A Hybrid Approach for Diabetic Retinopathy Detection Using Support Vector Machine

Diabetic Retinopathy is an eye disease that has been linked with diabetes mellitus. Diabetic Retinopathy is actually the main cause of blindness in working-age people, and almost 93 million people suffer from it. When such a disease is discovered in time, one's progress of vision impairments could be slowed down or managed at best. This makes it difficult because the disease is usually hardly ever symptomatic for early diagnosis and treatment. At the moment, the diagnosis of diabetic retina is currently a very long and cumbersome process that pertains to the assessment and evaluation by an ophthalmologist or other qualified medical practitioner of digital color fundus images of the retina in search anomalies resulting from the disease. Through the breakdown of the identification and decision-making procedures, an automated Diabetic Retinopathy screening approach will help in managing or controlling Diabetic Retinopathy progression. The study offers an automated classification method based on various lighting and field of view to analyze fundus images used in assessing the severity of Diabetic Retinopathy. This technique attains 89.2% sensitivity, 90.9% accuracy, 88.8% specificity, and 0.924 AUC for right classification of the images in five classes: from 0 (with no Diabetic Retinopathy) up to category 4 defined as proliferative Diabetic Retinopathy.