Vision Beyond the Haze: Advancements in Cataract Detection and Diagnosis

Eye diseases are among the most critical health issues impacting human life. One of the most frequently encountered eye problems is cataracts, due to which people suffer as they age. The cloudiness that forms on eye lenses is known as cataracts which can significantly damage vision. Diagnosing the disease in its acute stage rather than its onset should not blind the patient. This study presents an improved Cataract detection method using pre-trained convolutional neural networks (CNNs) on an Ocular Disease Intelligent Recognition (ODIR) dataset, Comparing the performance of ResNet50, ResNet101, and VGG19 architectures to find the most effective model for this project. The proposed approach uses deep learning to analyse fundus images, aiming to increase diagnostic accuracy, speed, and reliability. several preprocessing and data Augmentation techniques were utilized to prepare the data set, which provided robust training and improved model generalization. Samples were refined using pre-trained ImageNet weights and adjusted for binary classification of Cataract presence. Among the tested models, ResNet50 exhibited the best performance, achieving an overall accuracy of 99.08% It also showed good evaluation metrics, with a precision rate of 98.15% and 100% recall for cataract images, and 100% precision with 98.23% recall for normal images. The results highlight the potential of ResNet50 to provide accurate and effective cataract diagnosis, especially in areas where access to trained ophthalmologists is inaccessible. This study highlights the importance of using deep learning technologies to improve the early detection and treatment of cataracts, ultimately contributing to better patient outcomes and global eye health progress.