摘要
眼底图像中动静脉血管的宽度变化能够体现患者糖尿病的病程情况,人工确定血管宽度及其差值耗时费力且需要丰富经验。本文针对辽宁何氏眼科医院眼病筛查系统中受检者的眼底图像提出了动静脉宽度差变化率的概念,指出了其与糖尿病诊断之间的关系,并选取RGB、LAB、YCb Cr、Gaussian 4个颜色空间中不同的通道分量,定义了新的基于血管中心线像素和血管像素的特征向量,采用线性判别分析(LDA)分类器完成了对动静脉的全自动分类,继而实现了对动静脉血管宽度测量及相应差值变化率的计算。实验结果表明,与其他算法相比,本文方法在血管分割、动静脉分类、血管宽度及差值计算等方面都较准确,其诊断结果与临床诊断基本一致,具有一定的临床应用价值。
The change in vessel widths of arteries and veins in retinal images is associated with diabetes process of patient. Manual determination of retinal blood vessel width and it difference not only requires expertise, but also is a very tedious and time-consuming task. Aiming at the retinal images of the retinopathy screening program of He＇ s ophthalmology hospital, the concept of the change ratio of the difference between artery and vein widths is introduced in this paper, and its relationship with diabetes diagnosis is pointed out. Different channel components in the RGB, LAB, YCbCr and Gaussian color spaces are selected, and a new feature vector based on the centerline pixels and vessel pixels of the blood vessel is defined. The LDA classifier is adopted to complete the full-automated classification of the artery and vein blood vessels; then, the vessel width measurement and corresponding width difference change ratio calculation are achieved. The experiment results demonstrate that the proposed method is accurate in vessel segmentation, artery and vein classification, as well as the blood vessel width and corresponding width difference change ratio calculation compared with the well- established techniques in literatures. The diagnosis results of the proposed method are consistent with those of clinical diagnosis. The proposed method has clinical application value.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2016年第4期912-919,共8页
Chinese Journal of Scientific Instrument
基金
辽宁省教育厅优秀人才项目(LJQ2014011)
辽宁省教育厅一般项目(L2014041)
沈阳市科技攻关项目(F12-010-2-00)
沈阳工业大学第三批青年学术骨干教师(3029906)项目资助
作者简介
郭莹，2009年于大连理工大学获得博士学位，现为沈阳工业大学副教授，主要研究方向为图像处理、自适应滤波算法及其应用。E—mail：gy20072009@sina．com
马秀丽，2011年于北方民族大学获得学士学位，现为沈阳工业大学硕士研究生，主要研究方向为图像处理。E-mail：117071220@qq．com
