A Systematic Review of Deep Learning Methods Applied to Ocular Images

Abstract

Artificial intelligence is having an important effect on different areas of medicine, and ophthalmology has not been the exception. In particular, deep learning methods have been applied successfully to the detection of clinical signs and the classification of ocular diseases. This represents a great potential to increase the number of people correctly diagnosed. In ophthalmology, deep learning methods have primarily been applied to eye fundus images and optical coherence tomography. On the one hand, these methods have achieved an outstanding performance in the detection of ocular diseases such as: diabetic retinopathy, glaucoma, diabetic macular degeneration and age-related macular degeneration.  On the other hand, several worldwide challenges have shared big eye imaging datasets with segmentation of part of the eyes, clinical signs and the ocular diagnostic performed by experts. In addition, these methods are breaking the stigma of black-box models, with the delivering of interpretable clinically information. This review provides an overview of the state-of-the-art deep learning methods used in ophthalmic images, databases and potential challenges for ocular diagnosis