>Could I now baseline correct/detrend the data in each slice of each trial
>independently

Quite some time ago I wrote an AFNI plug-in for my own use that implements the
sort of detrending that you want to do. It operates on a single BRIK file
produced by concatenating the data files from the separate runs. Each run
within this file is detrended separately; the results can then be lumped
together for further analysis (keeping in mind the slight residual
discontinuities at the boundaries between runs).

I never released this plug-in, since I hadn't invested any effort in
comprehensive documentation, the code needed some cleaning up which I never
got round to, and I was sceptical as to its general usefulness. I think
"3dDeconvolve -concat" does a better job of everything that my old plugin used
to do. If the plug-in code might be useful for you to have, though, you're
welcome to it; contact me out-of-band.

--
Matthew Belmonte <belmonte@mit.edu>
Autism Research Centre
University of Cambridge
Douglas House
18b Trumpington Road
Cambridge CB2 2AH
UK