An Investigation on Diabetes-Based Eye Disease Detection Methods using Deep Learning and GAN Techniques

Abstract

Machine learning and deep learning play successful and influential roles in image detection and classification in many medical imaging diagnosis areas. Diabetes is becoming a significant health concern, and diabetic eye diseases (DEDs) will be the leading cause of vision loss worldwide. This paper presents a review on the state-of-the-art studies concerned with the detection, classification, segmentation, and grading of diabetic eye diseases, including the common four eye diseases: diabetic retinopathy (DR), diabetic macular edema (DME), glaucoma, and cataracts. We classify the model techniques into three main categories: classification-based, localization-based, and generative adversarial network (GAN) models. We investigate the current and commonly used datasets available for fundus images of diabetic eye diseases. We also review research that employed different GAN techniques to improve the fundus image dataset or increase the size of the images in the datasets. Finally, we illustrate the performance measures used in the previous studies for evaluating various models of diabetic eye diseases.