Hello,
SPM uses the mask you give it in the prompt (or the box or sphere) as
search volume
and indeed corrects on the voxel and cluster level using the volume of the
mask instead
of the "whole search volume", usually defined by the "mask.img" in your
analysis folder.
However, the procedure is designed to focus on one cluster (or clusters)
and the regions around it
assuming that you are indeed interested in clusters in your "ROI".
So, as fas as I can see it from other analyses here, the clusters reported
AFTER s.v.c. show overlap with your ROI (although
they - but please check) are not necessarily completely within the ROI.
If your cluster is 'cut' by the ROI the cluster size k is of course
reduced, and you cannot expect that the ROI svc works
on cluster parts that are not in your ROI.
So, statistically the answer on your last question is "no" - only voxels
within the are interpreted, and so the cluster
p-value coming out from this does not refer to the "original whole cluster"
of your whole brain analysis.
You also should check if your ROI definition makes sense if you eventually
report cluster outside the ROI...
If I am wrong, please somebody else correct me!
br,
Philipp
Max-Planck-Institute of Psychiatry
NMR Research Group
Kraepelinsr. 2-10
80804 Munich
Mail: [log in to unmask]
Phone: 0049-89-30622-413
|