Our automated tools for doing connectivity analysis
(e.g. 3dGroupInCorr) are mostly volume-based (one
seed correlated across the entire volume, not just
against some other seed).
What you are proposing is not a volumetric analysis.
Instead, you would extract all seed time series for
each subject (3dmaskave? 3dROIstats?), and then
correlate them across whichever pairs you care for
(maybe 1ddot could be used for all pairs at once).
Maybe then apply the Fisher z-transformation and
do the group analysis on the result.
How will you generate the seed time series, via
ROIs? Spheres? Voxel coordinates?
- rick