Dear Jack,
Hello again. Your proposition seems very sensible to me. In fact, Keith
Worsley has experimented with the same kind of thing (I think he mentions it in
his 1996 SVC paper). The problem with this kind of approach is that the gray
matter map tends to have rather rough edges, and this inflates the correction
necessary, beause of the high surface area. If you use some sort of ellipse or
sphere that contains the cortical mantle, but is continuous rather than
voxellated, then you will likely get a less conservative correction. Maybe
someone out there has the relevant resel counts for such a volume, for the MNI
brain?
You could combine the mask with the stats by using explicit masking, with the
full monty option in the stats, or by writing out an SPM as an image, then
combining it with the mask using ImCalc. Then you'd have the problem of
displaying the activation image with a structural scan. You could use the
sexier bits of spm_orthoviews, or some other software, to look at the results. I
wrote a slice display routine which displays slices of combined activation and
structural images, which works OK, but is still early in development:
ftp://ftp.mrc-cbu.cam.ac.uk/pub/imaging/Slice_overlay
See you,
Matthew
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|