Dr. Gaser (or whoever finds it in their heart to answer),
I am admittedly new to neuroimaging analyses, though I feel like I have an OK handle on the methods. I've doing some VBM using CAT12 with two groups (n=20; n=39). I read several articles in my field that used TFCE and I like the idea that there is a non-parametric way of handling MRI data. So, I ran a t-test in CAT12/SPM12 and then applied TFCE (smoothed images 6-6-6 FWHM, default TFCE values, permutations = 5000) through CAT12. The result is not quite what I had expected to see.
When controlling FDR at the 0.05 level, I end up with one large cluster (k = 133572) and many smaller. However, within that one large cluster, there are clearly some areas that are more significant than others (R insula, L SMA, L Mid Occipital; see attached image 1). If I am more stringent in my control (e.g., FDR < 0.005), I still end up with what appears to be a lot of background noise and areas that are more interest (see attached image 2).
My question is thus twofold:
1. Does it appear that I did something wrong in the process that gave me such large clusters, of which only small areas appear valuable?
2. Is there a way within CAT12/SPM to somehow additionally threshold out all of that background.
2. As my analysis is, right now, geared toward a whole-brain approach, is there a way to obtain just those areas without all of the other noise?
Thanks in advance for anyone/everyone's help
|