AFNI anatUAC as input to freesurfer

Software Support
,
1

Hello,

I wonder if it is recommended to use AFNI’s anatUAC T1 which is intensity uniform-ized, anisotropically smoothed and ceiling-capped (AFNI program: @SSwarper) as input to freesurfer or if it is always better to feed the original T1 to freesurfer. FS certainly does similar things but would it be better to supply “cleaner” data? probably it might make things worse.

Thanks

2

Hi-

The short answer is that I think I would probably use the original anatomical as input to FS’s recon-all. That program will have its own cleaning/uniformizing of the anatomical that is has been built to use (e.g., N4 normalization), and so the downstream parts of that program will likely play best with a simple input.

One caveat/consideration is whether your anatomical has obliquity. The TL;DR is that different software will use/apply the obliquity information differently (sometimes applying it and blurring unnecessarily; sometimes concatenating it well in a concatenated processing stream; or sometimes ignoring it), so it can often be simplest with an anatomical to remove it in a good way if present (the obliquity in an EPI can be left in to be dealt with well during pipeline processing, which is what afni_proc.py does). The ‘in a good way’ in the previous sentence can mean 1) not applying obliquity to unnecessarily blur the data with interpolation during regridding and 2) preserving the coordinate origin (x,y,z)=(0,0,0). But first, just check if your anatomicals do have any obliquity:

# simple, single dset
3dinfo -is_oblique DSET1

# multiple dsets
3dinfo -is_oblique -prefix DSET1 DSET2 DSET3 ...

… which outputs a 1 for “yes, it is oblique” and 0 for “no, it is not oblique”. If you have oblique anatomicals, I would recommend deobliquing before running either FS’s recon-all or AFNI’s @SSwarper; the recipe for the above ‘in a good way’ in AFNI would be (which is used in some processing examples linked below):

# copy to a BRIK/HEAD format
3dcopy  DSET_INPUT   tmp
# do deobliquing without regridding but preserving coord origin
3drefit -oblique_recenter tmp+orig.HEAD
3drefit -deoblique        tmp+orig.HEAD
# copy deobliqued resultt to some output name, and DSET_OUTPUT could be BRIK/HEAD or NIFTI
3dcopy  tmp+orig.HEAD     DSET_OUTPUT

Here are some further notes on running AFNI and FreeSurfer together:
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/tutorials/fs/main_toc.html
Importantly, if you run @SUMA_Make_Spec_FS after recon-all, you get some nice QC images and convenient tissue related maps. See particularly here:
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/tutorials/fs/fs_makespec.html#the-renumbered-ren-atlas-dsets
and
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/tutorials/fs/fs_makespec.html#the-fs-nii-gz-mask-dsets
I think the output brainmask fs_parc_wb_mask.nii.gz might be particularly useful as a -mask_ss fs_parc_wb_mask.nii.gz option with @SSwarper.

Here are some links to papers and online code repositories, where many projects have used both FS’s recon-all and AFNI’s @SSwarper (as well as afni_proc.py, etc.):
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/codex/main_toc.html
For example, this project, with GitHub links:
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/codex/fmri/2022_TaylorEtal.html
… or this other recent one on QC examples (that I have to add to the Codex, I see)

… that relates to this paper:

–pt

3

Hi Paul,

Thanks for your detailed reply. Will look into these materials.