Hi Harma
I will see if I can help a little bit (see below).
> 1. Am I correct in assuming that theoretically there shouldn't be any
> structural difference between the modulated and unmodulated images, only
> intensity differences? When I subtract the binarized modulated from the
> binarized unmodulated images all that seems to remain are some voxels around
> the edges of the brain and ventricles.
I would also expect the binarized images to be the same, so I will
have to leave the answer to this to someone more knowledgable.
> 2. When you calculate the modulated and unmodulated images in the same
> segmentation step I interpret them as being complementary to each other. So
> if I for example find significant differences in the modulated images
> between two groups this might result from a volumetric difference between
> those two groups in that region. But the difference might also stem from an
> initial difference in concentration that was already present before
> modulation. So if I would then have a look at my unmodulated images and I
> find no differences in that specific area I can be a bit more confident in
> assuming that the differences I found in the modulated comparison were due
> to volumetric differences. But if I find differences in the unmodulated
> images as well things become a bit more tricky. Does this sound reasonable
> or am I missing something?
Modulated images should be a quantitative representation of total
amount of GM. Therefore, if you only look at the modulated analysis,
you can interpret these as being due to volumetric differences in GM.
I don't think that looking at the unmodulated analysis affects this
interpretation. In fact, I think unmodulated analyses can be thought
of as reflecting registration errors: if every subject's GM
segmentation were perfectly aligned and not modulated, they would be
identical.
> 3. There has also been some discussion in the SPM mailing list about using
> total gray matter as a covariate or total intracranial volume (and there are
> a few more options). When you compare two groups (patient vs control) and
> you find a lot of regional differences that are all in the same direction
> (for example always patient > control) this will also create a global
> difference in gray matter. When you subsequently use total gray matter as a
> covariate you will mask all interesting regional differences. Therefor I am
> more inclined to use some measure of total intracranial volume as opposed to
> total gray matter, despite of the difficulties in determining CSF. Does
> anyone have any suggestions regarding this topic?
The issue of covariates is complex, and the short answer is that there
is probably no "right" answer—it depends very much on the questions
you are asking. However, an important item to note is that total
intracranial volume (TIV) isn't affected by changes in GM volume...and
thus, including it as a covariate, although it may account for SOME
variability, isn't really accounting for the global differences in
your groups. For example, in healthy aging, TIV remains stable over
time, but GM shows significant decrease (and CSF volume increases to
fill the gaps). Even including TIV as a covariate, any analysis would
be dominated by these striking global effects on GM.
It really comes down to what you think is relevant. If you are
interested in regional GM differences that cannot be explained by
global differences, GM as a covariate is appropriate. I would think
you could look at this as a "conservative" approach, given that you
are limiting yourself to effects that show up in addition to global
effects.
On the other hand, if one group of subjects really shows significantly
less GM across the whole brain than another group, this may be
informative, and including total GM as a covariate might mask this
interesting effect.
Hope this helps,
Jonathan
|