Daniel Glen Wrote:
-------------------------------------------------------
> This may have to do with differences in the
> resampling/interpolation to the template. Try
> resampling first with 3dresample to the grid of
> the template and nearest neighbor interpolation.

Hi Daniel,

resampling what to the grid of the template? I _think_ that all my files at this point are on the same grid.(or is there anything more to the grid than the Geometry String, Data Axes Orientation, and Data Axes Tilt?)

I start with a structural T2, which I align to the template. After this point, as far as I can tell, the aligned T2 is on the same grid as the template, which is the same grid as the atlas. (by the way, there is no intermediate T1 anymore, but that's a different story)

Next I mask the T2 (same grid), and then I threshold and clusterize it in GUI, and save cluster maps (same grid).

Then I calculate the overlaps between cluster maps and atlas ROIs, and I tried to do it in three ways: via whereami, using afni programs to isolate "my" and atlas ROIs and find intersections, and with a Matlab script that does the same (after cluster maps are converted to NIfTI via AFNI program, as Matlab easily loads NIfTI).

And whereami gives slightly different results than the other two. I agree that the magnitude of the difference suggests something like a side effect of resampling, but I do not understand what and why would be resampled, if the intersected files are on the same grid. Or - again - I think they are.

I should also say that yesterday Paul asked me offline to provide the files, which I did. So maybe he will be able to figure what's going on.