Convolutional Neural Network for glaucoma detection in fundus images

Villegas-Cubas J.E.; Capuñay-Uceda O.E.; Celi Arévalo E.K. · Proceedings of the LACCEI international Multi-conference for Engineering, Education and Technology (2024) · DOI: 10.18687/laccei2024.1.1.438

Resumen

Glaucoma is a silent disease and is the most common cause of irreversible blindness; early detection can prevent cases of blindness and improve the patient's quality of life. This research implements a convolutional neural network (CNN) for glaucoma detection in fundus images. The architecture of the proposed convolutional neural network consists of an input layer that receives an image, seven convolutional layers, five pooling layers, one flattened layer, two fully connected layers, and a two-class output layer. The proposed CNN was trained and tuned with 85 epochs using 3864 fundus images from the LAG dataset. The performance of the proposed CNN was evaluated with 990 images and 96.57% accuracy, 95.73% sensitivity and a specificity of 98.83% were obtained, which represent better performance compared to previous studies.

Instituciones: Universidad Nacional Pedro Ruiz Gallo