Diabetic Retinopathy Abnormality Classification And Identification Using Mathematical Morphological And Machine Learning Approach

Abstract

Diabetic Retinopathy (DR) is a leading cause of vision impairment and blindness among working-age adults globally. Early detection and precise classification of DR abnormalities are crucial for timely intervention and treatment. This paper proposes a hybrid approach combining Mathematical Morphological operations and Machine Learning techniques to identify and classify DR-related abnormalities from retinal fundus images. The proposed methodology involves pre-processing using morphological techniques such as dilation, erosion, opening, and closing to enhance features like microaneurysms, exudates, and hemorrhages. Feature extraction is performed on the enhanced images to capture texture, shape, and intensity characteristics. These features are then used to train supervised machine learning classifiers such as Support Vector Machines (SVM), Random Forests (RF), and K-Nearest Neighbors (KNN) for accurate classification of DR stages (No DR, Mild, Moderate, Severe, and Proliferative DR). The system is evaluated on publicly available datasets, and performance metrics such as accuracy, sensitivity, specificity, and F1-score are computed. The results demonstrate that the integration of morphological preprocessing significantly improves the classification accuracy of machine learning models, making the proposed approach effective and reliable for clinical decision support in DR screening programs