Thanks for the suggestions.

I was reading one of the example scripts ('@DDmb_analyze '), and it suggests '3dTcat' to concatenate all the runs and the concatenated output of '3dTcat' be used as input to 3dDeconvolve. The questions I have is, Should I detrend each run before I concatenate all the runs from different sessions and use the concatenated file as input to 3dDeconvolve?

The reason I am asking this is, the page for correlation analysis [afni.nimh.nih.gov]
says, i should detrend each run and concatenate across runs to construct the seed.1D.

But,the following lines from example script ('@DDmb_analyze') doesn't say if I should detrend before I concatenate runs across sessions and use the concatenated file as input to 3dDeconvolve?
3dTcat -prefix DDrall \
DDr1+orig'[2..109]' DDr2+orig'[2..109]' \
DDr3+orig'[2..109]' DDr4+orig'[2..109]'

Is it not a problem if I construct my interaction regressor by concatenating detrended data from runs across sessions(seed.1D) but not the input dataset (just concatenating runs across sessions with no detrending) to the 3dDeconvolve?

Thanks a lot for the help