为实现糖尿病视网膜病变（糖网）的自动筛查，建立了基于免散瞳眼底图像的糖网自动筛查方法。该方法包括视盘定位及提取、糖网白色病灶（硬性渗出、棉绒斑）自动检测以及微动脉瘤与视网膜内出血的自动检测。在此基础上设计并实现了基于免散瞳眼底图像的糖网自动筛查系统。利用已实现的系统对临床采集的7 687个样本共15 374幅眼底图像进行糖网自动筛查，对样本个体的检测结果为：灵敏度96.46%，特异性96.07%，平均处理时间57.87 s。测试结果表明，所构建的基于免散瞳眼底图像的糖网自动筛查系统满足英国糖尿病协会提出的糖网自动筛查标准（最低灵敏度80%，最低特异性95%）。

In order to realize automatic screening for diabetic retinopathy (DR), an automated diabetic retinopathy screening system based on non-dilated fundus images is proposed and studied. Firstly, the composition and key technology of the system including locating and extracting optic disc and automated detection of hard exudates, cotton wool spots, hemorrhages as well as microaneurysms were analyzed and reviewed. Then the system based on non-dilated fundus images was developed and tested on an image dataset collected in clinic which include 7 687 samples in total of 15 374 fundus images. With an exam-based criterion, sensitivity of 96.46%, specificity of 96.07% are achieved. Furthermore, the average time cost in processing a sample is 57.87 seconds. Results suggest that the performance of DR screening system meet the minimum standard of 80% sensitivity and 95% specificity for the detection of sight-threatening DR which was recommended by British Diabetic Association guidelines.