基于有限训练样本的融合网络模型用于盆腔危及器官自动分割的研究  

A fusion network model based on limited training samples for the automatic segmentation of pelvic endangered organs

作　　者：吴青南 王运来[3] 全红[2] 王俊杰[4] 谷珊珊[3] 杨薇[3] 葛瑞刚[3] 刘杰 鞠忠建[3] WU Qingnan;WANG Yunlai;QUAN Hong;WANG Junjie;GU Shanshan;YANG Wei;GE Ruigang;LlU Jie;JU Zhongjian(Department of Radiation Oncology,Peking University International Hospital,Beijing 102206,P.R.China;School of Physics Science and Technology,Wuhan University,Wuhan 430072,P.R.China;Department of Radiation Oncology,People's Liberation Army General Hospital,Beijing 100853,P.R.China;Department of Radiation Oncology,Peking University Third Hospital,Beijing 100191,P.R.China;Beijing Oriental Ruiyun Technology Corporation,Beijing 100020,P.R.China)

机构地区：[1]北京大学国际医院放疗科,北京102206 [2]武汉大学物理科学与技术学院,武汉430072 [3]中国人民解放军总医院放疗科,北京100853 [4]北京大学第三医院放疗科,北京100191 [5]北京东方瑞云科技有限公司,北京100020

出　　处：《生物医学工程学杂志》2020年第2期311-316,共6页Journal of Biomedical Engineering

基　　金：数字诊疗装备研发项目基金(2016YFC0105715);国家自然科学基金(61671204)。

摘　　要：将深度学习应用到医学影像中危及器官自动分割领域时,为解决训练样本不足时三维卷积神经网络优化出现的退化、梯度消失等问题,本研究将Dense Net与V-Net两个网络模型进行融合,开发一种用于三维计算机断层扫描(CT)图像自动分割的Dense V-Network算法,勾画女性盆腔危及器官。采用戴斯相似性系数(DSC)、豪斯多夫距离(HD)、杰卡德距离(JD)三个参数来定量评估分割效果。结果显示膀胱、小肠、直肠、股骨头和脊髓自动分割的DSC值均在0.87以上(平均值是0.9);JD值均在2.3以内(平均值是0.18);除小肠外,HD值均在0.9 cm以内(平均值是0.62 cm)。经验证,Dense V-Network网络可精准地勾画盆腔危及器官。When applying deep learning to the automatic segmentation of organs at risk in medical images, we combine two network models of Dense Net and V-Net to develop a Dense V-network for automatic segmentation of three-dimensional computed tomography(CT) images, in order to solve the problems of degradation and gradient disappearance of three-dimensional convolutional neural networks optimization as training samples are insufficient. This algorithm is applied to the delineation of pelvic endangered organs and we take three representative evaluation parameters to quantitatively evaluate the segmentation effect. The clinical result showed that the Dice similarity coefficient values of the bladder, small intestine, rectum, femoral head and spinal cord were all above 0.87(average was 0.9);Jaccard distance of these were within 2.3(average was 0.18). Except for the small intestine, the Hausdorff distance of other organs were less than 0.9 cm(average was 0.62 cm). The Dense V-Network has been proven to achieve the accurate segmentation of pelvic endangered organs.

关 键 词：深度学习 多模型融合 卷积神经网络 自动分割 危及器官勾画 

分 类 号：R816[医药卫生—放射医学] TP391.41[医药卫生—临床医学] TP18[自动化与计算机技术—计算机应用技术]