GrayscaleModelMaker geometry incorrect in 5.6.1 but works in preview

[pieper]

Summary

There's a large inaccuracy in the location of the model created by the GrayscaleModelMaker module in several recent releases. The volume in question is somewhat far from the origin and is slightly rotated.

In a recent preview it looks like this, as it should:

In 5.6.1, either the binary downloaded or a local build it looks like this:

Steps to reproduce

• Download this zip and extract nrrd file:

ProblemVolume.nrrd.zip

• Load as normal scalar volume

• Use Grayscale Model Maker with default settings

• Enable slice display in Models module

Note: the segmentations infrastructure does not appear to have this regression, so I have suggested that the users who reported this issue migrate to that method since this module is very old. Still, it would be good to know what is at the heart of this.

Environment

• Slicer versions: I tested some older versions of slicer.

  • 4.0.1 (2012-01-06) it works correctly
  • 4.6.0 (2016-10-13) it works correctly
  • 4.11.20200930 it fails
  • 5.0.2 (2022-05-06) if fails
  • 5.6.1 (2023-12-12) if fails
  • 5.7.0 (2024-03-25) it works correctly
    so something appears to have happened in between that has since been corrected in the current preview version.

• Operating system: My tests were on Mac, but the issue was reported to me on Windows

[jcfr]

Looking at the changes associated with SuperBuild/External_VTK.cmake, update to vtkWindowedSincPolyDataFilter may have fixed the problem:

• BUG: Improve surface smoothing accuracy in segmentations #7643

cc: @lassoan @pieper

[jcfr]

Applying this patch updating vtkWindowedSincPolyDataFilter to use the Hamming function allow to revert back to the "failing" case:

diff --git a/Modules/CLI/GrayscaleModelMaker/GrayscaleModelMaker.cxx b/Modules/CLI/GrayscaleModelMaker/GrayscaleModelMaker.cxx
index 7db9cec856..89bfe29ff2 100644
--- a/Modules/CLI/GrayscaleModelMaker/GrayscaleModelMaker.cxx
+++ b/Modules/CLI/GrayscaleModelMaker/GrayscaleModelMaker.cxx
@@ -185,6 +185,7 @@ int main(int argc, char* argv[])
       std::cerr << "Warning: Smoothing iterations of 1 not allowed for Sinc filter, using 2" << endl;
       Smooth = 2;
     }
+    smootherSinc->SetWindowFunctionToHamming();
     smootherSinc->SetInputData(mesh_LPS);
     smootherSinc->SetNumberOfIterations(Smooth);
     smootherSinc->FeatureEdgeSmoothingOff();

[jcfr]

Closing this issue as we identified which VTK update effectively fixed the problem.