Thanks for your prompt reply. I think we have been doing it step-by-step because we had a lot of fails at the initial registration stage and needed to optimise things. We have been doing the BET brain extraction step using -f -R and -c options optimised for our scans followed by first_flirt using the -b option for a brain extracted input. We were then doing the first segmentation on that output followed by boundary correction using a -p threshold for clean up which worked well with our scans.
This seems to be working well, but we had been using the old version of fsl (4.0.4). It is obviously more time consuming and we have upgraded to version 4.1.5 so is it something we could do with the run_first_all script now?