Dear SPMers,
I would be grateful for some advice. In a 15 run PET study in normals, I
identified a ROI (R) engaged during the active task C [C- (A + B)]. Using a
mixed effects model, I have demonstrated that in PATIENTS, the active task C
[C- (A + B)] is associated with activation in the contralateral hemisphere
in ROI R1. What I would like to do next is to look for a correlation between
performance of task C by individual patients and activity in ROI R1.
I have 2 behavioural measures taken on line, with values for each run, and 5
measures taken off line with a single value for each subject. R1 is the pars
opercularis which approximates neither to a sphere nor to a box.
I have looked for a correlation between the first 2 measures and activation
in R1 using multisubject covariates only and used a SVC, but the P values
are still very small.
1. In view of my specific hypothesis about R1, is it appropriate to be using
SPM, or should I be doing a ROI analysis outside SPM & averaging the voxel
values across my ROI?
2. If SPM is appropriate, how do I extract a physiological value for the
contrast [C - (A + B)] for each subject to use in a correlation analysis?
3. Reading through the archives, I wondered if the first eigenvariate is the
appropriate measure? If so, should I be using a value for a cluster within
my ROI/ an average of all voxels within the ROI or a peak voxel?
4. Practically speaking, how can I extract data from within my VOI only (
allowing for the fact that it will have to be an imported masked image).
5. What has happened to the y function in matlab? Has this been superceeded
by eigenvariates and Y?
Many thanks for your help.
|