All spectral domain optical coherence tomography (SD-OCT) machines have artifacts. Our job as physicians is to spot these artifacts in order to make an accurate diagnosis. Although tools and technologies can help the physician diagnose or monitor glaucoma, we still need to critically assess any test results to see if it can be used for treatment decisions.
SD-OCT plays a key role in the diagnosis and management of glaucoma. These high-resolution cross-sectional scans of the eye allow ophthalmologists to detect structural changes with axial resolutions of 5 to 7 μm at the peripapillary retina, optic nerve head, and macula. However, like any technology, optical coherence tomography (OCT) is imperfect, and the images it generates may sometimes contain artifacts. These artifacts can arise from errors in scan acquisition, subsequent analysis, or from ocular pathology unrelated to glaucoma. Regardless of the cause, artifacts represent false data and can be misleading to those unfamiliar with them. Therefore, it is important for ophthalmologists to be able to identify artifacts when they occur and to interpret OCT information within the context of the entire patient.
First, this chapter will discuss the types of artifacts that ophthalmologists may encounter when using OCT to diagnose and manage glaucoma. Second, the chapter will describe how these artifacts can give rise to the phenomenon of “OCT diseases” and will review relevant clinical examples of red and green disease. Lastly, the chapter will discuss future directions, such as three-dimensional parameters and OCT angiography, and the role that artifacts will play.
Key wordsartifact – decentration – poor signal – segmentation error – floor effect – OCT diseases
8 Artifacts and Masqueraders
8.1 Incidence of Artifacts in OCT Imaging
Artifacts are unintended and undesired, yet inescapable, features of optical coherence tomography (OCT) scans. These artifacts occur frequently in clinical practice. In fact, between 7.1 and 46.3% 1 , 2 of peripapillary retinal nerve fiber layer (pRNFL) thickness scans from spectral domain optical coherence tomography (SD-OCT) demonstrate some form of artifact. The SD-OCT instrument is also capable of generating high-density volume scans of the macula and optic nerve head (ONH). These cube scans carry frequent artifacts as well, with rates ranging between 6.0 and 90.9% 3 , 4 for the macula and 12.1 and 84.0% 5 , 6 for the ONH. Notice the wide range of percentages of reported artifact rates. These variabilities may arise from differences in scan protocols, scan locations, OCT machines, enrollment criteria, study methodologies, and artifact definitions used throughout the literature. These differences make it difficult to compare numbers between any two studies. The high frequency of artifacts may seem intimidating. However, by keeping in mind the principles of this chapter, the clinician will be able to avoid common pitfalls from OCT artifacts.
8.2 Etiologies of Peripapillary RNFL OCT Artifacts
Peripapillary RNFL thickness is the OCT parameter that is most commonly used when following glaucoma patients. Therefore, the following section describes common artifacts that present in scans of the peripapillary RNFL.
8.2.1 Artifacts from Errors in Scan Acquisition
Decentration artifact (Fig. 8‑1) is the most common type of artifact seen in peripapillary RNFL circular B-scans, occurring in 27.8% of Spectralis (Heidelberg Engineering, Heidelberg, Germany) B-scans, where a technician needs to manually center the RNFL circular scan over the optic nerve and when decentration is defined as imperfect alignment of the circular scan over the ONH by at least 10%. 2 Decentration artifacts are less of an issue in machines where the software automatically determines the optic nerve location (e.g., Cirrus, Carl Zeiss Meditec, Dublin, CA). Since the RNFL normally thins farther away from the ONH, decentration may cause artifactually increased or decreased measurements of RNFL thickness for a given sector or clock hour, depending on whether the measured region is closer or farther away from the optic disc, respectively. For instance, if the scan circle is shifted superiorly, then the superior quadrant will appear artifactually thin (farther from the nerve head) and the inferior quadrant will appear artifactually thick (closer to the nerve head). It is evident that superior decentration may make it appear to the clinician as though the patient has glaucomatous thinning in the superior quadrant, when in fact there was no change at all.
Poor signal (Fig. 8‑2 and Fig. 8‑6) is another common type of artifact seen in OCT scans, presenting in 5.1% of peripapillary RNFL circle scans taken with the Spectralis SD-OCT machine (Heidelberg Engineering, Heidelberg, Germany). 2 Poor signal often looks grainy or static-like, which makes it difficult to discern the layers of the retina. When the OCT attempts to segment these blurry or grainy images, it is difficult and sometimes impossible for the software to locate the boundaries of the retinal layers, thereby leading to false measurements.
Signal can be degraded by any of the ocular structures that the OCT light beam must pass in order to image the retina. This includes the cornea, anterior chamber, lens, and vitreous. The most common causes of poor signal include dry eyes 7 and cataract. 8 These conditions make it difficult for the OCT machine to take a clear picture of the back of the eye, just as it would be difficult for a camera to take a good photo of an object through frosted glass. Stein and associates sought to establish the effects of dry eye by taping normal subjects’ eyelids open to prevent blinking over the course of the testing period. They found significant reductions in both signal strength and RNFL thickness strictly by preventing blinking (Stratus OCT, version 3.0; Carl Zeiss Meditec, Inc., Dublin, CA). 7 Also using the Stratus OCT, Mwanza and associates found a 9.3 and 24.1% increase in peripapillary RNFL thickness and signal strength, respectively, in 45 glaucomatous and normal eyes after cataract removal. 8 These studies demonstrate that poor signal due to dry eyes and cataract may significantly impair accurate OCT measurement of RNFL thickness, and that strategies such as blinking and administration of artificial tears can improve signal quality if a scan needs to be retaken. Other less common causes of poor signal include anterior chamber cell or flare, posterior capsular opacification, and vitreous hemorrhage. In fact, it is even important to make sure that the OCT lens is cleaned, so as to avoid “smudge artifact” from dirty equipment degrading signal quality.
OCT machines automatically generate a quality score to reflect the signal strength of each scan. The range of scores differs among instruments, and each manufacturer defines a separate minimum acceptable value (Table 8‑1). For the Spectralis SD-OCT, quality scores range from 0 to 40, and a score less than 15 is considered poor. On the other hand, the Cirrus SD-OCT quality scores range from 0 to 10, and a score less than 6 is considered poor.
Motion artifact (Fig. 8‑3) presents in 0.2% of RNFL circular peripapillary B-scans taken with the Spectralis SD-OCT machine (Heidelberg Engineering, Heidelberg, Germany). 2 These artifacts result from subtle eye motions that occur during eye fixation, such as tremor, drifts, and microsaccades, or from larger movements, such as those due to head motion, heartbeat, respiration, or blinking. Eye-tracking technology can minimize errors from the former. Larger movements, however, can produce blurred, wavy images that move outside the boundary of the rectangular display box and preclude accurate segmentation.
Today, different OCT machines employ different methods of motion correction. Some machines have a second laser beam to detect eye movement. Other machines include the use of additional B-scans obtained during the examination as a framework for realigning individual B-scans. 9
So what can be done when there are wavy retinal layers or if the scan is cut off abnormally as a result of motion artifact? In the case of a tremor, an assistant can help to hold the patient’s head against the bar during image acquisition to minimize movement. Have the technician describe the importance of minimizing movement to the patient, so that he or she knows it will affect his or her diagnosis and management. Finally, it may be necessary to be patient and take multiple images, which can be difficult in a busy clinic. Having a well-trained technician can be one of the most important ways to minimize motion artifact.
In certain instances, the RNFL OCT fails to acquire data in an area of the scan (Fig. 8‑4). If this area of missing data is along the scan circle, the RNFL thickness measurements will be affected.
Cut edge (Fig. 8‑5) artifacts occur in 0.2% of RNFL circular peripapillary B-scans taken with the Spectralis SD-OCT machine (Heidelberg Engineering, Heidelberg, Germany). 2 A cut edge artifact refers to the abrupt truncation of the lateral edge of the RNFL. This may take place at one or both ends of the OCT scan and most likely results from patient movement.
8.2.2 Artifacts in Boundary Segmentation
The OCT instrument contains software that can identify specific layers of the retina, depending on the region and parameter of interest. In glaucoma patients, OCT is commonly used to measure and follow the thickness of the peripapillary RNFL over time. The OCT machine generates this measurement by first delineating the anterior and posterior borders of the RNFL and then calculating the distance between the two borders. These borders are marked by brightly colored (usually red, blue, or purple) lines. The anterior border represents the interface between the vitreous and the internal limiting membrane (ILM) of the retina, and the posterior border represents the interface between the ganglion cell nerve fibers and the ganglion cell bodies. OCT takes advantage of the higher reflectivity of the RNFL in order to locate these interfaces and distinguish the RNFL from its surrounding layers. Therefore, anything that obscures the boundary between the RNFL and the vitreous anteriorly or the RNFL and the ganglion cell layer posteriorly will interfere with accurate segmentation.
Anterior RNFL misidentification (Fig. 8‑6) presents in 3.2% of peripapillary RNFL scans taken with the Spectralis SD-OCT machine (Heidelberg Engineering, Heidelberg, Germany). 2 These occur when the segmentation algorithm fails to correctly identify the anterior border of the RNFL. On the printout, the top red line drifts anteriorly (upward) into the vitreous, resulting in falsely increased or decreased thickness measurements, respectively.
A common cause of this type of artifact is the presence of an epiretinal membrane (ERM) or posterior vitreous detachment (PVD). In one study, ERMs were found in 47.3% (26 out of 55) of peripapillary scans with artifacts and tended to increase measurements of RNFL thickness. 10 If there is suspicion for an ERM, a basic clinical examination with or without a macular OCT would confirm the presence of this pathology. Similarly, certain features of the vitreous can confuse anterior segmentation. Often in an area of vitreomacular traction (VMT), the hyaloid face appears as a hyperreflective line very near to the anterior boundary of the RNFL. In this case, the OCT may inaccurately segment the hyaloid face as the anterior RNFL. PVDs can induce other forms of artifact unrelated to segmentation errors, which is discussed further in the section Artifacts Due to Ocular Pathology Unrelated to Glaucoma.
Posterior RNFL misidentification (Fig. 8‑7) presents in 7.7% of peripapillary RNFL scans taken with the Spectralis SD-OCT machine (Heidelberg Engineering, Heidelberg, Germany). 2 These occur when the segmentation algorithm fails to correctly identify the posterior border of the RNFL, and this can lead to artifactually thick or thin RNFL measurements.
Glaucoma itself might contribute to this type of artifact, because glaucoma is associated with loss of RNFL reflectivity. Since OCT technology relies on differences in tissue reflectivity in order to construct the layers of the retina, loss of RNFL reflectivity means that the nerve fiber layer becomes less distinguishable from its surrounding layers. This change especially affects the interface between the RNFL and the less reflective deeper retinal layers. Theoretically, this would make glaucoma patients more prone to posterior misidentification artifact.
Incomplete segmentation artifact (Fig. 8‑8) presents in 0.6% of peripapillary RNFL scans taken with the Spectralis SD-OCT machine (Heidelberg Engineering, Heidelberg, Germany). 2 The anterior and posterior segmentation lines should be present from the left-most side of the image to the right-most side of the image. However, if the software cannot trace part of the scan, this is referred to as incomplete segmentation. This should be an easy artifact for the clinician to spot, as long as he or she looks at the scan.
Physiologically, the RNFL thickness can never be 0 μm. In fact, even in the most advanced glaucoma patient, RNFL thickness values cannot be less than around 50 μm, because the RNFL is partly comprised of nonneuronal tissue such as blood vessels and glial tissue. Despite this, the OCT machine will often report thickness values as low as 0 μm, which is below the floor of around 50 μm (Fig. 8‑7 and Fig. 8‑10). This is an easy error to spot because the clinician can just look at the line graph and note that any points that drop to 0 μm are clearly artifacts. That being said, the RNFL may be very thin in that area, so physicians should refer back to the scan to understand why it was segmented incorrectly.
8.2.3 Artifacts due to Ocular Pathology Unrelated to Glaucoma
To be an expert in glaucoma, the clinician should also be able to identify nonglaucomatous pathology and know how it can affect measurements. To this end, abnormalities of the globe, retina, vitreous, and so on, can dramatically affect measurements and potentially render the scan unusable.
In a study at the Duke Eye Center in 2015 examining 277 patients in the glaucoma service, the most common ocular pathology causing artifacts in RNFL OCT scans included an ERM and VMT. 10 In a larger study by Massachusetts Eye and Ear in 2015 examining a total of 2,313 eyes, 14.4% of patients had artifact caused by PVD. Although less frequent, other ocular pathologies leading to artifacts included peripapillary atrophy (PPA), staphyloma, and myelinated nerve fiber layer (MNFL). 2
Vitreomacular Traction/Posterior Vitreous Detachment
With normal aging, the vitreous may detach from the ILM. Initially, there may be a stage of VMT or partial PVD, which can progress to a complete PVD. The precise effects of PVD on RNFL artifact in the literature are varied. However, the rates of PVD-associated artifact is estimated to be approximately 14%. As the posterior hyaloid pulls off the ILM, some patients show focal artifactual thickening or “tenting” of the RNFL in the region of the VMT due to the physical pull from the vitreous (Fig. 8‑9a, b). 10 Subsequently, as the traction is released and as the PVD progresses from a partial to complete PVD, there can be relative RNFL thinning as the anterior RNFL surface may drop back more posteriorly. This is an important confounding factor for clinicians to be aware of in order to avoid incorrectly interpreting the decrease in RNFL thickness following the progression from partial PVD to complete PVD as a sign of glaucomatous changes to the RNFL. 11 It is important to know that the vitreous can create artifacts that can cause both falsely thickened and thinned RNFL. It is also important to know that this can change over time as the clinician follows a patient with sequential scans depending on the status of the vitreous traction. Another source of error during segmentation is if the detached or partially detached hyaloid face is misidentified as the anterior boundary of the RNFL (Fig. 8‑9b). 10 , 12 This is discussed further in the section Artifacts in Boundary Segmentation.
As an example, when monitoring a patient over time, if a focal decrease in RNFL thickness occurs at the same time as a newly diagnosed PVD, it would be prudent to carefully evaluate the scans and consider other diagnostic metrics prior to changing the patient’s therapy, as this could be due to artifact.