Non-Proliferative Diabetic Retinopathy Classification Based on Hard Exudates Using Combination of FRCNN, Morphology, and ANFIS

One of retinal eye disease caused by complications of diabetes mellitus is Diabetic Retinopathy. This disease consists of several levels. Severe diabetic retinopathy can cause blindness for the sufferer. The presence of hard exudate in the retinal fundus image is one symptom of diabetic retinopathy. That lesions are utilized to categorize two severity levels in diabetic retinopathy. Those are the severe and moderate NonProliferative Diabetic Retinopathy (NPDR). This research is using Faster Region-based Convolutional Neural Network (FRCNN) to remove the optic disk, mathematical morphology method to process hard exudates segmentation and Adaptive Neuro-Fuzzy Inference System (ANFIS) method to process the classification. The accuracy level of the classification system in this research was 83.54 %. The result of this research can be utilized as additional decision support for the ophthalmologist. This research is expected to help the ophthalmologist and the community in the prevention of diabetic retinopathy. So, this research is also expected to reduce the level of blindness caused by diabetic retinopathy.