Page 128 - Computational Retinal Image Analysis

CHAPTER

7
OCT layer segmentation

a
a
a
Sandro De Zanet , Carlos Ciller , Stefanos Apostolopoulos ,
b
Sebastian Wolf , Raphael Sznitman c
a RetinAI Medical AG, Bern, Switzerland
b Department of Ophthalmology, Inselspital, University Hospital, University of Bern, Bern,
c
Switzerland ARTORG Center, University of Bern, Bern, Switzerland
Retinal layer segmentation methods in OCT images has been one of the most re-
searched areas of ophthalmic medical image analysis of the last decade. Largely
driven by improved image quality and the need to quantify retinal thickness and
perturbations, a variety of different automated methods have been proposed and have
been validated across patients with different retinal diseases. In this chapter we ex-
plore the task of OCT layer segmentation and some of the key methods of the past
decade, highlighting some of the challenges that remain.

1 Anatomical description and clinical relevance

Volumetric OCT imaging provides remarkable capabilities in visualizing retinal tis-
sue. With OCT, normal retinal tissue exhibits multiple retinal layers that sit on top of
each other to form the core elements of the retina. These cover the vast majority of
the posterior part of the eye and have the important role of transforming light energy
into neural signals to be interpreted by the brain. Briefly, light that interacts with
light-sensitive cells known as rods and cones, convert photons into action potentials
that are then transmitted by bipolar and ganglion cells. The axons of ganglion cells
ultimately exit the eye toward the brain by way of the eye’s optic nerve. Using OCT
imaging, several important retinal layers are visible in healthy retinas and are sum-
marized in Table 1.
These interconnected layers form the neural retinal tissue that lies above the cho-
roid and choriocapillaris. In most cases, OCT volumes centered on the macula image
these layers with high resolution, contrast and intensity (see Fig. 1).
OCTs of pathological retinas, however, can additionally depict a variety of differ-
ent biomarkers such a subretinal fluid (SRF), intraretinal fluid (IRF), intraretinal cysts
(IRC), fibrovascular pigment epithelium detachments (PED), reticular pseudodrusen

Computational Retinal Image Analysis. https://doi.org/10.1016/B978-0-08-102816-2.00007-1 121
© 2019 Elsevier Ltd. All rights reserved.