基于形态特征和k均值聚类的黄斑检测与定位

Detecting and Locating the Macular Using Morphological Features and k-means Clustering

彩色眼底图像已经广泛地应用于眼科相关疾病的辅助诊断和筛查。眼底图像中的黄斑区域检测和中心定位是眼科疾病分级、诊疗的重要步骤。提出一种有效检测与定位黄斑的方法,通过分析黄斑的低亮度和趋于圆形的形态特征,可以不依赖视盘和血管信息,在二值化眼底图像中实现黄斑检测,确定黄斑区域。改进k均值聚类方法,引入图像的空间信息,优化聚类对象,获取黄斑的边缘信息,实现黄斑中心的有效定位。在公开的眼底图像数据库上验证方法的性能,具有较高的准确率。对正常和存在病变的眼底图像的黄斑中心有效定位,可达到96.11%和92.12%,平均准确率达到93.92%。实验表明,提出的基于形态特征和k均值聚类的黄斑检测与定位方法简单、高效,对眼科疾病的计算机辅助诊断有实用价值。

Color fundus images have been widely used in the diagnosis and screening of ophthalmic diseases.The macular detection and foveal location in fundus images are important steps in grading and diagnosis of ophthalmic diseases. An efficient method not relying on the optic and vascular information was proposed in this work for detecting and locating macular foveal. After a general analysis on morphological characteristics of macular,which was low brightness and round,the area of macular could be ensured in the binary images.Then,an improved k-means clustering method was proposed on the basis of spatial information of images and optimizes clustering objects to obtain the edge information of macular and achieve accurate position of the macula foveal. Experimental tests showed good performance in the public fundus images database. For the normal and the pathological changes of the fundus images,the effective location rate of the macula was 96. 11%and 92. 12% respectively,and the average accuracy reached 93. 92%. Thus the proposed method based on morphological features and k-means clustering proved a simple,efficient and useful tool for computer-aided diagnosis of ocular diseases.