Table 2 Segmentation results of IMT images on BraTs2015 evaluated by FCN-score. The gap between translated images and the real images can evaluate the generation performance of our method. Note that “all” represents mean accuracy of all pixels (the meanings of “all” are the same in the following tables). We achieve close segmentation results between translated-modality images and target-modality images.

From: MRI Cross-Modality Image-to-Image Translation