afni.nimh.nih.gov

That approach seems fine, if it is just to show the
average response over those ROIs. Note that one
small danger is over-modelling. A fast event design
cannot always handle replacing 1 or 2 basis functions
with 8. But you can see how it goes.

- rick