Hi Theresa,
> 1) Whether there are any methods/scripts to tabulate the volume (mm3 or
> no. of voexls) of significant clusters for quantitative interpretations?
I think you can do this with the standard SPM machinery, at least in
SPM5, the following should give the number of voxels within each
cluster in TabDat{:,4} (see help spm_list for more information)
[hReg,xSPM,SPM] = spm_results_ui;
TabDat = spm_list('List',xSPM,hReg);
To get this in a friendlier format, you might like to try a very
slightly modified version of spm_list, which I've put here:
http://www.cs.ucl.ac.uk/staff/gridgway/vbm/spm_list.m
which will let you follow the above commands with:
spm_list('TxtList', TabDat, 1, 'tabdat.txt');
Producing a tabdat.txt file which you can read into e.g. MS Excel as a
tab-delimited text file (don't merge multiple delimiters).
If any of the official SPM developers would like to include these
modifications in the next set of updates, please go ahead.
> 2) the volumes of all the voxels surviving the voxel level threshold,
> e.g. p < 0.001, uncorrected could be used to measure the volume of all
> regions showing activation for a given contrast.
I think you can do this fairly easily by first clicking the "save"
button in the "display" panel of the interactive window (at least,
that's what/where under SPM5 Updates 826), to create an image of the
significant voxels. If you want the total significant volume, you'd
then want to binarise this image using imcalc with expression 'i1>0',
and finally sum up the voxels, possibly multiplying by their volume in
millilitres, which you can do with this script:
http://www.cs.ucl.ac.uk/staff/gridgway/vbm/get_totals.m
> 3) Because I'm searching the whole brain, any voxel has the same chance of
> false positive or false negative, so I do not use cluster extent correction,
> is it correct?
I'm not too sure what you mean here... I know there are arguments
about not using cluster-based correction if the smoothness is strongly
non-stationary, as it probably will be for e.g. structural VBM data. I
believe that this is because the risk of false positives/negatives
would vary spatially depending on the smoothness.
Satoru Hayasaka has a toolbox (still in beta testing, I believe) that
does allow for non-stationary smoothness:
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0702&L=SPM&P=R3496
Does that answer your question?
Best,
Ged.
|