> Many thanks for having provided all of us with such detailed information on how to extract ROIs, which will be most helpful for many.
> 1) Let's say I am interested in extracting the first eigenvariate from a subcortical ROI using your instructions (combination of large fixed sphere + smaller sphere moving to nearest local maximum), but would want to be absolutely sure that I extract my VOI from a voxel within a given anatomical region (for instance, in my case, a sub-cortical structure such as the caudate). Would it make sense to add a region of interest in the form "New: Mask Image" between the larger sphere (which corresponds to the second entry) and the smaller sphere (which would now be shifted as fourth entry in the list of Region(s) of Interest)? As for the mask image specifics, I would then load a caudate mask as 'image file', leaving the 'Threshold' field as default. I would then modify the 'Mask Expression' field of the smaller sphere to i2&i3 (previously, it was set to i2 only, in your instructions) and then change the final Expression as follows: ' i1 & i2 & i3 & i4'
How about simply using an anatomical mask (New: Mask Image) instead of the larger sphere?
> 2) Second question -- as we use 'Thresholded SPM map' as second entry in the list of Region(s) of Interest, default settings include Threshold = 0.001 (uncorrected) and cluster 'Extent (voxels). Does this mean that, if there were no significant voxels at p=0.001 for the region of interest specified by the combination of spheres (±anatomical mask), no time series would be extracted? If so -- would it make sense to choose a more liberal value of 0.05 (uncorrected), to be sure to get signal possibly from every subject?
Yes - if I remember correctly it will give an error if there are no supra-threshold voxels. A liberal threshold would be a good idea to protect yourself from this. P < 0.05 sounds fine.
All the best