Error when running watershed

Support
1

I’m trying to use watershed segmentation (5.12.4 on Win10) and I get a segmentation fail message. This is what I get on detials:

Traceback (most recent call last):

File “C:\Users/JSG/AppData/Local/slicer.org/3D Slicer 5.12.4/bin/Python\slicer\util.py”, line 3614, in tryWithErrorDisplay

yield

File “C:\Users\JSG\AppData\Local\slicer.org\3D Slicer 5.12.4\lib\Slicer-5.12\qt-scripted-modules\SegmentEditorEffects\AbstractScriptedSegmentEditorAutoCompleteEffect.py”, line 303, in onPreview

self.preview()

File “C:\Users\JSG\AppData\Local\slicer.org\3D Slicer 5.12.4\lib\Slicer-5.12\qt-scripted-modules\SegmentEditorEffects\AbstractScriptedSegmentEditorAutoCompleteEffect.py”, line 567, in preview

self.computePreviewLabelmap(mergedImage, outputLabelmap)

File “C:/Users/JSG/AppData/Local/slicer.org/3D Slicer 5.12.4/slicer.org/Extensions-34645/SegmentEditorExtraEffects/lib/Slicer-5.12/qt-scripted-modules/SegmentEditorWatershedLib/SegmentEditorEffect.py”, line 111, in computePreviewLabelmap

labelImage = sitk.ReadImage(sitkUtils.GetSlicerITKReadWriteAddress(mergedLabelmapNode.GetName()))

             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File “C:\Users\JSG\AppData\Local\slicer.org\3D Slicer 5.12.4\lib\Python\Lib\site-packages\SimpleITK\extra.py”, line 384, in ReadImage

return reader.Execute()

       ^^^^^^^^^^^^^^^^

File “C:\Users\JSG\AppData\Local\slicer.org\3D Slicer 5.12.4\lib\Python\Lib\site-packages\SimpleITK\SimpleITK.py”, line 8534, in Execute

return \_SimpleITK.ImageFileReader_Execute(self)

       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

RuntimeError: Exception thrown in SimpleITK ImageFileReader_Execute: D:\a\SimpleITK\SimpleITK\Code\IO\src\sitkImageReaderBase.cxx:91:

sitk::ERROR: The file “slicer:0000023F71755F30#vtkMRMLLabelMapVolumeNode1” does not exist.

I created two segments and placed seeds in them. I am doing something wrong?

2

You are not doing anything wrong, this is a known bug in the extension, not a workflow issue. The exact same traceback (SimpleITK ReadImage failing on the slicer: pseudo file path) is filed in the SegmentEditorExtraEffects issue tracker as issue 71, reported against 5.12 and still open. Painting seeds in two segments is exactly how the effect is meant to be used. Until it gets fixed, the built-in Grow from seeds effect in Segment Editor covers the same seed-based workflow without using the extension’s code. You can subscribe to the issue on GitHub to get notified when a fix lands.

3

Thank you for letting me know about the bug and for the suggestion!

4

if you know some basics of python, it’s easy to run with sitk functions.

def watershed(self,
input_image: sitk.Image,
level: float,
fully_connected: bool,
mark_watershed_line: bool,
need_fill_holes: bool,
maximum_hole_volume: int,
return_intermediate: bool = False,
) → sitk.Image:

# 1. Hole Filling
if need_fill_holes:
    fill_holes_image = self.remove_holes(input_image, maximum_hole_volume, fully_connected)
    print('done with filling holes')
    slicer.app.processEvents()
else:
    fill_holes_image = input_image

# Generate binary mask for distance calculations
binary_mask = fill_holes_image != 0

# 2. Connected Components
components_image = sitk.ConnectedComponent(binary_mask, fully_connected)
print('have had components_image already when it says so')
slicer.app.processEvents()

# 3. Distance Map (CRITICAL FIX: useImageSpacing=True)
# Using binary_mask is safer than the labeled components_image
distance_map = sitk.SignedMaurerDistanceMap(
    binary_mask, 
    insideIsPositive=False, 
    squaredDistance=False, 
    useImageSpacing=True 
)
print('have had distance_map while it says so')
slicer.app.processEvents()

# 4. Compute the watershed
watershed_image = sitk.MorphologicalWatershed(
    distance_map,
    markWatershedLine=mark_watershed_line,
    level=level,
    fullyConnected=fully_connected,
)
print('doing morphologicalwatershed==> done')
slicer.app.processEvents()

# 5. Mask out the background (Cookie-cutter step)
masked_watershed_image = sitk.Mask(
    watershed_image, 
    sitk.Cast(binary_mask, watershed_image.GetPixelID())
)
print('doing Masking, done.')

if return_intermediate:
    return (
        masked_watershed_image,
        fill_holes_image,
        distance_map,
        components_image,
        watershed_image,
    )
else:
    return masked_watershed_image