眼底图像中自动定位黄斑中心凹的快速定位方法研究

Study on the method of fast positioning of automatically positioning central fovea of macula in the images of eye fundus

目的:研究自动定位眼底图像中黄斑中心凹位置的快速方法。方法:设计一种血管主干与视盘定位结合迭代阈值变换定位黄斑中心凹的方法,运用形态学变换与连通域计算提取出视网膜血管主干;在图像中选取图像中灰度最高的1%的点作为视盘感兴趣区(ROI)的候选,筛选出距离血管主干骨架最近的亮区作为视盘ROI,并通过形态学处理、阈值变换、Hough变换等步骤识别出视盘中心与视盘半径;最后在视盘的周围2.5倍视盘半径附近,建立黄斑ROI,并采用迭代阈值方法构建灰度“等高线”,确定黄斑中心凹位置中心。使用该方法对医院240张眼底图像进行测试。结果:在240幅眼底图像中,使用该定位方法进行测试,定位黄斑中心凹的准确率达97.1%。结论:实验结果表明,血管主干与视盘定位结合迭代阈值的自动定位黄斑中心凹的方法,可以定位眼底图像中黄斑中心凹的位置,且具有较好的准确性与适用性。

Objective: To investigate fast method of automatically positioning central fovea of macula in the images of eye fundus. Methods: A new method, which using iterative threshold changing combined with the localization of retinal vascular trunk and optic disc (OD), was proposed to locate on the central fovea of macula. First, the retinal vascular trunk was extracted by morphological transformation and connected domain computing. And then, 1% point of highest gray level of images was selected as the candidate of region of interest (ROI) of the OD, and the nearest bright area away main skeleton of blood vessel was further screened from them as ROI of OD, and the center and radius of OD were recognized through morphology processing, threshold transform and Hough transform and other steps. Finally, the ROI of the macula was established nearby 2.5 times of the radius of OD, and the isoheight of gray level were constructed by adopting iterative threshold value so as to determine the center of central fovea of macula. Used this method to test 240 images of eye fundus of our hospital. Result: In this test of 240 images of eye fundus by using this method, the accurate rate of positioning central fovea of macula reached to 97.1%. Conclusion: Based on experiment data, the proposed method could locate on the position of central fovea of macula in the images of eye fundus and it has better accuracy and applicability.