Hello! I'm excited to be exploring SPHARM, but unfortunately have hit a snag.

I am looking at manually traced mouse corpus callosum volumes in Slicer 4.4.0, saved as nii.gz. I am using SPHARM-PDM (revision d674cd6, released September 5, 2015; the only version I could get installed on my windows PC). The Shape Analysis Toolbox appears to run without error, and generates all of the expected files (including procalign.vtk etc). Yet, using Shape Population Viewer to check, the output resembles a salad, rather than a chunk of corpus callosum. Although I get no explicit error when SPHARM runs (e.g. it says 'completed' rather than 'completed with errors'), checking the log I encounter this error.

ERROR: In ..\..\..\..\Slicer-440\Base\QTCLI\vtkSlicerCLIModuleLogic.cxx, line 1981
vtkSlicerCLIModuleLogic (000000000E5B8CC0): Shape Analysis Module standard error:
Error: C:\D\C\S440-E-b\SPHARM-PDM-build\ITKv4\Modules\ThirdParty\VNL\src\vxl\core\vnl/algo/vnl_svd.txx: suspicious return value (3) from SVDC
Error: C:\D\C\S440-E-b\SPHARM-PDM-build\ITKv4\Modules\ThirdParty\VNL\src\vxl\core\vnl/algo/vnl_svd.txx: M is 3x3
Error: M = [ ...
Error: -1.#QNAN00000000 -1.#QNAN00000000 -1.#QNAN00000000
Error: -1.#QNAN00000000 -1.#QNAN00000000 -1.#QNAN00000000
Error: -1.#QNAN00000000 -1.#QNAN00000000 -1.#QNAN00000000 ]
Error: C:\D\C\S440-E-b\SPHARM-PDM-build\ITKv4\Modules\ThirdParty\VNL\src\vxl\core\vnl/algo/vnl_svd.txx: suspicious return value (3) from SVDC
Error: C:\D\C\S440-E-b\SPHARM-PDM-build\ITKv4\Modules\ThirdParty\VNL\src\vxl\core\vnl/algo/vnl_svd.txx: M is 3x3
Error: M = [ ...
Error: -1.#QNAN00000000 -1.#QNAN00000000 -1.#QNAN00000000
Error: -1.#QNAN00000000 -1.#QNAN00000000 -1.#QNAN00000000
Error: -1.#QNAN00000000 -1.#QNAN00000000 -1.#QNAN00000000 ]
Executing BatchMake: Error!

This occurs on multiple computers, with files in multiple locations (C drive, dropbox, thumbdrive etc.). Troubleshooting I have attempted is as follows:

• There are no islands or holes in my traces.
• I have tried smoothing my traces (using Label Map smoothing in 3D slicer) to no avail.
• I have checked the mirroring (for some bizarre reason flipped volumes were un-flipping being saved as gipl.gz; hence I am now using nii.gz files).
• I've opened my nii.gz files in other software just to make sure Slicer wasn't hiding an error from me.
• I have tried with a larger slice of the corpus callosum (taking 1mm either size), results still seemed scrambled. Things do look correct, however, if I trace the whole corpus callosum (resulting in a helmet-shaped object), but fail to work again if I split this into halves so I can look at asymmetry.
• Ramping sx, sy, and sz up to 1 or down to 0.01 does not improve things, so it isn't just the area I'm looking at is too physically small.
• Reasoning the high-frequency edge of the traces might be causing issues, I've tried with more (and in desperation, fewer) harmonics, to no avail.

I have attached my example shapes to see if others can replicate this error/find a solution.

I would appreciate any help to get this working - it'd be awful to have to give up when I'm so close!