Your steps look correct. The one thing that I'm not sure you mentioned is that your EPI datasets need to be coregistered/aligned to the anatomical datasets (I usually use align_epi_anat.py to do this). The outcome of this can be verified by looking at each individual's single-subject statistical results on the respective anatomical file. If you use afni_proc.py, it will generate the @ss_review_driver script that will take you through this check and some others.

As for which mask to use... You can use any of 1) the mask of the EPIs that get fed into 3dDeconvolve; 2) the average mask of all the anatomical images resampled to the resolution (which you're in the process of making); 3) or the (dilated) mask of the template image. I've seen all of these in publications, and I arranged them in this order as I think the first two are often very similar and the most helpful for identifying activation variations across the sample (and in some ways the "most honest" because these might be outside of the template brain) and the last one often makes for the nicest graphics.



Edited 1 time(s). Last edit at 08/22/2014 07:27PM by Peter Molfese.