A Data Mining Inspired Methodology towards the Identification of Diabetic Retinopathy

Abstract

The Biomedical image analysis technique used in most of the clinical diagnosis activities, which is one of the explorative areas that appeal intense significance among scientists. The retinal fundus images are utilized in clinical diagnosis extensively for the treatment and to observe various eye diseases. Diabetic retinopathy is one of the foremost sources for blindness. The major diagnostic sign of diabetic retinopathy is the damage of blood vessels due to various reasons in the eye and then establishment of lesions in the retina. The screening and detection of Diabetic Retinopathy can be performed using retinal fundus images. The identification and analysis of diabetic retinopathy (DR) by means of color fundus images involves experienced practitioners to recognize the existence of many small topographies with a detailed grading system, makes this a complex and time-consuming mission. In this paper, a novel systematized method for the discovery of exudates in retinal images to diagnose diabetic retinopathy. The color fundus images are characterized and analyzed to find microaneurysms on the retina and provides the severity. The algorithm is tested on datasets provided by ophthalmologists and Messidor dataset, which gave excellent and promising results.