Morphological approach for Retinal Microaneurysm detection

Abstract

Diabetic Retinopathy is an end-stage micro vascular complication of Diabetes where the retina is impaired as a result of fluids exuding out of blood pathways into the retina. Image analysis and computer vision techniques are employed to identify different Iesions associated with Diabetic Retinopathy. The existence of microaneurysms in the retina is the primal visible sign of Diabetic Retinopathy. Early automated microaneurysm detection can help detect the onset of Diabetic Retinopathy. The number of microaneurysms are counted for automatic ranking of the severity of the malady. In this paper, we have investigated morphological methods for the automatic detection of microaneurysms from digital color images of the retina. Retinal characteristic components such as vessels or optic disc and exudate Iesions which may cause false detection are also detected. Using a ground truth, from a DIARETBI database, microanurysms were detected with 80.41% sensitivity and 92.79% specificity.