您的位置:山东大学 -> 科技期刊社 -> 《山东大学学报(理学版)》

《山东大学学报(理学版)》 ›› 2021, Vol. 56 ›› Issue (9): 13-20.doi: 10.6040/j.issn.1671-9352.0.2020.655

•   • 上一篇    下一篇

ASR-UNet: 一种基于注意力机制改进的视网膜血管

易三莉1,2(),陈建亭1,贺建峰1,*()   

  1. 1. 昆明理工大学信息工程与自动化学院, 云南 昆明 650500
    2. 云南省计算机技术应用重点实验室, 云南 昆明 650500
  • 收稿日期:2020-11-24 出版日期:2021-09-20 发布日期:2021-09-13
  • 通讯作者: 贺建峰 E-mail:152514845@qq.com;120112624@qq.com
  • 作者简介:易三莉(1977—), 女, 博士, 讲师, 研究方向为数字图像处理、数字信号处理. E-mail: 152514845@qq.com
  • 基金资助:
    国家自然科学基金资助项目(82060329);云南省教育厅项目(2020J0052)

ASR-UNet: an improved retinal vessels segmentation algorithm based on attention mechanism

San-li YI1,2(),Jian-ting CHEN1,Jian-feng HE1,*()   

  1. 1. School of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, Yunnan, China
    2. Key Laboratory of Computer Technology Application of Yunnan Province, Kunming 650500, Yunnan, China
  • Received:2020-11-24 Online:2021-09-20 Published:2021-09-13
  • Contact: Jian-feng HE E-mail:152514845@qq.com;120112624@qq.com

摘要:

针对现有算法因视网膜图像中血管细小和光照等因素导致的分割精度低的问题, 在U-Net的基础上进行改进, 提出了一种能够较好地提取血管结构的算法模型ASR-UNet。首先, 在编码和解码阶段使用了SE-Resnet结构, 引入通道注意力机制对血管细微结构进行通道增强, 之后在跳跃连接部分使用了AG模块对血管细微结构进行空间增强, 提高网络模型对血管细微结构的分割能力。在公开数据集DRIVE和CHASE_DB1上验证了本文的算法, 在评价指标Acc上分别为0.9697和0.9657, 在敏感性上分别为0.8044和0.7673, 在特异性指标上为0.9859和0.9866。实验结果表明, 近年来的视网膜血管分割算法相比, 本文提出的算法在性能有更好的表现。

关键词: 图像分割, U-Net, 视网膜血管分割

Abstract:

Factors such as tiny blood vessels in the retina and light will cause the problem of low segmentation accuracy of existing algorithms. This paper proposes a new retinal vessel segmentation algorithm model based on U-Net segmentation model combined with attention mechanism. The improved measures of this paper are as follows: first, we use SE-Resnet to replace the common convolution used in U-Net, thereby introducing a channel attention mechanism to enhance the original features. After that, AG are added to the jump connection part to enhance the spatial characteristics, thereby improving the models ability to segment the microstructure of blood vessels. The algorithm of this paper is verified on the public datasets DRIVE and CHASE_DB1, and the results are 0.9697 and 0.9657 on Acc, 0.8044 and 0.7673 on sensitivity, and 0.9859 and 0.9866 on specificity. Experimental results show that the proposed algorithm has better performance than recent retinal blood vessel segmentation algorithms.

Key words: image segmentation, U-Net, retinal vessel segmentation

中图分类号: 

  • TP391

图1

SE模块结构"

图2

SE-Resnet模块结构"

图3

AG模块结构"

图4

ASR-UNet视网膜血管分割算法模型"

图5

彩色视网膜图像各通道对比"

图6

局部块"

表1

评价指标和公式"

Evaluation metric Description
Accuracy(Acc) $ \frac{{TP + TN}}{{TP + TN + FP + FN}} $
Sensitivity(Sn) $ \frac{{TP}}{{TP + FN}} $
Specificity(Sp) $ \frac{{TN}}{{TN + FP}} $
F1-score(F1) $ \frac{{2 \times TP}}{{2 \times TP + FP + FN}} $

图7

DRIVE测试数据集上分割结果"

表2

DRIVE数据库测试集结果对比"

方法出处 年份 Acc Sn Sp F1
Vega[21] 2015 0.9412 0.7444 0.9612 0.6884
Liskowski[22] 2016 0.9535 0.7811 0.9807 N/A
Wu[14] 2018 0.9567 0.7844 0.9847 N/A
Alom[23] 2019 0.9556 0.7792 0.9813 0.8171
Zhun Fan[15] 2019 0.9661 0.7957 0.9827 0.8033
Our method 2020 0.9697 0.8044 0.9859 0.8256

表3

CHASE_DB1数据库测试集结果对比"

方法出处 年份 Acc Sn Sp F1
Roychowdhury[24] 2015 0.9467 0.7615 0.9575 N/A
Liskowski[22] 2016 0.9628 0.7816 0.9836 N/A
Li[25] 2016 0.9527 0.7569 0.9816 N/A
Wu[14] 2018 0.9631 0.7538 0.9847 N/A
Alom[23] 2019 0.9634 0.7756 0.9820 0.7928
Our method 2020 0.9657 0.7673 0.9866 0.8105

表4

不同改进算法在DRIVE数据库测试集指标对比"

Model Acc Sn Sp F1
U-Net 0.9688 0.7915 0.9861 0.8186
U-Net+AGs 0.9689 0.7911 0.9862 0.8189
U-Net+SE-Resnet 0.9695 0.7758 0.9883 0.8197
Our method 0.9697 0.8044 0.9859 0.8256
1 KIRBAS C , QUEK F . A review of vessel extraction techniques and algorithms[J]. ACM Computing Surveys (CSUR), 2004, 36 (2): 81- 121.
doi: 10.1145/1031120.1031121
2 LIU I , SUN Y . Recursive tracking of vascular networks in angiograms based on the detection-deletion scheme[J]. IEEE Transactions on Medical Imaging, 1993, 12 (2): 334- 341.
doi: 10.1109/42.232264
3 CAN A , SHEN H , TURNER J N , et al. Rapid automated tracing and feature extraction from retinal fundus images using direct exploratory algorithms[J]. IEEE Transactions on Information Technology in Biomedicine, 1999, 3 (2): 125- 138.
doi: 10.1109/4233.767088
4 ZANA F , KLEIN J C . Segmentation of vessel-like patterns using mathematical morphology and curvature evaluation[J]. IEEE Transactions on Image Processing, 2001, 10 (7): 1010- 1019.
doi: 10.1109/83.931095
5 MENDONCA A M , CAMPILHO A . Segmentation of retinal blood vessels by combining the detection of centerlines and morphological reconstruction[J]. IEEE Transactions on Medical Imaging, 2006, 25 (9): 1200- 1213.
doi: 10.1109/TMI.2006.879955
6 CHAUDHURI S , CHATTERJEE S , KATZ N , et al. Detection of blood vessels in retinal images using two-dimensional matched filters[J]. IEEE Transactions on Medical Imaging, 1989, 8 (3): 263- 269.
doi: 10.1109/42.34715
7 HOOVER A D , KOUZNETSOVA V , GOLDBAUM M . Locating blood vessels in retinal images by piecewise threshold probing of a matched filter response[J]. IEEE Transactions on Medical Imaging, 2000, 19 (3): 203- 210.
doi: 10.1109/42.845178
8 KASS M , WITKIN A , TERZOPOULOS D . Snakes: active contour models[J]. International Journal of Computer Vision, 1988, 1 (4): 321- 331.
doi: 10.1007/BF00133570
9 ESPONA L, CARREIRA M J, ORTEGA M, et al. A snake for retinal vessel segmentation[C]//Iberian Conference on Pattern Recognition and Image Analysis. Berlin: Springer, 2007: 178-185.
10 STAAL J , ABRMOFF M D , NIEMEIJER M , et al. Ridge-based vessel segmentation in color images of the retina[J]. IEEE Transactions on Medical Imaging, 2004, 23 (4): 501- 509.
doi: 10.1109/TMI.2004.825627
11 RICCI E , PERFETTI R . Retinal blood vessel segmentation using line operators and support vector classification[J]. IEEE Transactions on Medical Imaging, 2007, 26 (10): 1357- 1365.
doi: 10.1109/TMI.2007.898551
12 FRAZ M M, REMAGNINO P, HOPPE A, et al. A supervised method for retinal blood vessel segmentation using line strength, multiscale Gabor and morphological features[C]//2011 IEEE International Conference on Signal and Image Processing Applications (ICSIPA). [S. l. ]: IEEE, 2011: 410-415.
13 RONNEBERGER O, FISCHER P, BROX T. U-net: Convolutional networks for biomedical image segmentation[C]//International Conference on Medical Image Computing and Computer-assisted Intervention. Cham: Springer, 2015: 234-241.
14 WU Y, XIA Y, SONG Y, et al. Multiscale network followed network model for retinal vessel segmentation[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham: Springer, 2018: 119-126.
15 FAN Z, MO J, QIU B, et al. Accurate retinal vessel segmentation via octave convolution neural network[EB/OL]. (2019-06-28)[2020-05-18]. https://arxiv.org/abs/1906.12193v1.
16 KOMATSU R , FUJII H , TAMURA Y , et al. Octave deep plane-sweeping network: reducing spatial redundancy for learning-based plane-sweeping stereo[J]. IEEE Access, 2019, 7, 150306- 150317.
doi: 10.1109/ACCESS.2019.2947195
17 SZEGEDY C, IOFFE S, VANHOUCKE V, et al. Inception-v4, inception-resnet and the impact of residual connections on learning[C]//Proceedings of the Thirty-first AAAI Conference on Artificial Intelligence. [S. l. ]: AAAI, 2017: 4278-4284
18 HU J , SHEN L , SUN G . Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 42 (8): 2011- 2023.
19 OKTAY O, SCHLEMPER J, FOLGOC L L, et al. Attention U-net: learning where to look for the pancreas[J/OL]. (2018-04-13)[2020-05-18]. https://arxiv.org/abs/1804.03999v2.
20 OWEN C G , RUDNICKA A R , MULLEN R , et al. Measuring retinal vessel tortuosity in 10-year-old children: validation of the computer-assisted image analysis of the retina (CAIAR) program[J]. Investigative Ophthalmology & Visual Science, 2009, 50 (5): 2004- 2010.
21 VEGA R , SANCHEZ-ANTE G , FALCON-MORALES L E , et al. Retinal vessel extraction using lattice neural networks with dendritic processing[J]. Computers in Biology and Medicine, 2015, 58, 20- 30.
doi: 10.1016/j.compbiomed.2014.12.016
22 LISKOWSKI P , KRAWIEC K . Segmenting retinal blood vessels with deep neural networks[J]. IEEE Transactions on Medical Imaging, 2016, 35 (11): 2369- 2380.
doi: 10.1109/TMI.2016.2546227
23 ALOM M Z , YAKOPCIC C , HASAN M , et al. Recurrent residual U-net for medical image segmentation[J]. Journal of Medical Imaging, 2019, 6 (1): 014006.
24 ROYCHOWDHURY S , KOOZEKANANI D D , PARHI K K . Iterative vessel segmentation of fundus images[J]. IEEE Transactions on Biomedical Engineering, 2015, 62 (7): 1738- 1749.
doi: 10.1109/TBME.2015.2403295
25 LI Q , FENG B , XIE L P , et al. A cross-modality learning approach for vessel segmentation in retinal images[J]. IEEE Transactions on Medical Imaging, 2015, 35 (1): 109- 118.
[1] 王磊,谢江宁. 基于层级分割的颜色恒常性算法[J]. 山东大学学报(理学版), 2016, 51(1): 101-105.
[2] 刘战杰1,马儒宁1,邹国平1,钟宝江2,丁军娣3. 一种新的基于区域生长的彩色图像分割算法[J]. J4, 2010, 45(7): 76-80.
[3] 徐光柱1,刘鸣2,任东1,马义德3,刘晓丽1. 基于脉冲耦合神经网络的多区域图像分割[J]. J4, 2010, 45(7): 86-93.
[4] 董吉文,杨 森,鲁守银 . 基于单目视觉的移动机器人导航方法[J]. J4, 2008, 43(11): 1-04 .
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] 刘汝军,曹玉霞,周 平 . 利用小反馈实现离散非线性混沌系统的反控制[J]. J4, 2007, 42(7): 30 -32 .
[2] 许万银. 一类拟线性Neumann问题的多重解[J]. J4, 2009, 44(10): 39 -42 .
[3] 杜瑞颖, 杨勇, 陈晶, 王持恒. 一种基于相似度的高效网络流量识别方案[J]. 山东大学学报(理学版), 2014, 49(09): 109 -114 .
[4] 石长光 . Faddeev模型中的多孤立子解[J]. J4, 2007, 42(7): 38 -40 .
[5] 董炯,曹小红. 算子立方的Weyl定理及其紧摄动[J]. 山东大学学报(理学版), 2016, 51(8): 15 -21 .
[6] . 银杏叶提取物中总黄酮含量的分析方法研究[J]. J4, 2009, 44(5): 40 -44 .
[7] 张凤霞1,李莹1,2,郭文彬1,赵建立1. 分块Hermite阵与斜Hermite阵的最大秩与最小秩[J]. J4, 2010, 45(4): 106 -110 .
[8] 毕晓冬 . 左拟正规带的自由积[J]. J4, 2008, 43(6): 83 -86 .
[9] 凌密然, 米据生, 马丽. 异构形式背景上的不确定推理[J]. 山东大学学报(理学版), 2014, 49(08): 28 -32 .
[10] 吴松丽1,2,陈桂友3*. 外-扰动与属性合取收缩规律挖掘-分离[J]. 山东大学学报(理学版), 2014, 49(06): 1 -5 .