Chapter 12 - Deep learning for ophthalmology using optical coherence tomography

Abstract

The majority of recent deep learning research has been the reapplication of established image-processing techniques to retinal images. Another prominent and substantially richer imaging modality used by ophthalmic professionals is optical coherence tomography (OCT), which provides the same topological information as retinal images, as well as 3D tissue morphologies. The most common approaches to analyze this data similarly reapply established algorithms to 2D OCT cross-sections and rarely make use of all the available information. Only recently have datasets been made available to the public, enabling the development of novel deep learning architectures and research. This chapter discusses the fundamentals of OCT imaging, traditional algorithmic approaches to ophthalmic computer-aided diagnostics, and how these new datasets are being used in modern deep learning research.