Hi Paul,
Thank you very much for your thorough reply. To sum it up: you would advice to detrend first, then to cut the time series corresponding to each video clip, demean, reorder and concatenate them and run the preprocessing on the resulting data series. For the preprocessing, the code from the NI paper can be used, but -regress_polort 0 should be added. The resulting output (errts* file) can then be used to compute the inter subject correlation amongst subjects, but I would use the option -polort -1 as the data has been preprocessed already. Is this correct?
There is one more thing that occurred to me: My a priori defined regions of interest are subcortical. Would it still be recommendable to do FS surface reconstruction using the FS command recon-all -all or would it be better to replace it with the -autorecon1 and -autorecon2 option instead?
Many thanks,
Stef