Hi all!
I've a methodological question about multi-subject RSN analysis with Dual Regression.
I've two groups (e.g. paients and controls). First I run group-based TICA on all subjects' data (patients and controls together) to get the “group-average” ICA maps. Then I need to relate group maps back to the individual subjects, so I run the dual regression script. After collection of the individual maps associated with group maps, I can perform a voxel-wise test between the two groups (e.g. controls > patients).
My question is: what if there is a "group-average" component comes from only the control group and this component is not presented in the patient group at all? This time, when we regress this spatial IC into each patient’s 4D data to find subject-specific timecourses, we can get "useless" timecourses, as this component is not presented in the patient group at all. Then we will probable get spatial maps having nothing to do with the original group-average map, and there will be a correspondence problem between patients' and controls' IC maps.
Could anybody help me to clarify this issue?
Thanks a lot,
Gábor
|