中文摘要：

眼底微动脉瘤是糖尿病视网膜病变最早期的症状,准确检测眼底图像中的微动脉瘤对糖尿病视网膜病变的筛查具有重要意义.提出一种基于相位一致性模型的微动脉瘤检测方法.首先采用相位一致性模型获取微动脉瘤候选者,然后通过构建灰度剖面图去除图像中血管片段等无关信息,从而筛选出真正的微动脉瘤.通过对ROC网站提供的50幅眼底图像进行实验,在图像水平上实现了灵敏度94％、特异性100％、准确率96％的检测效果.结果表明,该方法对图像的亮度、对比度不敏感,能够高效自动地检测出彩色眼底图像中的微动脉瘤.

英文摘要：

The presence of microaneurysms in the retina is the earliest clinical symptom of diabetic retinopathy （DR） , thus their reliable detection is essential in the DR screening system. Based on phase congruency, this paper proposes a new microaneurysms detection method. The first step aimed at obtaining microaneurysms candidate regions achieved by using phase congruency. Then the irrelevant information, such as the vessel fragments, was removed by constructing directional cross-section profiles. Through testing on 50 fundus images provided by ROC website, the method achieved a sensitivity of 94% , specificity of 100% , and accuracy of 96% at the image level, respectively. This method can accurately get microaneurysms in color fundus images, and it is insensitive to image brightness and contrast.