Hello Joe
> Well, the important thing to remember when reporting cluster stats is
> that the height threshold is arbitrary. So there is *no* statistical
> reason for picking one height over another. If you want to use 3.1,
> that's fine, but 2.3 or 4.5 is equally fine statistically.
You are right of course. But there are conventions among the scientific
community that one should stick to (or become an outsider). And the 3.1
level is conventional I'm afraid, especially for someone trying to report
results from some not very well known university ;-)
>
> To report your mega-cluster, you could simply find peaks in the
> different anatomical regions. Personally, I don't like to do this
> because cluster stats explicitly binarise height data (over the
> height threshold or not) and therefore peak information is not
> relevant to cluster stats but it's a common practice.
How do you get those local maxima? I don't know the appropriate procedure
> method is simply to describe the anatomical regions partly or wholly
> encompassed by the activation. A good anatomical description (and a
> figure) is far more informative than a table of relatively arbitrary
> peaks anyway.
Ok, here I would like to share my solution to the problem, just in case
someone would be in the same trouble.
1) I load thresholded zstat to MRIcro and using ROI>>export analyze image as
ROI I create ROI consisting of entire activated region.
2) With MRIcro I open ALL image and on that image I open previously created
ROI (it also might require Transferring ROI to new coordinates - those of
ALL image)
3) Pressing the Info button in the ROI menu of MRIcro I get the statistics
telling how many voxels in a given structure are covered by my activation
4) Knowing how many voxels given structure occupies I can calculate percent
of activation.
Best regards
Mike
|