摘要
目的:将一款基于深度学习的危及器官自动勾画软件系统DeepViewer应用于临床,实现自动勾画肿瘤患者治疗计划中危及器官的功能。方法:DeepViewer使用改进后的全卷积神经网络U-Net来实现自动勾画患者CT扫描部位所包含的危及器官,并使用Dice相似性系数(DSC)对比分析这22种危及器官自动勾画与手动勾画的差异。结果:11种危及器官DSC平均值在0.9以上,5种危及器官DSC平均值为0.8~0.9,5种器官DSC平均值为0.7~0.8,视交叉DSC平均值最低,为0.676。总体结果表明DeepViewer系统能够较准确地自动勾画出危及器官,特别是左、右肺、膀胱、脑干等器官,已基本满足临床需求。结论:DeepViewer软件系统可以实现放疗肿瘤患者危及器官的自动勾画,准确性较高。同时,DeepViewer系统勾画完毕后,可以通过网络系统自动传输RTStructure DICOM3.0文件,无需其他操作,能极大地提高临床医生工作效率,降低治疗计划流程中的勾画总时间。
Objective To realize the automatic segmentation of organs-at-risk(OAR)by DeepViewer system which was an automatic OAR segmentation software based on deep learning.Methods The OAR in CT images was automatically segmented by DeepViewer system via improved U-Net convolutional neural network.Dice similarity coefficient(DSC)was used to compare the differences between the automatic segmentations and manual segmentations of 22 OAR.Results There were 11 OAR with an average DSC above 0.9,5 OAR with an average DSC of 0.8-0.9,and 5 OAR with an average DSC of 0.7-0.8.The mean DSC of optic chiasma was the lowest(0.676).The overall results showed that DeepViewer system could be used to realize the automatic segmentation of OAR,especially left lung,right lung,bladder,brain stem and so on,which basically met the clinical requirements.Conclusion DeepViewer system can be used to automatically segment OAR in tumor patients,with a high accuracy.Meanwhile,RTStructure DICOM3.0 files can be transmitted automatically through the network after the segmentation by DeepViewer,thus greatly facilitating the workflow of clinicians and shortening the total segmentation time in treatment planning.
作者
汪志
常艳奎
吴昊天
张键
徐榭
裴曦
WANG Zhi;CHANG Yankui;WU Haotian;ZHANG Jian;XU Xie;PEI Xi(Center of Radiological Medical Physics,University of Science and Technology of China,Hefei 230025,China;Department of Radiation Oncology,the First Affiliated Hospital of Anhui Medical University,Hefei 230022,China;Anhui Wisdom Technology Co.,Ltd.,Hefei 230088,China)
出处
《中国医学物理学杂志》
CSCD
2020年第8期1071-1075,共5页
Chinese Journal of Medical Physics
基金
国家重点研发计划(2017YFC0107500)
安徽省自然科学基金(1908085MA27)
国家自然科学基金(11575180)
安徽省重点研究与开发计划(1804a09020039)。
关键词
深度学习
危及器官
自动勾画
放射治疗
deep learning
organ-at-risk
automatic segmentation
radiotherapy
