Dear Feng,
It is rather uncommon to acquire whole-brain data and then to just look at a certain region as you could have acquired data for the ventral parts of the brain with a much shorter TR then. But leaving this aside, in that case it's up to you which threshold to choose. You should definitely correct for multiple comparisons at some point, but you could do so
1) on peak/voxel level, based on an FWE-corrected height threshold of e.g. .05
2) on cluster level, based on an uncorrected height threshold of e.g. .001 combined with an FWE-corrected cluster threshold of e.g. .05
For 2) you just go with "none" for p value adjustment to control, followed by e.g. .001 and 0 for extent threshold. Then press SVC, define the small volume. The table should now show SVC-adjusted results (note that the glassbrain is NOT updated!), you would have to check whether any of pFWE-corr under "cluster-level" are smaller than .05. If so, you could state "Due to our a priori hypothesis whole-brain analysis was restricted to the right fusiform gyrus, which was defined anatomically based on atlas XYZ. For that purpose a small volume correction was applied on cluster level. Clusters were regarded as significant when falling below an initial uncorrected voxel threshold of .001 and an FWE-corrected cluster threshold of .05 adjusted for the small volume". You could certainly shorten the passage, but this way people should definitely be able to understand what you did. If the corrected cluster p-values are still > .05 you failed to find sig. results. p-values of the peak section are irrelevant in that case. And for the results section something like "the SVC resulted in one sig. cluster located in the anterior part of the fusiform cluster (peak coordinates x y z, T = 11.11, k = 20, corrected cluster p = .02).
For 1) it's a little tricky. You go with "FWE" for p value adjustment to control, followed by e.g. .05 plus 0 for extent threshold. Look at the bottom of the results table, FWEp tells you the T value which is required for a voxel to reach significance on peak/voxel level when looking at the whole brain. Now press SVC and define the small volume. Look at the bottom of the results table again, FWEp should have changed. This FWEp corresponds to the T value which is required for a voxel to reach significance on peak/voxel level when restricting the analysis to the small volume. Now go with another p value adjustment to control, but this time select "none" and then enter the T value you've just read on the output. Ignore the whole-brain results, click on SVC, look at the results table. Check for sig. results, this time ignore the p-values of the cluster-level section. You could state "Due to our a priori hypothesis whole-brain analysis was restricted to the right fusiform gyrus, which was defined anatomically based on atlas XYZ. For that purpose a small volume correction was applied on voxel level. Voxels were regarded as significant when falling below a corrected voxel threshold of .05 (FWE) adjusted for the small volume". And for the results section something like "the SVC resulted in two sig. clusters located in the anterior part of the fusiform cluster (peak coordinates x y z, T = 11.11, k = 20) and the ventrodorsocentromedianwhatsoever section (peak coordinates x y z, T = 22.22, k = 33)." Of course, you can add an (arbitrary) extent threshold of e.g. 10 voxels if you think that results are relevant only if they exceed a certain size.
Don't hesitate if you need some more information, I fear my description could be a little confusing. ;-)
Best,
Helmut
|