【摘要】 为了给患者下一步诊疗提供判断依据，研究了利用CT图像对肝脏肿瘤区域进行自动分割，提出一种新的深度神经网络Attention-ResUNet。该网络重新设计U-Net的编解码结构，在特征提取模块中结合残差模块来加强了特征映射，并利用通道注意力机制和空间注意力机制对特征重新标定，增强有效特征，使得特征能够高效传输。实验表明，Attention-ResUNet在3D-IRCADb数据集和LiTS数据集均能取得接近标注数据的分割性能。更多还原

【Abstract】 To automatically segment liver tumors using CT images and provide a basis for subsequent diagnosis and treatment, a new deep neural network called Attention-ResUNet is proposed, in which the encoder-decoder structure of U-Net is redesigned, the feature mapping is enhanced by incorporating residual module into the feature extraction module, and the extracted features are recalibrated by using the channel attention mechanism and the spatial attention mechanism. It enhances the effective features and improves the features transmission. Experiments show that Attention-ResUNet achieves segmentation performance close to annotated data on both 3D-IRCADb dataset and LiTS dataset.更多还原