Automated Eye Disease Prediction Using VGG19 Deep Learning Model for Cataract, Glaucoma and Normal Classification

Recent developments in artificial intelligence have brought new potential to the arena of medical diagnosis. Specifically, deep Ocular pathologies remain a predominant cause of irreversible vision impairment globally, where timely intervention is critical to halt disease progression and prevent blindness. This paper proposes an automated diagnostic system leveraging deep learning to classify prevalent eye diseases from retinal fundus images. The designed framework specifically discriminates between three distinct categories: Cataract, Glaucoma, and Normal (healthy) retinas. We take a transfer learning approach, using the pre-trained VGG19 model for powerful feature extraction and classification without having to start at the beginning and fund massive datasets. The experimental results show that the model is more accurately predictive in determining severe diabetic retinopathy degrees of diseased eyes while also highlighting the usefulness of deep convolutional neural networks in interpreting complex ophthalmic images. This work primarily aims to improve the ability of ophthalmologists to diagnose diseases by having a quick, inexpensive, and useful screening tool to improve the accessibility of eyewear in many clinical areas.