CC-DenseUNet: Densely connected U-Net with criss-cross attention for liver and tumor segmentation in CT volumes
Abstract—The automatic segmentation of liver and tumor is important for hepatic tumor surgery. In this paper, we propose a novel densely connected U-Net (CC-DenseUNet), which integrates criss-cross attention (CCA) module, to segment the liver and tumor in computed tomography (CT) volumes. The dense interconnections in CC-DenseUNet ensure the maximum information flow between encoder layers when extracting intraslice features of liver and tumors. Moreover, the CCA module is used in CC-DenseUNet to efficiently capture only the necessary and meaningful non-local contextual information of CT images containing liver or tumors. We evaluated the proposed CCDenseUNet on the Liver Tumor Segmentation Challenge and 3DIRCADb datasets. Experimental results show that our method outperformed the state-of-the-art methods in liver tumor segmentation and achieved a highly competitive performance in liver segmentation.
