In recent years, it has become rare to attend an ophthalmology meeting without hearing at least one talk on the capabilities and potential of optical coherence tomography angiography (OCTA). At retina meetings and in journals (peer-reviewed or otherwise), OCTA is often highlighted as the future of retinal imaging. However, in my own informal survey of retinal specialists across the country who have had varying degrees of first-hand experience with the commercially available spectral domain OCTA platforms in the United States, I find most clinicians are underwhelmed by the impact of OCTA in their practices.

Though it might be reasonable to think that the muted reaction is driven by lack of reimbursement (it doesn’t have a billing code separate from OCT itself), I have heard from retinal specialists that (1) their images rarely look like the images presented at the meetings, (2) OCTA takes a long time to look at, (3) images can be difficult to interpret, and (4) it lacks a clinical scenario where the technology proves critical.

This lack of a “killer app” for OCTA may be exacerbated by current limitations. Projection artifacts (the appearance of flow in deeper retinal layers from overlying, more superficial flow) and motion artifacts from eye movements can make interpretation of OCTA images difficult.¹ These artifacts are then exacerbated by incorrect automated segmentation algorithms, which often occur when the normal retinal architecture is distorted. These segmentation errors are common in eyes with cystoid macular edema, retinal pigment epithelial detachments (PED), and subretinal hyperreflective material. Unfortunately, these are the situations where OCTA can be most helpful.

Despite these limitations, I have found OCTA can be useful in the diagnosis and management of some retinal patients. For example, it can help identify subtle choroidal nevoascular membranes (CNVM) that are not always readily apparent on fluorescein angiography. It can enable specialists to follow the changes in retinal perfusion in diabetic or vein occlusion patients.

What follows are some practical tips I have learned using these devices during routine retinal care. I have been using the Zeiss Cirrus 5000 Angioplex for the past year, for both clinic and research patients. In addition, I demo’ed the AngioVue system from Optovue for several months this past year. The Topcon DRI-OCT Triton and the Heidelberg Engineering Spectralis OCTA systems are not yet commercially available in the United States. Munk et al published a helpful article comparing the different OCTA systems.²

KNOW WHERE YOU ARE AND WHAT YOU ARE LOOKING FOR

The structural (B-scan) image can be used to identify the correct x- and y-axes and orient the OCTA image. Then, by manually adjusting the automated segmentation, the z-axis can be shifted to capture the flow in a specific area of the retina, even if the remaining segmentation is off elsewhere in the retina. For example, in a patient with an irregular fibrovascular PED and concern for possible conversion to exudative AMD, the OCTA can be manually adjusted to look for flow in that area, even if the automated segmentation does not properly follow the contours of the PED. I have generally found it easiest to start with either the outer retina or the choriocapillaris slabs and slide the inner and outer limits (based on distance from the ILM or RPE) to isolate the area of retina I am concerned about.

MANUALLY ADJUST THE AUTOMATED SEGMENTATION TO IDENTIFY ARTIFACTS

This is the most critical component to extracting useful information from OCTA. As others have noted, even in a healthy retina, the segmentation does not follow the boundaries of the inner and outer retinal capillary plexus.³ In eyes with macular disease, this is even more of an issue. Not only does the z-axis need to be adjusted to capture the section of retina you are interested in (eg, just above or below a RPE detachment), you can then scroll more superficially to determine if what you are seeing in the outer retina is present, or if it is a projection artifact from the inner retina. If the same flow line is present superficially, and never disappears as you toggle back and forth between superficial and deeper layers, then it likely represents a projection artifact from the inner retina.

Another challenge is that the avascular RPE shows up as bright even though OCTA is supposed to be a difference map and show only flow. However, the RPE, which is hyperreflective on structural (B scan) OCT images, remains hyperreflective on OCTA due to a projection artifact from the overlying retinal vessels.

The automated segmentation algorithms use the RPE and ILM because their hyperreflectivity makes them easy to detect, and the various layers are based on offset distances from either the ILM or RPE. Each platform has a slightly different algorithm for these offsets. For example, on the Angioplex (Zeiss), the “avascular” slab has an outer boundary just above the RPE, and generally shows up black (hyporeflective) representing the lack of vascular flow in the outer retina. The AngioVue “outer retina” covers a similar segment of the retina as the “avascular” slab on the Zeiss platform, but includes the RPE, and therefore has a stippled reflectivity pattern.

The challenge occurs when the RPE is elevated or has migrated into the retina. In these cases, the automated segmentation often includes some slices through RPE, which can then appear as flow on the en face OCTA image. This needs to be manually adjusted and mentally accounted for during image interpretation (Figure 1).

USE THE STRUCTURAL (B SCAN) OCT IMAGE TO HELP

The traditional B scan structural images can be useful in several ways. First, they help orient where in the retina you are looking. For example, if you are concerned for a possible choroidal neovascular membrane, and there is a suspicious area on the OCTA en face image, move the B scan cross-sectional image to that section of the retina and confirm this is the same area that aroused your suspicion. Then, use the B scan image to see where the automated segmentation lines are drawn and focus on just the retinal layer you are concerned about.

In addition to toggling to the more superficial layers to rule out artifacts, the B scan image can also be used with the flow overlay to differentiate projection artifact from real flow. If there is a red area of flow in the inner retina, and a “shadow” of red deep to it, then the flow seen in the deeper retina in the en face image is likely a projection artifact.

SELECT THE APPROPRIATE SCAN SIZE

The 3 mm x 3 mm scan gives better resolution than the 6 mm x 6 mm or 8 mm x 8 mm patterns (Figure 2). The higher resolution scans have in practice proven better for CNVM detection. I will let my technicians know to use the 3 mm x 3 mm scan in cases where I am looking for CNVMs. The larger scan patterns are helpful for assessing the retinal perfusion status in the superficial and deep capillary plexus in patients with DR or RVO.

SELECTIVELY INCORPORATE OCTA INTO YOUR WORKFLOW

It is easy to have the technician add a scan pattern, which just takes a few seconds, when he or she is already acquiring a structural OCT on that patient on the same imaging platform. However, in a busy retina clinic, it can slow patient throughput if the structural B scan OCT images are being obtained on a different platform (eg, a Heidelberg Spectralis, which does not offer OCTA in the United States currently). It takes even more time to load and review these large voxel image sets, especially if the segmentation needs to be adjusted manually. Because of this, I do not obtain an OCTA on every patient at every visit, but will use it selectively to aid in diagnosis and management.

OCTA IS COMPLEMENTARY TO FLUORESCEIN ANGIOGRAPHY

I have not yet found OCTA to be a replacement for FA. OCTA shows flow, while FA highlights leakage. In addition, as research on OCTA progresses, so too does research on ultrawidefield angiography. Current OCTA platforms are primarily useful in the macula, and can be stitched together to image extramacular retinal vasculature,⁴ though still not nearly as peripheral or complete as ultrawidefield FA. Though swept-source OCTA offers faster scan speeds, and thus better resolution across a wider field of view with fewer motion artifacts, it is not broadly available in the United States. Ultimately, OCTA may replace FA, but it hasn’t yet.

Despite the drawbacks of OCTA, I have found it helpful to have OCTA in my practice, much like FA and ICG are useful in select cases. I am confident that as the research progresses and as the technology improves, OCTA will become an increasingly common part of routine retinal care. RP

REFERENCES

1. Spaide RF, Fujimoto JG, Waheed NK. Image artifacts in optical coherence angiography. Retina. 2015; 35(11):2163-2180.
2. Munk MR, Giannakaki-Zimmermann H, Berger L, Huf W, Ebneter A, Wolf S, Zinkernagel MS. OCT-angiography: a qualitative and quantitative comparison of 4 OCT-A devices. PLoS ONE. 12(5): e0177059.
3. Spaide RF, Curcio CA. Evaluation of segmentation of the superficial and deep vascular layers of the retina by optical coherence tomography angiography instruments in normal eyes. JAMA Ophthalmol. 2017;135(3):259-262.
4. Goldberg RA, Chen M, Leahy C, et al. OCT angiography montaging for wide-field visualization of retinal pathologies. Presented at: the ARVO Annual Meeting, 2017, Baltimore, MD.