Page 30 - Computational Retinal Image Analysis

20 CHAPTER 3 The physics, instruments and modalities of retinal imaging

an image of the retina we are using the camera in reverse: in this way, it has the
characteristics of a microscope in which the cornea and lens focus the light like a
microscope objective and the retina sits in the object plane like a biological sample.
A key distinction between the eye and a camera or microscope however is that
the eye is integrated within the human visual system and the brain, and has evolved
to provide a necessary-and-sufficient performance for the tasks we need to accom-
plish. The eye actually has a relatively poor optical performance, but the brain is
able to compensate for many of the shortcomings to provide a perceived wide-field
image quality that exceeds the objective quality of the eye. For example even a low-
cost camera system, such as is found in a modern mobile phone, is able to record
a 10-Megapixel image with a uniformly high angular resolution of 0.015° across a
field of 60°. Conversely the human eye is able to form images over a field of view of
almost 200°, but a comparable angular acuity to the low-cost camera can be achieved
for only a central field of view of 0.3° or about 0.001% of the total visible solid angle.
The limitations on the acuity of vision are determined by a combination of the opti-
cal aberrations of the eye and also the variation in concentrations of photoreceptors
(the rods and cones). For imaging the retina, where we use the optical system of the
eye as a microscope, it is fundamental physics and the quality of the eye optics that
determines our ability to record high-resolution images of small retinal features.
To get the best performance out of any electronic imaging system it is common to
compensate for systematic effects such as fixed pattern noise (using flat fielding) or
interpolation to correct for malfunctioning pixels. The human visual system demon-
strates similar capabilities: an image of the retina recorded with an ophthalmoscope
shows a network of blood vessels and an optic disc (corresponding to the ‘blind
spot’), yet the image perceived by the eye's owner is normally devoid of such pat-
terning. Indeed retinal disease can lead to extended areas of the retina not functioning
before the effects become overtly apparent—and this highlights the importance of
the early detection of disease using retinal imaging. Quantitative computer analysis
of retinal images is therefore important for both fundamental understanding of retinal
processes and for screening for disease. The aim of this chapter is to explain the main
physics that underpins the observable characteristics of retinal images that are impor-
tant for quantitative computer analysis. We use the established concepts of optics and
image formation as the basis for our discussion and the interested reader is invited to
consult one of the many excellent text books on optics, such as the accessible Optics
by Hecht [5] and the more advanced Principles of Optics by Born and Wolf [6], for
explanation on the underpinning physics.
In Section 2 we discuss the important principles for imaging the retina, where
the eye acts as a microscope, including the effects of aberrations, diffraction and
reflections. In Section 3 we review the main optical instruments or imaging mo-
dalities used for retinal imaging, including an in-depth review of Optical Coherence
Tomography for its increasing importance. We discuss how differing imaging modal-
ities affect the characteristics of retinal images in Sections 2.4 and 2.5. In Section 4
we provide a brief summary of the importance of polarization and birefringence in
retinal imaging.