha, beta, C)

• How accurate the labelmap of the lung lobes should be? Could you mention acceptable error in millimeters?

• I'm executing this on my laptop, do I need a more powerful hardware?

Thank you :)

[rjosest]

Hi Mauro,

Thank you for the question. Here are some answers

1. For feature strenght, I would suggest using FrangiXiao. A good set of params could be
   ComputeFeatureStrength -i case.nrrd -o ridge.nrrd -f RidgeSurface -m FrangiXiao --kappa 2 --alpha 0.2 --beta 0.5 --C 25 --std 1,1,1

2. The lobe labelamp is used as a reference to assign particles to the right fissure. I would say that you can tolerate a reasonable error (+- 10mm)

3. Your laptop should be fine. The process should not consume more than 8 GB in memory for contemporary CT scans.

I hope this helps

[mauigna06]

Thanks a lot @rjosest

For feature strenght, I would suggest using FrangiXiao. A good set of params could be
ComputeFeatureStrength -i my_CT.nrrd -o feature_strength_from_CT.nrrd -f RidgeSurface -m FrangiXiao --kappa 2 --alpha 0.2 --beta 0.5 --C 25 --std 1,1,1

This worked much better than my previous parameters. The fissures are more much visible now.

The lobe labelamp is used as a reference to assign particles to the right fissure. I would say that you can tolerate a reasonable error (+- 10mm)

I'm trying to have an automatic workflow for fissure completeness assessment . So my labelmap was created with TotalSegmentator and adapted to the conventions of CIP (i.e. labelmap with values from 4 to 8 with datatype ushort). In most parts the segmentation is below the error threshold you mentioned but not everywhere.

---

To execute the workflow I used docker like this:

docker run -d -ti --name CIP -v $USER_FOLDER/CIP_tests:/host -d acilbwh/chestimagingplatform:1.71
docker exec -ti CIP /bin/bash
python /ChestImagingPlatform/cip_python/nipype/workflows/fissure_phenotypes_workflow.py --in_fiss ridge.nrrd --in_lobes case_labelmap.nrrd --out_csv CaseId_fissurePhenotypes.csv --cid CaseId

I have the following crash:

root@614c89f03251:/host# nipype_display_crash /host/crash-20241126-222251-root-classify_particles-6a55bb95-928e-44b3-b936-33a99d1bd547.pklz


File: /host/crash-20241126-222251-root-classify_particles-6a55bb95-928e-44b3-b936-33a99d1bd547.pklz
Node: FissurePhenotypesWorkflow.classify_particles
Working directory: /tmp/tmptNRTu2/FissurePhenotypesWorkflow/classify_particles


Node inputs:

angle = <undefined>
args = <undefined>
dist = <undefined>
dist_thresh = <undefined>
environ = {}
fischer_thresh = <undefined>
ifp = /tmp/tmp535M0Q/cid_fissureParticles.vtk
ignore_exception = False
ilsm = <undefined>
irsm = <undefined>
lm = /host/remapped_lung_lobesV2.nrrd
olo = /tmp/tmp535M0Q/CaseId_leftObliqueParticles.vtk
orh = /tmp/tmp535M0Q/CaseId_rightHorizontalParticles.vtk
oro = /tmp/tmp535M0Q/CaseId_rightObliqueParticles.vtk
terminal_output = stream



Traceback: 
Traceback (most recent call last):
  File "/root/miniconda2/lib/python2.7/site-packages/nipype/pipeline/plugins/linear.py", line 39, in run
    node.run(updatehash=updatehash)
  File "/root/miniconda2/lib/python2.7/site-packages/nipype/pipeline/engine/nodes.py", line 394, in run
    self._run_interface()
  File "/root/miniconda2/lib/python2.7/site-packages/nipype/pipeline/engine/nodes.py", line 504, in _run_interface
    self._result = self._run_command(execute)
  File "/root/miniconda2/lib/python2.7/site-packages/nipype/pipeline/engine/nodes.py", line 630, in _run_command
    result = self._interface.run()
  File "/root/miniconda2/lib/python2.7/site-packages/nipype/interfaces/base.py", line 1043, in run
    runtime = self._run_wrapper(runtime)
  File "/root/miniconda2/lib/python2.7/site-packages/nipype/interfaces/base.py", line 1660, in _run_wrapper
    runtime = self._run_interface(runtime)
  File "/root/miniconda2/lib/python2.7/site-packages/nipype/interfaces/base.py", line 1694, in _run_interface
    self.raise_exception(runtime)
  File "/root/miniconda2/lib/python2.7/site-packages/nipype/interfaces/base.py", line 1618, in raise_exception
    raise RuntimeError(message)
RuntimeError: Command:
 ClassifyFissureParticles  --ifp "/tmp/tmp535M0Q/cid_fissureParticles.vtk" --lm "/host/remapped_lung_lobesV2.nrrd" --olo "/tmp/tmp535M0Q/CaseId_leftObliqueParticles.vtk" --orh "/tmp/tmp535M0Q/CaseId_rightHorizontalParticles.vtk" --oro "/tmp/tmp535M0Q/CaseId_rightObliqueParticles.vtk"
Standard output:
Reading lung particles...
Reading lung lobe label map...
Getting surface points...
Identifying particles region and isolating...
Standard error:
Segmentation fault (core dumped)
Return code: 139
Interface ClassifyFissureParticles failed to run. 



root@614c89f03251:/host# 

What makes me ask myself, why the values below appear undefined on the crash-report?

angle = <undefined>
args = <undefined>
dist = <undefined>
dist_thresh = <undefined>
fischer_thresh = <undefined>

[mauigna06]

Hi @rjosest. Thanks for the great work

The lobe labelamp is used as a reference to assign particles to the right fissure. I would say that you can tolerate a reasonable error (+- 10mm)

I'm trying to have an automatic workflow for fissure completeness assessment . So my labelmap was created with TotalSegmentator and adapted to the conventions of CIP (i.e. labelmap with values from 4 to 8 with datatype ushort). In most parts the segmentation is below the error threshold you mentioned but not everywhere.

I've checked again and my labelmap does have less than +- 10mm error everywhere

---

I could replicate the results from CIP papers.

• CIP paper figure below:
  

• My results below (i.e. particles plotted over fissures) using the CIP paper's workflow:
  
  

Only problem I have now is that, in last step, the workflow fails (in this line) and according to some debugging I did, it is due to these lists at runtime have too few elements because these ifs may be too strict. I'm temped to change the if clauses or edit the is_fissure function definition to be to be more compliant.

Do you have any advice?