Imaging the Retinal Vasculature

Abstract

Advances in retinal Imaging are enabling researchers and clinicians to make precise noninvasive measurements of the retinal vasculature in vivo. This includes capillary blood flow and regulation and oximetry as well as the complete mapping of perfused blood vessels. These advances promise to revolutionize our understanding of vascular regulation as well the management of retinal-vascular diseases. This review provides an overview of imaging and optical measurements of the function and structure of the ocular vasculature. We include general characteristics of vascular systems with an emphasis on the eye and its unique status. The functions of vascular systems are discussed, along with physical principles governing flow and its regulation. Vascular measurement techniques based on reflectance and absorption are briefly introduced, emphasizing ways of generating contrast. One of the prime ways to enhance contrast within vessels is to use techniques sensitive to the motion of cells, allowing precise measurements of perfusion and blood velocity. Finally, we provide a brief introduction to retinal vascular diseases.

1. INTRODUCTION TO VASCULAR SYSTEMS AND THE RETINAL VASCULATURE

1.1. Background and Purpose.

This review provides a snapshot of the research that uses optics, particularly imaging, to analyze the structure, function and health of the vasculature of the eye (Burns et al. 2019, Cuenca et al. 2020). We concentrate on the wide array of techniques for imaging the retinal vasculature (Figure 1), as well as providing a brief review of the vascular system. The review is oriented towards non-invasive or minimally invasive human studies, adding basic results and data from non-human imaging primarily to expand the scientific foundation of the review.

Figure 1.

Vascular imaging of the human retina. (a) Scanning laser ophthalmoscope images, montaged from the right eye of a normal subject. The central retinal artery enters the eye through the optic disc (red arrow) and rapidly divides to produce multiple branches (A) that course across the retina. Veins return from the retina (V), typically interleaved between arterioles. The fovea (Fov) is centered in this image, which extends about 50 degrees horizontally and 30 degrees vertically. (b) A foveal centered OCT image of the retina extending to the edge of the optic nerve (right). This shows the foveal pit where ganglion cells are displaced from the center of the fovea, as well as the increased thickness of the nerve fiber layer near the optic disc. (c) An OCT angiogram (OCT-A) cross-section of the superior retina. The yellow indicates areas of blood flow. OCT-A images of the (d)superficial), (e) intermediate, (f) and deep vascular plexuses. (g) En face OCT-A image of the choriocapillaris (region indicated by red segmentation lines in (c). While OCT-A does not resolve the choriocapillaris fine structure, this patient does have a flow void (red arrow). High resolution AOSLO flow maps of (h) the superficial, (i) intermediate, and (j) deep vascular plexuses, demonstrating the vascular branching pattern. Yellow arrows indicate sample locations where there are vertical connections between the plexuses.

The vasculature is essential to the function of all organs, but the eye is unique in several ways. First, the eye is one of the major sensory organs for humans. Mammalian retinas, particularly the photoreceptors, have some of the highest metabolic rates in the body (Wangsa-Wirawan & Linsenmeier 2003, Yu & Cringle 2001). Second, as part of the central nervous system (CNS), the retina is protected from the vascular contents by the blood retinal barriers. The inner blood retinal barrier consists of tight junctions between the endothelial cells that line the inside of retinal vessels. The outer blood retinal barrier is primarily the tight junctions between the retinal pigment epithelial (RPE) cells. Third, the retinal circulation must enter the eye through a relatively small region, the optic nerve head, located at the junction of two pressurized spaces: the globe of the eye and the CNS, which may have considerably different pressures. This provides a location of unique vulnerability for the vascular system. As a result, the eye may be more subject to vascular pathologies than other organs, with ischemic and neovascular retinal vascular conditions causing a major portion of the blinding disorders, e.g. diabetic retinopathy and age-related macular degeneration. Through the pupil of the eye, researchers have a unique ability to measure blood vessels and damage to them utilizing technologies ranging from direct visualization to adaptive optics scanning laser ophthalmoscopy (AOSLO) and ocular coherence tomography (OCT).

2. FUNCTION OF THE VASCULAR SYSTEM

There are critical differences between the larger arteries and veins (macrovasculature), and the small vessels (microvasculature). The macrovasculature primarily distributes blood products at high velocities and pressure to and from other parts of the body. The vessels of the microvasculature are often specialized according to the demands of the tissue they supply. The macrovasculature consists of blood vessels with diameters larger than 100–200 μms, and the microvasculature consists of vessels smaller than 100 μms. Thus, most vessels in the eye are considered part of the microvascular system.

2.1. Roles of the Macro- and Micro- vascular networks

The vascular system must deliver oxygen, nutrients, and precursors for metabolic processes; transport hormones and growth factors; remove byproducts of metabolism such as carbon dioxide and lactic acid; defend against infection through the delivery of immune cells and antibodies as well as cytokines and chemokines; and serve as a dynamic control system regulating flow based on tissue demands modulated by systemic parameters such as blood pressure. Throughout most of the body, the vascular system is divided into distinctly different conduits: the arteries, the veins, and the capillaries. The macrovascular arterial system has muscular walls suitable for transporting blood at high pressures and velocities. The initial portion of the arterial system, the aorta, has high elasticity that evens out the pulsatile blood flow and reduces the workload on the heart. Arteries then distribute the blood to the microvascular system. The arterioles are control conduits that regulate flow to individual regions of the body by dilation and constriction accomplished by a continuous layer of smooth muscles in the arteriolar walls. In the eye, the usual ophthalmic instruments visualize large and small arterioles but not capillaries, which are too small to be seen without contrast agents or advanced techniques. Capillaries have thin walls, typically a single endothelial cell thick. Rather than smooth muscle cells, capillaries have pericytes intermittently present within the capillary basement membranes, The pericytes are contractile and have may control flow for individual capillaries, although this is controversial (Hall et al. 2014, Hill et al. 2015). The thin walls of capillaries allow exchange of nutrients and waste products. The endothelial cells express molecular signals which inform the circulating immune cells of local inflammation/infection. The venular portion of the capillary system is specialized to allow the binding and egress of leukocytes from the vascular system, occurring mostly at the endothelial cell junctions but occasionally through the center of an endothelial cell (Bharadwaj et al. 2013, Joseph et al. 2020). Veins are the conduit for returning blood to the heart. The large diameter of the venular system means that almost two thirds of the systemic circulation blood is in the veins. This provides a reservoir for the body, and sympathetic stimulation (McDougal & Gamlin 2014) can reduce the volume through constriction, enabling the vascular system to function with a blood loss due to injury.

The vascular system communicates with the body at the surface of vessels. The majority of the vascular surface area is in the capillaries, with the aortic cross section being only 1% that of the integrated area of the capillaries. With the slow flow rates in the capillaries, about 1 mm/sec, and the short length of most capillaries, erythrocytes reside within a capillary for less than a second, yet almost all of the transfer of oxygen, carbon dioxide, and nutrients occurs during this time. Blood pressure drops rapidly over the capillary from as high as 35 mm Hg at the arteriole end to about 10 mm Hg at the venous end. The blood flow to most tissues is controlled according to local need, with flow in a muscle varying by as much as 20 to 30 fold but in the eye it decreases by about 40% during O2 breathing (Luksch et al. 2002) and increases by about 25% with hypercapnia (Venkataraman et al. 2008). These are small changes relative to many other tissues.

Blood flow in most of the vascular system is laminar. In laminar flow there is a radial distribution of velocity with the highest velocity in the center and a cell-free layer along the wall. This cell free layer arises from dynamical considerations (Caro et al. 2012). In turbulent flow cells move in multiple directions, not just parallel to the walls. Laminar flow has a low resistance to flow and therefore requires less energy to pump blood. In the eye, there is almost never turbulence except occasionally at locations of vessel crossings (Christoffersen & Larsen 1999) where it has been associated with branch vein occlusions.

The nature of laminar flow changes between large and small vessels. For large vessels, blood acts like a Newtonian fluid with viscosity linearly related to local forces, and the properties of the cells are not critical. In contrast, flow in small vessels becomes non-Newtonian because cells interact with each other and the vessel walls as the lumen shrinks.

For Newtonian flow, flow through a vessel is proportional to the pressure difference between two points times the conductance. This is captured by Poiseuille’s law,

$$V=\frac{\pi p{r}^4}{8\eta l}$$

where V is the volume flow (μl/min), r is the radius, p is the pressure differential, η is the viscosity of the blood, and l is the length. Thus, flow increases with the fourth power of the vessel radius and so small changes to the radius of a blood vessel have a large impact on blood flow. Blood viscosity, another important factor in flow, is largely determined by the hematocrit. This can go down, as anemia is common, or up with dehydration or polycythemia. In small vessels blood flow becomes non-Newtonian due to the increasing importance of cellular elements. Thus, as vessel diameters decrease, viscosity is first decreased and then increases (Pries et al. 1992) when vessel size is comparable to erythrocyte diameter (7.5 μm).

2.2. The Blood Brain Barrier and Unique Features of the CNS Vasculature

Blood vessels are lined by endothelial cells. In most of the body, there are intercellular clefts, slightly smaller than an albumin molecule, between the endothelial cells through which fluids move freely. These slits are about 1 thousandth of the endothelial surface area and allow rapid movement of most small water-soluble molecules and ions. Endothelial cells also possess caveolae through which larger molecules can move via endocytosis. Lipid soluble materials, as well as oxygen and carbon dioxide, can diffuse directly through the cell membrane, and therefore the rate of diffusion is many times faster than for water soluble substances such as glucose. Water moving largely through the clefts diffuses so quickly that the rate of exchange across the capillary membrane is 80 times the rate of water flow within the plasma flowing within the capillary.

The endothelial cells of the retina and most of the central nervous system are connected to one another by tight junctions unlike the slit pores found in most of the rest of the body. Water and small lipid soluble molecules such as oxygen, carbon dioxide, and alcohol readily pass through the blood retinal barrier, but electrolytes, proteins, and large hydrophilic organic molecules are blocked. In addition to poor access for large hydrophilic molecules, the blood retinal barrier possesses various drug efflux transporters including multidrug resistance proteins which exclude many lipophilic drugs (Loscher & Potschka 2005). The blood brain barrier helps stabilize neural function against alterations in the blood but can impair the immune response by blocking antibody access and penetration of therapeutic agents.

2.3. Blood Flow Regulation

In general blood flow is maintained at the minimum rate required to supply the requirements of that tissue. One of the primary problems of regulating blood flow to a tissue is stabilizing that blood flow despite the significant changes in blood pressure that occur both short term, during the activities of daily living such as postural changes and lifting heavy objects, and longer term changes such as the elevation of blood pressure with aging. Much of this control is affected by the smooth muscles of the areterioles with elevated blood pressure causing the smooth muscles to contract and conversely, lowering of blood pressure causing them to relax (Drummond et al. 2008). In each case the response stabilizes blood flow to a tissue. This also occurs in the eye where changes are primarily controlled by the arterioles (Pournaras et al. 2008), These changes also follow changes in the oxygenation state of the blood (Duan et al. 2017, Luksch et al. 2002, Riva et al. 1992, Venkataraman et al. 2008). Demand also changes with tissue activity: exercise for a muscle, a meal for the gastrointestinal tract, or a flickering light for the retina. This change in blood flow with activity is termed active hyperemia and is part of the vascular regulation in the eye (Kur et al. 2012, Newman 2013).

The response of the vascular system to metabolic demand requires communication between numerous cells. The endothelial cells seem to act both as sensors of local tissue needs and controllers of local blood flow, affecting vasoconstriction or vasodilation. Endothelial cells synthesize nitric oxide (NO) that can drive the initial dilation of small arterioles, and together with activity dependent signaling from neurons both directly and via glial cells (neurovascular coupling), allows rapid response of the vascular system (Kur et al. 2012). The endothelial cells also can sense shear stress produced by the blood flow on the endothelial glycocalyx. This stress increases with increasing blood flow, which increases the synthesis of NO by larger arterioles. The smooth muscle cells of the arterioles also communicate with each other, producing a lateral spread of the vasodilation. In systemic diseases such as diabetes (Garhofer et al. 2004), it is likely that there is continuing damage to the endothelium with disturbances to the regulatory control of tissue blood flow controlled by the endothelium. Direct neural inputs to the systemic and cerebral vasculature are also important in vascular regulation but are expected to have no specific role within the retinal vasculature as it is thought to lack innervation, but may play a role in the choroidal vasculature that is extensively innervated (Bron et al. 2001). The CNS is different in this regard from the retina as there is extensive neural innervation of the cerebral vasculature and sympathetic innervation plays a large role in the regulation of cerebral vascular resistance in response to hypertension and calcitonin gene-related peptide from the trigeminal innervations, which is important in the pathophysiology of migraine.

3. BASIC STRUCTURE AND FUNCTION OF THE RETINAL AND CHOROIDAL VASCULATURE

3.1. Unique Constraints on the Retinal Vasculature.

The retinal vasculature has several important constraints that may be unique. Chief among these is the fact that blood and blood vessels are highly scattering and lie on the light path and thus can degrade vision. It is generally accepted that primates evolved a foveal avascular zone (Figure 1) to minimize vascular scattering for the highest acuity part of the retina. There is probably a similar constraint for the rest of the retina, with a strong benefit in minimizing the number of blood vessels, while still supporting the very high metabolic demands of the retina (Snodderly & Weinhaus 1990). The result is that the retina seems to be critically organized (Chan et al. 2012, Stone et al. 1995), with pruning of vessels occurring during development (Pries & Secomb 2014, Simonavicius et al. 2012). The visual advantage of fewer vessels, together with the high metabolic activity of the retina, may be the reason the retina is highly sensitive to vascular insult, explaining its sensitivity to early diabetes and hypertensive disease.

3.2. Introduction to the eye’s vascular system.

The arterial supply to the eye and orbit is the ophthalmic artery. This artery is a branch of the intracranial portion of the internal carotid artery. Within the orbit it has a tortuous course which is thought to allow eye movements without impairing the blood supply. The first branch of the ophthalmic artery is the central retinal artery. This enters the optic nerve about 10 mm behind the globe and supplies small branches to the nerve during its forward course. At the disk it usually bifurcates twice, first vertically and then horizontally creating the major arterioles, the arcades of the retinal circulation (Figure 1a). The central retinal artery is a terminal artery and blockage of its blood supply by an embolus results in a near total loss of vision except when a cilioretinal artery is present, as it is in some individuals. Other branches of the ophthalmic artery include the two posterior ciliary arteries which branch into the short ciliary arteries supplying the choroidal circulation, and the long ciliaries which supply the anterior eye. Blood drains from the retina via the central retinal vein and from the choroid through the vortex veins.

The central retinal vein and its branches are susceptible to occlusions from thrombosis causing local or global loss of vision. After the central retinal vein leaves the eye it continues a central course within the optic nerve and leaves it about 10 mm posterior to the globe. The eye and the orbit are largely drained by the ophthalmic vein into the cavernous sinus although a fraction of the drainage is via the facial veins. Both of these drainage systems flow into the jugular system which lacks the valves found in veins elsewhere in the body, and typically when a person is in an upright position these veins are collapsed with an internal pressure of zero. The veins within the eye are not collapsed and have a pressure exceeding the intraocular pressure which in most cases maintains the flow in the retinal veins. This is readily seen with flow maps (Figure 1).

The central retinal artery and its main branches are greater than 100 μm in diameter, as are the central retinal vein and its main contributors. The choroidal arteries and veins range in size from over 200 μm for the large choroidal arteries and veins down to a few microns for the choriocapillaris (Branchini et al. 2013). Thus, the retinal vasculature falls largely within the size range of the microvasculature. Functionally the photoreceptors fall in the watershed region between the choroidal vasculature and the retinal vasculature. The high metabolic activity of the photoreceptors results in a minimum of the transretinal oxygen saturation curve at the photoreceptor ellipsoid zone (Braun et al. 1995, Linsenmeier & Zhang 2017), which implies that photoreceptor function depends on both the choroidal and the inner retinal circulation.

3.3. Retinal Plexuses

Within the retina there are up to four major capillary beds (Campbell et al. 2017, Hormel et al. 2020, Snodderly & Weinhaus 1990, Snodderly et al. 1992), (Figure 1c–g), with the number and location of the capillary beds varying with local retinal anatomy. Thus, near the fovea the number of capillary beds is reduced to a single layer. At the fovea there are no capillaries in most individuals, creating a foveal avascular zone (FAZ). The three intra-retinal capillary beds are layered within the retina, next to or within the nuclear layers, supplying cell bodies. The superficial vascular plexus (SVP) lies within the ganglion cell layers of the retina. Below the SVP are two deeper plexuses, the intermediate vascular plexus (IVP) which lies towards the top of the inner nuclear layer and the deep vascular plexus (DVP) which lies towards the bottom of the inner nuclear layer. This layering is dependent on the overall thickness of the retina, and peripherally, where retinal layers are thinner, there is a single capillary plexus (Campbell et al. 2017). Thus, each capillary plexus is located to provide support to the layers of the retina with the highest metabolic demands as evidenced by the number of mitochondria (Snodderly & Weinhaus 1990). In addition, each plexus has a slightly different vascular pattern (Figure 1) with the SVP showing a more linear projection from arteriole to venule (Figure 1h), and the DVP being more convoluted (Figure 1j). Capillary location is dependent on the local oxygen supply as well, being less common near arterioles and to a lesser extent venules (Arthur et al. 2019a), where oxygen can be supplied directly from the larger vessels. Near the fovea capillaries are present only where the inner retinal layers thicken to more than about 60 μm (Chui et al. 2014). A fourth plexus of radial peripapillary capillaries runs parallel to the bundles of ganglion cell axons (the nerve fiber layer, NFL) to meet the high metabolic demand of the unmyelinated axons. This layer of capillaries is most prominent near the optic nerve, where the axons are most numerous and the NFL thickest (Figure 1b).

3.4. Choroidal Vascular Layers

The choroidal circulation is arranged somewhat differently, with the largest vessels lying deep within the choroid (Haller’s layer), a layer of intermediate sized vessels (Sattler’s layer), and the choriocapillaris, a dense meshwork of large lumened capillaries lacking tight junctions adjacent to the retinal pigment epithelium (Figure 1g) (Olver 1990).

3.5. Vessel Branching

The normal vascular tree is altered in a number of retinal diseases, with clinicians typically describing the vascular pattern in terms of “tortuosity” when the vessels seem to cover longer distance and follow more curved paths between major branches, or as having “capillary dropout” when there are fewer capillaries and areas of ischemia. To move beyond this descriptive classification authors have often quantified the tortuosity (Cheung et al. 2011, Hughes et al. 2006, Witt et al. 2006), or used fractal analysis to capture the space filling aspects of the vasculature (Huang et al. 2016, Lorthois & Cassot 2010, Stosic & Stosic 2006, Zamir 1999). Because fractals are self-similar the use of fractal analysis implies that the branching pattern of vessels is similar from the largest vessels down to the smallest sizes. However, while fractal analysis can be relatively simple and generates a usable metric for changes in the vasculature, the pattern is likely not fractal, because there is a high cost for blood vessels, both in the metabolic cost and in light scattering. In general a vascular system will be space filling, since maintaining a minimum perfusion distance, while minimizing the total number of blood vessels, should dominate the spacing of the smallest vessels (Panico & Sterling 1995). This seems to be true in the retina since vessels appear around the fovea once the retina reaches a fixed thickness (Chui et al. 2014). For the macro-vasculature, Murray’s law predicts that at a branch point the sum of the radii of the daughter branches cubed is equal to the parent radius cubed as well as determining the branching angle. This may be achieved because shear stress during the cardiac cycle causes vascular remodeling (Painter et al. 2006). In the retinal micro-vasculature, the fixed relations predicted from Murray’s law are not met. At all levels measurements consistently have smaller exponents than the cubic relation predicted by Murray’s law (Riva et al. 1985, Stanton et al. 1995). While close to 3 for large vessels, the exponent is smaller for smaller vessels (Luo et al. 2015). The branching of vessels changes in diabetics, suggesting that remodeling is related to physical factors, such as blood viscosity and shear stress (Luo et al. 2015). Also space filling assumes a certain uniformity of metabolite consumption whereas the laminar structure of the retina agrees with the observation that vessels occur in retinal layers with high numbers of mitochondria (Snodderly & Weinhaus 1990).

4. MEASURING RETINAL VASCULAR STRUCTURE AND FUNCTION (FUNDUS PHOTOGRAPHY, ANGIOGRAPHY, OCT AND ADAPTIVE OPTICS)

The retina is unique in the CNS in that it allows direct, noninvasive, measurements of the structure and function of both vascular and neural structures. With the advent of OCT and cellular imaging with adaptive optics (AO), comes progress in investigating retinal vascular diseases and neurodegenerations such as Alzheimer’s disease and glaucoma. In this section we give very brief overviews of the techniques used. Due to rapid advances in optical measurements of the eye, we concentrate on introductions to techniques and provide references to many excellent reviews.

The advantage of being able to image the vasculature was utilized by researchers and clinicians since the 1800’s, and fluorescein angiography (Novotny & Alvis 1961) demonstrated retinal blood flow and loss of the blood retinal barrier. Currently retinal imaging is divided into two basic approaches, flood illuminated imaging and scanning. In flood illuminated imaging, a region of the fundus is exposed to light, often a bright flash, and the light returning from the fundus is recorded. Film has been replaced with high resolution CCD or CMOS sensors. The advantage of flood illuminated imaging is that each image can be collected in a brief time, e.g. a few milliseconds, reducing image smear from eye movements. The disadvantage is that light returning from all illuminated points in the eye is mixed across all points on the detector. Image contrast is significantly decreased, especially when light is scattered over a wide retinal area from aged lenses or pathological retinas. In contrast, in scanning imaging, discrete areas of retina are sequentially scanned, typically in a raster fashion, and the light captured by the detector is synchronized to capture each discrete retinal area, with the image built up over time. Scanning (Webb et al. 1980) overcame the loss of contrast from widely scattered light by imaging only a small region, typically a point or a line, of the retina. Each scanned area is either de-scanned onto a matched detector (Elsner et al. 1992, Webb 1986) or reimaged onto an electronically gated detector (Muller et al. 2013). Scanned techniques have the disadvantage that they are more susceptible to intraframe warping from eye motion. Scanned imaging systems have the advantage of ease of implementation of a confocal aperture, i.e. an aperture to limit light to a specific region in a plane conjugate to the target, thereby minimizing out of plane scattered light or light that is laterally scattered from unwanted locations. Light that is directly backscattered or conversely multiply-scattered is emphasized to probe multiple aspects of light tissue interactions, revealing structures not otherwise seen (Elsner et al. 1996, Elsner et al. 2000, Webb 1986). Scanned systems represent a large proportion of the retinal imaging market, including optical coherence tomography (OCT) and scanning laser ophthalmoscopy (SLO).

4.1. Generating Contrast of Vessels in Retinal Images

4.1.1. Absorption

Blood vessels are one of the most salient features in retinal images. Hemoglobin absorbs maximally in the short wavelength part of the spectrum, with secondary peaks in the green, yellow, and near infrared regions of the spectrum. Because of this difference in absorption with wavelength (Delori & Pflibsen 1989, Elsner et al. 1996, Van Norren & Tiemeijer 1986) retinal images depend strongly on the wavelengths used for imaging. In addition, because the absorption spectra of oxygenated and de-oxygenated hemoglobin differ, it is possible to measure the oxygen saturation of hemoglobin in the retinal veins and arteries, a technique known as oximetry (discussed below).

4.1.2. Fluorescein and Indocyanine Green Angiography

To enhance vessel contrast, a fluorescent dye is injected, typically into a vein in the arm, or ingested. The passage of dye over time is imaged to observe blockage, and/or staining of tissues from dye leakage suggesting a breakdown of the blood retinal barrier. Fluorescein angiography uses a short wavelength excitation wavelength, but indocyanine green angiography uses a near infrared excitation wavelength, useful in dark fundi despite a lesser degree of leakage. (Elsner & King 2015, Flower 1973). A key target for angiography is unwanted neovascularization, which usually leaks, and the relation to retinal structural alterations such as retinal pigment epithelial detachment (Wolf et al. 1993).

4.1.3. Backscatter, Specular Reflection

In contrast to flood illuminated imaging, which combines all illuminated light paths within the eye at the detector, confocal imaging allows maximizing the selection of only certain chosen light paths. By imaging a scanning point or line onto the eye, and then descanning the beam back through an aperture in a conjugate plane (confocal imaging) (Webb 1986), each point or line in the image is limited to light originating within a narrow light path. Thus, a confocal SLO can provide high contrast images of blood vessels over a range of wavelengths, but the image detected is dominated by the index of refraction differences on and within the vessel rather than absorption differences (Elsner et al. 1996).

4.1.4. Diffraction and Multiply Scattered Light

Scanning systems are readily used in a dark-field imaging mode where the directly backscattered light is blocked. This allows light that is forward scattered or diffracted in a forward direction that has been reflected off one or more interfaces, then passed out through the pupil, to be detected. This type of imaging is often called multiply scattered light imaging due to the multiple interactions between light and the fundus. (Chui et al. 2012a, Elsner et al. 1998, Elsner et al. 2001, Sapoznik et al. 2018). Blood cells, blood vessels, and other tissues that are highly scattering are readily imaged using multiply scattered light (Figure 2c). In pathology, often the usual low refractive differences between retinal cells is increased or new structures form, such as diabetic cysts, and these are readily imaged (Figure 2b).

Figure 2.

Examples of microvascular changes in hypertension and diabetes. (a) Flow map from AOSLO image montage, showing the patent vessels around the FAZ. The arteries (A) and veins (V) are clearly visible as well as the capillaries connecting them and the capillary free regions adjacent to large vessels (white arrows). Also visible is a remodeled capillary (yellow arrows) and small microaneurysms (orange arrows). (b) AOSLO structural and (c) flow images of a diabetic FAZ. Loops (yellow arrows) as well as non-perfused capillaries (red arrows) can be visualized. Vessel walls visualized using multiply scattered light imaging with AOSLO in (c) a normal subject and (d) and patient with controlled hypertension with thickened walls. (f) OCT in diabetic subject where vascular leakage has led to edema (thickening) with 3D disruption to retinal layers, and cysts (dark regions).

4.1.5. Coherence Gating to Improve Detection (OCT)

A widespread method to control the light making up the retinal image is to use coherence gating, sampling the interference of light between the imaging beam returning from the fundus and a reference beam. When using short coherence illumination, interference occur only over a limited range of distances in the fundus, providing depth information. Scanning the beam across the retina and analyzing the interference signal provides a depth resolved image of retinal reflectance, a technique known as optical coherence tomography (OCT) (Huang et al. 1991). OCT has become a dominant technique for analyzing the retina in vivo, evolving to provide improved imaging (Kashani et al. 2017, Spaide et al. 2018). As in confocal imaging, OCT is primarily sensitive to index of refraction variations, and thus can image blood cells and vessel walls well. The analysis of sequential images can be used to map particle flow through blood vessels (OCT- A, Figure 1c–g). Quantification of motion (doppler OCT) and analysis of changes across the spectrum (OCT oximetry) provide a growing platform for advanced imaging of vascular physiology as discussed below.

4.2. Adaptive Optics

Imaging the retina through the natural optics of the eye is degraded by optical aberrations for pupil sizes > 2–3 mm. Compensating for an individual’s aberrations across a larger pupil, such as with adaptive optics (AO), improves both image resolution and contrast, visualizing cellular level and other small features (Burns et al. 2019, Liang et al. 1997). Adaptive optics is a platform that can be used to improve the resolution of any of the imaging modalities mentioned previously, including OCT (Hermann et al. 2004, Wang et al. 2011), angiography (Pinhas et al. 2013), SLO (Burns et al. 2019) and flood-illuminated imaging (Marcos et al. 2017, Paques et al. 2018). While clinical OCTA viewing the inner retinal vasculature and choriocapillaris (Figure 1 h–i), resolution is limited. The addition of AO allows precise measures of inner retinal vessels (Figures 1 and 2), the combination of AO and OCT (AO-OCT) (Jonnal et al. 2016) allows measurement of details of the choriocapillaris (Kurokawa et al. 2017).

4.3. Measuring Moving Cells, Flowmetry, OCTA and Cell Imaging

Fluorescence techniques, by taking images at multiple times following injection of the fluorescent dye, allow measurement of moving cells (Flower 1973, Hodge & Clemett 1966). This has been used to measure blood flow (Wolf et al. 1991). Both FA and ICGA can demonstrate blood flow and vascular pathology but are invasive. Speckle imaging raised the possibility of non-invasive measures of flow. By using coherent light and taking long exposure images, the movement of blood causes local changes in interference due to differences in scattering during the exposure, decreasing speckle (Fercher & Briers 1981). Modern noninvasive techniques map retinal vessel perfusion and quantify blood flow based on this same principle. One approach to using variation over time is flowmetry (Petrig & Riva 1988). Flowmetry is a measure of the amount of flow within a volume of tissue, providing a measurement of perfusion through the tissue, but not precise information on velocity and direction. This approach has been used to image blood vessels using the retinal flowmeter (Michelson et al. 1995), stabilized SLO imaging (Ferguson et al. 2004), and high speed OCT and SLO (Mujat et al. 2019) as well as AOSLO imaging that allows resolution of the smallest capillaries and their alterations based on motion (Chui et al. 2012b, Tam et al. 2011) (Figure 1h–j, Figure 2a).

The most widely used flowmetry approach is OCT angiography (OCTA) (Makita et al. 2006), which has been implemented using a number of processing techniques (Kashani et al. 2017, Spaide et al. 2018). OCT-A makes multiple measurements of the same retinal region using OCT. The processing algorithm then compares the light returning to the detector using amplitude, phase, or both. If cells move between acquisitions then the OCT signal at that depth level will change, leading to a depth-resolved map of retinal and choroidal blood vessels (Figure 1). The striking advantage of OCTA is that images can be generated rapidly, use widely available instrumentation, and provide a full 3D measurement of the vasculature. While most instruments do not have the resolution to actually resolve individual capillaries, OCTA detects the motion through the capillaries, demonstrating the presence of vascular perfusion in 3D on a scale determined by the pixel size of the instrument and the optics resolution. (Figure 1).

4.1. Quantifying Blood Motion

While flowmetry provides information about where blood is moving, with some implementations also showing how rapidly, velocimetry provides an estimate of both velocity and direction, first achieved using doppler techniques (Riva et al. 1972). Laser doppler velocimetry (LDV) uses the fact that scattering of light from the moving cells shifts the wavelength of light due to the Doppler effect. A discrete portion of a blood vessel is illuminated with coherent light. The interference between moving blood cells and the nearby structures generates a signal at the difference frequency between the illumination and shifted wavelengths, which is measured as a temporal variation in the intensity of the signal. To obtain absolute measurements, independent of the exact angle of the vessel relative to the illumination, simultaneous measurement from two positions (bidirectional LDV) are used (Riva et al. 1979). LDV has been used extensively to measure changes in blood flow (Pournaras & Riva 2013, Pournaras et al. 2008, Riva et al. 2005). By summing flow into and out of the larger retinal arteries and veins, an estimate of total blood flow in the eye is obtained (Grunwald et al. 1992).

OCT also depends on the comparison between a reference beam and light scattered from the retina, making possible doppler OCT to provides knowledge of both velocity and the 3D position of the retinal vessels being measured (Leitgeb et al. 2004). Similar to LDV, these can be bidirectional, measuring the absolute velocity of blood flow as well as the 3D morphometry of the retina (Werkmeister et al. 2008).

While LDV and Doppler OCT allow precise measurements of blood velocity in larger retinal vessels, they are limited in their ability to measure small vessels, primarily by the relatively large point spread function of most systems. However, with adaptive optics it is possible to make precise measurements of blood flow in small vessels, first performed by measuring the speed of leukocytes moving through vessels (Martin & Roorda 2005). This approach revealed the preferred paths for leukocytes as well as measuring the pulsatility of flow in capillaries (Tam & Roorda 2011). However, the motion of leukocytes is not necessarily representative of flow of erythrocytes since both the cells and the vessel walls must deform as leukocytes move through the narrower capillaries, slowing the flow. Measuring the smaller, faster moving (approximately 1 um/msec), erythrocytes is more difficult since imaging at slow speeds is susceptible to aliasing. While there are direct approaches to increasing the frame rate (Burns et al. 2019, Grieve et al. 2006) for an AOSLO, these are limited. Momentarily stopping the slow scan and measuring cells as they flow along a vessel across the scanning line allows accurate measurements of the velocity over the cardiac cycle (Zhong et al. 2008) and the velocity profile of speeds of cells within vessels (Zhong et al. 2011). The use of a line scanning camera also increases the speed of imaging (Gu et al. 2018). If imaging small retinal regions, flood illuminated cameras achieve high frame rates (Rha et al. 2006) and can measure flow within capillaries (Bedggood & Metha 2012). Capillary velocities can also be measured using an AOSLO with two beams, such that a vessel is imaged twice within the imaging time of a single video frame (de Castro et al. 2016, Warner et al. 2020) allowing resolution of the entire velocity profile across the cardiac cycle (figure 3). The actual flow through the network is complex at the level of individual capillaries and will require not only precise measurements within a plexus, but also across plexuses (Figure 3).

Figure 3.

AOSLO blood flow measurements in retinal microvasculature. (a) Montage of the superficial vascular plexus, where AOSLO measurements provide the direction of blood flow (red arrows) and velocity, including capillaries, arterioles, and venules. Yellow circles indicate the regions where there are vertical connections between plexuses, which are oriented mostly laterally across the retina. (b) Blood flow measurements in a small arteriole. The variation in velocity represents the cardiac cycle.

4.2. Oximetry

Spectral differences in light absorption by blood provides a major contrast mechanism for clinical retinal imaging, revealing not only pathology but also oxygenation properties. The spectral absorption function for oxygenated and de-oxygenated hemoglobin differ, making it possible to measure the ratio of oxygenated to de-oxygenated hemoglobin, also called the saturation level (Beach 2014, Delori 1988). Early measures of blood oxygenation were limited by the difficulty in getting an accurate measure of reflectance from a blood vessel, since light scatter in the eye is high, As a result absolute measurements were variable (Delori 1988). More recently with improved imaging methods and modeling of light scatter (Beach 2014), the use of oximetry is seeing a renaissance (Stefansson et al. 2019). Visible light OCT techniques (Shu et al. 2017, Yi & Li 2010) address the issue of scattered light directly, comparing the spectrum of light returning from the top and bottom of a vessel. The power of visible light OCT can be seen in animal imaging where the oxygen saturation of blood has been measured throughout the entire vascular tree, including in capillaries (Pi et al. 2020). Another approach is photoacoustic imaging, which builds on the absorption of a flash of light causing motion in the absorbing medium. This motion is measured using high resolution acoustic imaging (Jiao et al. 2010, Liu et al. 2013) or OCT. For non-human investigations there are also techniques that can measure phosphorescence decay. (Chamot et al. 2003, Shahidi et al. 2010) Together with the use of fluorescent microbeads to measure blood velocity, these techniques provide the ability to measure the local metabolic state of the retina and are a powerful research tool (Blair et al. 2016, Shahidi et al. 2010).

4.3. Retinal Oxygen Delivery and Consumption

While blood velocity is often directly measured, flow is computed from the velocity and diameter of the blood vessels. This generally uses a model based on Poiseuille’s law. However, for the microvasculature, flow deviates from Poiseuille’s law (Riva et al. 1985, Stanton et al. 1995) particularly for smaller vessels (Zhong et al. 2011) as discussed above. But Poiseulle’s law is adequate to allow the whole retinal oxygen delivery of the entire retina to be estimated from the flow through the large arteries and veins at the optic nerve (Grunwald et al. 1992, Leitgeb et al. 2014, Wang et al. 2008).

By combining flow measurements with the difference in oxygen saturation between the arteries and veins, an estimate of oxygen consumption can be made. To measure delivery of the most crucial factor, oxygen, and measurement of oxygen extraction by the retinal tissue, we need hemoglobin saturation measures at the arteriolar/venular levels. While absorption oximetry has issues in reproducibility of absolute values, it can show responses to physiological interventions and can be used in both humans and animals to estimate total oxygen consumption (Felder et al. 2015, Werkmeister et al. 2015). This will be increasingly important as techniques for oximetry improve, as discussed above. For the delivery of the other nutrients, as well as removal of carbon dioxide and organic acids such as lactic acid, volumetric flow seems adequate. Ultimately the ability to combine all measures to quantify regional metabolism will allow better understanding of vascular control and the impact of diseases.

5. NEUROVASCULAR COUPLING IN THE RETINA AND FUNCTIONAL HYPEREMIA

The eye provides a unique opportunity to measure properties of the CNS non-invasively. This allows testing of basic aspects of neurovascular coupling, which underlies techniques such as fMRI, as well as to gaining insights into the vascular components of neurodegenerative diseases. This section provides a brief overview of functional measurements based on the vascular system.

5.1. Whole Retina Responses to Visual Stimuli

The ability to measure aspects of vascular function directly enables precise measurements of the response of the retina to changes in neural activity or to physiological manipulations. The retina increases blood flow in response to visual stimuli (Linsenmeier & Zhang 2017, Newman 2013, Riva et al. 1992, Riva et al. 2005). This mechanism, termed functional hyperemia, is modulated by the response of both Mueller cells and astrocytes to neural activity, with the subsequent release of vasoactive signals (Newman 2013).

The most common visual stimulus used to stimulate changes in blood flow is full-field flicker (Riva et al. 2005). The luminance flicker response peaks at from 8–10 Hz (Riva et al. 2005) and has a temporal band-pass function, while chromatic flicker responses are maximized at lower temporal frequencies and have a more low-pass temporal characteristic (Riva et al. 2001), consistent with psychophysical sensitivity to chromatic and achromatic stimuli and the underlying ganglion cell populations. Focally, blood flow would depend on the nature of the visual stimuli, as well as the sensitivity of cell and plexus being supplied. The retina can respond to localized stimuli, and the responses can be measured down to the capillary level (Warner et al. 2020).

6. IMPACT OF DISEASE ON VESSEL STRUCTURE AND FUNCTION

This section describes results and pathophysiology of some common retinal vascular diseases. The field of noninvasive retinal imaging is rapidly expanding, as OCT-A and AOSLO are increasingly able to make precise measurements of vascular function in patients. A parallel development that is increasingly important in clinical measures of retinal vascular diseases is the use of ultra-wide field imaging with field extents greater than 200 degrees to improve detection of peripheral vascular abnormalities (Quinn et al. 2019, Witmer & Kiss 2013). The coupling of this approach to OCT-A (Wei et al. 2020, Zhang et al. 2015) allows precise measurements of vascular function over wide regions of the retina.

6.1. Systemic Hypertension

Hypertension affects the retinal vasculature by causing thickening of the arteriolar vascular walls (Figure 2e) through autoregulatory processes which protect the vulnerable capillary bed from exposure to damagingly high systemic blood pressures. These changes increase the thickness of the arteriolar walls (Hillard et al. 2016, Koch et al. 2014, Rosenbaum et al. 2016) as well as increased vascular tortuosity (Cheung et al. 2011, Malek et al. 2015). With uncontrolled hypertension there can be numerous sequalae, including hemorrhages, cotton wool spots that represent the backup of axonal transport due to a local vascular infarcts, and retinal edema due the high pressures exceeding the regulatory capacity of the retinal arterioles (Wong & Mitchell 2007).

6.2. Diabetes and Retinal Vascular Diseases

Diabetes is the major cause of visual loss in the working age population, due to retinal microvascular damage (Elsner & King 2015). Diabetes causes an increased loss of vascular endothelial cells, which initially are replaced by both replication from circulating stem cell precursors (Park et al. 2017), but eventually there is a loss of endothelial cells and pericytes contained within the capillary basement membrane (Park et al. 2017). These alterations ultimately cause capillary occlusion, generally starting in the periphery and later in a patchy fashion in the posterior pole (Fu et al. 2016). The retinal areas that have lost capillaries become hypoxic and release VEGF, which is responsible for both vascular leakage and neovascularization. Macular edema can result, reducing visual acuity. Neovascularization also occurs, and if uncontrolled results in a profound loss of vision from hemorrhage and retinal traction.

Advances in retinal imaging have allowed visualization of early loss of capillaries, changes in macular thickness, and capillary remodeling at a stage prior to the clinical detection of even background diabetic retinopathy (Burns et al. 2014). Flow maps visualize perfused vessels on OCTA (Spaide et al. 2018), but finer lateral detail is provided with AOSLO, which also has confocal and multiply scattered light images that show nonperfused vessels (Figure 2b, 2c). Both techniques provide resolution of capillary loss in the deep vs superficial capillary plexus, but OCT-A maps a wider area (Figure 1d–g). Both retinal capillaries and the choriocapillaris are damaged by diabetes, and with the failure of the outer blood retinal barrier (Xu & Le 2011), leakage (Arthur et al. 2019b) and deposition of lipids and proteins from blood (Arthur et al. 2018) are visualized in the deeper layers of human retinas with OCT. Flow maps from AOSLO or OCTA do not document leakage as in angiography or the exudates and cysts imaged with structural OCT (Figure 2f) but can show decreased perfusion as opposed to just thickness increases. This is important, as diabetic retinas are managed clinically by retinal thickness on structural OCT, based on the incorrect assumption that edema reliably leads to increased thickness, while failing to account for neural degeneration simultaneously decreasing thickness. OCT-A maps the lack of perfusion in the choriocapillaris, known as flow voids (Spaide et al. 2018, Zhang et al. 2018), and larger choroidal vessels, as well as remodeling in retinal vessels and enlargement of the FAZ (Hormel et al. 2020). The earlier than expected damage seen on SLO, OCT, AOSLO, and OCTA is changing the thinking about the timeline of damage to vessels and deficits of visual function, as well as the more widespread prevalence than seen from FA of macular cysts that decrease vision (Beausencourt et al. 2000).

There is vascular dysregulation and alterations to neurovascular coupling in diabetes (Pournaras et al. 2008). Blood flow in diabetic retinopathy often increases initially then declines later (Palochak et al. 2019, Pournaras et al. 2008), but how alterations in flow occur within local areas of the retina including the different plexuses remains undetermined (Scarinci et al. 2016). Regardless of approach, flow measurements in retinal arterioles and capillaries will provide a better understanding of the effects of diabetes on the ocular microvasculature at a local level, particularly the macula. Given the damage to endothelial cells and pericytes (Arthur et al. 2018, Hammes et al. 2002, Xu & Le 2011), it is widely assumed that diabetes damages the autoregulatory pathways controlling retinal blood flow.

6.3. Neurodegenerations

6.3.1. Glaucoma

Glaucoma, also known as glaucomatous optic neuropathy, is a multi-factorial family of diseases characterized by a progressive decrease in sensitivity and visual field due to a loss of ganglion cells. Cell death can be secondary to damage to the unusually long unmyelinated axons, about 2 cm, occurring in the passage through the disk from a higher pressure environment, the eye, to a lower pressure one, the CNS. The optic nerve head, but not the vulnerable optic tract and its blood supply, is visualized by retinal imaging, and vascular dysregulation has been hypothesized as a key driver (Orgul 2007). Blood flow is decreased and neurovascular coupling impaired (Gugleta et al. 2013).

Recent advances with OCT-A show that glaucoma reduces optic nerve head vascular density both in the superficial disc tissue and deep in the lamina cribrosa. A decrease of capillary density, particularly in the superior vascular plexus where the ganglion cells reside, occurs in glaucoma (Hormel et al. 2020). Since ganglion cells are the major oxygen consumers in the outer retina, OCT-A is more sensitive to ganglion cell loss, whereas structural OCT measures of nerve fiber thickness incorporate not only neural cells, but also damaged glia, and blood vessels in thickness measurements.

6.3.2. Alzheimer’s Disease

The relation of Alzheimer’s Disease (AD) to retinal vascular integrity is increasingly recognized, with the eye representing a unique opportunity for investigating vascular integrity (Frost et al. 2013). While the hallmark of AD has been the presence of plaques and tangles in the CNS, there is also increasing interest in the ability to measure vessel dropout (Brown 2010) and changes based in blood flow and retinal thickness and the regulation of flow (Attwell 2019).

6.4. Age-Related Macular Degeneration

Age-related macular degeneration, a progressive retinal degeneration of individuals aged 55 years and older, is a multi-factorial family of diseases that is the chief cause of irreversible vision loss in many industrialized countries (Elsner & King 2015). Inflammatory processes and the build-up over decades of debris from phagocytosis lead to loss of the RPE, with the role of damaged choriocapillaris either primary or secondary, seen on histology (Lutty et al. 2020).

The AMD produces two pathways to retinal damage. In the atrophic form of AMD there is a relatively slow progressive loss of RPE, choriocapillaris and Sattler’s layer, seen on OCT as thinned layers that readily transmit light (Elsner et al. 2020). The proximity of the large choroidal vessels to the photoreceptors may underlie their surprising survival in regions of atrophy. Documentation of decreased perfusion in the choriocapillaris has been shown for patients initially with flowmetry (Xu et al. 2010) and later with the measurement of flow-voids using OCT-A (Hormel et al. 2020).

In the neovascular form of AMD, there is a disastrous attempt at wound-healing by new blood vessels that leak or bleed, including choroidal, subretinal, and even retinal vessel nets (Hartnett et al. 1996), that were not appreciated fully on FA, particularly through hemorrhage. SLO with multiply scattered light imaging quickly documents the 3D disruptions from neovascularization (Elsner et al. 1998, Elsner et al. 2020), compared to the invasive and time-consuming FA. OCT and OCT-A document 3D aspects of neovascularization, providing detailed information about the location of fluid within the retinal and subretinal layers, and therefore the risk to neurons (Elsner et al. 2020). OCT and OCT-A have also improved recognition of polypoidal macular degeneration, a degeneration of the choroidal vasculature, which accounts for a major portion of neovascularization in the macula in African Americans and in Asian populations (Elsner & King 2015, Hormel et al. 2020). The ability to follow the natural history of CNV with imaging can guide treatment, and help avoid overtreatment that could damage neurons, showing that choroidal new vessels exist within regions of soft drusen for extended periods of time without extensive growth, hemorrhage, or fluid leakage and occur in various forms (Hormel et al. 2020).

6.5. ROP

Retinopathy of prematurity occurs in part because high oxygen exposure in the neonatal care unit, often required for survival of the infant, prevents the peripheral retina from becoming normally vascularized. With greater maturity the infant is given normal oxygen exposure but the peripheral retina, without vascularization is now hypoxic, and begins the synthesis of high levels of VEGF (Aiello et al. 1995). This drives a process of neovascularization at the edge between the vascularized and avascular portions of the retina often progressing to extensive retinal detachment and scarring. Portable OCT imaging and wide-field imaging has helped in the classification of the stages of the pathology produced by this process and in evaluation of its treatment by anti-VEGF agents (Chan-Ling et al. 2018).

7. CONCLUSIONS AND FUTURE DIRECTIONS.

The last decade has produced important advances in retinal imaging. Many imaging techniques have been commercialized and improved the monitoring and treatment of patients with retinal diseases of a fundamental vascular nature including diabetic retinopathy, ROP, retinal vein occlusions, and exudative age-related macular degeneration. Coupled with the existing anti-VEGF agents and newer agents that broaden the spectrum of available anti-angiogenic molecules, the future is bright for this field and for the many patients affected by these diseases