IW3D_invert convergence warning when using intermediate outputs from sswarper2

[Nori]

Hello.

I would like to ask about this warning:
++ 3dNwarpCat: AFNI version=AFNI_25.2.05 (Jul 20 2025) [64-bit] ++ Authored by: Zhark the Warper *+ WARNING: IW3D_invert: iterations failed to converge :-( ++ Output dataset ./anatQQ.subj5_full_INVWARP.nii.gz ++ total CPU time = 15.7 sec Elapsed = 7.1

I warped the atlas from the MNI space to the native space using intermediate files from sswarper2 following the instructions on this page:

Convert ROI from MNI to native space

Hi, heatherb- Thanks, and that is a good suggestion for people using @SSwarper, which is generally our currently recommended way for people to align a subject anatomical to a template, such as the MNI. However, @SSwarper includes both linear affine and nonlinear warp components, and I only see linear affine pieces included in your inverse transformation. Also, the initial question was about warping and ROI, specifically, which has the particular constraint of wanting to use an “NN” interpolan…

When I ran the 3dNwarpCat, I got that warning, although the output atlas seems to be sufficiently warped to native space.
How can I resolve this warning?
I would appreciate any advice.

[ptaylor]

Hi-

That warning is probably OK to ignore, if the results look good on the other side. There might have been a tiny spot that didn't converge, possibly even at the edge or outside the brain.

Note that this discussion provides a more modern model for using+inverting @SSwarper or the newer/recommended sswarper2:

use Schaefer atlas in subject space

Hi, Maya- I'll assume this refers to volumetric alignment (the S-Y atlas is defined on surface meshes as well, which transfer to subject space meshes directly if using SUMA-standardized ones). This process will likely involve running nonlinear alignment between each subject's anatomical volume and the template on which the Schaefer-Yeo atlas is defined. If the errts itself is living in the anatomical volume space, then that would be all that is necessary (that is, if the "align" block were u…

The difference is in the order of concatenating the nonlinear warp with the affine transform part and inverting the warp. 3dQwarp's warps are diffeomorphic, and hence invertible; but if concatenation happens first, that beast might not be smoothly invertible in some places. So, it is preferable to invert each piece first, and then concatenate those parts. Namely, for applying the inverse warp to an integer-valued dataset, use:

# create the inverse transform from sswarper2 and apply it to atlas in template space
3dNwarpApply                                                        \
      -nwarp  "INV(${odir}/anatQQ.${subj}.aff12.1D)                 \
               INV(${odir}/anatQQ.${subj}_WARP.nii)"                \
      -ainterp NN                                                   \
      -master "${dset_in_native_anatomical_space}"                  \
      -source "${dset_in_template_space}"                           \
      -prefix DSET_OUT

(If your "source" dataset is not integer valued, you can leave out the -ainterp NN or even keep it and replace "NN" with something like "wsinc5".)

--pt

[Nori]

Hi,
thank you for your reply and giving me information about newer approach.
I will try the newer approach and check if the result looks good to me.

Nori