An Automated Deep Learning Framework For Fundus Disease Detection Using Retinal Images

Eyecare practitioners will explain how retinal disease can be defined as any disease that develops in the retina, the backside of the globe’s “light”-sensitive region. Some diseases which fall in this category are Glaucoma, cataract, diabetic retinopathy to mention a few, and these diseases are viewed with great concern to the world as they cause great consequences regarding loss of vision as their effects are irreversible. Therefore, the world is in urgent need for efficient and effective diagnosing systems. This project introduces an automatic methodology for identifying and classifying the various stages of retinal disease in images using Machine Learning (ML) and Deep Learning (DL) approaches. This approach leverages advanced Convolutional Neural Networks (CNNs) to classify three key retinal diseases: Glaucoma, Cataract, and Diabetic Retinopathy. Trained on diverse, large-scale datasets, the model effectively identifies complex patterns in retinal images, ensuring robust and adaptable disease categorization. The use of VGG16, achieving 88% accuracy, and MobileNet, achieving 93% accuracy, enhances the diagnostic process, making it more accurate and accessible for early detection and treatment planning.