Abstract:
Diabetic Retinopathy is a disease of retina which affects patients with diabetes mellitus and it is a main reason for blindness. It is a disease in which the retinal blood vessels swell, this damages the retina of the eye and may lead to blindness if it doesn’t early and accurately detected.
The most effective treatment is early detection through regular screenings but fundus images are boor quality and the diagnosis of diabetic retinopathy depends on ophthalmologist, this make the process difficult by eye which can lead to mistake decision, so Automatic screening of these images would help the doctors to easily detect the patient's condition in more accurate way.
This research present an automatic detection of diabetic retinopathy using fundus image ,The proposed algorithm is performed using 131fundus image from mecca eye hospital (79 images were diagnosed normal and 52 images were diagnosed with edma), acquired image undergo preprocessing to be enhanced and then segmentation stage done by using k.means , Extraction of statistical and shape feature to fed it to the final classification stage to determine image as normal or edma.
After applying the above methods, it observed that the experimental result that the proposed system achieves better accuracy for Retinopathy with combined feature (shape and statistic) than each type of feature individually.
