Automated Glaucoma Detection in Retinal Fundus Images Using Machine Learning Models

Abstract

Glaucoma, a neurodegenerative eye disorder, stands as a major global health concern, ranking second in causing blindness. The urgency for early detection is paramount to mitigate its irreversible effects on vision. This research presents an innovative model adept at analyzing retinal fundus images, aiming to assist ophthalmologists in early diagnosis. By harnessing the capabilities of Convolutional Neural Networks (CNN), Support Vector Machines (SVM), and k-Nearest Neighbors (KNN), our model demonstrated an impressive 86% accuracy in identifying diabetic retinopathy from retinal images. Focusing on glaucoma detection, we utilized the RIGA dataset, comprising 2,664 images, which are categorized into non-glaucoma (1,488 images) and glaucoma (1,176 images). Our model showcased a commendable testing accuracy of 94%. This endeavor not only amplifies the potential of machine learning in glaucoma prediction but also signifies a step forward in creating a user-centric, efficient diagnostic tool. Such innovations are pivotal in enhancing global eye care and reducing the prevalence of vision impairment due to glaucoma.