U-Net改进及其在新冠肺炎图像分割的应用

CT成像已成为检测新型冠状病毒肺炎（COVID-19）最重要的步骤之一。针对手动分割患者胸部CT图像中毛玻璃混浊区域繁琐的问题提出了一种自注意力循环残差U型网络模型来实现COVID-19患者肺部CT图像的自动分割，辅助医生诊断。在U-Net模型的基础上引入了循环残差模块和自注意力机制来加强对特征信息的抓取从而提升分割精度。在公开数据集上的分割实验结果显示，该算法的Dice系数、敏感度和特异度分别达到了85.36%、76.64%和76.25%，与其他算法相比具有良好的分割效果。

Improved U-Net and its application in COVID-19 image segmentation

Abstract: Computed tomography (CT) has become one of the most important steps in the detection of the corona virus disease 2019 (COVID-19). Aiming at the cumbersome problem of manually segmenting the ground-glass opacity area in the chest CT image, a U-shaped network model with recurrent ResNet and self-attention is proposed to realize the automatic segmentation in the lung CT images of COVID-19 patients and assist doctors in diagnosis. The improved U-Net with recurrent ResNet and self-attention is introduced to enhance the capture of feature information, thereby improving the accuracy of segmentation. The segmentation experiment on the public data set show that the Dice coefficient, sensitivity and specificity of the proposed algorithm reach 85.36%, 76.64% and 76.25%, respectively. Compared with other algorithms, the proposed algorithm has better segmentation performance.