Output data space

I have used QSIPrep to preprocess some diffusion data; I have not been able to clearly find in the documentation if the data gets registered to a template space by default, and if yes, which that space is.


Hi @jhlegarreta,

The output data are in subject space aligned to ACPC. There are also anatomical outputs in MNI space, but your DWI derivatives are aligned to ACPC without any template warping. It is mentioned here: Quick Start — qsiprep version documentation


Hi @Steven,

thanks for the reply and link.

There are also anatomical outputs in MNI space,

Which are those outputs?

I did not specify the --dwi-only or --output-spaces flags; so, according to the help
All outputs will be registered to the T1w image, but if the anatomical outputs (and I understand T1w being one) are in MNI space, then the diffusion data is also in MNI space as I see it.

Can this be clarified and possibly documented/disambiguated, please?

Thanks for the support.

It’s just the anat transforms that have anything to do with MNI (and should be self-explained by file names). All other outputs are in the T1 ACPC space. Anatomical inputs should not be in MNI - only raw images should be input to QSIPrep.

I think some of the confusion here comes the fact that the AC-PC alignment changes the coordinates from the native scanner coordinates to a new system where 0,0,0 is where the midline intersects the anterior commissure. This is the same 0,0,0 as in MNI space, so the brains will look somewhat aligned if you open the MNI template and the AC-PC image, as long as it’s an adult brain.

I loaded into a viewer

  • the T1w that heudiconv produced,
  • the <sub-id>_<desc-label>_space-T1w_dwiref.nii.gz that qsiprep produced,
  • the <sub-id>_<desc-label>_space-T1w_desc-preproc_dwi.nii.gz that qsiprep produced,
  • the icbm152_2009c_nonlinsymm MNI template,

and it is exactly what @mattcieslak describes.

I’ll try to make a PR to see if this can be explained better.