Dear all:
We are interested doing descriptive statistics to compare activations
between two groups of subjects for regions like the superior temporal sulcus
(STS), for example.
We already did a fixed-effect group analysis using SPM99.
We consider using the Marsbar SPM toolbox, and after having read the Marsbar
tutorial, we have many related questions concerning finding the right ROI:
- Should we use a functional or structural ROI (for the STS region)?
It seems that the STS region vary a great deal in position between subjects,
which would suggest to use a subject's own activation pattern to define the
ROI (functionally defined ROI).
- Is this right?
- Is the STS region an area that is (well enough) naturally defined by brain
structure (where functional areas are closely linked to the position of
large an invariant sulci) or not?
- Is the STS a good candidate for anatomical region, the same way the
putamen is?
Since our datasets consists in 8 subjects (4 in each of the two groups) and
we have a single run, sacrificing a subject's run to get independent data
for the definition of a functional ROI seems too big a sacrifice.
- Is that right?
- How to evaluate the importance of the sacrifice ?
If however we use the same data (run) for the ROI definition AND analysis,
we will be biased to have high signal for that run.
- How to evaluate the importance of the bias for this run ?
- Is it possible (resonnably acceptable) to use same data to extract ROI and
do the analysis ?
- If we opt for using structural ROIs, does AAL contains the STS region as
one of its predefined regions?
- If not, then could we define ourselves a "larger" structural regions
enclosing the STS region to account for the between subject variability. If
we do so, what is the consequence (meaning) of the statistics we will obtain?
I'll appreciate any clarifying answers.
Thanks
Yves Roy
|