高级搜索

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

基于多尺度残差双域注意力网络的乳腺动态对比度增强磁共振成像肿瘤分割方法

刘侠 吕志伟 李博 王波 王狄

刘侠, 吕志伟, 李博, 王波, 王狄. 基于多尺度残差双域注意力网络的乳腺动态对比度增强磁共振成像肿瘤分割方法[J]. 电子与信息学报, 2023, 45(5): 1774-1785. doi: 10.11999/JEIT220362
引用本文: 刘侠, 吕志伟, 李博, 王波, 王狄. 基于多尺度残差双域注意力网络的乳腺动态对比度增强磁共振成像肿瘤分割方法[J]. 电子与信息学报, 2023, 45(5): 1774-1785. doi: 10.11999/JEIT220362
LIU Xia, LÜ Zhiwei, LI Bo, WANG Bo, WANG Di. Segmentation Algorithm of Breast Tumor in Dynamic Contrast-Enhanced Magnetic Resonance Imaging Based on Network with Multi-scale Residuals and Dual-domain Attention[J]. Journal of Electronics & Information Technology, 2023, 45(5): 1774-1785. doi: 10.11999/JEIT220362
Citation: LIU Xia, LÜ Zhiwei, LI Bo, WANG Bo, WANG Di. Segmentation Algorithm of Breast Tumor in Dynamic Contrast-Enhanced Magnetic Resonance Imaging Based on Network with Multi-scale Residuals and Dual-domain Attention[J]. Journal of Electronics & Information Technology, 2023, 45(5): 1774-1785. doi: 10.11999/JEIT220362

基于多尺度残差双域注意力网络的乳腺动态对比度增强磁共振成像肿瘤分割方法

doi: 10.11999/JEIT220362
基金项目: 国家自然科学基金(61172167),黑龙江省青年科学基金(QC2017076)
详细信息
    作者简介:

    刘侠:男,博士,教授,研究方向为医学图像处理、模式识别、机器学习

    吕志伟:男,硕士生,研究方向为医学图像分割、机器学习

    李博:男,硕士生,研究方向为医学图像分割、机器学习

    王波:男,博士,副教授,研究方向为医学图像处理、模式识别、机器学习

    王狄:女,硕士生,研究方向为医学图像分割、机器学习

    通讯作者:

    王波 hust_wb@hrbust.edu.cn

  • 中图分类号: R737.9; R445.2; TP391.41

Segmentation Algorithm of Breast Tumor in Dynamic Contrast-Enhanced Magnetic Resonance Imaging Based on Network with Multi-scale Residuals and Dual-domain Attention

Funds: The National Natural Science Foundation of China (61172167), The Youth Science Foundation of Heilongjiang Province (QC2017076)
  • 摘要: 针对乳腺肿瘤大小形态多变、边界模糊以及前景与背景间严重类不平衡的问题,该文提出一种多尺度残差双域注意力融合网络。该网络以多尺度卷积构成的多尺度残差块作为基本搭建模块,通过提取多尺度特征和优化梯度传播通道提高其识别不同尺寸目标的能力,同时融入双域注意力单元,提高网络的边缘识别和边界保持能力。另外该文提出一种混合自适应权重损失函数改善网络优化方向,缓解正负样本极度不均衡的影响。实验结果表明,该文所提方法的平均骰子相似系数(Dice)值达到0.8063,较U形网络(UNet)提高5.3%,参数量下降73.36%,具有更优的分割性能。
  • 图  1  多尺度残差双域注意力融合网络

    图  2  多尺度残差块

    图  3  多尺度卷积

    图  4  双域注意力单元

    图  5  空间、通道注意力模块和全域卷积

    图  6  各网络分割效果对比

    图  7  各网络获取的指标累积分布曲线

    图  8  各网络获取的指标箱线图

    表  1  多尺度残差块在UNet基础网络中的表现

    UNetMSRUNet_CMSRUNet_SMSRUNet_SS
    DSC0.7533±0.09370.7729±0.06340.7717±0.06880.771±0.0653
    IOU0.6115±0.12070.6336±0.08550.6324±0.09340.6311±0.0883
    TPR0.7036±0.18010.7343±0.12730.7304±0.09450.7284±0.1125
    PPV0.8484±0.07830.8334±0.06160.822±0.05220.8313±0.06
    ACC0.9975±0.0010.9977±0.00090.9976±0.00110.9976±0.0009
    HD114.7123±52.768299.5655±48.170498.8848±48.301299.8627±48.4836
    总参数31044162732147470915866869058
    参数占比(%)10023.5822.8422.13
    下载: 导出CSV

    表  2  多尺度残差块在UNetPP基础网络中的表现

    UNetPPMSRUNetPP_CMSRUNetPP_SMSRUNetPP_SS
    DSC0.7654±0.09040.773±0.0750.7757±0.07260.7773±0.0824
    IOU0.627±0.1210.6349±0.1010.6382±0.09810.6417±0.1103
    TPR0.7182±0.16170.7675±0.1430.7241±0.11510.7281±0.1353
    PPV0.8436±0.0510.8033±0.09560.8449±0.04330.8479±0.0403
    ACC0.9976±0.0010.9976±0.00090.9977±0.0010.9978±0.0011
    HD102.8907±50.246498.5523±44.385296.3007±49.406496.6875±50.0719
    总参数36173186886560286190748381666
    参数占比(%)10024.5123.8323.17
    下载: 导出CSV

    表  3  不同注意力机制表现对比

    方法MSRNetMSRAUNetMSRDANet
    DSC0.771±0.06530.7784±0.07230.7839±0.0665
    IOU0.6311±0.08830.6419±0.09870.6485±0.0902
    TPR0.7284±0.11250.7203±0.11530.7273±0.1089
    PPV0.8313±0.060.8549±0.02050.8593±0.0267
    ACC0.9976±0.00090.9978±0.0010.9978±0.0009
    HD99.8627±48.483696.1847±47.104995.5195±42.2526
    下载: 导出CSV

    表  4  不同损失函数表现对比

    方法UNet_CDMSRNet_CDMSRDANet_CD
    DSC0.7541±0.07630.7767±0.06750.7848±0.0736
    IOU0.6119±0.10430.6391±0.09330.6507±0.1007
    TPR0.7104±0.12780.7209±0.12010.7364±0.1313
    PPV0.8394±0.09230.8537±0.02960.8545±0.035
    ACC0.9975±0.00090.9978±0.00090.9978±0.0009
    HD108.9456±44.196598.6967±50.869393.7069±45.4647
    方法UNet_HLMSRNet_HLMSRDANet_HL
    DSC0.7722±0.07920.788±0.06530.8063±0.0538
    IOU0.6345±0.10780.654±0.08830.6781±0.0752
    TPR0.716±0.14270.7449±0.11660.7765±0.0939
    PPV0.8556±0.04010.8479±0.03230.8452±0.0338
    ACC0.9977±0.00090.9978±0.00090.998±0.0007
    HD106.7584±44.964587.6175±47.164377.9541±37.3557
    下载: 导出CSV

    表  5  各网络分割指标对比

    SegNetUNetCPFNetAUNet
    DSC0.7399±0.09570.7533±0.09370.7602±0.05830.7653±0.0758
    IOU0.5946±0.12370.6115±0.12070.6159±0.07540.6249±0.1031
    TPR0.6678±0.15480.7036±0.18010.6955±0.07070.7178±0.1472
    PPV0.8489±0.00220.8484±0.07830.8429±0.07210.8445±0.0734
    ACC0.9975±0.00110.9975±0.0010.9975±0.00110.9976±0.0009
    HD117.6852±66.3663114.7123±52.7682109.1354±51.5033100.1426±47.0054
    总参数29444738310441628936928631745286
    参数占比(%)94.85100287.88102.26
    下载: 导出CSV
    UNetPPUNet3+MSRNetMSRDANetMSRDANet_HL
    DSC0.7654±0.09040.7697±0.08280.771±0.06530.7839±0.06650.8063±0.0538
    IOU0.627±0.1210.6315±0.11120.6311±0.08830.6485±0.09020.6781±0.0752
    TPR0.7182±0.16170.7096±0.14080.7284±0.11250.7273±0.10890.7765±0.0939
    PPV0.8436±0.0510.8577±0.03340.8313±0.060.8593±0.02670.8452±0.0338
    ACC0.9976±0.0010.9977±0.0010.9976±0.00080.9978±0.00090.998±0.0007
    HD102.8907±50.246495.8319±45.031299.8627±48.483695.5195±42.252677.9541±37.3557
    总参数3617318626975234686905882715628271562
    参数占比(%)116.5286.8922.1326.6426.64
    下载: 导出CSV
  • [1] SUNG H, FERLAY J, SIEGEL R L, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA:A Cancer Journal for Clinicians, 2021, 71(3): 209–249. doi: 10.3322/caac.21660
    [2] WILD C P, WEIDERPASS E, STEWART B W, et al. World Cancer Report: Cancer Research for Cancer Prevention[M]. Lyon: IARC Press, 2020: 25–27.
    [3] ROMEO V, CAVALIERE C, IMBRIACO M, et al. Tumor segmentation analysis at different post-contrast time points: A possible source of variability of quantitative DCE-MRI parameters in locally advanced breast cancer[J]. European Journal of Radiology, 2020, 126: 108907. doi: 10.1016/j.ejrad.2020.108907
    [4] WANG Shuai, SUN Kun, WANG Li, et al. Breast tumor segmentation in DCE-MRI with tumor sensitive synthesis[J/OL]. IEEE Transactions on Neural Networks and Learning Systems, 2021.
    [5] CARVALHO E D, SILVA R R V, MATHEW M J, et al. Tumor segmentation in breast DCE-MRI slice using deep learning methods[C]. 2021 IEEE Symposium on Computers and Communications (ISCC), Athens, Greece, 2021: 1–6.
    [6] ZHANG Jun, SAHA A, ZHU Zhe, et al. Hierarchical convolutional neural networks for segmentation of breast tumors in MRI with application to radiogenomics[J]. IEEE Transactions on Medical Imaging, 2019, 38(2): 435–447. doi: 10.1109/TMI.2018.2865671
    [7] MENCATTINI A, RABOTTINO G, SALMERI M, et al. Breast mass segmentation in mammographic images by an effective region growing algorithm[C]. 10th International Conference on Advanced Concepts for Intelligent Vision Systems, Juan-les-Pins, France, 2008: 948–957.
    [8] SHAN Juan, CHENG Hengda, and WANG Yuxuan. A novel automatic seed point selection algorithm for breast ultrasound images[C]. IEEE 2008 19th International Conference on Pattern Recognition (ICPR), Tampa, USA, 2008: 1–4.
    [9] PRAKASH R M, BHUVANESHWARI K, DIVYA M, et al. Segmentation of thermal infrared breast images using K-means, FCM and EM algorithms for breast cancer detection[C]. 2017 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS), Coimbatore, India, 2017: 1–4.
    [10] LIU Hui, LIU Yiping, ZHAO Zuowei, et al. A new background distribution-based active contour model for three-dimensional lesion segmentation in breast DCE-MRI[J]. Medical Physics, 2014, 41(8): 082303. doi: 10.1118/1.4886295
    [11] ZHOU Zhuhuang, WU Weiwei, WU Shuicai, et al. Semi-automatic breast ultrasound image segmentation based on mean shift and graph cuts[J]. Ultrasonic Imaging, 2014, 36(4): 256–276. doi: 10.1177/0161734614524735
    [12] LIU Liangliang, CHENG Jianhong, QUAN Quan, et al. A survey on U-shaped networks in medical image segmentations[J]. Neurocomputing, 2020, 409: 244–258. doi: 10.1016/j.neucom.2020.05.070
    [13] AL-FARIS A Q, NGAH U K, ISA N A M, et al. Breast MRI tumour segmentation using modified automatic seeded region growing based on particle swarm optimization image clustering[M]. SNÁŠEL V, KRÖMER P, KÖPPEN M, et al. Soft Computing in Industrial Applications. Cham: Springer, 2014: 49–60.
    [14] CHAKRABORTY J, MUKHOPADHYAY S, SINGLA V, et al. Detection of masses in mammograms using region growing controlled by multilevel thresholding[C]. 2012 25th IEEE International Symposium on Computer-Based Medical Systems (CBMS), Rome, Italy, 2012: 1–6.
    [15] SHARMA J, RAI J K, and TEWARI R P. A combined watershed segmentation approach using K-means clustering for mammograms[C]. 2015 2nd International Conference on Signal Processing and Integrated Networks (SPIN), Noida, India, 2015: 109–113.
    [16] 冯宝, 陈业航, 刘壮盛, 等. 结合MRF能量和模糊速度的乳腺癌图像分割方法[J]. 自动化学报, 2020, 46(6): 1188–1199. doi: 10.16383/j.aas.c180759

    FENG Bao, CHEN Yehang, LIU Zhuangsheng, et al. Segmentation of breast cancer on DCE-MRI images with MRF energy and fuzzy speed function[J]. Acta Automatica Sinica, 2020, 46(6): 1188–1199. doi: 10.16383/j.aas.c180759
    [17] XIAN Min, ZHANG Yingtao, CHENG H D, et al. Automatic breast ultrasound image segmentation: A survey[J]. Pattern Recognition, 2018, 79: 340–355. doi: 10.1016/j.patcog.2018.02.012
    [18] RONNEBERGER O, FISCHER P, and BROX T. U-Net: Convolutional networks for biomedical image segmentation[C]. 18th International Conference on Medical Image Computing and Computer-Assisted Intervention, Munich, Germany, 2015: 234–241.
    [19] ZHOU Zongwei, SIDDIQUEE M M R, TAJBAKHSH N, et al. Unet++: Redesigning skip connections to exploit multiscale features in image segmentation[J]. IEEE Transactions on Medical Imaging, 2020, 39(6): 1856–1867. doi: 10.1109/TMI.2019.2959609
    [20] SCHLEMPER J, OKTAY O, SCHAAP M, et al. Attention gated networks: Learning to leverage salient regions in medical images[J]. Medical Image Analysis, 2019, 53: 197–207. doi: 10.1016/j.media.2019.01.012
    [21] HUANG Huimin, LIN Lanfen, TONG Ruofeng, et al. UNet 3+: A full-scale connected UNet for medical image segmentation[C]. ICASSP 2020–2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, Spain, 2020: 1055–1059.
    [22] FENG Shuanglang, ZHAO Heming, SHI Fei, et al. CPFNet: Context pyramid fusion network for medical image segmentation[J]. IEEE Transactions on Medical Imaging, 2020, 39(10): 3008–3018. doi: 10.1109/TMI.2020.2983721
    [23] BENJELLOUN M, EL ADOUI M, LARHMAM M A, et al. Automated breast tumor segmentation in DCE-MRI using deep learning[C]. 2018 4th International Conference on Cloud Computing Technologies and Applications (Cloudtech), Brussels, Belgium, 2018: 1–6.
    [24] JIAO Han, JIANG Xinhua, PANG Zhiyong, et al. Deep convolutional neural networks-based automatic breast segmentation and mass detection in DCE-MRI[J]. Computational and Mathematical Methods in Medicine, 2020, 2020: 2413706. doi: 10.1155/2020/2413706
    [25] WANG Hongyu, CAO Jiaqi, FENG Jun, et al. Mixed 2D and 3D convolutional network with multi-scale context for lesion segmentation in breast DCE-MRI[J]. Biomedical Signal Processing and Control, 2021, 68: 102607. doi: 10.1016/j.bspc.2021.102607
    [26] SINGH V K, ABDEL-NASSER M, AKRAM F, et al. Breast tumor segmentation in ultrasound images using contextual-information-aware deep adversarial learning framework[J]. Expert Systems with Applications, 2020, 162: 113870. doi: 10.1016/j.eswa.2020.113870
  • 加载中
图(8) / 表(6)
计量
  • 文章访问数:  279
  • HTML全文浏览量:  180
  • PDF下载量:  77
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-03-31
  • 修回日期:  2022-07-15
  • 网络出版日期:  2022-07-26
  • 刊出日期:  2023-05-10

目录

    /

    返回文章
    返回