Categorization of Diabetic Retinopathy Applying Ensemble Model

Abstract

Regardless of the advancement in the science and medical field, diabetes still emerges as a big threat to humanity. The only medicament that will help in reducing the effects of this disease to minimum is early detection and taking prophylactic steps toward it. One of the complications that affect eyes is diabetic retinopathy (DR). A diabetic causes impairment in the blood vessels of the tissue that is light sensitive at the retina. As per survey conducted by International Diabetes Federation in the year 2015, it was mentioned that it affects approximately 410 million people, worldwide. India is a commorancy to approximately 70 million people with this disease. Diabetic eye disease or DR is the most common impediment of diabetes. Round about 2.6% of global blindness is caused by this disease. In unmitigated terms, just about 3–4.5 million individuals in India are anticipated to fall victim to vision threatening diabetic retinopathy (VTDR). The choices for treatment of VTDR demand costly devices and medications, and a regular follow-up from diagnosis is required to the last day of your life. In deliberation of the fact that 70% of the Indian citizenry depend on necessitous expenses for their healthcare benefit, one person with VTDR in a household is enough to drive a menage to below poverty line. Therefore, all undertakings should be initiated exigently to prevent individuals with diabetes to move into the vicious cycle of diabetes, blindness, and poverty. This proposed work will classify the stages of diabetic retinopathy with an approximate precision of 97.8% which is an improvement from the previous model with a margin of around 10%. We have used pretrained models in our ensemble learning process.