Deep Learning in Retinal Image Analysis: A Review

Abstract

Retinal image analysis holds an imperative position for the identification and classification of retinal diseases such as Diabetic Retinopathy, Hypertensive Retinopathy, Glacuma, Age Related Macular Degeneration, Retinal Detachment, and other retinal disease. Automated identification of retinal diseases is a big step towards early diagnosis and prevention of exacerbation of the disease. A number of state-of-the-art methods have been developed in the past that helped in the automatic segmentation and identification of retinal landmarks and pathology. However, the current unprecedented advancements in deep learning and modern imaging modalities in ophthalmology have opened a whole new arena for researchers. This paper is a review of deep learning techniques applied to 2-D fundus and 3D-OCT retinal images for automated identification of retinal landmarks, pathology, and disease classification. The methodologies are analyzed in terms of sensitivity, specificity, Area under ROC curve, accuracy, and F score on publically available datasets which include DRIVE, STARE, CHASE-DB1, DRiDB, NIH AREDS, ARIA, MESSIDOR-2, E-OPTHA, EyePACS-1 DIARETDB and OCT image datasets.