Glaucoma Detection Using Convolutional Neural Network

One of the quintessential examples in ophthalmology is the early detection of glaucoma, a disease that accounts for significant irrecoverable vision loss when caught later. Conventional diagnostic methods are based on visual inspection, which may be time-consuming and open to subjective interpretation. We suggest a deep learning method using the Convolutional Neural Network (CNN) model to solve this. With the ORIGA data (a labeled fundus image collection where each has been identified as containing glaucoma), we trained CNN to learn about complex spatial patterns that may discriminate against the disease. We can see a reduction in loss and improvement in accuracy through the graphs; our model is learning well. A confusion matrix with highly accurate favorable rates also shows the effectiveness of CNN, and its predictions individually were accurate. Our CNN model reduces loss by up to 49.3% and improves validation accuracy by nearly 3.97%, which is highly competitive compared to standard logistic regression and machine learning techniques.