The increase of diabetic retinopathy patients and diabetic mellitus worldwide yields lot of challenges to ophthalmologists in the screening of diabetic retinopathy. Different signs of diabetic retinopathy were identified in retinal images taken through fundus photography. Among these stages, the early stage of diabetic retinopathy termed as microaneurysms plays a vital role in diabetic retinopathy patients. To assist the ophthalmologists, and to avoid vision loss among diabetic retinopathy patients, a computer-aided diagnosis is essential that can be used as a second opinion while screening diabetic retinopathy. On this vision, a new methodology is proposed to detect the microaneurysms and non-microaneurysms through the stages of image pre-processing, candidate extraction, feature extraction, and classification. The feature extractor, generalized rotational invariant local binary pattern, contributes in extracting the texture-based features of microaneurysms. As a result, our proposed system achieved a free-response receiver operating characteristic score of 0.421 with Retinopathy Online Challenge database.

中文翻译：

---

使用纹理描述符检测眼底图像中的微动脉瘤的辅助观察者系统。

全世界范围内糖尿病性视网膜病患者和糖尿病患者的增加对眼科医师在糖尿病性视网膜病的筛查中提出了很多挑战。在通过眼底照相拍摄的视网膜图像中发现了糖尿病性视网膜病变的不同体征。在这些阶段中，称为微动脉瘤的糖尿病性视网膜病的早期阶段在糖尿病性视网膜病患者中起着至关重要的作用。为了帮助眼科医生，并避免糖尿病性视网膜病患者的视力丧失，计算机辅助诊断是必不可少的，可以在筛查糖尿病性视网膜病时用作第二意见。因此，提出了一种通过图像预处理，候选提取，特征提取和分类等步骤检测微动脉瘤和非微动脉瘤的新方法。特征提取器 广义旋转不变局部二进制模式，有助于提取微动脉瘤的基于纹理的特征。结果，我们提出的系统通过视网膜病变在线挑战数据库获得了0.421的自由响应接收器操作特性评分。