Glaucoma Detection Based-on Convolution Neural Network and Fundus Image Enhancement

The number of glaucoma sufferers is expected to increase to around 111.8 million in 2040 due to aging and population growth. It is necessary to develop technology to detect glaucoma automatically with a high degree of accuracy and efficiency. This research made the glaucoma classification system by comparing CNN architecture (AlexNet, GoogleNet, ResNet-18, ResNet-50, and ResNet-101) and implementing the CLAHE method as a pre-processing step. The evaluation system based on accuracy shows that the implementation of CLAHE can improve accuracy by about one to two percent. The average accuracy on Resnet-101 reached 88.24%, followed by average accuracy of ResNet-50, ResNet-18, AlexNet, and GoogleNet architectures with 87.82 %, 86.01%, 85.68%, and 85.63% accuracy, respectively. From all experiments, the best accuracy result is 89.59% on the ResNet-50 architecture experiment with a batch size of 16.