Sorry about the vague description. What I'm doing is taking a statistical image and converting it into 1's and 0's based on whether a given voxel had a significant overall F-test. Actually, I was able to co-register this just fine to the T1 (as you said, it is forced into alignment). I guess my only remaining question is what comments anyone would have on the ROI analyses. Basically, I have a hand-drawn ROI and would like to limit which voxels are utilized in computing the median to those that also showed a significant overall F-test. As mentioned in my previous message, the reason behind this is that there is a great deal of inter-subject variability in percent signal change values so I thought this would be a good method for focusing even further on voxels of interest.

Thank you all for your time.