Dear Chih-Hao,
Assuming you refer to two separate fMRI sessions acquired with one EPI sequence each, and with subjects moving during the break between the two sessions:
MRI images are no exact copies of the objects/subjects under investigation, instead the geometry is just reconstructed. Distortions in magnetic field strength will lead to geometric distortions in the reconstructed image. Different positions inside the scanner will result in somewhat different geometric distortions. However, the usual realignment procedure assumes rigid-body transformations to be sufficient to realign one EPI image onto another. If EPI images differ to a large extent one would need non-rigid transformations to map one image or session onto another. So basically you will have to check your data whether the output from the combined preprocessing looks reasonable or not (I guess it will). It's not just motion between the sessions, other parameters might have been changed as well (e.g. exact orientation of the field of view, encoding directions, possibly leading to systematic differences with regard to geometric distortions, signal drop).
Due to the rigid-body transformations there's another practical limitation, the realignment procedure can't correct for motion during the acquisition of a particular volume, and it can't correct for corresponding signal artefacts associated with "fast motion" like that.
Best
Helmut
|