[MINC-users] Issue with itk_convert_xfm

[Andrew Wood]

Hi Vlad,

I think I didn't explain my issue correctly. Let me
add -invert_transformation to my mincresample invocation and try again.


Here are the commands that I think illustrates a problem:
$ make_phantom phantom.mnc
$ param2xfm -translation 10 10 10 trans.xfm

# Apply the xfm
$ mincresample -use_input_sampling -*invert_transformation* -transform
trans.xfm phantom.mnc out.mnc

# Do the same transform in nifti space
$ itk_convert_xfm trans.xfm trans.mat
$ mnc2nii phantom.mnc phantom.nii
$ antsApplyTransforms --input phantom.nii \
--transform trans.mat \
--output out.nii \
--reference-image phantom.nii

# Open result in a minc+nifti-friendly viewer, or convert back to minc
$ nii2mnc out.nii thru_nifti.mnc
$ register out.mnc thru_nifti.mnc



As far as I can tell, out.mnc and thru_nifti.mnc should be identical, but
they clearly are not.

Best,
Andrew

On Thu, Jan 14, 2016 at 4:39 PM, Vladimir S. FONOV <vladimir.fonov at gmail.com
> wrote:

> antsApplyTransforms is basically equivalent to mincresample
> -invert_transformation
>
> On Thu, Jan 14, 2016 at 3:18 PM, Andrew Wood <andrew at biospective.com>
> wrote:
>
> > Hi all,
> >
> > I came across a set of transforms generated by ANTs which I've been
> trying
> > to convert and use on minc volumes. There seems to be a bug in one or
> more
> > of the tools I'm using, where applying the transforms to nifti volumes
> does
> > not give the same result as applying them to minc volumes.
> >
> > Likely culprits:
> > - itk_convet_xfm
> > - mnc2nii or nii2mnc
> > - antsApplyTransforms
> > - mincresample
> >
> > Somewhere along the way, several components of the transforms get
> inverted.
> > Here is an example:
> >
> > $ make_phantom phantom.mnc
> > $ param2xfm -translation 10 10 10 trans.xfm
> >
> > # Apply the xfm
> > $ mincresample -use_input_sampling -transform trans.xfm phantom.mnc
> out.mnc
> >
> > # Do the same transform in nifti space
> > $ itk_convert_xfm trans.xfm trans.mat
> > $ mnc2nii phantom.mnc phantom.nii
> > $ antsApplyTransforms --input phantom.nii \
> > --transform trans.mat \
> > --output out.nii \
> > --reference-image phantom.nii
> >
> > # Open result in a minc+nifti-friendly viewer, or convert back to minc
> > $ nii2mnc out.nii thru_nifti.mnc
> > $ register out.mnc thru_nifti.mnc
> >
> > The result of this for me is that the z component of the transform was
> > inverted, but the x and y components are OK.
> >
> > From experimenting with various simple transforms, it seems that the
> > following components are inverted:
> >    - z translation
> >    - z rotation
> >    - x,y,z scale
> >    - x shear
> >
> > Has anybody else noticed or dealt with this before?
> >
> > Thanks,
> > Andrew
> > _______________________________________________
> > MINC-users at bic.mni.mcgill.ca
> > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
> >
>
>
>
> --
> Best regards,
>
>  Vladimir S. Fonov ~ vladimir <dot> fonov <at> gmail <dot> com
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>