Dear Raphael,
One additional comment, as you wrote "I got results, but almost every blob would only appear if I do not use the FWE correction."
Based on the figure you've used FWE correction on voxel level, but you don't have to use this threshold (depending on patient group and sample sizes it might be "too conservative"). You should definitely account for multiple testing of course, but this could also be achieved with an uncorrected voxel threshold of say .001 in combination with a corrected cluster threshold of say .05 FWE, probably taking into account non-stationarity. For the latter adjust spm_defaults.m by setting defaults.stats.rft.nonstat = 1; When hitting results / whole-brain the calculation is going to take longer, the cluster-level p corrected is still called the same, but it now takes into account local smoothness. This can lead to a significant cluster with 80 voxels in one region and another larger but non-significant cluster in another. This you have to be careful when checking the results and it might take a while and/or need some editing/photoshopping to get proper figures.
Best,
Helmut
|