Dear FSL team,
we have both resting and task-related fMRI data from the same group of subjects. previously we have used ICA on the resting component of the scan to define resting state networks. we then identified some ROI within these networks and, having calculated network integrity using a dual regression approach, we correlated this with BOLD activation to a picture stimulus during during the task part of the scan in the same spatial areas. we found that under certain subject conditions and for certain picture types, there was a strong correlation between resting network integrity and later activation in response to task.
as there were additional areas which activated in response to the task, we wanted to explore further whether there would also be correlations between this activation and what was happening in those same areas during the earlier resting scan. we have therefore defined additional ROI from the activation data. the difficulty is that not all of these ROI are included spatially in any of the resting networks we identified with ICA. so two questions:
1) would it be legitimate to use our new ROI maps, defined spatially from activation data, as "probes" for an additional dual regression (instead of the usual group ICA maps) using the resting data as the raw material (even though these new ROI are not present within any of our resting networks and we not identified with the assistance of the ICA)
2) if yes, would it be necessary to still include in the combined spatial map file, all the "noise" components that WERE identified from the original ICA, so as to ensure that any connections we identify as relating to our ROI represent useful information, rather than reflecting noise in those areas during the resting scan?
any advice appreciated
thanks
joe starke
research assistant
C3NL, imperial college, hammersmith
|