| dc.description.abstract | Hypertension causes disease complications that can lead to death. Hypertensive retinopathy causes damage to the retina or blood circulation around the retina due to high blood pressure which can lead to blindness. Hypertensive retinopathy can be detected by fundoscopic examination by ophthalmologists, which are very limited in number and are not evenly distributed in Indonesia. In addition, the analysis of the funduscopy is also still manual and processed for a long time, thus allowing for errors in carrying out further follow-up on patients. Therefore, a method is needed to help detect hypertensive retinopathy through fundus images automatically. The Deep Belief Network (DBN) method used in this study and retinal fundus image as input data in the image processing process. Several steps were taken before the identification step, namely pre-processing in the form of resizing, green channel, CLAHE, average filter, background exclusion, segmentation thresholding process, feature extraction process in the form of GLCM (Gray Level Co-occurrence Matrix). The data used in this study were obtained from the STARE (Structure Analysis of the Retina) dataset site, DR HAGIS, kaggle.com site, Prima Vision Eye Hospital and North Sumatra University Hospital with a total of 800 data, 640 data for training, 80 data for validation and 80 data for the testing process. The results of system testing in this study in identifying hypertensive retinopathy obtained a validation accuracy of 95%, testing accuracy of 92.5%, sensitivity 90.4%, and specificity 94.7%. | en_US |
