Total segmentator module returned with non-zero exit status 1 on to 5.12.1

Support
1

I’m having problems with the execution of the total segmentator module, I thought it was something installation, as version 5.10 was running I updated to 5.12.1 but it keeps the same error only with this module. I saw that some people have already reported problems with it! Trying to run lung nodules or lung segment airway and vessels I received the same error message

IMG_9238
IMG_92381920×1440 784 KB

2

My workaround for this problem was to uninstall Slicer, delete the app files from AppData, and reinstall Slicer inside the C drive (not in any subfolder, just inside the drive itself)

3

I tried this, but nothing done! Already do the same error

4

Please follow these instructions: GitHub - lassoan/SlicerTotalSegmentator: Fully automatic total body segmentation in 3D Slicer using "TotalSegmentator" AI model · GitHub

5

I would like to share the solution that worked for me, as it may help other users and possibly identify a compatibility issue.

I was using:

  • 3D Slicer 5.12.2 (Windows)
  • TotalSegmentator extension 2.13.0
  • CPU inference (AMD GPU, no CUDA)

The following tasks worked correctly:

  • total

However, the following tasks consistently failed:

  • lung_nodules
  • lung_vessels

The errors were:



RuntimeError: Could not find requested nnUNetTrainer_MOSAIC_1k_QuarterLR_NoMirroring

and



RuntimeError: Could not find requested nnUNetTrainerSkeletonRecall

I tried all of the following without success:

  • Force reinstall of the TotalSegmentator extension
  • Re-downloading the model weights
  • Deleting and downloading the datasets again
  • Restarting Slicer multiple times

The downloaded models were correct, and the standard total task always worked.

Checking the installed package versions revealed:



TotalSegmentator: 2.13.0
nnunetv2: 2.8.1

The issue was resolved by downgrading nnunetv2 to 2.6.2 from the Slicer Python Console:



slicer.util.pip_install("nnunetv2==2.6.2")

After restarting Slicer:



TotalSegmentator: 2.13.0
nnunetv2: 2.6.2

Everything worked correctly:

  • :white_check_mark: total
  • :white_check_mark: lung_nodules
  • :white_check_mark: lung_vessels (vessels and airways)

This suggests that TotalSegmentator 2.13.0 is currently not fully compatible with nnUNetv2 2.8.1 for tasks that rely on custom nnUNet trainers, while it works correctly with nnUNetv2 2.6.2.

Hopefully this information helps other users experiencing the same issue and assists the developers in tracking the version compatibility problem.

Thank you for the excellent extension.

6

Thanks for the information. It looks like the issue was resolved in TotalSegmentator 2.14.0.

SlicerTotalSegmentator will need to be updated to a newer version: