Machine Learning-Based Diabetic Retinopathy Detection: Exploring Techniques and Methods

Diabetic retinopathy, a prevalent eye disorder and a complication of diabetes, poses a risk of vision loss. Signs of diabetic retinopathy including subtle or entirely absent and eventual blindness are a possible outcome. This paper examines the various researches on the features of exudates, blood vessels, and microaneurysms, key components of diabetic retinopathy. These features enable the classification of retinopathy stages into healthy, slightly non-proliferative, temperately non-proliferative, strictly non-proliferative, and proliferative. In this work, three models are used for finding diabetic retinopathy stages include AlexNet, Inception v3, and EfficientNet. This study critically analyzes all proposed techniques, evaluating them based on complexity, interoperability, and accuracy. The Kaggle dataset is used for the implementation and validation of the proposed techniques.