Hello,

I had a question regarding comparing tlrc'ed datasets across subjects. In this case, I have two groups containing 12 subjects each, and have extracted functional time series data for all. The brains were registered to the same atlas (TT_N27), and for each I extracted (adwarped) functional data contained in its own cortical grey matter mask (constructed from Freesurfer data). The masks were also adwarped to tlrc'ed anat.

I was hoping to make direct voxel-to-voxel comparisons across these brains, by compariing the extracted tlrc xyz coordinate values (using option -xyz w/ 3dmaskdump). As you can see from the attached image though, the overlap between voxel coordinates seems to be less than ideal. Very few coordinates are found in all 12 datasets, and the majority of xyz values are unique to one or just a few datasets.

One source of these discrepancies is certainly that each brain has its own GM mask so one would not expect perfectly matching cortical volumes. Also, due to different folding the mapping to tlrc atlas is not perfect.

However, I was thinking that since the functional data and masks were resampled to 6mm voxel res, it may be that the absolute coordinates depend on the centers of the (6mm)^3 voxels, so that if the voxels between brains are shifted in tlrc space relative to each other, this could cause matching voxels to be off by 1-2mm (sub-voxel res), impeding voxel comparisons.

So, my question is simply whether these results seem surprising and whether there is a better way to go about making voxel-wise comparisons across subjects in this way. Any thoughts?

Thanks!

John S.