计算机科学

计算机科学 ›› 2020, Vol. 47 ›› Issue (8): 213-220.doi: 10.11896/jsjkx.190600026

所属专题： 医学图像

• 计算机图形学&多媒体 • 上一篇    下一篇

GNNI U-net：基于组归一化与最近邻插值的MRI左心室轮廓精准分割网络

高强1, 高敬阳1, 赵地2   

  1. 1 北京化工大学信息科学与技术学院 北京 100029
    2 中国科学院计算技术研究所 北京 100190
  • 出版日期:2020-08-15 发布日期:2020-08-10
  • 通讯作者: 高敬阳(gaojy@mail.buct.edu.cn)
  • 作者简介:Vic_black@163.com
  • 基金资助:
    北京市自然科学基金(5182018);国家重点研究发展计划(SQ2017ZX106047);北京市自然科学基金重点项目(4161004);北京市科技计划项目(Z171100000117001, Z161100000216143)

GNNI U-net:Precise Segmentation Neural Network of Left Ventricular Contours for MRI Images Based on Group Normalization and Nearest Interpolation

GAO Qiang1, GAO Jing-yang1, ZHAO Di2   

  1. 1 College of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, China
    2 Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China
  • Online:2020-08-15 Published:2020-08-10
  • About author:GAO Qiang, born in 1994, postgra-duate, is a member of China Computer Federation.His main research interests include deep learning, medical image analysis.
    GAO Jing-yang, born in 1966, professor, Ph.D supervisor, is a member of China Computer Federation.Her main research interests include medical image analysis based on deep learning, big data analysis of genomics based on machine learning.
  • Supported by:
    This work was supported by the Natural Science Foundation of Beijing, China(5182018), National Basic Research Program of China(SQ2017ZX106047), Natural Science Foundation of Beijing, China(4161004) and Beijing Science and Technology Project(Z171100000117001, Z161100000216143).

PDF (PC)

1155

摘要： 心血管疾病已成为威胁人类健康的头号杀手。目前, 医生们通过左心室MRI成像技术对左心室轮廓进行手工标注来计算心脏的各项功能参数, 以监测和预防心血管疾病, 但此方法的标注工作量大、耗时且繁琐。目前, 深度学习在许多医疗影像分割领域取得了显著的成功, 但在左心室轮廓分割领域仍有提升的空间。文中提出了一种基于组归一化与最近邻插值的MRI左心室轮廓精确分割网络——GNNI U-net(U-net with Group Normalization and Nearest Interpolation), 该网络利用组归一化方法构建了能够快速、准确提取特征信息的卷积模块, 基于最近邻插值法构建了用于特征信息还原的上采样模块。在Sunnybrook与LVSC两个左心室分割数据集上采用了中心裁减ROI提取的预处理方法, 并对GNNI U-net进行了充分的对比实验。所提网络在Sunnybrook数据集上获得了Dice系数为0.937以及Jaccard系数为0.893的精度。在LVSC数据集上获得了Dice系数为0.957以及Jaccard系数为0.921的精度。GNNI U-net在左心室轮廓分割领域取得了比现有卷积网络分割方法更高的Dice系数精度。最后, 进一步讨论并验证了组归一化操作卷积模块能够加速网络的收敛并提高分割精度;采用最近邻插值法的上采样模块对左心室轮廓这类较小目标的分割效果更好, 能够在一定程度上加速网络的收敛。

关键词: GNNI U-net, 组归一化, 最近邻插值, 左心室分割

Abstract: Cardiovascular disease has become the No.1 killer of human health.At present, doctors manually label the left ventricle contour by left ventricular MRI imaging technology to calculate various functional parameters of the heart to prevent cardiovascular disease, which is not difficult but time consuming and cumbersome.Deep learning has achieved remarkable success in many areas of medical image segmentation, but there is still room for improvement in the field of left ventricular contour segmentation.We proposes a convolutional neural network based on group normalization and nearest interpolation up-sampling which is called GNNI U-net (U-net with Group normalization and Nearest interpolation) for MRI left ventricular contour precise segmentation.We constructe a convolution module with group normalization operation based on group normalization method for fast and accurately feature extraction.Based on nearest neighbor interpolation method, we constructe an up-sampling module for feature restoring.We conducte a pre-processing method for Center cropping ROI extraction and a detailed controlled experiment of GNNI U-net on the Sunnybrook left ventricular segmentation dataset and the LVSC left ventricular segmentation dataset.The experimental results show that the GNNI U-net obtains the accuracy of Dice coefficient of 0.937 and the accuracy of Jaccard coefficient of 0.893 on the Sunnybrook dataset, and it obtains the accuracy of the Dice coefficient of 0.957 and the accuracy of the Jaccard coefficient of 0.921 on the LVSC dataset.The GNNI U-net network achieves higher Dice coefficient accuracy than other convolutional network segmentation methods in the field of left ventricular contour segmentation.Finally, according to the experimental results, we discusse and verifie that the convolutional module of the normalization operation can accelerate the convergence of the network and improve the accuracy, and the up-sampling module using the nearest neighbor interpolation method is more friendly to the smaller target segmentation such as the left ventricle contour, and our model can accelerate network convergence to a certain extent.

Key words: GNNI U-net, Group normalization, Left ventricle segmentation, Nearest interpolation

中图分类号: 

  • TP391.7
[1]MOZAFFARIAN D, BENJAMIN E J, GO A S, et al.Heart Di-sease and Stroke Statistics-2015 Update:A Report From the American Heart Association[J].Circulation, 2015, 131(4):e29-e322.
[2]LOW W Y, LEE Y K, SAMY A L.Non-communicable diseases in the Asia-Pacific region:Prevalence, risk factors and community-based prevention[J].International Journal of Occupational Medicine & Environmental Health, 2014, 28(1):1-7.
[3]LISET VZQUEZ R, MARLY G F C, FRANCISCO P R, et al.Left ventricle segmentation in cardiac MRI images using fully convolutional neural networks[C]∥Medical Imaging 2017:Computer-Aided Diagnosis.International Society for Optics and Photonics, 2017:324-336.
[4]TRAN P V.A Fully Convolutional Neural Network for Cardiac Segmentation in Short-Axis MRI[J].arXiv:1604.00494 .
[5]TAN L K, LIEW Y M, LIM E, et al.Convolutional neural network regression for short-axis left ventricle segmentation in cardiac cine MR sequences[J].Medical Image Analysis, 2017, 39:78-86.
[6]PETITJEAN C, DACHER J N.A review of segmentation me-thods in short axis cardiac MR images[J].Medical Image Analysis, 2011, 15(2):169-184.
[7]PENG P, LEKADIR K, GOOYA A, et al.A review of heartchamber segmentation for structural and functional analysis using cardiac magnetic resonance imaging[J].Magma, 2016, 29(2):155-195.
[8]TIAN J X, LIU G C, GU S S, et al.Deep Learning in Medical Image Analysis and Its Challenges[J].Acta Automatica Sinica, 2018, 44(3):401-424.
[9]FENG C L.Research on Key Algorithms for Segmentation and Visualization of Cardiac Tissue[D].Shenyang:Northeastern University, 2014.
[10]LONG J, SHELHAMER E, DARRELL T.Fully Convolutional Networks for Semantic Segmentation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 39(4):640-651.
[11]RONNEBERGER O, FISCHER P, BROXT.U-Net:Convolu-tional Networks for Biomedical Image Segmentation[C]∥International Conference on Medical ImageComputing and Compu-ter-Assisted Intervention.Springer International Publishing, 2015:234-241.
[12]LITJENS G, KOOI T, BEJNORDI B E, et al.A survey on deep learning in medical image analysis[J].Medical Image Analysis, 2017, 42:60-88.
[13]ZHOU Z, SIDDIQUEE M R, TAJBAKHSH N, et al.UNet++:A Nested U-Net Architecture for Medical Image Segmentation[J].arXiv:1807.10165v1.
[14]ZENG G, YANG X, LI J, et al.3D U-net with Multi-level Deep Supervision:Fully Automatic Segmentation of Proximal Femur in 3D MR Images[C]∥International Workshop on Machine Learning in Medical Imaging.Springer, Cham, 2017:274-282.
[15]GARCIA-GARCIA A, ORTS-ESCOLANO S, OPREA S, Et al.A Review on Deep Learning Techniques Applied to Semantic Segmentation[J].arXiv:1704.06857v1.
[16]ZHANG Q L, ZHAO D, CHI X B, et al.Review for Deep Learning Based on Medical Imaging Diagnosis[J].Computer Science, 2017, 44(S2):1-7.
[17]WU Y, HE K.Group Normalization[J].arXiv:1803.08494v3.
[18]IOFFE S, SZEGEDY C.Batch Normalization:Accelerating Deep Network Training by Reducing Internal Covariate Shift[J].ar-Xiv:1502.03167v3.
[19]NOH H, HONG S, HAN B.Learning Deconvolution Network for Semantic Segmentation[J].arXiv:1505.04366v1.
[20]BADRINARAYANAN V, KENDALL A, CIPOLLA R.Seg-Net:A Deep Convolutional Encoder-Decoder Architecture for Scene Segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017:39(12):2481-2495.
[21]LIN T Y, DOLLR, PIOTR, et al.Feature Pyramid Networks for Object Detection[J].arXiv:1612.03144v2.
[22]FONSECA C G, BACKHAUS M, BLUEMKE D A, et al.The Cardiac Atlas Project--an imaging database for computational modeling and statistical atlases of the heart[J].Bioinformatics, 2011, 27(16):2288-2295.
[1] 康牧,李永亮.
一种基于弹性模型的图像放大算法
Image Zoom Algorithm Based on the Spring Model
计算机科学, 2009, 36(10): 292-295.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发