基于反注意力机制U-Net网络的胃部肿瘤分割  

Gastric tumor segmentation by U-Net based on reverse attention mechanism

作　　者：王萍 徐凯成 张一弛 王海玲 蔡清萍 卫子然 胡尊琪 WANG Ping;XU Kaicheng;ZHANG Yichi;WANG Hailing;CAI Qingping;WEI Ziran;HU Zunqi(School of Continuing Education,Shanghai University of Engineering Science,Shanghai 201620,China;School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China;Department of Gastrointestinal Surgery,Shanghai Changzheng Hospital,Shanghai 200003,China)

机构地区：[1]上海工程技术大学继续教育学院,上海201620 [2]上海工程技术大学电子电气工程学院,上海201620 [3]上海长征医院肠胃外科,上海200003

出　　处：《中国医学物理学杂志》2022年第9期1133-1139,共7页Chinese Journal of Medical Physics

基　　金：上海市科学技术委员会科研计划项目(18411952800)。

摘　　要：首先利用全局与局部注意力对肿瘤进行定位,然后在模型中加入反注意力机制,将显著特征从原特征图中消除,并保留肿瘤的边缘轮廓信息。此外还在模型中使用深度监督,监督各个深度解码层的训练,有效抑制模型梯度消失现象,提高分割的准确性。本研究使用的是上海长征医院的胃部CT数据集,并将提出的模型与U-Net、Attention U-Net和ET-Net的实验对比。研究结果表明,相较于传统的U-Net网络模型,基于反注意力机制的U-Net模型在胃部肿瘤分割中性能得到了较大的提高,证明了该网络模型的有效性。The global and local attention mechanisms are used to localize tumor,and a reverse attention mechanism is added to the model to remove the salient features from the original feature map while retaining the edge contour information.In addition,deep supervision is also applied to supervise the training of each deep decoding layer,which effectively suppresses gradient disappearance and enhances segmentation accuracy.The gastric CT data set used in the study is from Shanghai Changzheng Hospital.The performance of U-Net model with reverse attention mechanism in gastric tumor segmentation has been greatly improved when compared with the traditional U-Net networks(U-net,Attention U-net and ET-Net),which proves the effectiveness of the proposed model.

关 键 词：胃部肿瘤分割 深度学习 图像处理 反注意力机制 U-Net网络 

分 类 号：R318[医药卫生—生物医学工程] TP391[医药卫生—基础医学]