BUG: Jacobian determinant filter results depended on index space orientation #5390
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
github-actions
bot
added
type:Bug
dzenanz
reviewed
Jun 10, 2025
Modules/Filtering/DisplacementField/test/itkDisplacementFieldJacobianDeterminantFilterTest.cxx
Outdated
Show resolved
Hide resolved
Modules/Filtering/DisplacementField/test/itkDisplacementFieldJacobianDeterminantFilterTest.cxx
Show resolved
Hide resolved
|
Thanks, I'll update on both of these points |
|
@hjmjohnson Why are we merging main into branches. I thought our standard work flow was to rebase, to avoid an ugly commit history to extraneous merges? |
hjmjohnson
force-pushed
the
fix_jac_determinant
branch
from
9dd0051 to
5f0446b
Compare
It was a mistake. It tried using "gh pr update-branch" which did not do what I wanted. I have re-pushed the changes without the merge. |
The gradient is calculated in a neighborhood defined in index space. DisplacementFieldTransform accounts for this by multiplying rows of the gradient by the direction matrix. DisplacementFieldJacobianDeterminantFilter did not, hence its determinants were different for the same physical transformation under different voxel index ordering. DisplacementFieldJacobianDeterminantFilter is now fixed to use the direction matrix. The test is updated to check that the determinant for a given displacement field is consistent at the same point in physical space, regardless of how the index space is oriented. More details in issue InsightSoftwareConsortium#5358.
hjmjohnson
force-pushed
the
fix_jac_determinant
branch
from
5f0446b to
b084fc0
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixing #5358
I updated the ITK test to check the determinants after permuting or flipping the index space, and updating the header to compensate. I think I've done this correctly, at least I can now get the same answer as with an identity transform.
I did some more tests in some sample data I made for debugging ANTs jacobians (linked in the issue).
This PR makes the filter more comparable to what you would get by iterating over the displacement field voxels and calling the
ComputeJacobianWithRespectToPositionfunction at each point. It's not exactly the same, because the finite difference sampling scheme is not the same, although the filter allows users to set custom weights, so maybe it could be made more similar.One other thing (which I can fix): in the test, the seeming typo variable name "dispacementfield" was used throughout one function, I thought it might be intentional, so I didn't change it. But happy to update if needed.