Hi AFNI peeps -

I've done my analysis in MNI space; if I draw an ROI using the Draw Dataset plugin, and select some anatomy with the TT_Daemon in the dropdown (fusiform gyrus, in my case), draw it, and then open a "whereami" window and click around, I notice a couple of things. First, the ROI generated looks really weird; second, the results that "whereami" reports for the TT_Daemon atlas are often not fusiform gyrus. In short, it has drawn a weird thing that even it does not seem to believe is fusiform gyrus.

If I do these same steps using the CA_ML_18_MNIA atlas that Daniel made available a while back, the drawn ROI looks much more like the fusiform gyrus, and is mostly reported as fusiform gyrus as I click around, except in the more posterior parts or on the edge of the masked region, where it reports the FG is 1-5mm away.

Questions:

1. I thought that, behind the scenes, the coordinates from all of these atlases were being translated into the space of the active dataset using the appropriate magic (Brett transform, for example, in the case of MNI > TLRC); so that even though my dataset is in MNI space, that if I selected an ROI from a TT template that stuff would just magically work and results would be sensible. That seems to not be the case. Does that mean that I shouldn't use Draw Dataset / whereami to define ROIs or report on locations unless the atlas is in the same space as the dataset?

2. If the answer to 1 is "No, you can't use an atlas from another space" then does that mean my option is to xform my group analysis results into the TT template (or some other template) and then do this ROI stuff?

Long ago I received some sage advice in this thread:

[afni.nimh.nih.gov]

The summary was that I should do the analysis using a template brain from and in the space with a good atlas. I didn't change anything at that time because our analysis takes weeks to run. If only I had listened then... Anyway, in the spirit of the advice:

4 How one can tell whether an atlas is good or not? Is it just whether or not it has the structures you care about in it? Or whether you want the probability that a given location is in a certain structure, vs. just getting a value assigned?

5. I can't find details of what MNI Anatomical actually is; but presumably it's closer to MNI space than TLRC is. So if I want to use the MNI_A atlas like CA_ML_18_MNIA, which has most of the macro-structures I want, is it less bad to shift the stats from MNI to MNI_anat using @Shift_Volumes than it would be to shift them to TLRC, for instance? [EDIT: it appears from this reference

[cercor.oxfordjournals.org]

that maybe MNI anatomical is the same as MNI, just shifted over a bit? In which case, the bad things about xforming the stats into a new space would not apply?]

Okay, that's enough. Thanks for any help, suggestions, pointers.