Edge Constraint and Location Mapping for Liver Tumor Segmentation from Nonenhanced Images

Figure 2

Our ECLMS model is composed of localization network and dual-branch segmentation network. The tumor location maps generated by the localization network, which is used as the guide information to be fused into the segmentation network. The main decoding branch and edge decoding branch in dual-branch segmentation network can be mutually restrained and jointly optimized in the direction of ground truth.