Figure 5
Flowchart of our approach for cross-modality segmentation. First, we input a given-modality image to our IMT network to generate a translated-modality image. For instance, given a T1 image, \(\widehat{T}2\) images can be generated with our method. Second, two identical given-modality images and one corresponding translated-modality image are fed to channels 1, 2, and 3 and segmented by FCN networks. Under the standard FCN-32s, standard FCN-16s, and standard FCN-8s settings, we output our segmentation results. The red box indicates our translated images.
