SISTEM DETEKSI RETINOPATI DIABETIK MENGGUNAKAN SUPPORT VECTOR MACHINE

Abstract

Diabetic Retinopathy is a complication of Diabetes Melitus. It can be a blindness if untreated settled as early as possible. System created in this thesis is the detection of diabetic retinopathy level of the image obtained from fundus photographs. There are three main steps to resolve the problems, preprocessing, feature extraction and classification. Preprocessing methods that used in this system are Grayscale Green Channel, Gaussian Filter, Contrast Limited Adaptive Histogram Equalization and Masking. Two Dimensional Linear Discriminant Analysis (2DLDA) is used for feature extraction. Support Vector Machine (SVM) and k-Nearest Neighbour (kNN) are used for classification. The test result performed by taking a dataset of MESSIDOR with number of images that vary for the training phase, otherwise is used for the testing phase. Test result show the optimal accuracy are 84% for 2DLDA-SVM and 80% for 2DLDA-kNN. Keywords : Diabetic Retinopathy, Support Vector Machine, Two Dimensional Linear Discriminant Analysis, k-Nearest Neighbour, MESSIDOR