Inquiry on FID for Evaluating Unpaired 3D Image Translation

[aeinkoupaei]

I'm interested in using FID to assess the quality of 3D medical images translated from CT to MRI.
Since the images are unpaired, with no exact correspondence between CT and MRI slices, comparing them directly poses a challenge. The translated images often blend content from the source domain (CT) with the style of the target domain (MRI), further complicating evaluation.
Could you guide me on how to apply FID in a standardized and meaningful way for Evaluating Unpaired 3D Image Translation?