How to understand the axialization process by fat_proc_axialize_anat?

AFNI Message Board
1

Hi Paul,

The FATCAT tutorial says that :

This takes the NIFTI T2w dset and axializes it with respect to the reference dset (here, from the MNI 2009 templates, which was manually AC-PC aligned and regridded …

Is the MNI 2009 template not AC-PC aligned? Why re-AC-PC align it?
Is this understanding correct: it’s a easy way (not need to manually assign anatomical markers) to automatically produce near AC-PC aligned T2w? As the reference is AC-PC aligned.

Thanks,
2086

2

Hi, 2086-

The TORTOISE folks (https://tortoise.nibib.nih.gov/) had a pipeline procedure where they were performing manual AC-PC alignment of their anatomical volume before using it as a reference in DIFFPREP/DR_BUDDI processing. In general, some kind general alignment of the brain to have standard viewing planes makes sense: the final output of DIFFPREP/DR_BUDDI is not warped to standard space, and we use directional color encoding (DEC maps) to view structures; it would be good to have “red” (left-right) features be approximately similar across a group, etc. Additionally, if one is going to use TORTOISE’s DR-TAMAS for templatizing, having the DWIs start out in more similar/near-overlapping spaces might likely be beneficial.

AC-PC alignment is a common clinical way to achieve this goal; there are others that seem reasonable (e.g., alignment to a standard template brain). This is the general procedure I refer to as “axialization”, of which AC-PC alignment is one example. AC-PC alignment focuses on aligning the subcortical structures of the brain (which, I was told, tend to be more uniform across people than cortical ones); in some cases, one could conceivably prefer to not focus on the subcortex and have cortical structures more

The motivation of developing fat_proc_axialize_anat was to automate this axialization process, and have it be uniformly performed across a group. Again, in working with the TORTOISE folks, they were quite keen to have AC-PC-like alignment, where subcortical structures were emphasized in the axialization process. Hence, they AC-PC aligned the MNI template manually, and I generated a region-based mask to emphasize the subcortical part of the brain, which is utilized via the “-extra_al_wtmask …” option.

You are not required to use either the AC-PC alignment MNI template brain as a target, nor the subcortical weightmask, in your axialization process.

–pt

3

Hi Paul,

Thanks for your detailed answer. I understand the role of having a uniform DTI reference now.

I was following the FATCAT tutorial, using the same axializing reference, and fat_proc_axialize_anat kept out data uniform, that’s good. But when I tried to write this part in paper, I found I was not clear about this process. So could I write like this:

We axialized our data by rotating to an AC-PC aligned reference to have a uniform looking across subjects

But I still not understand, is the MNI 2009 template not AC-PC aligned? Why they re-AC-PC align it manually?

Thanks,
2086

4

Hi, 2086-

I would say something like:
“Since subject brains can be acquired at arbitrary angles within the FOV, we axialized our data by applying a rigid-body transform to an AC-PC aligned reference, in order to have more uniform viewing planes across subjects and directionally encoding color (DEC) maps for identifying structures. This was performed using AFNI’s fat_proc_axialize_anat, using an AC-PC aligned version of the MNI template as a reference base, with an additional subcortical weight mask when calculating alignment.”
… which can be modified if you use a different target, if you include different options and/or don’t use the weight mask, etc. (NB: you can put your code/commands in an appendix or supplement for specificity).

AC-PC alignment depends on identification of specific structures in the brain. The TORTOISE folks did the AC-PC alignment of the MNI template, and I think in their estimation a small pitch adjustment was required.

–pt

5

Hi Paul,

Thank you very much!!!

Best,
2086