动静脉血管自动分类方法及其管径测量  

Artery/vein automatic classification in retinal images and vessel diameter Measurement

作　　者：薛岚燕[1,2] 曹新容[1] 林嘉雯[1] 郑绍华[1] 余轮[1] Xue Lanyan Cao Xinrong Lin Jiawen Zheng Shaohua Yu Lun(College of Physics and Information Engineering, Fuzhou University, Fuzhou 350108, China Institute of Computer and Information, Fufian Agriculture and Forestry University, Fuzhou 350002, China)

机构地区：[1]福州大学物理与信息工程学院,福州350108 [2]福建农林大学计算机与信息学院,福州350002

出　　处：《仪器仪表学报》2017年第9期2307-2316,共10页Chinese Journal of Scientific Instrument

基　　金：福建省自然科学面上基金(2016J01297);福建省中青年教师教育科研项目(JAT160398;JAT160070)资助

摘　　要：视网膜动静脉管径以及动静脉比值可以反映高血压患者脑卒中发病的风险,因此对视网膜血管直径的量化分析有助于病情的风险评估和防治工作。提出了一种视网膜动静脉自动分类和血管直径的自动测量方法。首先,对视网膜血管网络进行分割,并获取中心线;其次选取了不同颜色空间中的不同通道分量,提出了基于中心线像素和血管像素的特征向量、血管宽度和中心光反射的特征向量,采用K均值聚类实现感兴趣测量区域内动静脉的自动分类;最后统计血管横截面的灰度曲线分布,利用高斯曲线进行拟合,根据半高度全宽获取动静脉宽度。分别对REVIEW和DRIVE数据库进行实验,验证了本方法的有效性。The ratio and the diameter of artery/vein of the retinal can reflect the risk of stroke of hypertension, thus, quantitative analysis of the diameter of the vessel can contribute to risk assessment and prevention of the disease. A artery/vein automatic classification method in Retinal Images and Vessel Diameter Measurement are introduced in this paper. Firstly, the vessel eenterline is obtained after segmenting the retinal vessel. Secondly, different channel components in different color spaces are selected, the eigenvectors of centerline pixels, vessel pixels, vessel widths and central light reflection are defined. K-means clustering algorithm is adopted to classify the artery/vein in the measurement region of interest. Finally, the distribution curve of the vessel cross section is fitted using the Gaussian curve and the width of the artery/vein can be obtained according to Half Height Full Width. Experiments with Review public database and Drive public database prove the effectiveness of the proposed method.

关 键 词：眼底图像 血管特征向量 动静脉分类 血管直径测量 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TH79[自动化与计算机技术—计算机科学与技术]