Insights into Fundus Images to Identify Glaucoma Using Convolutional Neural Network

Glaucoma, an eye disease, is a multi-factorial neuro-degenerative disease that vitiates vision over time and which may cause permanent vision impairment. In recent years, machine learning has been used with the idea of using algorithms to find patterns and/or mark extrapolations based on a collection of data. The detection of glaucoma has been achieved with various deep learning (DL) models so far. This paper presents the Convolutional Neural Network (CNN) approach for the diagnosis of glaucoma with remarkable performance. In this approach, glaucoma and healthy images can the differentiated because it forms patterns that can be detected with the CNN. The fundus images are used as image modality which includes publically available retinal image datasets as IEEE DataPort, Drishti-GS and Kaggle Dataset. The analysis is performed for selected datasets, it is observed that, the IEEE DataPort dataset gives better results than others and obtained values of accuracy, sensitivity and specificity are 95.63%, 100% and 91.25% respectively.