Abstract— Diabetic retinopathy (DR) is one of the complications in the retina caused by diabetes. The symptoms shown by patients with DR, among others mikroaneurysms, hemorrhages, hard exudate and soft exudates. These symptoms at a certain intensity can be an indicator of phase (severity) of diabetic retinopathy. DR severity levels are divided into four classes namely: Normal, Non-Proliferative Diabetic Retinopathy (NPDR), Proliferative Diabetic Retinopathy (PDR), and Macular edema (ME) .The system built in this thesis is the detection of diabetic retinopathy level of images obtained from STARE (Structured Analysis of the Retina). There are four main stages to resolve the problems of the pretreatment, extraction of anatomical structures, feature extraction and classification. Pretreatment methods are used including gray image (grayscale), a Gaussian filter, Histogram retinal image with wavelet de noising and Masking. The retinal image using neural network trained with backpropagation algorithm for classification. The resulting performance of this approach is the sensitivity 100% , sfesificity 95%, accuracy 96%. Keywords: Diabetic retinopathy, Neural Network, Backpropagation, STARE.
Copyrights © 2017