@auto_tlrc with Schaefer 2018 atlas

I’d like to use the @auto_tlrc function to standardize my anatomical image to the Schaefer 2018 atlas.

First off, I align the epi to the anatomical image:
align_epi_anat.py -anat “$sub”.struc+orig -epi “$sub”.image+orig -epi_base 10

and then use @auto_tlrc:
@auto_tlrc -no_ss -base $directory/Schaefer2018_400Parcels_7Networks_order_FSLMNI152_2mm.+tlrc -input $sub.struc_al+orig

As showed in the attached screenshot, I’m having issues with @auto_tlrc output. I normally use ‘-base TT_N27+tlrc’ and never had issues. I’m probably missing some steps here on how to make that atlas usable by afni.

I’d appreciate any help that points to documentation or any help solving this issue.

thanks much,

Don’t align to an atlas; align to a template that is aligned to an atlas (the dataset with labeled regions). The N27 dataset is a template. That version of the Schaefer-Yeo atlas is aligned to the 2006 version of the MNI template supplied with FSL. We have a new version of the Schaefer-Yeo atlas aligned to the 2009 asymmetric version of the MNI template, supplied with AFNI, and you can try it out with the link below.


thanks, I get now why it wasn’t working earlier. I used the the MNI152_T1_2009c template with @auto_tlrc, but when using 3dROIstats later to get the beta coefficient for each voxel/node of the Schaefer_17N_400, I see a message in the terminal saying that my dataset has 8530021 voxels/nodes while the mask has 16777216. am I using the wrong template? thanks

The resolution of your functional data is likely much lower than that of the atlas. You need to resample the atlas to be the same grid of the dataset. For example here, you might use:

3dresample -master funcdset.nii -inset atlasdset.nii.gz -rmode NN -prefix atlas_rs.nii.gz

This assumes the functional dataset is in the same template “space” as the atlas by aligning to the appropriate template, in this case, MNI_2009c. Your example shows using @auto_tlrc, but that only does an affine transformation. It doesn’t do a nonlinear alignment to align the data much better than that of an affine alignment alone. For that, you would probably use @SSwarper to compute the transformation of your anatomical data to the MNI 2009c template space. That program computes both affine and nonlinear alignment and skullstrips the dataset too. The EPI to anatomical alignment can be combined with that anatomical to template transformation. afni_proc.py is probably the easiest way to do all the combinations needed (motion correction, obliquity, EPI-anat, anat-template).

3dresample worked well. I’m going to follow your suggestion and use afni_proc.py for alignments. thank you