MNI Spatial Normalization Trasformation

Dear afni experts,

what kind of spatial normalization is used to perform spatial normalization in mni?

thanks,

Piero

Hi, Piero-

Can you please clarify/rephrase this? I don’t understand.

–pt

Sorry, i mean which transformation (affine, other?)

thanks

P.

Hi-

I’m afraid I still don’t understand.

Would you like to know what kind of spatial normalization (=alignment method) we would recommend for registering your data to an MNI template? If so, we recommend strongly using nonlinear alignment to align a high res (~1 mm iso) anatomical dset (typically a T1w for when processing FMRI; perhaps a T2w when processing DTI) to an MNI template.

In general, when aligning 2 brains of different subjects (which includes subject to template alignment), linear affine alignment is not “strong” enough or “warpy” enough to get good alignment of brain structures.

Tools for nonlinear alignment include 3dQwarp—the main workhorse program—but then also convenient wrappers for that program that include more features. We would recommend using @SSwarper for skullstripping (ss) plus nonlinear alignment for human subjects, esp. since these outputs can be integrated directly into afni_proc.py for FMRI processing. For animal subjects, and analogous multi-utility program for skullstripping/nonlinear alignment/mapping templates is called @animal_warper-- it too can be integrated into FMRI processing (there are multiple macaque demos https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/nonhuman/macaque_demos/main_toc.html). But I assume you are processing humans because you are asking about MNI (unless you have veeeeeery evolved other species???). Note that @SSwarper takes specific multi-volume templates, derived from the original standard ones, and for MNI, you would use MNI152_2009_template_SSW.nii.gz as a reference base.

There are a lot of other important considerations, such as cost function, etc. I would recommend some of these notes/comments/thoughts on alignment in general:
https://www.youtube.com/watch?v=PaZinetFKGY&list=PL_CD549H9kgqJ1GDXAs1BWkgEimAHZeNX
One also has to consider several other points about performing alignment preferably; such as concatenating transforms before applying them to the EPI data (so it gets warped as little as possible) and more… Again, much of that is covered in the video playlist.

–pt