I’m working on finding changes in resting-state connectivity between the subgenual anterior cingulate (sgACC) and left dlPFC following TMS. I’ve been asked to identify a subset of the sgACC to use as the sgACC ROI to calculate connectivity effect sizes for pilot data that will then be used for power estimates for a grant app. When the entire sgACC is used, findings aren’t significant; the theory that’s been presented to me is that it (i.e., BA25) is too large.
I was initially asked to choose the sgACC ROI based on the region that had the biggest change in connectivity with the dlPFC following TMS, but that struck me as blatantly circular. I considered finding the region with the highest baseline connectivity, since I’d be focusing on change, but it occurred to me that connectivity would most likely decrease from baseline just as a function of regressing to the mean. I’m now being asked to concatenate the timeseries data from before and after TMS, average it, and choose clusters of high connectivity that way, but that still seems circular to me. Then again, I’m very new to fMRI analysis and could be wrong. Does this make sense as a strategy? Is there an acceptable way to limit the ROI based on this dataset that isn’t dredging?