Hello AFNI people -

I have an existing processing pipeline that works in MNI space. I want to get WamI reports on the overlap of the various cluster activations that emerge form this analysis. As a result of an earlier message board conversation:

[afni.nimh.nih.gov]

I realize that not all atlases are available in all spaces. Since the TLRC atlases are the most numerous, and report information in a format that I want (Brodmann areas) one suggestion has been to simply convert my statistical results to TLRC space using 3dWarp. And indeed, this works, I get nice reports, etc.

After some thinking and discussion, though, I realize I'm confused about what's going on behind the scenes. For example, isn't the Talairach brain significantly smaller than the MNI reference brain? And so warping results from one to another wouldn't preserve the information present in the larger structure? For instance, if you have to squash two voxels into one voxel the process is inherently information lossy.

To make this a little more concrete, in this screenshot the top image are my (original) group results, from subjects warped/aligned to MNI152, and overlaid on a MNI reference image. The bottom are those same results, except both anatomical and stats have been warped into TLRC space.

[www.dropbox.com]

This p/cluster threshold produces two clusters of the following sizes:

MNI: 191/81
TLRC: 116/38

3dinfo on both stats files shows voxels sizes in each case are 2.5mm isotropic. So there's significant voxel attrition after the MNI->TLRC transformation. This writeup:

[www.nil.wustl.edu]

Suggests that temporal lobe in particular, and regions further from the center of the brain, are more affected, which would make sense with these results. But I'm not confident I've digested the nuances of this message appropriately.

That's a lot of text. Here's some questions:

1. Am I doing something obviously stupid here?
2. Is there a better way to get the relationship between my cluster activations and these TLRC atlases?
3. Would it change anything to redo the stats from soup to nuts, in TLRC space? or
4. Is this just something you live with?

Advice, pointers to resources, and hard liquor are welcome.

S