Desikan-Killiany/Destrieux to MNI152NLin2009cAsym (LPS, no sform)

I’m looking for advise on how to prepare the Desikan-Killiany atlas for inclusion in qsiprep.
This entails transforming the fsaverage aparc+aseg segmentation into MNI152NLin2009cAsym space (LPS+ without sform).

What I tried was:

  1. Run mri_label2vol using the MNI152NLin2009cAsym atlas like so:
    mri_label2vol --temp mni_1mm_t1w_lps.nii.gz --seg aparc+aseg.mgz --reg $FREESURFER_HOME/average/mni152.register.dat --o aparc_aseg_mni152.nii.gz

  2. Since mni152.register.dat is based on a MNI152 template in LAS format, I manually flipped the data along the Y dimension using Matlab functions.

While the result looked resonable, this dis not take care of the sform issue. I tried reading the header from one of the atlases included with qsiprep with niftiinfo in Matlab and write back the data to disk using that, which did not lead to a valid brain image, so whatever format is used in qsiprep does not seem compatible with the Matlab functions.

I also have my doubts about using mni152.register.dat since I beleive this is based on an earlier MNI template version.

I have a feeling the solution is much cleaner/simpler than what I am doing. Any hints/tips/suggestions are welcome!

2 Likes

Unfortunately I can’t assist as I am also facing the same issue adding an atlas to QSIPrep. If anyone knows the solution that would be great.

1 Like

Did you ever figure this out?

Hi. I try to do same things(using custom atlas), I already resampled my atlas to the LPS+ and same FOV. But I didn’t understand how to zero-out the sform? Is it necessary and how to do it? would you mind to help me?