We have a dataset where most of BOLD data have a corresponding fieldmap to run sdc with fmriprep. I have noticed that syn-sdc is much much worse than fieldmap-sdc. So I am thinking to use a field map acquired from the same subject for another fMRI scan (during the same session, or on different sessions) to correct for distortions rather than using syn-sdc. Has anybody tried this before? Is it ok to borrow a field map from the same session (or even from another session from 1 year ago) to run sdc in fmriprep?
My limited understanding is that fieldmaps are acquired before each fMRI scan to have the exact head position that will be in fMRI, but after all the head can move during fMRI itself, so I am not sure to what degree a field map obtained from the same subject at another time can be used for the current scan.
A more extreme idea is the use of a field map from another subject, how bad would that idea be?