Cascaded Sematic Segmentation for Kidney and Masses

Ruize Shi, Bing Han

Abstract: The segmentation for kidney and related masses is meaningful for the clinical surgery and treatment. In this paper, we use a cascaded method to locate the kidney area and segment the kidney as well as masses accurately based on 3D U-Net architecture. The region based segmentation is used in the training process to make result consistent with the evaluation criteria. The test result with Dice value is 0.9736, 0.8416 and 0.7971 for Kidney + Masses, Masses and Tumor respectively.