[1]张福全,周永康,杨悦,等.可分离卷积与注意力机制结合的肝癌靶区CT图像自动分割[J].中国医学物理学杂志,2025,42(7):918-922.[doi:DOI:10.3969/j.issn.1005-202X.2025.07.011]
　ZHANG Fuquan,ZHOU Yongkang,YANG Yue,et al.Automatic liver cancer target area segmentation in CT image using separable convolution and attention mechanism[J].Chinese Journal of Medical Physics,2025,42(7):918-922.[doi:DOI:10.3969/j.issn.1005-202X.2025.07.011]
点击复制

可分离卷积与注意力机制结合的肝癌靶区CT图像自动分割()

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

卷:
42
期数:
2025年第7期
页码:
918-922
栏目:
医学影像物理
出版日期:
2025-07-25

文章信息/Info

Title:
Automatic liver cancer target area segmentation in CT image using separable convolution and attention mechanism
文章编号:
1005-202X（2025）07-0918-05
作者:
张福全; 周永康; 杨悦; 陈芷涵; 高宇楠; 林小惟
复旦大学附属中山医院放射治疗科，上海 570228
Author(s):
ZHANG Fuquan;  ZHOU Yongkang;  YANG Yue;  CHEN Zhihan;  GAO Yunan;  LIN Xiaowei
Department of Radiotherapy, Zhongshan Hospital, Fudan University, Shanghai 570228, China
关键词:
肝癌; 深度学习; 卷积神经网络; 靶区分割; 注意力机制
Keywords:
liver cancer;  deep learning;  convolutional neural network;  target area segmentation;  attention mechanism
分类号:
R318；R811.1
DOI:
DOI:10.3969/j.issn.1005-202X.2025.07.011
文献标志码:
A
摘要:
目的：开发一种可分离卷积与注意力机制结合的肝癌靶区CT图像分割方法。方法：基于U-Net卷积神经网络的肝癌靶区分割方法，为增强分割模型的特征表达能力，将注意力模块与U-Net模型相结合，提高与分割任务相关性更大的特征通道权重；在网络模型编码阶段引入本文提出的可分离卷积，补充下采样过程中损失的特征信息。结果：在复旦大学附属中山医院50例肝癌患者数据集上进行靶区分割，实验结果表明，相比于已有方法，本文提出方法的分割平均戴斯相似系数比3D U-Net提高4.54%。结论：基于可分离卷积和注意力机制的U-Net卷积神经网络对肝癌靶区达到更好的分割精度，有望在临床应用中提高医生的工作效率。
Abstract:
Abstract: Objective To propose an approach integrating separable convolution and attention mechanism for automaticallysegmenting target area in radiotherapy for liver cancer from CT images. Methods The novel liver cancer segmentationmethod based on the U-Net convolutional neural network integrated attention block and U-Net model to increase the weightsof feature channels with greater relevance to segmentation tasks, thereby enhancing the feature expression ability, and aseparable convolution was used in the U-Net encoding stage for supplementing feature information lost during thedownsampling process. Results Target area segmentation was conducted on a dataset of 50 liver cancer patients at ZhongshanHospital, Fudan University. Experimental results showed that the proposed method improved the average Dice similaritycoefficient by 4.54% as compared with 3D U-Net. Conclusion The U-Net convolutional neural network based on separableconvolution and attention mechanism achieves better segmentation accuracy for liver cancer target area, which is expected toimprove the work efficiency of doctors in clinic.

相似文献/References:

[1]林拥华,章宝燕,李新丰,等.肝癌介入治疗中旋转数字减影血管造影三维重建技术的应用价值[J].中国医学物理学杂志,2015,32(06):851.[doi:doi:10.3969/j.issn.1005-202X.2015.06.019]
　[J].Chinese Journal of Medical Physics,2015,32(7):851.[doi:doi:10.3969/j.issn.1005-202X.2015.06.019]
[2]陶源,王佳飞,杜俊龙,等.基于卷积神经网络的细胞识别[J].中国医学物理学杂志,2017,34(1):53.[doi:10.3969/j.issn.1005-202X.2017.01.011]
　[J].Chinese Journal of Medical Physics,2017,34(7):53.[doi:10.3969/j.issn.1005-202X.2017.01.011]
[3]朴俊杰,徐寿平,段学章,等.CyberKnife中Fixed/Iris准直器对肝癌患者治疗计划的影响及评价[J].中国医学物理学杂志,2017,34(8):762.[doi:DOI:10.3969/j.issn.1005-202X.2017.08.002]
[4]李柱,乔安意,王翀,等. 超声引导下射频消融术治疗肝癌疗效及预后影响因素分析[J].中国医学物理学杂志,2017,34(10):1058.[doi:DOI:10.3969/j.issn.1005-202X.2017.10.018]
　[J].Chinese Journal of Medical Physics,2017,34(7):1058.[doi:DOI:10.3969/j.issn.1005-202X.2017.10.018]
[5]裴宇. 多层螺旋CT灌注参数评价肝癌经动脉化疗栓塞术后疗效[J].中国医学物理学杂志,2018,35(2):191.[doi:DOI:10.3969/j.issn.1005-202X.2018.02.014]
　PEI Yu. Application of multi-slice spiral CT perfusion parameters in the evaluation of curative effect of TACE for hepatocellular carcinoma[J].Chinese Journal of Medical Physics,2018,35(7):191.[doi:DOI:10.3969/j.issn.1005-202X.2018.02.014]
[6]门阔,戴建荣. 利用深度反卷积神经网络自动勾画放疗危及器官[J].中国医学物理学杂志,2018,35(3):256.[doi:DOI:10.3969/j.issn.1005-202X.2018.03.002]
　MEN Kuo,DAI Jianrong. Automatic segmentation of organs at risk in radiotherapy using deep deconvolutional neural network[J].Chinese Journal of Medical Physics,2018,35(7):256.[doi:DOI:10.3969/j.issn.1005-202X.2018.03.002]
[7]李雅宁,陆世培,彭应林,等. BodyFIX联合腹压板在肝癌大剂量放射治疗中的应用[J].中国医学物理学杂志,2018,35(3):265.[doi:DOI:10.3969/j.issn.1005-202X.2018.03.004]
　LI Yaning,LU Shipei,PENG Yinglin,et al. Application of BodyFIX combined with abdominal pressure plate in stereotactic body radiotherapy for liver cancer[J].Chinese Journal of Medical Physics,2018,35(7):265.[doi:DOI:10.3969/j.issn.1005-202X.2018.03.004]
[8]邓金城,彭应林,刘常春,等. 深度卷积神经网络在放射治疗计划图像分割中的应用[J].中国医学物理学杂志,2018,35(6):621.[doi:DOI:10.3969/j.issn.1005-202X.2018.06.001]
　DENG Jincheng,PENG Yinglin,LIU Changchun,et al. Application of deep convolution neural network in radiotherapy planning image segmentation[J].Chinese Journal of Medical Physics,2018,35(7):621.[doi:DOI:10.3969/j.issn.1005-202X.2018.06.001]
[9]兰川胜,程云章,蔡克尧. 肝癌经皮穿刺射频消融治疗中消融边缘的测量与控制[J].中国医学物理学杂志,2018,35(8):972.[doi:DOI:10.3969/j.issn.1005-202X.2018.08.020]
　LAN Chuansheng,CHENG Yunzhang,CAI Keyao. Quantitative measurement and real-time control of ablation margins during percutaneous radiofrequency ablation therapy of liver cancer[J].Chinese Journal of Medical Physics,2018,35(7):972.[doi:DOI:10.3969/j.issn.1005-202X.2018.08.020]
[10]查雪帆,杨丰,吴俣南,等. 结合迁移学习与深度卷积网络的心电分类研究[J].中国医学物理学杂志,2018,35(11):1307.[doi:DOI:10.3969/j.issn.1005-202X.2018.11.013]
　ZHA Xuefan,YANG Feng,WU Yunan,et al. ECG classification based on transfer learning and deep convolution neural network[J].Chinese Journal of Medical Physics,2018,35(7):1307.[doi:DOI:10.3969/j.issn.1005-202X.2018.11.013]

备注/Memo

备注/Memo:
【收稿日期】2025-02-23【基金项目】上海市“科技创新行动计划”扬帆计划（23YF1438600）【作者简介】张福全，研究方向：人工智能、放射治疗、深度学习，E-mail:Zhang.fuquan@zs-hospital.sh.cn【通信作者】林小惟，研究方向：人工智能、放射治疗、深度学习，E-mail:lin.xiaowei@zs-hospital.sh.cn
更新日期/Last Update: 2025-07-25