Dear Kris,
Kris Boksman wrote:
> Hello Friends and Colleagues,
>
> I'm hoping someone might be able to advise about the global scaling
> feature of SPM99. I'm looking at comparing the activation elicited by a
> cognitive task between a clinical and a nonclinical group of study
> participants. I've set up the design to subtract the baseline task from
> the "activation" task, etc., and I am clear on how to perform the
> subsequent analyses.
>
> One task my advisor has suggested is to compare the baseline conditions to
> examine whether or not activation in the more "restful" conditions differ
> between the two groups of participants, and to determine if any particular
> areas of the brain are more active in one group than in the other when the
> brains are at relative "rest". It just occurred to me that perhaps by
> using the global scaling and modeling the baseline condition explicitly
> (instead of implicitly as SPM99 can do for you) any comparison I perform
> between the two sets of baselines could be uninformative, since I've
> globally scaled them in order to do the first analysis. (task A - task B)
>
> So my question is, would it be advisable to conduct a new analysis that
> includes only the scans for the baseline but this time to avoid using the
> global scaling? Intuitively this seems right (to me!), but I'm really not
> sure if it is the most appropriate method.
>
> Advise would be greatly appreciated, (and thanks in advance!)
>
I suspect you have a greater problem than that of global normalisation with
the comparison you suggest.
In modalities where one gets a quantitative (or at least linearly related)
estimate of flow/perfusion (e.g. PET) it may be of some interest to compare
resting state scans between groups, although even there I am quite pensive
about attributing detected differences to differences in neural activity.
When using T2* weighted fMRI there are numerous sources of variation of the
observed signal, neural activity being a possibly quite small contributor.
The way we usually designs fMRI studies aims at eliminating as much as
possible these other sources of variation such that we can interpret any
changes as being due to changes in neural activity. Failure to do so, e.g. if
we scan a subject in two sessions where in the first session the subject
rests with eyes closed and in the second session the subject views a
reversing checkerboard, typically leads to nonsensical results. The reason
for failure in that example being that between session changes are of the
same order as the functional changes. By including both conditions on both
sessions the functional changes becomes orthogonal (~independent) to the
session changes and the results become interpretable. In your case where you
have different subjects any detected differeces are almost certainly due
other sources than neural activity, e.g. residual anatomical differences.
Perhaps if you have anatomical scans for both groups you could have a go at a
bit of morphometry (where VBM is probably the easiest point to start).
It may be that I am being to much of a sceptic here (considering the quote
below) and any opposing views are more than welcome.
>
> Kristine
Good luck Jesper
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|