Diagnosis of Diabetic Retinopathy by Using Image Processing and Convolutional Neural Network

Diabetic retinopathy is a serious eye disease that originates from diabetes mellitus and is the most common cause of blindness in the developed countries. This study describes the use of image processing and deep learning to diagnose diabetic retinopathy from retinal fundus images. For retinal fundus images enhancement approach, a practical method which contains HSV, V transform algorithm and histogram equalization technics was used. Finally, Gaussian low-pass filter was applied to the retinal fundus image. After the image processing, the classification was made using the Convolutional Neural Network. The performance of the proposed method was assessed using 400 retinal fundus images in the Kaggle Diabetic Retinopathy Detection database. In experiments, classification work has been done for each stage of the image processing. The classification study performed after image processing. Twenty experiments were done for every stage and average values were found. In this experiment, the accuracy was 97%, the sensitivity was 96.67%, the specificity was 93.33%, the precision was 97.78%, the recall was 93.33%, and the F-score was 93.33%. The obtained results show that the proposed method is very efficient and successful to diagnose diabetic retinopathy from retinal fundus images.