OCT Angiography (OCTA) in Retinal Diagnostics
Optical coherence tomography angiography (OCTA) is an imaging modality which can be applied in ophthalmology to provide detailed visualization of the perfusion of vascular networks in the eye. Compared to previous state of the art dye-based imaging, such as fluorescein angiography, OCTA is non-invasive, time-efficient, and it allows for the examination of retinal vasculature in 3D. These advantages of the technique combined with the good usability in commercial devices led to a quick adoption of the new modality in the clinical routine. However, the interpretation of OCTA data is not without problems: Commonly observed image artifacts and the quite involved algorithmic details of OCTA signal construction can make the clinical assessment of OCTA exams challenging. In this article we describe the technical background of OCTA and discuss the data acquisition process, common image visualization techniques, as well as limitations and sources of artifacts of the modality. Examples of clinical cases underline the increasing importance of the OCTA technology in ophthalmology and its relation to dye-based angiography.
KeywordsOCT angiography (OCTA) Optical coherence tomography angiography (OCTA) Fluorescence angiography Fluorescein angiography (FA) Indocyanine green angiography (ICGA) Retina Vascular plexus
Using dense volume scans, it is possible to obtain OCTA images that are similar to fluorescence angiography images, which are the clinical gold standard. In contrast to fluorescence angiography, OCTA has the advantage of not requiring any dye injection. Moreover, while fluorescence angiography provides only two-dimensional images of the fundus, OCTA enables the visualization of structure and blood flow within the vitreous, the retina, and the choroid, separately (see Sects. 6.2.2 and 6.2.3). Using appropriately adjusted segmentation boundaries, it is also possible to examine the distinct capillary networks of the retina (with vessel diameters as small as approx. 8 μm) . The definition of the separating boundaries has evolved since the introduction of OCTA in the clinical practice and is described in Sect. 6.2.4.
Various OCTA algorithms have been proposed and utilized in research and in clinical devices for OCTA image construction (see Sect. 6.2.1). Therefore OCTA images from different devices vary in appearance [4, 5], which may result in different clinical diagnostic interpretations. While each unique OCTA algorithm is subject to slightly different limitations that are attributed to its overall approach, there are certain confounding factors and/or limitations that impact all algorithms and are innate characteristics of this imaging modality . These factors include, but are not limited to, reduced light penetration in deeper layers and image artifacts projected from more superficial layers to deeper ones. Artifacts can originate from image acquisition, eye motion, image processing, and display strategies . Section 6.3 describes some of the major artifacts related to OCTA as well as state-of-the-art countermeasures.
Section 6.2.5 briefly introduces OCTA metrics, which are intended for quantitative evaluation of OCTA data. Such numerical aggregates of the image data enable an objective analysis of disease progression and statistical conclusions in larger studies of diseases.
With fluorescence angiography, namely Fluorescein Angiography (FA) and Indocyanine Green Angiography (ICGA) , dynamic phenomena such as dye leakage, pooling, and staining can additionally be observed. These phenomena cannot be observed with OCTA because no motion of blood cells is involved. While these phenomena are also used in clinical diagnosis , retinal pathology can also be obscured by leakage or hemorrhage. In contrast, OCTA can generate high contrast, well-defined images of the microvasculature below areas of leakage or hemorrhage . Therefore, dye-based angiography and OCTA are giving complementary information. To illustrate the similarities and differences of OCTA with respect to the gold standard dye-based angiography, Sect. 6.4 provides side-by-side comparisons for clinical cases of diabetic retinopathy, retinal vein occlusion, macular telangiectasia, and age-related macular degeneration.
6.2 Technical Foundation for Clinical OCTA Imaging
OCT systems typically produce section images as shown in Fig. 6.1. Such images are commonly referred to as B-Scans. As can be seen in Fig. 6.1a, the B-Scans show a grainy pattern, also known as speckle pattern. These speckles are inherent to the interferometric OCT measurement. If two B-Scans are taken from the very same location of the retina (cf. Fig. 6.1a, b), the speckle pattern at locations of static tissue basically stays the same. In contrast, at locations of perfused blood vessels, the speckle pattern changes over time. The basic principle of OCTA is therefore to analyze the temporal variation of the OCT signal in order to derive an image of the perfused retinal vasculature.
To allow for the creation of images similar to fundus images from standard angiography, volume scans are performed. This means, that multiple adjacent B-Scans are acquired to cover extended regions of the retina. Eventually, these B-Scans are combined to form a three-dimensional sample of the retinal structure and blood flow.
6.2.1 OCTA Signal Processing and Image Construction
In OCTA imaging, several OCT B-Scans of the same retinal cross section are acquired repeatedly in short succession. Within this retinal cross section, at locations of static tissue, the microscopic configuration of illuminated scattering particles in the beam focus is well preserved over sequential acquisitions and consequently yields a consistent OCT signal over time. Contrarily, at locations where directed motion is present in the sample, like in retinal blood vessels, the scattering particles are continuously replaced by other particles in subsequent acquisitions. This continuous exchange of microscopic particles modulates the OCT signal and introduces an additional source of variability to the repeated measurements. Overall, the smaller the remaining portion of conserved scattering particles in the beam waist, the higher the variability in the OCT signal of sequential acquisitions. Maximum OCT signal variability is observed when the scattering particles are completely replaced by others in subsequent measurements. Clinical OCTA imaging on current commercially available OCT device hardware is typically operating in this regime, as the physiological blood flow speeds in most of the perfused retinal vasculature  significantly exceeds the velocity limit (i.e. typically only few mm/s) determined by the product of OCT beam waist diameter and scan repetition rate. Hence, it is unlikely to observe the same red blood cell configuration in two successive B-Scans. This, however, is a necessary requirement for deriving quantitative blood flow velocity measurements from the OCT signal variation in repeated scans (cf. Chap. 7). Accepting this current technical limitation of clinical OCT devices, OCTA algorithms for signal construction rather focus on reliably differentiating locations of significant blood flow from static tissue, instead of measuring blood flow velocity quantitatively. In this context, OCTA image construction is a quasi-binary classification problem with the goal to optimally distinguish significant flow from static tissue at each location of the retina.
Different algorithmic strategies have been suggested in the past for optimally addressing this classification task. While some algorithms are using exclusively either amplitude or phase of the complex OCT signal, others are combining information from both. As the microscopic pattern of illuminated particles in the sample and its detailed configurational change according to motion is practically inaccessible to measurement, its influence on the resulting OCT signal is typically rather considered probabilistically as a stochastic contribution to the overall measurement. OCTA algorithms are thus quantifying the amount of variability in the random realizations of the measured OCT signal from repeated acquisitions in one or the other way. Practically, for instance the temporal correlation  or the overall variance  of the signal or other more involved statistical parameters  with an expected relation to the OCT signal variability are assessed in different approaches. These statistical parameters, potentially after additional post-processing and contrast enhancement, are subsequently taken as the resulting OCTA signal in arbitrary units. As an alternative to these statistical parameter based OCTA signal construction methods, underlying probabilistic statistical models for the random OCT signal from sample locations with and without directed flow can be derived from theory and experiment [9, 10, 11]. Based on these underlying models and the repeated OCT signal observations, a probability for being static (versus in flow) can be assigned to each measured location. This holds the advantage of yielding easily interpretable probability values and no further contrast enhancement of the resulting OCTA signal is needed.
6.2.2 OCTA Data Visualization
In addition to en face images, section images are used for review of the spatial relationship of retinal structure and blood flow. The section images may have the same orientation as the originally acquired B-Scans (Fig. 6.2c) but may also be arbitrarily oriented within the volume; for instance Fig. 6.2d shows a section orthogonal to the original B-Scans. To provide a direct visual correlation of structural and flow information, structural OCT section images and the corresponding OCTA blood flow information at the same location can be superimposed; see Fig. 6.2c1–c3, d.
6.2.3 Projection Methods
The previous examples were given to illustrate the difference when reading the en face images. For OCTA analytics based on en face images there are further aspects that need to be considered. When using the mean projection as a measure for vessel density within several slabs of varying thickness, it is not possible to directly compare the results because voxels were given different weights in the different slabs. Also averaging of mean projections from different slabs will not give the same result as the computation of the mean projection of the combined slab.
6.2.4 Retinal Vascular Plexuses
In order to accurately detect and manage retinal vascular conditions, it is important to precisely discern the different retinal vascular plexuses. It is also important that slabs enable a continuous representation of the retinal and choroidal vasculature so that possible vascular abnormalities are not missed during image review. Currently, conflicting definitions of the axial location of boundaries between retinal plexuses make the direct comparison of en face images from different devices difficult .
6.2.5 Quantification of OCTA Data
For objective assessment of disease progression and its documentation, and to enable comparisons to normative data, a concise summary of the image data in terms of numerical measurements is desirable. Such numeric parameters, describing the structure of the vasculature network as derived from the OCTA images, are also referred to as “OCTA metrics” or “OCTA analytics”. Clearly, this must not be confused with OCT-based flowmetry (subject of Chap. 7), where physical blood flow velocity is measured quantitatively.
Typical examples of OCTA metrics parameters, quantifying static structural aspects of the eye’s vasculature, are various vessel density measures (vessel/perfusion density, binarized vessel density or vessel area density, skeletonized vessel density or vessel length index). These parameters are suitable for capturing dropout of vasculature that occurs in diseases like diabetic retinopathy, retinal vein occlusion or glaucoma [17, 18, 19, 20]. A quantification of flow void area is also possible, for instance for assessing the choriocapillaris structure .
Besides vessel density, other parameters that summarize morphological features of vessel branches and vessel network structure, including complexity measures such as vessel tortuosity, fractal dimension, or branching point densities and vessel diameter statistics, are in common use [17, 18, 19, 22, 23]. These measures aim at capturing pathological alterations of vessel shape and spatial arrangement, as occurring for example in diabetic retinopathy, macular telangiectasia, or neovascularization in age related macular degeneration.
These characteristics may be analyzed for the whole scan area, or alternatively as aggregates over sectors defined by specific grids (e.g. ETDRS grids), which are usually adapted to the eye anatomy and allow for the detection of spatially localized changes over time or deviations from the statistics of normal reference data.
In currently available approaches, vessel density measures are interpreted as a two-dimensional density, i.e. the fraction of area of a slab projection occupied by detected vessels (typically after applying a thresholding operation). As long as the instrument is able to axially resolve the different layers of capillaries that can be anatomically distinguished, quantitative OCTA parameters can be derived for each of them independently. To suppress the influence of larger vessels, two approaches are commonly used: Either slab projections of vessels are reduced to their centerlines (i.e. “skeletonization”), or larger vessels are simply masked out. This emphasizes thinner capillaries in the analysis.
Further quantitative parameters derived from OCTA data are area and shape measurements of specific vasculature regions, in particular the foveal avascular zone (FAZ) [19, 22] or segmented neovascular lesions .
There is high interest to use results from quantitative OCTA parameters as endpoints in clinical studies [25, 26, 27]. For this purpose, OCTA metrics need to be both repeatable and reproducible. Therefore, initial studies mostly focused on analyzing the robustness of the measurements. Errors in scan geometry, evaluation grid placement, layer segmentation, and parameters such as variable signal strength can negatively impact the measurement precision. Furthermore, data from instruments of different vendors are not directly comparable [28, 29]. This is due to differences in resolution as well as signal generation and postprocessing algorithms such as filtering, artifact suppression and thresholding . Also differences in the slab definitions (cf. Sect. 6.2.4) and layer segmentation results of different devices need to be carefully taken into account when comparing images or quantitative analysis results across devices.
6.3 Image Artifacts and Countermeasures
6.3.1 Projection Artifacts
6.3.2 Segmentation Artifacts
Considering that slabs are mainly defined by automatically segmented retinal layer boundaries, careful review of the segmentation is critical for correct interpretation of the en face projections. Segmentation failures are especially common in diseases where the appearance and shape of retinal layer is altered. For instance, intraretinal fluid, large pigment epithelial detachments, choroidal neovascularization, and certain atrophies often cause segmentation errors in state-of-the-art OCTA devices. A manual correction of such errors is cumbersome, if the correction is based on individual B-Scans within the dense volume. To facilitate and speed up the process of correcting compromised slab boundaries, interactive segmentation correction tools have been introduced. These tools propagate manual corrections of only a few B-Scans to the remainder of the volume, by fusing automatic segmentation results with these user-provided hints . These countermeasures are merely a workaround until more robust segmentation methods for the vascular networks are available.
6.3.3 Motion Artifacts
Adequate compensation of eye motion is one of the most critical aspects of OCTA acquisition. In order to detect temporal changes in the OCT signal related to blood flow in capillaries , eye motion needs to be detected and compensated for very accurately. Therefore sampling schemes and different eye tracking implementations play a crucial role for each device’s overall performance of blood flow visualization at the capillary level.
Eye movements affect the acquisition of OCTA data in two distinct ways. First of all, the detection of blood flow related changes in the OCT signal from repeated B-Scans requires spatial overlap of these scans (see Sect. 6.2.1). This means, that the scans must overlap within the lateral width of the probing beam, which is typically in the order of 15 μm. Secondly, larger eye movements can lead to geometrical distortion or missing data in the OCTA volume scans. The volume scans are typically acquired within several seconds. On this time scale, it is very likely that larger eye movement occurs due to saccades, change of fixation, or change of head pose.
Slow eye drifts within the B-Scan plane can be compensated for by image registration of the successive B-Scans, effectively removing the temporal signal change in regions of static tissue (cf. Fig. 6.1). Larger eye motion (saccades) or eye motion perpendicular to the B-Scan plane can strongly deteriorate the OCTA signal, because of insufficient spatial overlap of the B-Scan samples. This leads to missing data and geometrical distortions in the OCTA volume scans, if no countermeasures are taken. The goal of any motion artifact compensation is to ensure that OCTA data acquired from visit to visit is devoid of errors that may reduce the precision of quantitative change analyses.
There are two major approaches to mitigate motion artifacts in OCTA. One approach is to acquire several independent OCTA volume scans and combine the information in post-processing [34, 35, 36]. The second approach is using real-time tracking.
The post-processing approach has the advantage of relatively short acquisition times. In practice, often only two volume scans with perpendicular orientation of the fast scanning axis are used . In effect, the resulting volume is obtained by interpolation and averaging of the different input volumes. Larger data gaps from one volume are normally filled up with information from the other volume, but in general, there is no guarantee to obtain distortion-free results .
The real-time tracking approach employs accurate measurements of the eye motion in real-time (see also Chap. 3). B-Scans which are affected by too strong motion are re-acquired. During periods of slow eye motion (eye drifts) the real-time eye motion measurements can also be used to actively control the OCT scanners to keep the beam on the nominal scanning path. Due to the data filtration in the event of strong eye motion, the real-time tracking approach is sometimes slow in acquisition. However, using real-time tracking, the obtained volumes are geometrically accurate and uniformly sampled without gaps from missing data.
6.3.4 Lateral and Axial Resolution
The lateral resolution of an OCT system is determined by both the optical point spread function (PSF) as well as sampling density, i.e. the digital resolution. The OCTA signal from one voxel can be seen as a mixture of contributions from scatterers within the support of the combined (optical and digital) PSF. The larger the PSF, the more scatterers contribute to the mixture so that is becomes more challenging to separate the individual contributions, in particular the components due to flowing scatterers (blood cells) from the static components (tissue) according to their discriminating statistics.
The optical PSF is influenced by the imaging system as well as the imaged eye. Poor adjustment of the instrument’s focus as well as aberrations in the eye widens the PSF and lead to suboptimal signal separation. Similarly, using wide-field optics with smaller numerical aperture as well as less dense digital sampling in favor of covering larger fields of view compromises the ability to resolve small capillary details. The effect of the sampling density on the visibility of small capillaries is illustrated in Fig. 6.6, compare a and b.
The axial resolution in OCT is independent of the lateral resolution and is determined by the spectral bandwidth of the light source . However, using split-spectrum approaches  for OCTA processing, an algorithmic trade-off can be made between axial resolution and signal to noise ratio. Splitting the spectrum allows to obtain, from a single scan, several B-Scan sections of lower axial resolution (due to the lower bandwidth of the spectral sub-bands), which are then used as additional samples with independent shot noise contributions for improving the OCTA signal. This loss of resolution may impede the ability to separate axially closely spaced but distinct vascular layers, in comparison to algorithms that maintain the axially high optical resolution of the underlying OCT signals (cf. Fig. 6.6c, d) .
6.4 Clinical Application of OCTA
6.4.1 Diabetic Retinopathy
Diabetic retinopathy (DR) is classified into different stages according to stereoscopic color fundus photographs. The typical features of DR are microaneurysms, intraretinal hemorrhages, intraretinal microvascular abnormalities, venous beading, cotton wool spots, hard exudates and neovascularization.
The gold standard imaging technique for the assessment of macular perfusion is fluorescein angiography . It is able to show leaking microaneurysms and capillary non-perfusion. However, it produces only two-dimensional images in which fluorescence signals of the superficial and deep capillary networks overlap and are difficult to distinguish, especially when the dye leaks [38, 39]. In this context, OCTA is a useful imaging technique for the assessment of retinal vasculature in the different capillary networks.
6.4.2 Retinal Vein Occlusion
The ophthalmoscopic signs are venous dilation, tortuosity, intraretinal hemorrhage, retinal edema and hemorrhage in the vicinity of the occluded vein. Fluorescein angiography typically reveals delayed filling in the distribution of the involved retinal vessels. The veins are dilated and tortuous. There is leakage from capillaries with dye accumulating in the substance of the retina or within cystoid spaces. Fluorescein angiography has been the gold standard for visualizing retinal non-perfusion areas which appear as dark areas in the images . Retinal neovascularization can be identified by leakage of the fluorescein dye. Fluorescein angiography has been critically important for the diagnosis and the treatment decision.
OCTA shows the areas of vascular non-perfusion, the dilated tortuous venous segments, the microvascular abnormalities and the neovascularization. However, one of the most sight threatening complications is macular edema. In this context, it is crucial to verify the segmentations used for OCTA visualization . The difficulty with en face imaging is that selective enlargement of retinal layers occurs and these layers are typically not segmented correctly. The resultant en face vascular images may not show the correct representations of the actual flow characteristics.
6.4.3 Macular Telangiectasia
Idiopathic perifoveal or juxtafoveolar retinal telangiectasia are retinal capillary ectasia limited to the perifoveal area without any apparent specific cause .
Initially, idiopathic macular telangiectasia were divided into four groups according to the Gass classification . Subsequently, Yannuzzi proposed a new classification in which type 1 was defined as aneurysmal telangiectasia and type 2 as perifoveal idiopathic macular telangiectasia .
Type 1 is closely related to Coats disease, or more specifically a milder form of Coats previously known as Leber miliary aneurysms. It generally involves only one eye, and both the peripheral retina and macula can be affected. Fluorescein angiography reveals telangiectasia and multiple capillary, venular and arteriolar aneurysms with late leakage. Type 2 are bilateral, temporal and symmetrical, however , there have been reports of unilateral, asymmetric, and asymptomatic cases . It has been hypothesized that the primary involvement is the alteration of Müller cells and secondary the vascular and tissue remodeling. Fluorescein angiography displays the capillary telangiectasia with dilated and blunted retinal venules at right angles into the temporal parafoveolar area.
6.4.4 Age Related Macular Degeneration
Several studies have shown the utility of OCTA in the diagnosis and monitoring of age related macular degeneration (AMD). The advent of OCTA offered the opportunity to non-invasively visualize the neovascular networks and correlate the OCTA appearance with the standard imaging techniques and observe new findings.
In this article the basic principles, major sources of artifacts and the clinical application of OCTA were discussed. The discussion of the clinical examples showed that OCTA provides diagnostic value in several vascular diseases of the eye. However, the current state of OCTA has not yet fully replaced the gold standard dye-based angiography because of important limitations.
Ongoing endeavors to improve OCTA are addressing these shortcomings.
This includes targeting greater fields of view, which is especially important in DR. Improvements to automatic segmentation algorithms in the context of pathological alterations are necessary for reliable results. Based on these future improvements it is expected that robust metrics and sensitive monitoring of disease progression can be achieved.
- 6.Johnson RN, Fu AD, McDonald HR, Jumper JM, Ai E, Cunningham ET, Lujan BJ. Fluorescein angiography: basic principles and interpretation. In: Retina. 5th ed. Amsterdam: Elsevier Inc; 2012.Google Scholar
- 16.Rocholz R, Teussink MM, Dolz-Marco R, Holzhey C, Dechent J, Tafreshi A, Schulz S. SPECTRALIS optical coherence tomography angiography (OCTA): principles and clinical applications. Heidelberg Engineering Academy. 2018. https://academy.heidelbergengineering.com/course/view.php?id=505.
- 19.Jia Y, Bailey ST, Hwang TS, McClintic SM, Gao SS, Pennesi ME, Flaxel CJ, Lauer AK, Wilson DJ, Hornegger J, Fujimoto JG, Huang D. Quantitative optical coherence tomography angiography of vascular abnormalities in the living human eye. Proc Natl Acad Sci U S A. 2015;112(18):E2395–402.PubMedPubMedCentralCrossRefGoogle Scholar
- 22.Lu Y, Simonett JM, Wang J, Zhang M, Hwang T, Hagag AM, Huang D, Li D, Jia Y. Evaluation of automatically quantified foveal avascular zone metrics for diagnosis of diabetic retinopathy using optical coherence tomography angiography. Invest Ophthalmol Vis Sci. 2018;59(6):2212–21.PubMedPubMedCentralCrossRefGoogle Scholar
- 23.Coscas F, Cabral D, Pereira T, Geraldes C, Narotamo H, Miere A, Lupidi M, Sellam A, Papoila A, Coscas G, Souied E. Quantitative optical coherence tomography angiography biomarkers for neovascular age-related macular degeneration in remission. PLoS One. 2018;13(10):e0205513.PubMedPubMedCentralCrossRefGoogle Scholar
- 24.Zhang Q, Chen CL, Chu Z, Zheng F, Miller A, Roisman L, Rafael de Oliveira Dias J, Yehoshua Z, Schaal KB, Feuer W, Gregori G, Kubach S, An L, Stetson PF, Durbin MK, Rosenfeld PJ, Wang RK. Automated quantitation of choroidal neovascularization: a comparison study between spectral-domain and swept-source OCT angiograms. Invest Ophthalmol Vis Sci. 2017;58(3):1506–13.PubMedPubMedCentralCrossRefGoogle Scholar
- 30.Rabiolo A, Gelormini F, Sacconi R, Cicinelli MV, Triolo G, Bettin P, Nouri-Mahdavi K, Bandello F, Querques G. Comparison of methods to quantify macular and peripapillary vessel density in optical coherence tomography angiography. PLoS One. 2018;13(10):e0205773.PubMedPubMedCentralCrossRefGoogle Scholar
- 33.Lee K, Zhang H, Wahle A, Abràmoff MD, Sonka M. Multi-layer 3D simultaneous retinal OCT layer segmentation: just-enough interaction for routine clinical use. Cham: Springer International Publishing; 2018. p. 862–71.Google Scholar
- 35.Kraus M, Mayer MA, Bock R, Potsaid B, Manjunath V, Duker JS, Hornegger J, Fujimoto JG. Motion artifact correction in OCT volume scans using image registration. Invest Ophthalmol Vis Sci. 2010;51(13):4405.Google Scholar
- 41.Salz DA, de Carlo TE, Adhi M, Moult E, Choi W, Baumal CR, Witkin AJ, Duker JS, Fujimoto JG, Waheed NK. Select features of diabetic retinopathy on swept-source optical coherence tomographic angiography compared with fluorescein angiography and normal eyes. JAMA Ophthalmol. 2016;134(6):644–50.PubMedPubMedCentralCrossRefGoogle Scholar
- 45.Wilkinson CP, Ferris FL III, Klein RE, Lee PP, Agardh CD, Davis M, Dills D, Kampik A, Pararajasegaram R, Verdaguer JT. Global Diabetic Retinopathy Project, G., Proposed international clinical diabetic retinopathy and diabetic macular edema disease severity scales. Ophthalmology. 2003;110(9):1677–82.PubMedCrossRefGoogle Scholar
- 51.Gass JDM. Stereoscopic atlas of macular diseases: diagnosis and treatment (2 Volume Set). St. Louis, MO: Mosby; 1997.Google Scholar
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.