Klasifikasi Tingkat Keparahan Diabetic Maculopaty Melalui Citra Retina Menggunakan Deep Residual Network (Resnet)

Abstract

Diabetic maculopathy (DM) is a microvascular complication that causes damage to the blood vessels in the center of the retina (macula) resulting in the build-up of fluid that causes impaired central vision. This disease is characterized by a thickening of the retina due to the build-up of lipid residue fluid that have hardened or known as macular edema. Eye doctors (ophthalmologists) usually perform examinations to identify DM disease using Optical Coherence Tomography (OCT) and Fluorescein Angiography (FA). Furthermore, retinal images received from FA and OCT will be analyzed manually to determine the severity of DM. Therefore, it needs a method that can help ophthalmologists in classifying the severity of DM. The method applied in this research is Deep Residual Network (ResNet). The system design process begins with preprocessing in the form of augmentation, resizing, grayscaling, and contrast stretching. Then the image classification process is carried out by applying the trained model. Based on the result of research using ResNet, the system can classify the severity of diabetic maculopathy with accuracy of 72% in the model without augmentation and 85% in the model with augmentation.