> I also tried creating a new image, and summing the voxels within it, however
> that gave a different response to both the actual volume, and volume given
> by SVC. I'll use one of my ROI's in one of my paradigms as an example. The
> actual volume (given by both Marsbar and Analyze) is 1685mm^3; the volume
> given by SVC in SPM is 1236mm^3; and the volume I obtain using your method
> is 1488mm^3.
That's odd... how exactly did you generate this last image and derive
its volume? Use imcalc, select first the results directory mask.img,
and then the ROI (sorry, I probably didn't say this carefully last
time), and use the expression '(i1>0).*(i2>0)' then to get the volume
of this in mm^3, you'd want something like:
vol = spm_vol('new_mask.img');
img = spm_read_vols(vol);
Nvx = sum(img(:) > 0)
vxv = abs(det(vol.mat));
Nmm = Nvx * vxv
If I do this, with a 1mm isotropic ROI mask, and a 2mm isotropic
analysis mask, then I exactly reproduce the mm^3 output shown in the
> as I'm unsure whether using SVC, the whole ROI volume was searched for
> activation? Reading through your previous email on SVC, I'm assuming the
> whole volume ROI volume is searched through, and as such I should use the
> actual volume (i.e., 1685mm^3) when determining percentage of activated
> voxels within ROI...
First up, as I mentioned before, the SVC will not search voxels in the
ROI if they are outside the original analysis mask. Its 1236 is
therefore the volume it searched, and not your 1685, which ignores the
analysis mask. (however, you should hopefully be able to reproduce
1236 using the above method). Secondly, since this is a big
discrepancy, I think you must be using something like absolute or
relative threshold masking in your analysis, which you perhaps don't
want if you really want to search all of your SVC ROI. So check that.
And/or check whether your ROI goes outside the brain/GM, which you
maybe don't want either.
If you want to compare results with your ROI on different methods
and/or subjects, then possibly you might want to use your volume
anyway, rather than the potentially different volumes for each set of
results (since the analysis mask.img could be different, even without
threshold masking, since constant voxels are masked out). Your results
would then be a percentage of the *maximum* possible number of voxels
searched, not the number that the SVC actually looked at. Does that
sound a reasonable compromise?
Lastly, note that I still think my understanding of the code in the
following thread is correct, which means that SVC will additionally be
ignoring voxels that are below your initial whole-brain statistical
threshold (though these voxels are counted for the search volume
calculation), so you'll probably want to make sure you choose a lax
uncorrected threshold for this, before proceeding with SVC.
Sorry if any of this is unclear,