Abstract:

Diabetes is a disease that affects the structure of the eye and causes vision loss. It causes the formation of many types of damage in the eye structure. These injury on the retina images are a sign of different diseases. The most known of these diseases is diabetic retinopathy disorder. This disorder is very important in early diagnosis and treatment of injury detection. In the study, a computer-supported detection system based on the Regional-Evolutive Neural Networks was proposed to detect the damage found on the retina images. The aim of this recommended system is to support the diagnosis and treatment of specialists working in the field of eye diseases. The retina images used in the study were obtained from the STARE, DIARETDB0 and DIARETDB1 databases. 70% of the images in the used databases are assigned to training and 30% to test images. The Regional-Evractual Neural Networks are designed in an area selector that allows the selection of more than one area through a single image in order to label the training images due to the need for a lot of data in the training stage. Retina images are trained on the pre-educated network of cifar-10 which is commonly used in deep learning applications. At the end of the training, the results of testing in the STARE, DIARETDB0 and DIARETDB1 databases were 91, 98.53 and 93, respectively. It has been successfully identified with 01 accuracy.

Keywords: