remove fsl deps, containerization wip #12
Conversation
There was a problem hiding this comment.
A couple notes on the replacements for FSL utils that don't depend on NIfTI-specifics.
TBH I would try to break this out into separate PRs. There's a lot going on here, with reformatting, removing the extensions and tests, replacing FSL and adding a Dockerfile.
gradunwarp/core/unwarp_resample.py
Outdated
| outputOrient = get_fslorient(self.input_nii) | ||
| if outputOrient == b"NEUROLOGICAL": |
There was a problem hiding this comment.
Simpler:
| outputOrient = get_fslorient(self.input_nii) | |
| if outputOrient == b"NEUROLOGICAL": | |
| if np.linalg.det(self.input_nii.affine) > 0: |
You don't really care about all the logic of consistent or inconsistent or unknown. You just need the affine according to the NIfTI 3 methods, which both FSL and nibabel implement. And this also doesn't tie you to NIfTI. It will work just fine with an MGH file.
There was a problem hiding this comment.
It might be cleaner to restrict it to the 3x3 determinant - the 4th row being 0 0 0 1 (as it always should be, but we don't check that here) should cause it to ignore the translation values, but I prefer not to rely on assumptions when I don't have to.
Co-authored-by: Chris Markiewicz <[email protected]>
Co-authored-by: Chris Markiewicz <[email protected]>
|
Some of this seems to have been incorporated into #15 or otherwise separated, closing due to merge conflicts. |
FSL was only used for fslorient / fsval which I believe can be handled with nibabel.
Removing it would make the package pure-python (with the non-necessary c-extension removed earlier too #9 ).
Maybe if someone with knowledge about nifti orientation can double-check that the changes makes sense, that would be great.