An Online Platform for Early Eye Disease Detection using Deep Convolutional Neural Networks

India has a blind population of about 12 million people, compared to 39 million people worldwide, and the sad fact is that 85 percent of these instances are curable. It has been found that the highest leading causes of eye blindness are due to Diabetic Retinopathy, Glaucoma and Cataract. Hence an automatic or self-diagnosing method using deep learning model is proposed in this paper to detect all the three diseases within a minute with high prediction rate. There have been various categorization systems developed throughout the years, all of which have improved significantly in the last decade or so. Logistic regression, SVM (Support Vector Machine), Decision tree, KNN (K- Nearest Neighbors), Random Forest, and Back propagation seem to be few of the numerous categorization models available. We propose a DCNN-based expert system in this study similarly to a human brain with input, neurons, hidden layers, and output for diagnosing three diseases through an online platform. Fundus images of both healthy and affected patients are acquired for this study under good lighting conditions so that any concealed characteristics may be recognised. The fundus images are then subjected to image processing techniques, including grayscale, resize, and power transform. Finally, a deep CNN with one hidden layer, 16 input neurons, and two output neurons that are either healthy or affected is generated. The detection accuracy was 91% for DR, 90 % for cataract and 86% for glaucoma affected images. A user-friendly, interpretable, online Graphical User Interface (GUI) was developed with the system.