基于k均值聚类和自适应模板匹配的眼底出血点检测方法

doi:10.3969/j.issn.0258-8021.2015.03.002

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (4979 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要眼底出血点是糖尿病视网膜病变的早期症状，准确检测眼底图像中的出血点，对于构建糖尿病视网膜病变的自动筛查系统具有重要意义，本研究提出了一种基于k均值聚类和自适应模板匹配的出血点检测方法。首先利用HSV空间亮度校正以及对比度受限自适应直方图均衡化方法对眼底图像进行预处理，然后使用k均值聚类分割出候选目标，最后利用自适应归一化互相关模板匹配与支持向量机（SVM）分类器对候选目标进行筛选，从而得到真正的出血区。采用DIARETDB数据库的219幅眼底图像进行实验，本方法在图像水平的灵敏度为100%，特异性为80%，准确率为924%，在病灶水平的灵敏度为89%，阳性预测值为873%。结果表明本方法能够实现眼底图像中出血点的自动检测。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	肖志涛¹赵北方¹ 张芳^1*耿磊¹吴骏¹

关键词 ：眼底图像, 出血点, k均值聚类, 自适应模板匹配, 支持向量机

Abstract：Hemorrhages are early symptoms of diabetic retinopathy (DR), the accurate detection of hemorrhages in fundus images is an important contribution for building automatic screening system of DR, a novel algorithm based on kmeans clustering and adaptive template matching was proposed in this work. Firstly, HSV brightness correction and contrast limited adaptive histogram equalization were applied to fundus images. Then, the candidate hemorrhages were extracted by using kmeans clustering. At last, adaptive template matching with normalized crosscorrelation and SVM classifier were used to screen the candidates, and the hemorrhages were detected. The approach was evaluated on 219 fundus images from the databases of DIARETDB. Using an image criterion, we achieved 100% sensitivity, 80% specificity and 924% accuracy. With a lesion criterion, we reached a sensitivity of 89% and a positive predictive value of 873%. The results show that hemorrhages in fundus images can be detected automatically using this method.

Key words： fundus images hemorrhages k means clustering adaptive template matching support vector machine

基金资助:天津市科技支撑计划重点项目(13ZCZDGX02100)；天津市应用基础与前沿技术研究计划一般项目（15JCYBJC16600）

引用本文:

肖志涛¹赵北方¹ 张芳^1*耿磊¹吴骏¹

链接本文:

http://cjbme.csbme.org/CN/10.3969/j.issn.0258-8021.2015.03.002 或 http://cjbme.csbme.org/CN/Y2015/V34/I3/264

［1］Ravishankar S, Jain A, Mittal A. Automated feature extraction for early detection of diabetic retinopathy in fundus images ［C］ //International Conference on Computer Vision and Pattern Recognition. Miami: IEEE, 2009: 210-217.
［2］Jaafar HF, Nandi AK, AlNuaimy W. Automated detection of red lesions from digital color fundus photographs ［C］ //33rd Annual International Conference of the Engineering in Medicine and Biology Society. Boston: IEEE, 2011: 6232-6235.
［3］Nutnaree K, Smith G, Bunyarit U. Automated retinal hemorrhage detection using morphological top hat and rulebased classification ［C］ //3rd International Conference on Intelligent Computational Systems. Singapore: IEEE, 2013: 39-43.
［4］Li Tang, Niemeijer M, Abramoff MD. Splat feature classification: detection of the presence of large retinal hemorrhages ［C］ //8th International Symposium on Biomedical Imaging: From Nano to Macro. Chicago: IEEE, 2011: 681-684.
［5］Niemeijer M, Ginneken B, Staal J, et al. Automatic detection of red lesions in digital color fundus photographs ［J］. IEEE Transactions on Medical Imaging, 2005, 24(5): 584-592．
［6］Hatanaka Y, Nakagawa T, Hayashi Y, et al. CAD scheme to detect hemorrhages and exudates in ocular fundus images ［J］. SPIE Medical Imaging: ComputerAided Diagnosis, 2007, 6514(2): 1-8.
［7］Hatanaka Y, Nakagawa T, Hayashi Y, et al. Improvement of automated detection method of hemorrhages in fundus images ［C］ //30th Annual International Conference of the Engineering in Medicine and Biology Society. Vancouver: IEEE, 2008: 5429-5432.
［8］Saleh MD, Eswaran C. An automated decisionsupport system for nonproliferative diabetic retinopathy disease based on MA and HA detection ［J］. Computer Methods and Programs in Biomedicine, 2012, 108(1): 186-196.
［9］García M, Sánchez CI, López MI, et al. Automatic detection of red lesions in retinal images using a multilayer perceptron neural network ［C］ //30th Annual International Conference of the Engineering in Medicine and Biology Society. Vancouver: IEEE, 2008: 5425-5428.
［10］García M, López MI, lvarez D, et al. Assessment of four neural network based classifiers to automatically detect red lesions in retinal images ［J］. Medical Engineering & Physics, 2010, 3210): 1085-1093.
［11］Kande GB, Savithri TS, Subbaiah PV. Automatic detection of microaneurysms and hemorrhages in digital fundus images ［J］. Journal of Digital Imaging, 2010, 23(4): 430-437.［12］Bae JP, Kim KG, Kang HC, et al. A study on hemorrhage detection using hybrid method in fundus images ［J］. Journal of Digital Imaging, 2011, 24(3): 394-404.
［13］Devaraj D, Nagaveena. Detection of red lesion in diabetic retinopathy using adaptive thresholding method ［J］. International Journal of Engineering Research & Technology, 2013, 2(4): 1889-1892.
［14］Krishna K, Murty MN. Genetic Kmeans algorithm ［J］. IEEE Transactions on System, Man and Cybernetics, 1999, 29(3): 433-439.
［15］Vlachos M, Dermatas E. Multiscale retinal vessel segmentation using line tracking ［J］. Computerized Medical Imaging and Graphics, 2010, 34(3): 213-227.
［16］Lewis JP. Fast template matching ［C］ //Canadian Image Processing and Pattern Recognition Society. Quebec: IEEE, 1995: 120-123.