Machine Learning for Diabetic Retinopathy: Gaussian SVM Model for Accurate Analysis in Retinal Fundus Images

The most common ocular ailment in people with diabetes that causes vision loss in this population is diabetic retinopathy. Patients are protected from vision loss by diabetic retinopathy. The findings of this study point to the need for an early diagnosis system that can help the physician in analyzing retinal images. The main goal is to classify severity level from any given retinal images automatically. To accomplish this, an initial image processing phase is utilized to distinguish and isolate blood vessels, micro-aneurysms, and hard exudates. This process enables the retrieval of particular attributes that can be applied by a support vector machine in order to ascertain the severity of retinopathy for every retinal image. The efficacy of this idea has been evaluated using a database consisting of 300 retinal images that have been categorized based on the severity of diabetic retinopathy. The outcome yielded a maximum sensitivity of 95.95% and an accuracy of 92.66%. An assessment of the algorithm’s resilience to variations in its parameters has also been conducted. The suggested framework will enhance the efficiency and accuracy of disease detection compared to existing frameworks