Penerapan Convolutional Neural Network Untuk Klasifikasi Tingkat Keparahan Retinopati Diabetik Pada Penderita Diabetes Melitus

Abstract

Diabetic retinopathy is a disease that can interfere with retinal blood vessels that cause blindness for people with diabetes mellitus. If the disease is treated too late then the patient can experience blindness. Proper care and examination can help prevent the increasing severity of diabetic retinopathy. Manual examination by an ophthalmologist in diagnosing this disease requires a relatively long time, so we need a system to classify the severity of diabetic retinopathy. The system designed in this study uses the Convolutional Neural Netwok method to classify the severity of diabetic retinopathy. The severity of Diabetic Retinopathy is divided into 5 classes namely NO DR, Mild, Moderate, Severe, and Proliferative DR. Research on the Application of Convolutional Neural Netwok to Classify the Severity of Diabetic Retinopathy in Diabetes Mellitus Patients using 64 x 64 x 3 images with RGB channels. The image pre-processing stage is done by resizing the image. The CNN architecture used consists of 5 blocks where each block contains a batch normalization layer, convolution layer, max pooling layer using the learning rate parameter of 0.0005. The results of the evaluation of the 652 model test data showed the best accuracy of 91.10%.