2.765

2022影响因子

(CJCR)

  • 中文核心
  • EI
  • 中国科技核心
  • Scopus
  • CSCD
  • 英国科学文摘

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

一种边界增强的医学图像小样本分割网络

贾熹滨 郭雄 王珞 杨大为 杨正汉

贾熹滨, 郭雄, 王珞, 杨大为, 杨正汉. 一种边界增强的医学图像小样本分割网络. 自动化学报, xxxx, xx(x): x−xx doi: 10.16383/j.aas.c220994
引用本文: 贾熹滨, 郭雄, 王珞, 杨大为, 杨正汉. 一种边界增强的医学图像小样本分割网络. 自动化学报, xxxx, xx(x): x−xx doi: 10.16383/j.aas.c220994
Jia Xi-Bin, Guo Xiong, Wang Luo, Yang Da-Wei, Yang Zheng-Han. A few-shot medical image segmentation network with iterative boundary refinement. Acta Automatica Sinica, xxxx, xx(x): x−xx doi: 10.16383/j.aas.c220994
Citation: Jia Xi-Bin, Guo Xiong, Wang Luo, Yang Da-Wei, Yang Zheng-Han. A few-shot medical image segmentation network with iterative boundary refinement. Acta Automatica Sinica, xxxx, xx(x): x−xx doi: 10.16383/j.aas.c220994

一种边界增强的医学图像小样本分割网络

doi: 10.16383/j.aas.c220994
基金项目: 国家重点研发项目中国和韩国政府间联合研究项目(2019YFE0107800), 国家自然科学基金(62171298, 82071876) 资助
详细信息
    作者简介:

    贾熹滨:北京工业大学信息学部教授. 主要研究方向为视觉信息认知与计算, 智能医学图像分析和诊断, 情感计算. 本文通信作者. E-mail: jiaxibin@bjut.edu.cn

    郭雄:北京工业大学信息学部硕士研究生. 主要研究方向为计算机视觉, 医学图像分割, 小样本学习. E-mail: guox@emails.bjut.edu.cn

    王珞:北京工业大学信息学部讲师. 主要研究方向为图像检索, 深度学习, 医学影像处理以及多模态数据融合. E-mail: wangluo@bjut.edu.cn

    杨大为:首都医科大学附属北京友谊医院副教授. 2020年获得首都医科大学博士学位. 主要研究方向为肝脏疾病影像研究. E-mail:yangdawei@ccmu.edu.cn

    杨正汉:首都医科大学附属北京友谊医院教授. 1999年获得北京医科大学博士学位. 主要研究方向为腹部疾病影像诊断, 肝细胞癌及癌前病变的早期影像诊断, MRI新技术的开发与应用. E-mail: yangzhenghan@vip.163.com

A Few-shot Medical Image Segmentation Network With Iterative Boundary Refinement

Funds: Supported by National Key Research and Development Program of China and South Korea Intergovernmental Joint Research Project (2019YFE0107800), National Natural Science Foundation of China (62171298, 82071876)
More Information
    Author Bio:

    JIA Xi-Bin Professor at Faculty of Information Technology, Beijing University of Technology. Her research interest covers visual information cognition and computing, intelligent medical image analysis and diagnosis and emotional calculation. Corresponding author of this paper

    GUO Xiong Master student at the Faculty of Information Technology, Beijing University of Technology. His research interest covers computer vision, medical image segmentation and few-shot learning

    LUO Wang Lecturer at the Faculty of Information Technology, Beijing University of Technology. His research interests include image retrieval, deep learning, medical image processing and multi-modal data fusion

    YANG Da-Wei Vice professor of Beijing Friendship Hospital, Capital Medical University. He received his Ph.D degree from Capital Medical University in 2020. His main research interest is imaging diagnosis and research on liver disease

    YANG Zheng-Han Professor of Beijing Friendship Hospital, Capital Medical University. He received his Ph.D degree from Beijing Medical University in 1999. His research interests include imaging diagnosis of abdominal diseases, early imaging diagnosis of hepatocellular carcinoma and precancerous lesions, and development and application of new MRI technology

  • 摘要: 精准的医学图像自动分割是临床影像学诊断和影像三维重建的重要基础.但医学图像数据的目标对象间对比度差异小、受器官运动影响大, 加之标注样本规模小, 因此在小样本下建立高性能的医学分割模型仍是目前的难点问题. 针对主流原型学习小样本分割网络对医学图像边界分割性能差的问题, 提出一种迭代边界优化的小样本分割网络(Iterative boundary refinement based few-shot-segmentation network, IBR-FSS-Net). 以双分支原型学习的小样本分割框架为基础引入类别注意力机制和密集比较模块, 对粗分割掩码进行迭代优化, 引导分割模型在多次迭代学习过程中关注边界, 从而提升边界分割精度. 为进一步克服医学图像训练样本少且多样性不足的问题, 使用超像素方法生成伪标签, 扩充训练数据以提升模型泛化性. 在主流的ABD-MR和ABD-CT医学图像分割公共数据集上进行实验, 与现有多种先进的医学图像小样本分割方法进行了广泛的对比分析和消融实验. 结果表明, 该方法有效提升了未见医学类别的分割性能.
    1)  本文责任编委 XXX Recommended by Associate Editor BIAN Wei
  • 图  1  基于注意力机制的边界优化模块

    Fig.  1  Boundary refinement module based on attention mechanism

    图  2  基于注意力机制的边界优化模块

    Fig.  2  Boundary refinement module based on attention mechanism

    图  3  原型网络结构图

    Fig.  3  Prototype network structure diagram

    图  4  密集比较模块

    Fig.  4  Dense comparison module

    图  5  基于超像素的样本扩充方法

    Fig.  5  Self-supervised learning process based on superpixel algorithm

    图  6  数据集划分示意图

    Fig.  6  ABD-MR dataset partition diagram

    图  7  医学图像小样本分割网络的Baseline模型

    Fig.  7  Baseline model of few-shot medical image segmentation network

    图  8  核磁共振成像图像中的四种器官样例的预测掩码

    Fig.  8  Prediction masks for four organ samples in magnetic resonance images

    图  9  电子计算机断层扫描图像中的四种器官样例的预测掩码

    Fig.  9  Prediction masks for four organ samples in computed tomography images

    表  1  在 ABD-CT 与 ABD-MR 数据集上不同模型的量化分割结果(%)

    Table  1  Quantitative segmentation results of different models on ABD-CT and ABD-MR datasets (%)

    方法 ABD-CT数据集 ABD-MR数据集
    脾脏 左肾 右肾 肝脏 平均值 脾脏 左肾 右肾 肝脏 平均值
    SE-Net 0.23 32.83 14.34 0.27 11.91 51.80 62.11 61.32 27.43 50.66
    PANet 25.59 32.34 17.37 38.42 29.42 50.90 53.45 38.64 42.26 46.33
    ALPNet 60.25 63.34 54.82 73.65 63.02 67.02 73.63 78.39 73.05 73.02
    GCN-DE 56.53 68.13 75.50 46.77 61.73 60.63 76.07 83.03 49.47 67.30
    RP-Net 69.85 70.48 70.00 79.62 72.48 76.35 81.40 85.78 73.51 79.26
    ADNet 75.92 75.28 83.28 80.81 78.82
    PoissonSeg 52.33 50.11 47.02 58.74 52.05 52.85 50.58 53.57 61.03 54.51
    AAS-DCL 66.36 64.71 69.95 71.61 68.16 74.86 76.90 83.75 69.94 76.36
    IBR-FSS-Net 71.73 73.78 72.02 78.13 73.92 75.12 82.19 85.64 75.89 79.71
    下载: 导出CSV

    表  2  不同组件的量化分割结果(%)

    Table  2  Quantified segmentation results for different components (%)

    组件 脾脏 左肾 右肾 肝脏 平均值
    Baseline 62.33 63.65 66.87 64.18 64.26
    Baseline+Concat 60.62 65.55 68.53 66.56 65.32
    Baseline+BRM 66.63 72.10 74.83 69.17 70.68
    Baseline+3Concat 61.20 68.72 70.38 66.95 66.81
    Baseline+3BRM 75.12 82.19 85.64 75.89 79.71
    下载: 导出CSV

    表  3  边界优化模块的数量的影响(%)

    Table  3  The effect of the number of boundary refinement modules (%)

    组件 脾脏 左肾 右肾 肝脏 平均值
    Baseline 62.33 63.65 66.87 64.18 64.26
    Baseline+BRM 66.63 72.10 74.83 69.17 70.68
    Baseline+2BRM 69.88 79.98 82.12 73.56 76.39
    Baseline+3BRM 75.12 82.19 85.64 75.89 79.71
    Baseline+4BRM 68.57 77.23 78.82 69.60 73.56
    Baseline+5BRM 64.13 70.55 72.69 66.42 68.45
    下载: 导出CSV

    表  4  不同特征提取网络的量化分割结果(%)

    Table  4  Quantified segmentation results of different feature extraction networks (%)

    骨干网络 脾脏 左肾 右肾 肝脏 平均值
    VGG-16 52.09 63.83 64.48 57.88 59.57
    U-Net 69.66 78.94 80.46 72.15 75.30
    Res U-Net 71.82 78.24 81.10 73.41 76.14
    Attention U-Net 73.96 79.14 83.51 73.60 77.55
    ResNet-50 71.23 78.19 82.57 73.68 76.42
    ResNet-101 75.12 82.19 85.64 75.89 79.71
    下载: 导出CSV

    表  5  度量网络组合方式的影响(%)

    Table  5  The effect of the combination of metric network (%)

    原型网络拼接方式 脾脏 左肾 右肾 肝脏 平均值
    Prototypical-Net 70.92 80.61 83.70 74.48 77.43
    DCM 71.53 81.36 83.44 74.81 77.79
    DCM + Prototypical-Net 72.97 81.49 83.68 74.83 78.24
    Prototypical-Net +DCM 75.12 82.19 85.64 75.89 79.71
    下载: 导出CSV

    表  6  与使用半监督、对比学习、弱监督的医学分割模型在 ABD-CT 与 ABD-MR 数据集上结果对比(%)

    Table  6  Comparison of results with medical segmentation models ssing semi-supervised, comparative learning, and weakly supervised on ABD-CT and ABD-MR datasets (%)

    方法(比例) ABD-CT数据集 ABD-MR数据集
    脾脏 左肾 右肾 肝脏 平均值 脾脏 左肾 右肾 肝脏 平均值
    MagicNet(30%) 91.42 86.19 84.64 93.89 89.04
    CVCL(部分) 95.40 94.60 94.60 96.70 95.33
    C-CAM(0%) 74.16 81.00 84.75 72.68 78.15
    IBR-FSS-Net(5%) 71.73 73.78 72.02 78.13 73.92 75.12 82.19 85.64 75.89 79.71
    下载: 导出CSV
  • [1] Zhang B, Zhang L, Zhang L, Karray F. Retinal vessel extraction by matched filter with first-order derivative of Gaussian. Computers in Biology and Medicine, 2010, 40(4): 438-445 doi: 10.1016/j.compbiomed.2010.02.008
    [2] Vinyals O, Blundell C, Lillicrap T, Wierstra D. Matching networks for one shot learning. In: Proceedings of the PMLR on Advances in Neural Information Processing Systems. Barcelona, Spain: PMLR, 2016. 29
    [3] Finn C, Abbeel P, Levine S. Model-agnostic meta-learning for fast adaptation of deep networks. In: Proceedings of the PMLR on International Conference on Machine Learning. Sydney, Australia: PMLR, 2017. 1126−1135
    [4] Sung F, Yang Y, Zhang L, Tao X, Philip H S T, Timothy M. Learning to compare: Relation network for few-shot learning. In: Proceedings of the IEEE on Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018. 1199−1208
    [5] Li G, Jampani V, Sevilla-Lara L, Sun D, Kim Jonghyun, Joongkyu Kim. Adaptive prototype learning and allocation for few-shot segmentation. In: Proceedings of the IEEE on Conference on Computer Vision and Pattern Recognition. Virtual Event: IEEE, 2021. 8334−8343
    [6] Yu Q, Dang K, Tajbakhsh N, et al. A location-sensitive local prototype network for few-shot medical image segmentation. In: Proceedings of the IEEE on Image Processing. Anchorage, Alaska, USA: IEEE, 2021. 262−266
    [7] 孙君梅, 葛青青, 李秀梅, 赵宝奇. 具有边界增强功能的医学图像分割网络. 电子与信息学报, 2022, 44(5): 1643-1652 doi: 10.11999/JEIT210784

    Sun Jun-Mei, Ge Qing-Qing, Li Xiu-Mei, Zhao Bao-Qi. A medical image segmentation network with boundary enhancement. Journal of Electronics, 2022, 44(5): 1643-1652. doi: 10.11999/JEIT210784
    [8] Yuan Y, Chen X, Wang J. Object-contextual representations for semantic segmentation. In: Proceedings of the European Conference on Computer Vision. Virtual Event: Springer, 2020. 23-28, 173−190
    [9] Kim T, Lee H, Kim D. Uacanet: Uncertainty augmented context attention for polyp segmentation. In: Proceedings of the IEEE Conference on ACM International Conference on Multimedia. Chengdu, China: ACM, 2021. 2167−2175
    [10] Zhang C, Lin G, Liu F, Rui Y, Chunhua S. Canet: Class-agnostic segmentation networks with iterative refinement and attentive few-shot learning. In: Proceedings of the IEEE Conference on Conference on Computer Vision and Pattern Recognition. Long Beach, USA: IEEE, 2019. 5217−5226
    [11] Ouyang C, Biffi C, Chen C, Kart T, Qiu H, Rueckert D. Self-supervision with superpixels: Training few-shot medical image segmentation without annotation. In: Proceedings of the European Conference on Computer Vision. Virtual Event: Springer, 2020. 762−780
    [12] Tang H, Liu X, Sun S, Yan X, Xie X. Recurrent mask refinement for few-shot medical image segmentation. In: Proceedings of IEEE on International Conference on Computer Vision: Seoul, South Korea: IEEE, 2021. 3918−3928
    [13] Fan D P, Ji G P, Zhou T, Chen G, Fu H, Shen J, et al. Pranet: Parallel reverse attention network for polyp segmentation. In: Proceedings of the Medical Image Computing and Computer Assisted Intervention. Lima, Peru: Springer, 2020. 263−273
    [14] Shaban A, Bansal S, Liu Z, Essa I, Byron B. One-shot learning for semantic segmentation. In: Proceedings of the IEEE on Computer Vision and Pattern Recognition. Hawaii, USA: IEEE, 2017.
    [15] Wang K, Liew J H, Zou Y, Zhou D, Feng J. Panet: Few-shot image semantic segmentation with prototype alignment.In: Proceedings of the IEEE on International Conference on Computer Vision. Seoul, South Korea: IEEE, 2019. 9197−9206
    [16] Tian Z, Zhao H, Shu M, Yang Z, Li R, Jia J. Prior guided feature enrichment network for few-shot segmentation. Pattern Analysis and Machine Intelligence, 2020. 44(2): 1050-1065
    [17] Zhang G, Kang G, Yang Y, Wei Y. Few-shot segmentation via cycle-consistent transformer. In: Proceedings of the Advances in Neural Information Processing Systems. Montreal Canada: NeurIPS, 2021. 21984−21996
    [18] Roy A G, Siddiqui S, Polsterl S, Navab N, Wachinger C. “Squeeze & excit” guided few-shot segmentation of volumetric images. Medical Image Analysis, 2020, 59: Article No. 101587
    [19] Sun L, Li C, Ding X, Huang Y, Chen Z, Wang G, et al. Few-shot medical image segmentation using a global correlation network with discriminative embedding. Computers in Biology an Medicine, 2022, 140: Article No. 105067
    [20] Tang H, Liu X, Sun S, Yan X, Xie X. Recurrent mask refinement for few-shot medical image segmentation. In: Proceedings of the IEEE on International Conference on Computer Vision. Montreal, Canada: IEEE, 2021. 3918−3928
    [21] Hansen S, Gautam S, Jenssen R, Kampffmeyer M. Anomaly detection-inspired few-shot medical image segmentation through self-supervision with supervoxels. Medical Image Analysis, 2022, 78: Article No. 102385
    [22] Shen X, Zhang G, Lai H, et al. PoissonSeg: Semi-supervised few-shot medical image segmentation via poisson learning. In: Proceedings of the IEEE International Conference on Bioinformatics and Biomedicine. Houston, USA: IEEE, 2021. 1513−1518
    [23] Wu H, Xiao F, Liang C. Dual Contrastive learning with anatomical auxiliary supervision for few-shot medical image segmentation. In: Proceedings of the European Conference on Computer Vision. Tel Aviv, Israel: Springer, 2022. 417−434
    [24] Chen D, Bai Y, Shen W, Li Q, Yu L, Wang Y. MagicNet: Semi-supervised multi-Organ segmentation via magic-cube partition and recovery. In: Proceedings of the IEEE on Computer Vision and Pattern Recognition. Oxford, UK: IEEE, 2023. 23869−23878
    [25] Liu P, Zheng G. Context-aware voxel-wise contrastive learning for label efficient multi-organ segmentation. In: Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention. Singapore: Springer, 2022. 653−662
    [26] Chen Z, Tian Z, Zhu J, Li C, Du S. C-cam: Causal cam for weakly supervised semantic segmentation on medical image. In: Proceedings of the IEEE on Computer Vision and Pattern Recognition. New Orleans, USA: IEEE, 2022. 11676−11685
    [27] Chen S, Tan X, Wang B, Hu X. Reverse attention for salient object detection. In: Proceedings of the European Conference on Computer Vision. Munich, Germany: Springer, 2018. 234−250
    [28] 陈琼, 杨永, 黄田琳, 冯媛. 新型全向立体视觉系统设计. 数据与计算发展前沿, 2022, 3(6): 17-34

    Chen Qiong, Yang Yong, Huang Tian-Lin, Feng Yuan. A survey on few-shot image semantic segmentation. Frontiers of Data and Domputing, 2022, 3(6): 17-34.
    [29] Snell J, Swersky K, Zemel R. Prototypical networks for few-shot learning. In: Proceedings of the Conference on Neural Information Processing Systems. Long Beach, USA: NeurIPS, 2017. 30
    [30] Landman B, Xu Z, Igelsias J, Styner M, Langerak T, Klein A. Miccai multi-atlas labeling beyond the cranial vault-workshop and challenge.In: Proceedings of the Medical Image Computing and Computer Assisted Intervention. Munich, Germany: Springer, 2015. 12
    [31] Kavur A E, Gezer N S, Baris M, Aslan S, Conze P H, Groza V. CHAOS challenge-combined (CT-MR) healthy abdominal organ segmentation. Medical Image Analysis, 2021, 69: Atricle No. 101950
    [32] Irving B. MaskSLIC: regional superpixel generation with application to local pathology characterisation in medical images. In: Proceedings of the IEEE on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE, 2016. 09518
    [33] Lin T Y, Maire M, Belongie S, Hays J, Perona P, Ramanan D, et al. Microsoft coco: Common objects in context.In: Proceedings of the European Conference on Computer Vision. Zurich, Switzerland: Springer, 2014. 740−755
    [34] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. In: Proceedings of the IEEE on Computer Vision and Pattern Recognition. Columbus, USA: IEEE, 2014. 1409. 1556
    [35] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation. In: Proceedings of the Medical Image Computing and Computer-Assisted Intervention. Munich, Germany: Springer, 2015. 234−241
    [36] Xiao X, Lian S, Luo Z, Li S. Weighted res-unet for high-quality retina vessel segmentation. In: Proceedings of the International Conference on Information Technology in Medicine and Education. Hangzhou, China: IEEE, 2018. 327−331
    [37] Oktay O, Schlemper J, Folgoc L L, Lee M, Heinrich M, Misawa K. Attention U-net: Learning where to look for the pancreas. In: Proceedings of the IEEE on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018. 1804. Atricle No. 03999
    [38] Kervadec H, Bouchtiba J, Desrosiers C, Granger E, Dolz J, Ayed I.B. Boundary loss for highly unbalanced segmentation.In: Proceedings of the International Conference on Medical Imaging with Deep Learning. Shanghai, China: PMLR, 2019: 285−296
  • 加载中
计量
  • 文章访问数:  304
  • HTML全文浏览量:  395
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-12-23
  • 录用日期:  2023-07-22
  • 网络出版日期:  2023-10-15

目录

    /

    返回文章
    返回