【摘要】 豹纹状眼底常出现在近视眼与老年性退化的眼底中,它是临床上诊断视网膜-脉络膜病变的重要参考.为了辅助近视眼等疾病的临床诊断,提出基于脉络膜血管提取的盘周豹纹状眼底自动分级算法.首先根据所提取的感兴趣区域的直径将眼底图尺寸归一化,以适用于不同分辨率的彩色眼底图;然后基于主成分分析法对视盘中心进行定位,并以视盘中心为中心将眼底图分为上、下、鼻和颞4个象限;再提出描述脉络膜血管透见程度的3类特征:红色通道中的亮度均值、脉络膜血管面积比例以及脉络膜血管密度,并依次对4个象限进行特征提取;最后应用基于信息增益率的C4.5决策树算法将每个象限脉络膜血管透见程度分为0~3级,并将4个象限的分级结果累计,将豹纹状眼底自动分为无、轻度、中度和重度4个级别.用文中算法测试了130幅眼底图,豹纹状眼底自动分级的平均一致率可达84.7%;实验结果表明,该算法能较有效地实现豹纹状眼底的自动分级,为诊断与豹纹状眼底相关的疾病提供量化描述依据.更多还原

【Abstract】 Tessellated fundus is very common in myopic and aged eyes, and it is an important clinical marker for retinochoroidal changes. In order to assist the clinical diagnosis of the disease such as myopia, an approach for automatic tessellated fundus grading based on choroidal vessel extraction is proposed. Firstly, in order to process retinal images with different resolutions, image-size normalization based on the diameter of acquired region of interest(ROI) is performed in preprocessing. Next, optic disc center is located based on principle component analysis. A fundus is divided into 4 quadrants centered on optic disc center. Three types of features are extracted to describe the visibility of choroidal vessels, and the severity of tessellated fundus in every quadrant is classified into 0~3 grades using C4. 5 decision tree. Finally, the grading results of 4 quadrants are combined to classify a tessellated fundus into four classes: healthy, mild, moderate, and severe. The proposed approach was tested using 130 retinal images and the success rate of tessellated fundus grading can reach 84.7%. Our preliminary work illustrates the effectiveness of the proposed automatic tessellated fundus grading approach. Automatic grading of tessellated fundus can provide quantitative basis of ocular diseases related with tessellated fundus.更多还原