SymmetricEigenAnalysisImageFilter incorrect python wrapping

[chatti]

The python implementation of SymmetricEigenAnalysisImageFilter is returning 6 values per pixel instead of 3, and more than 3 of the 6 values is non-zero. Since there are only 3 eigenvalues for a 3D Hessian (3x3) eigenvalue decomposition, this functionality must be incorrect.

I asked about this on the forum and the comment was that it was probably wrapped incorrectly.
"Maybe the problem is in wrapping specification? The test instantiates the filter with <SSRT<3>, FixedArray<3> >, whereas the wrapping instantiates with <SSRT<3>, SSRT<3> >."

https://discourse.itk.org/t/computing-hessian-eigenvalues-in-python-how-to-interpret-output-of-symmetriceigenanalysisimagefilter/6517

Description

Steps to Reproduce

1. Use a test image (ideally a CT scan) with size >= 100^3
2. Run this code:

import itk
#Load the sample image, ideally a CT scan
hessian_image = itk.hessian_recursive_gaussian_image_filter(
            itk_image, sigma=2.0
        )
#hessian_image size = (6, 188, 217, 259)
eigen_values_image = itk.SymmetricEigenAnalysisImageFilter(hessian_image)
#eigen_values_image size = (6, 188, 217, 259)

eigen_values_array = itk.GetArrayFromImage(eigen_values_image, keep_axes=True)
print( eigen_values_array[:,100,100,100] )

Expected behavior

You should see an array printed out that has 6 values. Similar to this, but the numbers obviously will not be the same
array([ -3.0928779 , 1.86561006, 0. , -11.97670084,
6.26783045, 0. ])

Actual behavior

Reproducibility

Unknown

Versions

5.3.0

Environment

python:
4.9.18 (main, Feb 13 2024, 10:56:47)
[GCC 12.2.0]

[github-actions]

Thank you for contributing an issue! 🙏

Welcome to the ITK community! 🤗👋☀️

We are glad you are here and appreciate your contribution. Please keep in mind our community participation guidelines. 📜
Also, please check existing open issues and consider discussion on the ITK Discourse. 📖

This is an automatic message. Allow for time for the ITK community to be able to read the issue and comment on it.

[dzenanz]

Taking a second look, itk.SymmetricEigenAnalysisImageFilter is the object-oriented syntax. You should have a .Update(), or use snake case like for hessian_recursive_gaussian_image_filter.

[chatti]

SymmetricEigenAnalysisImageFilter

I'm still getting the same problems when I use this code:
hessian_image = itk.hessian_recursive_gaussian_image_filter( itk_image, sigma=sigma ) eigs = itk.symmetric_eigen_analysis_image_filter(hessian_image)

results in
[ -0.56114415 0.56627178 0. -28.24620925 5.85857641 │··························································· 0. ]

When I query a single pixel

[chatti]

Taking a second look, itk.SymmetricEigenAnalysisImageFilter is the object-oriented syntax. You should have a .Update(), or use snake case like for hessian_recursive_gaussian_image_filter.

Object oriented approach doesn't work either

   ```eig_filter = itk.SymmetricEigenAnalysisImageFilter.New(hessian_image)
    eig_filter.Update()
    eig_image = eig_filter.GetOutput()```

[ -0.56114415 0.56627178 0. -28.24620925 5.85857641 0. ]

[dzenanz]

It is a bit strange that your original code worked (due to syntax).

But I guess that problem is in the wrapping.

[chatti]

It is a bit strange that your original code worked (due to syntax).

But I guess that problem is in the wrapping.

Yeah, it ran without error but the values at that pixel were different.