Dear John,
Unwarping based on a field map corrects geometric distortions in the EPI data, and thus it should facilitate coregistration of structural and functional data. This doesn't have to do anything with the following preprocessing steps though. Of course, if you rely on a more advanced and time-consuming normalisation algorithm like Dartel it might be a waste of time to generate a high-resolution template based on the structural data and corresponding deformation fields for normalisation purpose if the coregistration between structures and functions is poor anyway.
Whether a field map is useful or not is going to depend on your data and the geometric distortions. In any case subjects should not move between the field map sequence and the EPI sequence, thus possibly several field maps for different EPI runs would be needed. Also keep in mind that a field map can only account for deformations due to B0 inhomogenities, it can't recover signal in regions with signal loss, and it also doesn't correct for intensity inhomogenities due to B1 effects, which all affect the quality of the coregistration. If you have doubts about the quality of the coreg step you might want to turn to boundardy-based registration as implemented in FSL, it has been reported to be more accurate (see Greve & Fischl, 2009, Neuroimage, https://dx.doi.org/10.1016/j.neuroimage.2009.06.060 ), although this might also depend on the exact settings.
But maybe I misunderstood you, maybe you want to use Dartel to create tissue probability maps / templates just based on the EPI data (so without relying on structural volumes?).
Best
Helmut
|