Optical Coherence Tomography Angiography in Healthy Subjects and Diabetic Patients

The traditional multimodal imaging, based on fluorescein angiography (FA), indocyanine green angiography (ICGA), and spectral-domain optical coherence tomography (OCT), leads to a better understanding of the pathophysiologic features of retinal and choroidal diseases [1]. It allows obtaining information about the normal retinal and choroidal anatomy, nearly comparable to histological findings. Moreover, dye angiographies (both FA and ICGA) provide essential dynamic information on the perfusion of different retinal and choroidal vascular layers, including the transit time from the arm to the eye.

FA is the generally accepted “gold standard” in imaging the fundus, due to its capability to visualize the morphology of the retinal capillary bed and its perfusion. Moreover, FA allows the clinician to detect and analyze one main clinical sign: leakage from abnormal and/or (retinal or choroidal) new vessels. Although the fluorescence of the injected dye permits an improved visualization of the retinal capillaries, it is well known that not all the different layers of the retinal capillary network can be visualized with this bidimensional examination method. FA images of the retina correspond to the anatomical arraignment of the superficial retinal vessels, whereas the deeper retinal capillaries are not visualized in the angiogram [2, 3]. Comparative findings suggest the deeper capillary network in the retina is not visualized well by FA, possibly because of light scattering of the retina [4]. Therefore, even if FA is the leading imaging modality for visualizing retinal vessels, one of the two major capillary networks does not appear to be imaged well, despite the fact that the retina is a nearly transparent structure [5].

Optical coherence tomography angiography (OCT-A) allows a clear, depth-resolved visualization of the retinal [5] and choroidal microvasculature [6] by calculating the decorrelation of the signal between static and nonstatic tissue. Given that the main moving elements in the eye fundus are contained in vessels, determining a vascular decorrelation signal permits a three-dimensional visualization of the retinal and choroidal vascular network [7]. Moreover, OCT-A does not require the administration of intravenous dye, reducing the risk of potential adverse events [8].

Generally, the blood-supplying artery to the retina is the central retinal artery, which, following the inferior margin of the optic nerve, enters the eye at the level of the optic nerve head. The central retinal artery divides to form two main branches, and each of these divides again to form the superior nasal and temporal and the inferior nasal and temporal arteries, which supply the four quadrants of the retina. The retinal venous vessels are distributed in a similar pattern.

Many anatomical variations in this division and distribution may be observed in a normal fundus. The major arterial and venous branches and the successive divisions of the retinal vasculature are present in the retinal nerve fiber layer close to the internal limiting membrane. The retinal arterial circulation is a terminal system, with no arteriovenous communication with other arterial systems. Thus, the perfusion of a specific retinal quadrant comes exclusively from a given retinal artery and vein that supply that area (although several anatomical variations were described). Any impairment or blockage in blood supply therefore causes an ischemia or infarction. As the large arteries extend within the retina toward the periphery, they divide to form successive levels of arteries with progressively smaller diameters until they reach the ora serrata. The retinal arteries branch dichotomously or at right angles from the original vessel. The arterioles coming from the retinal arteries form an extensive capillary network in the inner retina as far as the external border of the inner nuclear layer (INL), either toward the periphery and/or in the macular area.

Many macular arterioles, branching from the temporal (superior and inferior) retinal arteries, dive into the retina, forming the macular capillary bed with two distinct capillary plexuses: the superficial capillary plexus (SCP) and the deep capillary plexus (DCP). The SCP invests the ganglion cell layer with one or more layers of capillaries, while the DCP ordinarily brackets the INL with a layer of capillaries on either side. The SCP is imaged on OCT angiograms by starting with the inner border of the ganglion cell layer to the inner border of the inner plexiform layer (IPL). The DCP is commonly visualized by setting the inner boundary at the outer IPL and the outer boundary at the outer plexiform layer. An intermediate capillary plexus (ICP) is also described between the IPL and the INL. Due to the short distance between the ICP and the outer component of the DCP, they are commonly considered as a unique entity. Generally, no vessels extend deeper than the INL. The outer retinal layers and photoreceptors receive oxygen and nutrients from the choriocapillaris. A cilioretinal artery, originating from the short posterior ciliary artery, in less than 20% of the cases enters the retina on the temporal side of the optic nerve and reaches the macular area, ending in a capillary plexus in the retinal vasculature.

The FA images for the current review were obtained by the use of a confocal imaging system (Spectralis HRA2; Heidelberg Engineering, Heidelberg, Germany) (Fig. 1). As a confocal device, this system captures only light emitted in a predetermined plane, and consequently eliminates artifacts (due to reflection and diffraction) and superimposed images.

The macular capillary bed, including the fine perifoveal anastomotic arcade, may be visualized in case of clear media and good imaging contrast. This arcade will precisely delimitate the foveal avascular zone (FAZ). A fine focusing in some rare cases could help to appreciate some differences between the SCP and the DCP. Imaging them may therefore require successive focusing. Fine structures, including pathologic conditions such as neovascularizations, are clearly visualized with a similar approach.

The OCT angiograms shown in the current review were acquired with a Spectralis OCT2 device (Heidelberg Engineering), a spectral-domain OCT system able to acquire 85,000 A-scans per second with an axial resolution of 3.9 µm, a transverse resolution of 11 µm, and a lateral resolution of 6 µm. The ocular light power exposure was within the American National Standards Institute safety limit [9].

For this review, relevant imaging findings were selected from a sample of 156 eyes of 86 diabetic patients (35 females, 40.7%) and 96 eyes of 48 healthy subjects (28 females, 57.1%). The mean age of the diabetic group was 58 years (range 20–78) and that of the control group was 44 years (range 19–76). Diabetes was classified as type I in 29 patients (33.7%) and type II in 57 patients (66.3%).

The mean duration of the disease in the type I diabetic group was 12 ± 4 years (range 6–17), and the mean HbA_1c level was 7.1 ± 0.8% (range 6.4–8.8). The mean duration of the disease in the type II diabetic group was 9 ± 4 years (range 5–16), and the mean HbA_1c level was 7.4 ± 1.3% (range 6.1–8.9).

The imaging sessions with both the healthy subjects and the diabetic patients were performed in accordance with the Declaration of Helsinki after approval by the Paris Institutional Ethics Committees. Full written informed consent was obtained from all patients prior to the OCT-A assessment.

C-scans allow the visualization of arteries clearly distinguishable from veins by the presence of a surrounding hypointense halo due to the absence of efferent vessels coming directly out of the walls. The SCP appears as a fine capillary network with a hyperintense signal. The whole 360° of the perifoveal arcade is clearly visible (Fig. 2a, b). The DCP is shown on C-scans taken at the level of the INL (Fig. 2c, d). A dense capillary network, different from the superficial one, becomes clearly visible and develops all around the perifoveal area. OCT-A is the first in vivo examination method that allows a fine visualization of the DCP. It appears as a very dense, regularly anastomosed network with sinuous arborization. Arterioles and venules are not distinctly shown.

Conventional FA and OCT-A cannot be closely compared. This is mainly due to the bidimensional nature of FA, in which all the vascular structures included in the whole retinal thickness are simultaneously shown. In contrast, OCT-A, due to its depth-resolved nature, allows evaluating layer by layer the entire retinal and choroidal tissues for a detailed vascular analysis. In a limited area (in this case, 15 × 10°; Fig. 3) a significant difference is evident between traditional FA (on the left) and OCT-A (on the right).

A large part of the blood supply to the eye comes from the choroid, which originates from the ophthalmic arteries. The left and right ophthalmic arteries in most individuals arise as the first major branch of the internal carotid, usually where the latter breaks through the dura mater to exit the cavernous sinus. The posterior ciliary arteries, which form the blood supply to the choroid, and the central retinal artery, which enters the eye via the optic nerve, are branches of the ophthalmic artery. Other branches of the ophthalmic artery supply the lachrymal gland, extraocular muscles, and lids. The choroid is vascularized by two arterial systems: the short posterior ciliary arteries, which supply the posterior choroid, and the long posterior ciliary arteries, which supply the anterior portion of the choroid (as well as the iris and ciliary body). Several short posterior ciliary arteries (approx. 15–20) penetrate the sclera in a circular pattern surrounding the optic nerve, with the distance between these vessels and the nasal side of the nerve being closer than that on the temporal side. The circle of Zinn, an annular artery surrounding the optic nerve, is formed with the anastomoses of these arteries within the sclera. The branches from the circle of Zinn contribute to the pial circulation, the optic nerve at the level of the lamina cribrosa, and the nerve fiber layer of the optic disk. The short posterior ciliary arteries (other branches originating from the circle of Zinn) enter the choroid to provide the arterial blood supply to the posterior uvea. These arteries divide rapidly to terminate in the choriocapillaris, an exceptionally dense capillary network that nourishes the posterior choroid up to the level of the equator of the eye. The two long posterior ciliary arteries penetrate the sclera on either side of the optic nerve. The long posterior ciliary arteries begin to branch just anterior to the equator and contribute to the circulation of the iris and ciliary body. Just anterior to the equator, some branches of these vessels course down into the choroid and branch to terminate in the choriocapillaris from the ora serrata back to the equator of the eye.

These arteries continue to branch and ultimately form the extensive choriocapillaris adjacent to the acellular Bruch membrane located on the basal side of the retinal pigment epithelium (RPE). Their luminal diameter is nearly 20 µm in the macular region and 18–50 µm in the periphery.

Large draining vessels collect blood from the entire choroid (posterior to the equator). Drainage is homogeneously distributed toward the four quadrants, but sometimes the presence of large dilated vessels at the posterior pole may indicate the prevalent draining direction. Venous collecting vessels from the choriocapillaris emerge in the eye through the vortex veins (generally at a number of 4–7). In addition to the choroid, the vortex veins also drain the ciliary body and iris circulation. The vortex veins usually exit the sclera at the equator or up to 6 mm posterior to this location after forming an ampulla near the inner sclera. The vortex veins drain into the superior and inferior ophthalmic veins, which leave the orbit and enter the cavernous sinus.

ICGA has dramatically advanced our understanding and interpretation of choroidal imaging in ophthalmology. A normal angiogram is difficult to define because of the numerous changes that can occur with aging or that are related to differences in pigmentation. Anatomical variants are common in the arrangement of choroidal blood vessels and in the circulatory dynamics of filling and drainage. Arteries may emerge from different sites (usually perimacular and peripapillary) at variable intervals (Fig. 4). Veins show an unusual path and sometimes drain into the posterior pole. Due to the undulating nature of the vascular branch pathways, their layered distribution, and the variable caliber of the vessels, imaging them may require successive focusing.

OCT-A, in addition to depth-resolved information on the retinal vessels, may provide further insight into the choroidal blood flow. Nevertheless, the information yielded by segmenting different vascular layers deeper than Bruch’s membrane is still limited and not fully understood.

Starting from Bruch’s membrane for a 20-µm distance toward the choroidoscleral interface, different C-scans shaped on the Bruch membrane’s profile have a relatively homogeneous, grayish aspect. This aspect seems to be composed of a large number of tiny hyperintense or hypointense dots. This homogeneous pattern could correspond to the very richly anastomosed vascular layer of the choriocapillaris. No vascular channels are clearly detectable at this level, probably due to the limited resolution of OCT-A devices (Fig. 5a, b).

Different thin C-scans of 30 µm thickness each, i.e., deeper than the choriocapillaris, allow the analysis of Sattler’s layer (medium choroidal vessel layer). This layer is clearly visible on OCT-A B-scans with a quite continuous hyperintense signal that is mixed with some hypointense structures. C-scans show many hypointense, linear (black, tubular) entities resembling the medium vessel network on an almost continuous hyperintense, grayish background (Fig. 5c, d). The reason why we are not able to appreciate a fine hyperintense vascular network is that the attenuation of its own decorrelation signal induced by the structures mentioned above (RPE and choriocapillaris) makes it is indistinguishable from noise.

Even deeper down, C-scan segmentation allows the visualization of large choroidal vessels (Haller’s layer). The B-scan section shows alternative areas of hypointense (black, tubular) and hyperintense (grayish, diffuse) signals corresponding to these vessels, whose caliber is much larger than that of Sattler’s layer (Fig. 5e, f). C-scans show that the signal in this layer is discontinuous with multiple interruptions. In this case also the aspect is due to the signal attenuation from the structures mentioned above (RPE, choriocapillaris, and Sattler’s layer).

It has already been stated that FA is an invasive bidimensional examination method that cannot adequately resolve the deeper retinal networks (Fig. 6). Conversely, the major advantage of OCT-A is its capability to resolve the vascular layers of the retina in three dimensions.

In most cases, the in-built software of the different OCT-A devices distinguishes the retinal vasculature into two plexuses: the SCP and the DCP. The inner component of the DCP has also been termed the middle capillary plexus [10] or ICP [11]. Since the ICP is qualitatively and functionally distinct from the SCP and the DCP, a potential relevance of the ICP for clarifying the origin of middle retinal ischemic entities has been reported, therefore providing new insight into diabetic retinopathy (DR) [10]. Nevertheless, several concerns about the capability of the current OCT-A technologies to differentiate the ICP from the DCP have been raised [11, 12]. These are mainly due to failures in segmentation strategies and shadowgraphic artifacts, which cause the projection of the SCP onto the ICP and the DCP. Some recent advances in projection-resolved OCT-A algorithms substantially reduced these issues and allowed visualizing the three retinal vascular plexuses distinctly [13]. Projection-resolved OCT-A revealed vascular abnormalities in each of the different capillary layers, thereby making it possible to distinguish eyes with DR from healthy eyes and severe DR from mild DR with higher accuracy than with the standard double-layered segmentation.

In diabetic maculopathy, macular ischemia and enlargement of the FAZ can variably affect the SCP and the DCP [14]. Moreover, disorganization of the retinal inner layers [15] and paracentral acute middle maculopathy [16], which commonly occur in DR, are likely manifestations of ischemia at the level of the SCP and the DCP, respectively. Capillary nonperfusion can be associated also with photoreceptor and outer retinal disruption on structural OCT, which could be a specific manifestation of DCP ischemia [11, 16].

OCT-A provides several qualitative findings of diabetes-induced vascular abnormalities such as microaneurysms, intraretinal microvascular abnormalities (IRMA), venous beading, and neovascularizations. Microaneurysms are the most common features of DR since the very early stages. They are generally visible as roundish or fusiform focal vessel dilations. In some cases, due to the inconstant and nonlaminar blood flow inside these structures, they might not be appreciable on OCT-A. It has been demonstrated that the capability of OCT-A to detect microaneurysms is substantially inferior to that of FA [17].

OCT-A allows a quantitative assessment of vascular perfusion in different retinal layers [18-23]. Several studies (Table 1) have tried to quantify the diabetic capillary damage in different stages of the disease. Carnevali et al. [18] showed how OCT-A is able to disclose early vascular alterations in patients with diabetes, albeit without any biomicroscopic signs of DR. Agemy et al. [19] demonstrated a statistically significant difference in capillary perfusion density between the SCP and the DCP. The vascular perfusion damage seemed progressively increasing from the early stages of nonproliferative DR to proliferative DR (PDR). Similar findings were also documented for 13 eyes suffering from DR that underwent split-spectrum amplitude-decorrelation angiography followed by fractal dimension analysis to quantify macular perfusion [20]. Our group, in a recently published study, did not highlight any statistically significant difference in total vascular surface between the SCP and the DCP both among diabetic patients and among healthy subjects. This might mean that both layers are simultaneously affected. Moreover, a statistically significant difference in vascular density between healthy subjects and DR patients was reported for each different vascular layer [23].

DR is the leading cause of blindness in working-age individuals in the developed world, affecting approximately 75% of patients with diabetes mellitus after 15 years. Two of the early changes in diabetic eyes are loss of pericytes and proliferation of endothelial cells, leading to the development of microaneurysms. Pericyte loss impairs the blood-retinal barrier, thereby leading to venous dilation and beading. The gold standard for screening for DR is dilated biomicroscopic fundus examination, where microaneurysms in the posterior pole are typically the first sign on ophthalmoscopy. Although FA is more sensitive than examination to detect early DR, it is invasive, costly, and time-consuming and therefore is not appropriate as a screening test for DR. In a recent study, de Carlo et al. [17] demonstrated that the FAZ was increased in diabetic eyes compared with control eyes, and remodeling of the FAZ was also more prevalent in diabetic patients. Additionally, areas of capillary nonperfusion were noted more commonly in diabetics. Although 21% of the diabetic eyes demonstrated vascular tortuosity, this microvascular abnormality was noted in 25% of the control eyes as well, suggesting that some degree of vascular tortuosity may be a variant of normal anatomy and will probably not be useful as an OCT-A screening parameter for diabetic retinal vascular changes. Moreover, OCT-A allowed detecting microvascular changes in diabetic eyes before visualization on clinical examination. Remodeling and enlargement of the FAZ and areas of capillary nonperfusion were commonly noted in these eyes even before microaneurysms, which are currently believed to be the first clinical sign of DR [21] (Fig. 7).

Mild nonproliferative DR is a pathological condition that is mainly characterized by the presence of microaneurysms. OCT-A may allow identifying some subclinical retinal lesions, such as nonperfused areas, that can be hidden on biomicroscopic examinations of the eye fundus. The SCP and the perifoveal vascular arcade may appear rarefied and discontinuous (Fig. 8). A few variably sized areas of nonperfusion, both at the level of the SCP and at the level of the DCP, can be visible; in these areas the decorrelation signal coming from the vascular network is less intense or absent. Microaneurysms, which are the typical lesions of mild nonproliferative DR, are appreciable on OCT-A as roundish or fusiform, decorrelated structures. As already described, not all microaneurysms are visible on OCT-A [4].

Severe nonproliferative DR is a pathological condition characterized by the presence of more than 20 intraretinal hemorrhages in each of the 4 quadrants, definite venous beading in 2 quadrants, prominent IRMA in 1 quadrant, and no signs of proliferative retinopathy (Fig. 9). Intraretinal hemorrhages may block or attenuate the signal coming from decorrelated structures, thereby resulting in a sort of masquerading process on perfused vessels. The venous beading is clearly evaluable on OCT-A as contiguous focal dilations of retinal vessels; in some cases the blood flow may substantially vary when venous beading occurs, causing the partial visualization of vascular dilations, since they are not homogeneously perfused.

IRMA (Fig. 10a, b) are highly decorrelated vascular complexes in proximity to nonperfused areas. These are characterized by the presence of focal or diffuse dilations and a completely disrupted vascular architecture. They may extend from the inner to the outer retinal vascular layers, and sometimes exceed the borders of the INL as true intraretinal neovascularizations. All these features may be clearly identified on OCT-A, which can also highlight the presence of nonperfused areas at different depths. IRMA might be difficult to distinguish from intraretinal neovascularizations. A careful analysis of the macular pathological features in case of severe nonproliferative DR could reveal the initial signs of a proliferative retinopathy and lead to prompt treatment.

The main feature of PDR is neovascularization, which is typically located at the vitreoretinal interface. PDR may be associated with vitreous hemorrhage or tractional retinal detachment, which are among the major determinants of visual impairment in DR. On biomicroscopic examination, new retinal vessels appear as irregular vascular networks located on the retinal surface or growing into the vitreous cavity. FA is the gold-standard imaging modality for visualizing retinal neovascular structures; they appear as highly hyperfluorescent lesions in the early phases and generally show intense leakage during the late phases. OCT-A also showed the capability of detecting new retinal vessels in PDR as hyperintense vascular structures [24]. The depth-resolved nature of OCT-A not only allows highlighting the origin of a neovascularization, but it is also capable of defining a topographical relationship to nonperfused (or abnormally perfused) areas (Fig. 10, 11). Finally, B-scan angiograms easily localize lesions on the retinal surface or in the vitreous cavity (Fig. 10d).

Although retinal damage seems to be the main cause of visual loss in diabetes, there is also some evidence from histologic, angiographic, tomographic, and laser Doppler flowmetry studies that highlights the simultaneous involvement of the choroidal vasculature [25]. The choroid, especially the choriocapillaris, is the principal source of metabolic exchange for the avascular fovea and for the outer layers of the retina, including the RPE and photoreceptors. Choroidal changes in diabetic eyes include microaneurysms, dilation of the choriocapillaris, vascular remodeling, increased vascular tortuosity and capillary obstruction, vascular dropout, and new choroidal vessels [26].

Several studies have evaluated OCT-A in assessing diabetes-induced choriocapillaris vascular impairment. Agemy et al. [19] demonstrated a statistically significant reduction in vascular density in diabetic versus healthy eyes. Carnevali et al. [18] did not find any difference between normal subjects and type I diabetic patients at the level of the choriocapillaris using OCT-A. More recently, Nesper et al. [27] described a novel OCT-A index of retinal and choroidal nonperfusion: the percent area of nonperfusion. This index was significantly correlated with disease stage when considering the retinal vascular layers, but no significant correlation was found for the choriocapillaris. Our unpublished data show a substantial increase in choriocapillaris flow void areas among healthy eyes and eyes with mild nonproliferative DR. The mean vascular density ranged from 85% for diabetics to 89% for healthy controls (p < 0.05) (Fig. 12). Although promising, these results need further investigation in order to evaluate the role of OCT-A in assessing the functional impact of diabetes on the choroidal vasculature.

OCT-A might be a useful noninvasive imaging modality for diabetic patients, since it allows detecting microvascular changes even before the ophthalmoscopic appearance of typical DR signs. Furthermore, an automated quantification of macular perfusion could translate subjective qualitative findings into objective data, enabling ophthalmologists to integrate reliable functional data about retinal vascularity into clinical practice. Future studies will be focused on the possibility of defining cutoffs for disease severity, risk assessment, and monitoring of treatment outcome.

G. Coscas receives consulting fees from Allergan, Bayer, Heidelberg Engineering, and Novartis; F. Coscas receives consulting fees from Allergan, Bayer, and Novartis; M. Lupidi, J. Chhablani, and C. Cagini have nothing to disclose.

G. Coscas and M. Lupidi contributed equally to this review.