“…In recent years, GAN-based methods were demonstrated to offer state-of-theart performance in numerous synthesis tasks, including data augmentation as well as multi-modal synthesis [24], [34], [71], [72]. Important applications of GAN models include CT to PET [73], [74], MR to CT [75]- [77], unpaired crossmodality [78]- [81], 3T-to-7T [82], [83], and multi-contrast MRI synthesis [24], [30]- [42].…”