Diabetic Retinopathy Diagnosis System Based on Retinal Biomarkers Using EfficientNet-B0 for Android Devices

Currently, Diabetic Retinopathy has become one of the main causes of vision loss for the working-class population (20-65 years) and the disease is an asymptomatic disease which makes detection of the disease without medical assistance more difficult. Along with the development of the era, many new methods have been found in diagnosing diseases, one of which is Computer Aided Diagnosis (CAD) which relies on computer technology to assist medical personnel in diagnosing diseases. This research uses CAD with machine learning to classify Diabetic Retinopathy using transfer learning on the EfficientNet-B0 model which is implemented in a program for mobile devices for android. The program was designed with the aim of helping medical personnel to be able to diagnose Diabetic Retinopathy in areas that do not have sufficient medical facilities using only a smartphone, 20D lens, and several other drugs supports. The results showed that the EfficientNet-B0 model for classifying Diabetic Retinopathy was able to produce an accuracy of 91.85 percent in classifying three classes namely no DR, Non-proliferative DR, and Proliferative DR. This research was done to contribute in the growth of Artificial intelligence involvement in the medical world and also to help areas that lack the medical equipment to have better treatment.