北京生物医学工程

糖尿病视网膜病变筛查中的眼底图像质量控制

Quality control of retinal image for screening of diabetic retinopathy

作者：许莉莉梁歌杨智

单位：首都医科大学生物医学工程学院(北京 100069) 中国人民解放军火箭军总医院眼科(北京 100088)

关键词：眼底图像；小波变换；最小二乘支持向量机；质量控制

分类号：R318.04

出版年·卷·期（页码）：2019·38·2（166-170）

摘要：

目的糖尿病性视网膜病变（diabetic retinopathy,DR,以下简称糖网病）筛查中，有相当比例的图像因聚焦不清或曝光不佳不可用于临床诊断，浪费了医疗资源，因而，有必要对眼底图像进行质量监控。本文提出一种基于小波变换和最小二乘支持向量机(least square support vector machine, LS-SVM)的眼底图像质量控制算法。方法首先，对眼底图像进行2层静态小波变换，计算8个子图像的能量值作为特征向量，再利用LS-SVM对眼底图像进行质量评判。本文将某糖网病筛查项目中现场采集的146幅图像，分为训练集和测试集，LS-SVM使用训练集进行学习后，对测试集的97幅图像进行分类测试。结果训练后的LS-SVM能够对测试集很好地分类，鉴别出模糊的眼底图像。以线性函数为核的LS-SVM分类正确率为100%，以高斯径向基函数为核的LS-SVM的分类正确率为97.9%。结论以2层静态小波分解子图像的能量值为输入特征向量的LS-SVM能够很好地鉴别出本文使用的眼底图像是否模糊。

Objective In diabetic retinopathy (DR) screening, a significant portion of the acquisitions of retinal images is of low image quality due to out-of-focus, under-exposure or over-exposure. These acquisitions of retinal images cannot provide sufficient information for diagnosis, a waste of medical resources. Therefore, it is necessary to monitor the quality of fundus images for DR screening. In this paper, we propose a wavelet transform and support vector machine based method to identify the unusable fundus images. Methods First, a stationary wavelet decomposition is applied to the fundus images. Then, we use the channel energy values as the feature inputs to the least square support vector machine (LS-SVM). There are 146 images from a screening project divided into training set and test set. LS-SVM uses the training set for learning and the 97 images of the test set are classified by trained LS-SVM. Results The trained LS-SVM can identify the blurred images of the test set. The accuracy of linear kernel based LS-SVM for the test set is 100%, and that of LS-SVM based on Gaussian radial basis function is 97.9%. Conclusions Based on a limited number of the testing data, the proposed method is capable of identifying the unusable acquisitions of retinal images.

参考文献：

[1] 中华医学会糖尿病分会糖尿病慢性并发症调查组．全国住院糖尿病患者慢性并发症及其相关危险因素10年回顾性调查分析[J]．中国糖尿病杂志, 2003, 11 (4): 232-237.

Investigation Group for Chronic Diabetic Complications, Chinese Diabetes Society, Chinese Medical Association. Chronic diabetic complications and related macro-vascular diseases of in-patients with diabetes in mainland of China—a national retrospective analysis in recent 10 years[J]. Chinese Journal of Diabetes, 2003, 11 (4): 232-237.

[2] 刘卫，张勇进. 糖尿病性视网膜病变[J]. 国外医学眼科学分册, 2005, 29 (5): 352-357.

[3] 李春芳, 丁小燕．糖尿病性视网膜病变的研究进展[J]. 新医学,2004, 35 (1): 9-10.

[4] Massin P, Aubert JP, Eschwege E, et al. Evaluation of a screening program for diabetic retinopathy in a primary care setting Dodia (Dépistage ophtalmologique du diabète) study[J]. Diabetes & Metabolism, 2005, 31(2): 153-162.

[5] 段彦华, 杨春兰, 刘冰. 糖尿病视网膜病变自动筛查方法研究进展[J]. 北京生物医学工程, 2016,35(3): 314-317.

Duan YH, Yang CL, Liu B. Research development of automatic diabetic retinopathy screening[J]. Beijing Biomedical Engineering, 2016, 35(3): 314-317.

[6]Mookiah MRK, Acharya UR, Chua CK, et al. Computer-aided diagnosis of diabetic retinopathy: a review[J]. Computers in Biology & Medicine, 2013, 43(12): 2136-2155.

[7] Raja SS, Vasuki S. Screening diabetic retinopathy in developing countries using retinal images[J]. Applied Medical Informatics, 2015 ,36(1)：13：22.

[8]付浩祥, 周金长, 蔡多贤.分析比较不同筛查方法诊断糖尿病视网膜病变的价值[J].中国现代药物应用，2018, 12(12): 70-71.

[9] 高玮玮，程武山，沈建新，等.免散瞳眼底图像在糖尿病视网膜病变自动筛查中的应用[J].激光生物学报，2015，24(4): 335-340, 381.

Gao WW，Cheng WS，Shen JX, et al. Research of non-dilated fundus images for automated diabetic retinopathy screening[J]. Acta Laser Biology Sinica, 2015, 24(4): 335-340, 381.

[10]全婵娟，王肖，杨新怀，等.免散瞳眼底彩色照在糖尿病视网膜病变社区远程筛查中的应用[J]. 中山大学学报（医学版），2018, 39(2): 298-302.

Quan CJ, Wang X, Yang XH, et al. Application of non-mydriatic digital fundus camera in remote screen of diabetic retinopathy for community residents[J]. Journal of Sun Yat-Sen University (Medical Sciences), 2018, 39(2): 298-302.

[11]李建军, 徐亮, 彭晓燕, 等.远程眼科单张眼底像质量标准(征求意见稿)[J].眼科，2015，24(1)：11-12.

[12]李亚标，王宝光，李温温．基于小波变换的图像纹理特征提取方法及其应用[J]．传感技术学报，2009，22(9)：1308-1311.

Li YB, Wang BG, Li WW. A method of image texture feature analysis based on wavelet decomposition and its application[J]. Chinese Journal of Sensors and Actuators, 2009, 22(9): 1308-1311.

[13]肖玉飞，刘祖润，李目．基于小波包能量熵与SVM的模拟电路故障诊断[J]．电子测量技术，2011，34(6)：110-113.

Xiao YF, Liu ZR, Li M. Analog circuit fault diagnosis based on wavelet packet energy entropy and SVM[J]. Electronic Measurment Technology, 2011, 34(6):110-113.

[14] 张健沛，赵莹，杨静．最小二乘支持向量机的半监督学习算法[J]. 哈尔滨工程大学学报, 2008, 29(10)：1088-1092.

Zhang JP, Zhao Y, Yang J. Semi-supervised learning algorithm with a least square support vector machine[J]. Journal of Harbin Engineering University, 2008, 29(10):1088-1092.

[15]张学工. 关于统计学习理论与支持向量机[J].自动化学报, 2000, 26 (1): 32-42.

Zhang XG. Introduction to statistical learning theory and support vector machines[J]. Acta Automactica Sinica, 2000, 26 (1):32-42.

[16]阎威武, 邵惠鹤. 支持向量机和最小二乘支持向量机的比较及应用研究[J]．控制与决策，2003，18（3）：358-360.

Yan WW. Shao HH. Application of support vector machines and least squares support vector machines to heart disease diagnoses[J]. Control and Decision, 2003,18(3): 358-360.

[17]许莉莉, 师炜, 郭学谦,等. 基于最小二乘支持向量机的心音分类识别研究[J]．中国医疗设备,2017，32(4): 38-41.

Xu LL, Shi W, Guo XQ, et al. Heart sound recognition based on least squares support vector machines[J]. China Medical Devices, 2017，32(4): 38-41.

[18]顾燕萍，赵文杰，吴占松. 最小二乘支持向量机的算法研究[J]. 清华大学学报(自然科学版), 2010,50(7): 1063-1066.

Gu YP, Zhao WJ, Wu ZS. Least squares support vector machine algorithm[J]. Journal of Tsinghua University(Science and Technology), 2010,50(7): 1063-1066.

[19]朱向荣，单杨，李高阳，等.基于最小二乘支持向量机的国公酒中橙皮苷含量测定[J]. 光谱学与光谱分析，2009，29(9): 2471-2474.

Zhu XR, Shan Y, Li GY, et al. Determination of hesperidin content in guogongjiu medicinal wine based on NIR spectrometry and least squares support vector machines [J]. Spectroscopy and Spectral Analysis, 2009, 29(9): 2471-2474.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录