The fsfast stream defaults to using the middle time point of each run as a template for both motion correction and registration. The raw data are then sampled into a common space prior to analysis so that they are in the same space when the analysis combines them across run. It is possible to use the first time point of the first run as the template for all of the runs (so motion correction is done to this template and registration is done only once). You can also keep the analysis in the native functional space. doug

ps. I'm surprised that bbregister is failing. If you upload the subject data and a functional run to our filedrop (url below) I'll take a look.

SHAHIN NASR wrote:

Hi In monkey fmri, we usually have to do the functional-anatomical registration manually since bbregister doesn't work very well. It was easy in version 4.5 since we usually used just one run for registration. In version 5.1 we can use per-run option to improve registration but I am wondering how can we use it in monkey fmri? Should we register each run separately? Then how can we combine those individual registration files to generate the final registration file.