Automated Detection and Classification of Diabetic Retinopathy using Machine Learning and Ensemble Techniques

Abstract

Diabetic retinopathy is an eye condition caused by chronic diabetes. Diabetes is a serious consequence because it can harm blood vessel tissue that is sensitive to light. People of working age are predominantly affected by this ailment. At first, diabetic retinopathy may show no symptoms. On the other hand, it can eventually result in blindness. The suggested approach makes use of the Local Binary Pattern feature extraction technique. The K Nearest Neighbour, Random Forest, and Logistic Regression algorithms are given the features that were extracted. Stacking, voting, and averaging are ensemble strategies that combine them during training and testing. Images of diabetic retinopathy, which are categorized as mild, moderate, no diabetic retinopathy, proliferating diabetic retinopathy, and severe, are gathered for this project from the Kaggle dataset. The experimental results show that averaging produces good results of 74%.