Kidney cancer is one of the most common malignancies of the urinary system. To detect kidney cancer, we must identify kidney tumors, which can vary significantly in size, shape, and biological behavior, ranging from benign lesions to aggressive malignant tumors that require timely diagnosis and treatment. Accurate identification and segmentation of kidney tumors on CT or MRI scans are essential for clinical decision-making, surgical planning, and prognosis assessment.
We have recently been developing a framework called Green U-shaped Learning (GUSL) for kidney and tumor segmentation. This is a two-stage framework. In Stage 1, we performed segmentation of the kidney organ, and in Stage 2, we cropped the kidney region as the region of interest (ROI) for tumor segmentation. We employ our GUSL framework on both stages, which could achieve fine-to-coarse feature extraction and coarse-to-fine residual correction.
