[1]刘剑,张峥,牛兵,等.基于以数据为中心的多任务学习及切片尺寸对肺结节分割性能的影响[J].中国医学物理学杂志,2025,42(10):1306-1320.[doi:DOI:10.3969/j.issn.1005-202X.2025.10.007]
　LIU Jian,ZHANG Zheng,NIU Bing,et al.Effects of data-centric multi-task learning with larger patch sizes on pulmonary nodule segmentation performance[J].Chinese Journal of Medical Physics,2025,42(10):1306-1320.[doi:DOI:10.3969/j.issn.1005-202X.2025.10.007]
点击复制

基于以数据为中心的多任务学习及切片尺寸对肺结节分割性能的影响()

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
42
期数:
2025年第10期
页码:
1306-1320
栏目:
医学影像物理
出版日期:
2025-10-29

文章信息/Info

Title:
Effects of data-centric multi-task learning with larger patch sizes on pulmonary nodule segmentation performance
文章编号:
1005-202X（2025）10-1306-15
作者:
刘剑1; 张峥2; 牛兵3; 康帅3; 任娟3; 王雷4; 徐凯1
1.上海交通大学机械与动力工程学院， 上海 200241； 2.同济大学附属上海市肺科医院放射科， 上海 200433； 3.上海睿触科技有限公司， 上海 201600； 4.同济大学附属上海市肺科医院肿瘤科， 上海 200433
Author(s):
LIU Jian1;  ZHANG Zheng2;  NIU Bing3;  KANG Shuai3;  REN Juan3;  WANG Lei4;  XU Kai1
1. School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200241, China 2. Department of Radiology, Shanghai Pulmonary Hospital, Tongji University, Shanghai 200433, China 3. Shanghai Simple Touch Technology Co., Ltd., Shanghai 201600, China 4. Department of Oncology, Shanghai Pulmonary Hospital, Tongji University, Shanghai 200433, China
关键词:
肺结节; 多任务学习; 深度学习; 分割性能
Keywords:
Keywords: pulmonary nodule multi-task learning deep learning segmentation performance
分类号:
R318；R445.3；TP181
DOI:
DOI:10.3969/j.issn.1005-202X.2025.10.007
文献标志码:
A
摘要:
针对现有公开数据集缺乏肺部关键器官、组织标注的问题，收集863例胸部CT扫描图像，采用计算机视觉算法与放射科医生手动校正的半自动化方法，构建首个含肺血管、气道及肺结节标注的综合数据集。在此基础上，提出基于多任务学习的肺结节分割模型，通过加入肺血管（肺动静脉）、气道标注增强模型学习肺部特征能力，降低肺结节假阳性问题，提高泛化能力，并采用较大图像块的方式进一步优化提升模型性能。训练得到的VAAN_128模型综合表现最优，在肺结节分割任务中Dice系数为0.694，假发现率为0.210，并且能同时提供准确的肺血管、气管分割结果，协助制定更精确的诊疗方案。基于VAAN_128模型，设计开发穿刺手术导航定位系统软件，在临床操作中辅助医生准确定位肺结节，区分关键组织，提升术前规划效率，为肺部疾病早期诊断、病情监测等提供精准、高效技术支持，对于临床导航系统的路径规划以及未来肺部成像研究具有重要意义。
Abstract:
Abstract: Given the lack of annotations for key lung organs and tissues in existing public datasets, this study collected 863 cases of chest CT scan images and constructed the first comprehensive dataset containing annotations of pulmonary vessels, airways, and nodules using a semi-automated method that combines computer vision algorithms with manual corrections by radiologists. On this basis, a lung nodule segmentation model based on multi-task learning is proposed. By incorporating annotations of pulmonary vessels (pulmonary arteries and veins) and the trachea to enhance models ability to learn lung features, the proposed model reduces the false discovery rate in lung nodule detection, and improves generalization ability. Additionally, the use of larger image patches further optimizes model performance. The trained VAAN_128 model achieves the best performance, with a Dice coefficient of 0.694 and a false discovery rate of 0.210 for lung nodule segmentation. Moreover, it simultaneously provides accurate segmentation results of pulmonary vessels and the trachea, assisting in the formulation of more precise diagnosis and treatment plans. Based on the VAAN_128 model, a software system for navigation and localization in biopsy procedures is developed. In clinical practice, this system can assist physicians in accurately locating lung nodules, distinguishing critical tissues, and improving preoperative planning efficiency. This provides precise and efficient technical support for early diagnosis and disease monitoring of lung diseases, and is of great significance for path planning in clinical navigation system and future lung imaging research.

相似文献/References:

[1]唐思源,刘燕茹,杨敏,等.基于CT图像的肺结节检测与识别[J].中国医学物理学杂志,2019,36(7):800.[doi:DOI:10.3969/j.issn.1005-202X.2019.07.011]
　TANG Siyuan,LIU Yanru,YANG Min,et al.Detection and recognition of pulmonary nodules based on CT images[J].Chinese Journal of Medical Physics,2019,36(10):800.[doi:DOI:10.3969/j.issn.1005-202X.2019.07.011]
[2]张倩雯,陈明,秦玉芳,等.基于3D ResUnet网络的肺结节分割[J].中国医学物理学杂志,2019,36(11):1356.[doi:DOI:10.3969/j.issn.1005-202X.2019.11.021]
　ZHANG Qianwen,CHEN Ming,QIN Yufang,et al.Lung nodule segmentation based on 3D ResUnet network[J].Chinese Journal of Medical Physics,2019,36(10):1356.[doi:DOI:10.3969/j.issn.1005-202X.2019.11.021]
[3]曹宇,邢素霞,逄键梁,等.基于改进的VGG-16卷积神经网络的肺结节检测[J].中国医学物理学杂志,2020,37(7):940.[doi:DOI:10.3969/j.issn.1005-202X.2020.07.026]
　CAO Yu,XING Suxia,PANG Jianliang,et al.Detection of pulmonary nodules based on improved VGG-16 convolution neural network[J].Chinese Journal of Medical Physics,2020,37(10):940.[doi:DOI:10.3969/j.issn.1005-202X.2020.07.026]
[4]王乾梁,石宏理.基于改进YOLO V3的肺结节检测方法[J].中国医学物理学杂志,2021,38(9):1179.[doi:10.3969/j.issn.1005-202X.2021.09.024]
　WANG Qianliang,SHI Hongli,et al.Pulmonary nodule detection based on improved YOLO V3[J].Chinese Journal of Medical Physics,2021,38(10):1179.[doi:10.3969/j.issn.1005-202X.2021.09.024]
[5]刘雲,王一达,张成秀,等.基于深度学习结合解剖学注意力机制的肺结节良恶性分类[J].中国医学物理学杂志,2022,39(11):1441.[doi:DOI:10.3969/j.issn.1005-202X.2022.11.019]
　LIU Yun,WANG Yida,ZHANG Chengxiu,et al.Classification of benign and malignant pulmonary nodules by deep learning with anatomy-based attention mechanism[J].Chinese Journal of Medical Physics,2022,39(10):1441.[doi:DOI:10.3969/j.issn.1005-202X.2022.11.019]
[6]王新宇,赵静文,刘翔,等.人工智能在肺结节筛查和肺癌诊断中的应用[J].中国医学物理学杂志,2023,40(9):1182.[doi:DOI:10.3969/j.issn.1005-202X.2023.09.020]
　WANG Xinyu,ZHAO Jingwen,LIU Xiang,et al.Applications of artificial intelligence in lung nodule detection and lung cancer diagnosis[J].Chinese Journal of Medical Physics,2023,40(10):1182.[doi:DOI:10.3969/j.issn.1005-202X.2023.09.020]
[7]吴书裕,周露,王琳婧,等.基于BIRADs多任务学习模型的乳腺肿块分类[J].中国医学物理学杂志,2023,40(10):1220.[doi:DOI:10.3969/j.issn.1005-202X.2023.10.005]
　WU Shuyu,ZHOU Lu,WANG Linjing,et al.Breast mass classification based on BIRADs multi-task learning model[J].Chinese Journal of Medical Physics,2023,40(10):1220.[doi:DOI:10.3969/j.issn.1005-202X.2023.10.005]
[8]温帆,杨萍,张鑫,等.基于特征增强的多分支U-Net肺结节分割[J].中国医学物理学杂志,2023,40(11):1343.[doi:DOI:10.3969/j.issn.1005-202X.2023.11.005]
　WEN Fan,YANG Ping,ZHANG Xin,et al.Pulmonary nodule segmentation using multi-branch U-Net based on feature enhancement[J].Chinese Journal of Medical Physics,2023,40(10):1343.[doi:DOI:10.3969/j.issn.1005-202X.2023.11.005]
[9]刘涌涛,王宝珠,郭志涛.基于改进YOLOv7网络模型的肺结节检测算法[J].中国医学物理学杂志,2023,40(12):1509.[doi:DOI:10.3969/j.issn.1005-202X.2023.12.009]
　LIU Yongtao,WANG Baozhu,GUO Zhitao.Lung nodule detection algorithm using improved YOLOv7 network model[J].Chinese Journal of Medical Physics,2023,40(10):1509.[doi:DOI:10.3969/j.issn.1005-202X.2023.12.009]
[10]牛树国,周福兴,颜克松,等.不同CT阈值下实性成分占比对小肺癌浸润性预测的影响[J].中国医学物理学杂志,2024,41(3):323.[doi:DOI:10.3969/j.issn.1005-202X.2024.03.009]
　NIU Shuguo,ZHOU Fuxing,YAN Kesong,et al.Predictive value of consolidation/tumor ratio at different CT thresholds for invasiveness in small lung cancer[J].Chinese Journal of Medical Physics,2024,41(10):323.[doi:DOI:10.3969/j.issn.1005-202X.2024.03.009]

备注/Memo

备注/Memo:
【收稿日期】2025-03-10 【基金项目】国家自然科学基金（82141101，82102766）；上海市卫生健康委协同创新项目（2020CXJQ02）；上海市肺科医院重点培育项目（fkzr2001） 【作者简介】刘剑，博士研究生，研究方向：医疗机器人、影像学、机器学习，E-mail: maoforest@163.com 【通信作者】徐凯，博士，教授，研究方向：医疗机器人、特种工业机器人、服务业机器人、机器人学理论，E-mail: k.xu@sjtu.edu.cn；王雷，博士，副主任医师，研究方向：肿瘤学与影像学，E-mail: wangleixxxn@163.com
更新日期/Last Update: 2025-10-29