Exploring Deep Learning Techniques for Automated Cataract Detection in Digital Camera Imagery

Cataracts, a prevalent eye condition impacting vision, demand accurate and prompt identification for effective risk reduction and blindness prevention. Traditional methods of cataract diagnosis rely heavily on manual examination by ophthalmologists, which can be time-consuming and subject to human error. With the advent of digital imaging technology, there is an opportunity to enhance the efficiency of cataract detection. The deep neural network models VGG19 and ResNet-50 are recommended in this paper for automatic cataract identification on digital camera images. To optimize network performance, adjustments to loss and activation functions were made, incorporating minimal kernels, reduced training parameters, and layers. This resulted in a ResNet-50 model with significantly decreased computational cost and average operating time compared to the cutting-edge models. The proposed ResNet 50 model achieved an average accuracy of 93.41%, outperforming the VGG 19 model, which had an accuracy of 91.13%. This research contributes to the evolution of cataract detection methodologies, emphasizing the potential of deep learning models, particularly ResNet-50, in improving diagnostic accuracy and operational efficiency for enhanced eye health outcomes.