Detection of micro-aneurysms in low-resolution color retinal images

Diabetic retinopathy is the most common cause of blindness in the industrial countries. The best way to prevent visual losses from diabetes is early detection. From photographic and angiographic images, an experienced physician can find a lot of useful information to assess the development and progression of diabetic retinopathy, but this procedure is time-consuming. Automatic detection of indicators of diabetic retinopathy is a more efficient method. The goal of this project is to develop and optimize an algorithm to detect micro-aneurysms and small round hemorrhages in low-resolution color retinal images. Our algorithm is adapted from a micro-aneurysm detector developed for high-resolution angiographic films. It consists of pre-processing, morphological filtering, thresholding, region growing and feature analysis. This algorithm has been tested on retinal images of 640 x 480 resolution. Experimental results show that it is possible to set up parameters to achieve 100% global sensitivity with up to 80% global specificity, which means that all affected individuals could be directed to specialists without too much overload