Computer-Aided Diagnosis of Cataract Disease through Retinal Images

A Cataract is the most common cause of sight impairment around the globe. Therefore, determining the presence and severity of cataracts is critical for analysis and movement monitoring. This research proposes a framework for the automatic classification of cataracts through retinal fundus images. Moreover, this research provides an automatic structure for diagnosing and analyzing eye disease. Further-more, image augmentation is utilized to increase the number of images in the dataset. In addition, pre-processing is used to remove the noise in the images. Convolutional Neural Networks, GoogleNet, and AlexNet are proposed in this research for the classification of normal and cataract-affected images. A total of 449 cataract and normal fundus images are acquired for testing and training the models. The dataset was increased to 1159 images using a data augmentation technique. Exploratory outcomes demonstrate that the GoogleNet attained 97% accuracy, 90% sensitivity, and 90% specificity. On the other hand, AlexNet attained 90% accuracy, 88% sensitivity, and 85% specificity. Outcomes show that GoogleNet beats the best in class cataract detection approaches and can assist doctors in real-time cataract disease detection applications.