Automatic Detection of Microaneurysms in RGB Retinal Fundus Images

Abstract

In this study, an efficient and fast-working method to detect microaneurysm lesions, first symptom of diabetic retinopathy, is described. The proposed method is based on mathematical morphology, object pixel classification and connected component analysis. The proposed algorithm responses in 4.8 seconds for 2048x1536 pixel images. This shows this system runs faster than other microaneurysm detection systems. The sensitivity and specificity of this system is 69.1% and 99.3% specificity, respectively.

Keywords: Connected component analysis, Diabetic retinopathy, Microaneurysm, Mathematical morphology.