<![CDATA[Cataract is a condition that causes clouding of the lens of the eye, leading to blindness and poor vision. According to the WHO, around 18 million people suffer from cataract-related blindness, making it one of the leading causes of blindness globally. Prompt and accurate diagnosis is essential to prevent more serious outcomes. This research aims to develop a deep learning model that utilises Convolutional Neural Networks (CNN) in categorising cataract severity into four groups: hypermature, normal, immature and mature. This model is expected to provide a more efficient and accurate alternative to traditional methods in diagnosing cataracts. To achieve this, we implemented transfer learning using three popular CNN architectures: VGG16, VGG19, and ResNet50. Experiments were conducted using a dataset of pre-labelled eye images for training. Model performance was evaluated by calculating F1-score, recall, accuracy, and precision using a confusion matrix. The results showed that VGG19 produced 88% accuracy and F1-score of 0.87, while VGG16 had the best accuracy. On the other hand, ResNet50 showed the lowest accuracy with 63% and F1-score of 0.59. These findings highlight the importance of selecting the right CNN architecture for cataract diagnosis, while underlining the potential application of deep learning in ophthalmology.]]>
Copyrights © 2025
