i have seen a paper where No Adaptation of abdominal MRI to CT achieve average dice score of 58.2, can you clarify why the performance vary so much with your paper

The paper is Unsupervised Bidirectional Cross-Modality Adaptation via Deeply Synergistic Image and Feature Alignment for Medical Image Segmentation, which is published at IEEE TRANSACTIONS ON MEDICAL IMGAGING in 2020
