ORIGINAL STUDY

The Prevalence of Optical Coherence Tomography Artifacts

in High Myopia and its Influence on Glaucoma Diagnosis
Linda Yi-Chieh Poon, MD,*† Chi-Hsun Wang, MSc,*
Pei-Wen Lin, MD,*† and Pei-Chang Wu, MD, PhD*†

Précis: Optical coherence tomography (OCT) artifacts occur much

more frequently in highly myopic eyes compared with non-highly
G laucoma is a progressive optic neuropathy with char-
acteristic atrophy of the optic nerve resulting from the
loss of retinal ganglion cells and their axons. Optical
myopic eyes. A longer axial length is predictive of having
OCT artifacts. coherence tomography (OCT) facilitates structural analysis
of the peripapillary retina, optic disc, and macula, and has
Purpose: To investigate the types and prevalence of artifacts on become an indispensable tool in the diagnosis of glaucoma
OCT scans in patients with and without high myopia. over the past decades.1
Materials and Methods: Patients were divided into 4 groups based However, these machines are not without limitations.
on whether they had glaucoma and/or high myopia. All peripapil- OCT artifacts can be present in 19.9–46.3%2,3 of RNFL
lary retinal nerve fiber layer (RNFL) scan images were individually scans, which can lead to erroneous measurements of the
inspected for the presence of artifacts. retinal nerve fiber layer (RNFL) thickness, misleading
clinicians to an incorrect glaucoma diagnosis. Recognition
Results: Two hundred twenty-six patients were enrolled. The
prevalence of OCT artifacts was 18.6% in non-high myopes and of imaging artifacts is thus critical for accurate inter-
51.9% in high myopes (P < 0.001). Outer RNFL border mis- pretation of the OCT examination.
identification was the most common type of artifact for non-high Ocular pathologic features are possible sources of OCT
myopes, whereas retinal pathology-related artifact was the most artifacts.2 Patients with high myopia are known to have a
common in high myopes. Univariable regression analysis showed higher prevalence of peripapillary retinal changes including
that a longer axial length [odds ratio (OR) 1.815, P < 0.001], a peripapillary atrophy (PPA) and peripapillary retinoschisis,4,5
higher pattern standard deviation (OR 1.194, P < 0.001), and which could lead to a higher frequency of OCT artifacts
thinner RNFL (OR 0.947, P < 0.001) were predictive factors for occurring in this group of patients.
the presence of OCT artifacts. The diagnostic capability of global
Past studies that investigated OCT artifacts in peri-
RNFL thickness before and after manual correction of segmen-
tation errors did not differ for both non-high myopes [area under papillary RNFL scans2,3 are limited by a lack of clearly
the receiver operating curve 0.915–0.913 (P = 0.955)] and high defined refractive status and likely did not include much of
myopes [area under the receiver operating curve 0.906–0.917 highly myopic patients. There was one study6 that eval-
(P = 0.806)]. uated the influence of myopia on OCT segmentation errors
and glaucoma diagnosis which found a segmentation
Conclusion: The prevalence of OCT artifacts was the highest in
artifact rate of 44% on 3.45 mm pericapillary RNFL
patients with both high myopia and glaucoma. The most common
type of OCT artifact is different for non-high myopes and high scans. Glaucoma diagnostic capability of RNFL thickness
myopes. Physicians need to be aware of a higher likelihood of OCT improved from 0.827 to 0.888 [area under the receiver
artifacts, particularly in those with a longer axial length, worse operating curve AUC)] after manual correction of seg-
visual field, and thinner RNFL thickness. mentation artifacts; however, the study only included
patients with myopia of < −3.0 Diopter (D).6 The risk of
Key Words: artifacts, glaucoma, high myopia, optical coherence glaucoma is, however, much higher in patients with high
tomography, retinal nerve fiber layer myopia compare with those with mild or moderate
(J Glaucoma 2023;32:725–733) myopia.7,8 Therefore, it would be important to know how
reliably OCT can help clinicians diagnose glaucoma in a
group of highly myopic patients.
Received for publication March 15, 2023; accepted July 10, 2023. The purpose of this study is to evaluate the prevalence
From the *Department of Ophthalmology, Kaohsiung Chang Gung rate of different types of imaging artifacts on OCT peri-
Memorial Hospital, Kaohsiung; and †School of Medicine, Chang papillary RNFL scans in normal patients and in patients
Gung University, Taoyuan, Taiwan, Republic of China. with either glaucoma, high myopia, or both. The study also
Contents of this study were previously presented at the 9th World
Glaucoma E-Congress, June 30–July 3, 2021. aims to further investigate factors associated with the pres-
This study was supported by a grant for Kaohsiung Chang Gung ence of artifacts and assess how glaucoma diagnosis may be
Memorial Hospital (CMRPG8H0861). The funding organization had influenced by OCT artifacts.
no role in the design or conduct of this research.
Disclosure: The authors declare no conflict of interest.
Reprints: Pei-Chang Wu, MD, PhD, Department of Ophthalmology, MATERIALS AND METHODS
Kaohsiung Chang Gung Memorial Hospital, 123 Dapi Road, Niaosong
District, Kaohsiung 83301, Taiwan (R.O.C.) (e-mail: [email protected]). This is a prospective study that consecutively enrolled
Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, patients who visited the glaucoma service of Kaohsiung
Inc. This is an open access article distributed under the Creative Chang Gung Memorial Hospital between the period of Jan.
Commons Attribution License 4.0 (CCBY), which permits unre-
stricted use, distribution, and reproduction in any medium, provided
2019 to Jan. 2021. The study protocol was approved by the
the original work is properly cited. Chang Gung Medical Foundation Institutional Review
DOI: 10.1097/IJG.0000000000002268 Board (IRB number 201800926B0) and was conducted in

J Glaucoma  Volume 32, Number 9, September 2023 www.glaucomajournal.com | 725

Poon et al J Glaucoma  Volume 32, Number 9, September 2023

accordance with the tenets of the Declaration of Helsinki. maculopathy, vascular occlusive retinopathy, proliferative
Informed consent was obtained from all of the patients. diabetic retinopathy, or non-glaucomatous optic neuro-
All of the patients underwent a comprehensive eye pathy, which could result in VF loss not attributable to
examination that included history, visual acuity, automated glaucoma. If both eyes of a patient were eligible for inclu-
refraction, intraocular pressure, axial length measurement sion in the study, one eye was selected randomly using an
(IOLMaster, Carl Zeiss Meditec), visual field (VF) testing online randomization tool (https://www.randomlists.com).
(Swedish Interactive Threshold Algorithm standard 30-2 test; All of the patients’ peripapillary RNFL scan images were
Humphrey VF Analyzer 750i, Carl Zeiss Meditec Inc.), slit lamp reviewed by 2 experienced examiners (L.Y.-C.P., C.-H.W.) to
biomicroscopy, dilated fundus examination, and OCT peri- assess for the presence of image artifacts. The prevalence and
papillary RNFL scan (Spectralis OCT, Heidelberg Engineering). types of artifacts were recorded for analysis. With mod-
The peripapillary RNFL scan was obtained using a circular scan ifications based on the classifications used by past studies,2,3,9
with a diameter of 12 degrees centered on the optic disc. we divided OCT image artifacts on RNFL scans into 3 cate-
Patients were divided into 4 groups based on their refractive gories: software algorithm failure, retinal pathology-related
status and whether they had glaucoma or not: non-high myope artifacts, and image acquisition artifacts. Under the category of
controls, non-high myope glaucoma patients, highly myopic software algorithm failure, 3 types of OCT artifacts were
patients, and highly myopic patients with glaucoma. Patients identified, and included inner RNFL border misidentification,
were classified as having high myopia if they had a spherical outer RNFL border misidentification (Fig. 1), and complete
equivalent (SE) refraction of ≤ – 6.0 Diopter (D) or an axial segmentation failure. The category of retinal pathology-related
length of ≥ 26 mm if the patient had previously undergone artifacts included artifacts due to the presence of PPA, peri-
cataract surgery or laser refractive surgery. Diagnosis of glau- papillary retinoschisis, and posterior vitreous detachment and/
coma was based on a glaucomatous disc appearance with a or epiretinal membrane (PVD/ERM) (Fig. 2). Truncated
typical glaucomatous VF defect (paracentral defect, nasal step, image, low signal, and motion artifacts fall under the category
arcuate scotoma, generalized depression, or altitudinal defect) of image acquisition artifacts (Fig. 3). Definitions of each type
that is repeatable and compatible with the disc appearance. A of artifact are summarized in Table 1. Only segmentation
glaucomatous VF defect was defined as a cluster of 3 or more errors that occurred in the absence of retinal pathology or
contiguous test locations on the same side of the horizontal acquisition error are considered segmentation algorithm fail-
meridian in the pattern standard deviation plot that was ure. Segmentation errors that occurred due to the presence of
depressed at the P < 0.05 level with at least 1 at the P < 0.01 level. retinal pathology or acquisition error are counted as only
Only reliable VFs, as defined by a fixation loss of <20%, false- the retinal pathology or acquisition error itself. However,
positive rate of <15%, and false-negative rate of <15%, were more than 1 type of OCT artifact may be present on a single
included in the analysis. Non-high myope controls were those RNFL scan such as having both motion artifact and trunca-
with normal VF, SE of > −6.0 D and <+3.0D, and axial length tion artifact or having both PVD/ERM and PPA.
of <26 mm. In those with OCT artifacts resulting in segmentation
Eyes were excluded from analysis if they had a history errors, manual correction of the segmentation error was
of traumatic eye injury, history of vitreoretinal surgery, performed to assess the influence of OCT artifacts on RNFL

FIGURE 1. Examples of artifacts categorized under software algorithm failure. A, Outer retinal nerve fiber layer (RNFL) border
misidentification (thick red arrow). B, Inner RNFL border misidentification (thick red arrow).

726 | www.glaucomajournal.com Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc.
J Glaucoma  Volume 32, Number 9, September 2023 Types and Prevalence of Artifacts on OCT Scans

FIGURE 2. Examples of artifacts categorized under retinal pathology pathology-related artifacts. A, Peripapillary atrophy (in between thin
red arrows). B, Posterior vitreous detachment/epiretinal membrane (thick red arrow). C, Peripapillary retinoschisis (in between thin red arrows).

thickness measurements and its glaucoma diagnostic RESULTS

capability. Two hundred twenty-six scans of 226 patients were
Statistical analyses were performed using the SPSS enrolled in the study, with the patients having an average
Statistics software, version 26 (IBM Corp.). For continuous age of 49.0 ± 13.5 years. The study included 61 non-high
variables that are normally distributed, comparison across myope control, 57 non-high myope glaucoma patients, 62
groups was performed using the one-way ANOVA, and are high myopes without glaucoma, and 46 high myopes with
expressed as mean ± SD. Parameters that are not normally glaucoma. The baseline demographics of the study pop-
distributed are expressed as median (interquartile range) and ulation are summarized in Table 2. Eighteen (8.0%) eyes
are compared using the Kruskal-Wallis test. Comparisons of were pseudophakic and 17 (7.5%) eyes had previously
the artifact prevalence rate were performed using the χ2 test. undergone laser refractive surgery. The spherical equivalent
Logistic regression analysis was used to determine factors of the non-highly myopic patients was −2.56 ± 2.16 D, and
associated with the occurrence of OCT artifacts. The var- −9.71 ± 3.14 D in the highly myopic patients.
iance inflation factor was used to assess for multicollinearity For the entire population, OCT artifact was present in
between variables used for the regression analysis model. 78 out of 226 RNFL scans (34.5%). Out of the 78 scans with
The glaucoma diagnostic capability of OCT RNFL thick- OCT artifacts, 71.8% (56 out of 78) of patients were highly
ness was assessed by using the AUC analysis. Comparison myopic, whereas 28.2% were non-highly myopic (P < 0.001).
of AUC value between non-high myopes and high myopes is The most common category of OCT artifacts for the entire
performed using MedCalc Statistical software version population was retinal pathology-related artifact, which was
20.012 (MedCalc Software Ltd). A P value of <0.05 was present in 39 out of 226 (17.3%) scans. Segmentation algo-
considered statistically significant. rithm failure occurred in 15.9% (36 out of 226) of scans.

Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. www.glaucomajournal.com | 727
Poon et al J Glaucoma  Volume 32, Number 9, September 2023

FIGURE 3. Examples of artifacts categorized under image acquisition artifacts. A, Truncated image (thick red arrow). B, Low signal. C,
Motion artifact.

Image acquisition artifact was the least common and 118 scans, 11.0%), and occurred predominantly in glaucoma
occurred in 13 out of 226 (5.8%) scans. patients (Table 3). Motion artifacts occurred more fre-
When divided into the non-high myope and high quently in high myopes compared with non-high myopes
myope groups, the prevalence rate of OCT artifact was [10 out of 108 scans (9.3%) vs. 1 out of 118 scans (0.8%),
found to be 18.6% in non-high myopes and 51.9% in high respectively, P = 0.004]. Relatively low and similar preva-
myopes (P < 0.001). Table 3 shows the prevalence rate for lence rates between high myopes and non-high myopes were
all of the artifacts analyzed in this study. The most common found for PVD/ERM, peripapillary retinoschisis, truncated
category of OCT artifact is different for non-high myopes image, and low signal artifacts (Table 3). Overall, OCT
and high myopes. In highly myopic patients, the most artifact is relatively uncommon in a non-highly myopic
common category of OCT artifact is retinal pathology- normal population (4.9%, Fig. 4); however, the prevalence
related artifacts with PPA as the most frequent cause of of OCT artifacts becomes much higher with a patient having
OCT artifacts (25 out of 108 scans, 23.1%). In the non-high glaucoma (33.3%), high myopia (43.5%), or both (63.0%)
myopes, the most common category of OCT artifact is (Fig. 4).
software algorithm failure with outer RNFL border mis- To identify predictive factors associated with the
identification as the most frequent type of artifact (13 out of presence of OCT artifacts on RNFL scans, logistic

728 | www.glaucomajournal.com Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc.
J Glaucoma  Volume 32, Number 9, September 2023 Types and Prevalence of Artifacts on OCT Scans

TABLE 1. Definition of the Optical Coherence Tomography Artifacts Examined in this Study
Type of artifacts Definition
Software algorithm failure
Inner RNFL border misidentification Incorrect segmentation of the inner RNFL border
Outer RNFL border misidentification Incorrect segmentation of the outer RNFL border
Complete segmentation failure No segmentation line along the RNFL borders
Retinal pathology-related artifacts
PPA Incorrect segmentation of the RNFL borders due to the scan circle overlapping
peripapillary atrophy
PVD/ERM Misidentification of the inner RNFL border due to the presence of posterior vitreous
detachment or epiretinal membrane
Peripapillary retinoschisis Incorrect segmentation of the RNFL borders due to the presence of retinoschisis
Image acquisition artifacts
Truncated image Vertically displaced image resulting in part of the inner or outer retina not within the
acquisition window of the scan
Low signal Retinal layers indistinguishable from the background noise along a portion or the
entire length of the image
Motion artifact Wavy or undulating appearance of the scan image due to movement during the scan
resulting in indiscernible retinal layers
PPA indicates peripapillary atrophy; PVD/ERM, posterior vitreous detachment/epiretinal membrane; RNFL, retinal nerve fiber layer.

regression analysis was performed. The results of the in non-high myopes remained relatively unchanged from an
regression analysis are summarized in Table 4. Univariable AUC value of 0.915–0.913 (P = 0.955). In high myopes, the
regression analysis found that axial length, VF pattern AUC increased slightly from 0.906 to 0.917, but the differ-
standard deviation, global RNFL thickness, and RNFL ence was also not statistically significant (P = 0.806).
thickness in the superotemporal and inferotemporal sectors
were significant predictive factors associated with having
artifacts on OCT RNFL scans. Age, IOP, and scan quality DISCUSSION
score were not associated with the presence of OCT arti- Although several OCT parameters have been proposed
facts. In the multivariable model, only axial length [odds and investigated for glaucoma diagnosis,10 RNFL thickness
ratio 1.837, 95% CI (1.430–2.360), P < 0.001] remained remains one of the most commonly used and reported
significantly associated with the presence of OCT artifact. parameters in OCT for monitoring and detection of glau-
In OCT scans that had artifacts (78 out of 226 scans), coma;1 however, OCT-based diagnosis for glaucoma using
segmentation errors were manually corrected to assess the RNFL thickness relies considerably on the accuracy of the
influence of OCT artifacts on RNFL measurements and RNFL measurements, which can be significantly influenced
their diagnostic capability. After inspection and correction by the presence of OCT artifacts.6,11 Although it was often
of the segmentation errors, it was found that 7 out of 78 assumed in the past studies that high myopia can be a cause
scans (8.9%) could not be segmented, all of which contained of OCT artifacts on RNFL scans,2,3,6 the past studies have
motion artifacts and precluded the generation of RNFL not clearly demonstrated this relationship. In this study, we
thickness data and were not included in the calculation of comprehensively analyzed the frequency of artifacts on
AUC value. Despite manual correction of the segmentation RNFL scans of the Spectralis OCT in a population that
errors, the diagnostic capability of global RNFL thickness included non-high myope controls, non-high myope

TABLE 2. Baseline Demographics of the Study Population

High myopia without High myopia with
Normal (n = 61) Glaucoma (n = 57) glaucoma (n = 62) glaucoma (n = 46) P
Age* (y) 48.0 (33.0–56.5) 61.0 (54.0–69.0) 43 (32.0–49.0) 53 (42.8–57.0) < 0.001
IOP* (mmHg) 15.7 ± 3.2 13.3 ± 3.0 15.6 ± 3.0 14.2 ± 3.5 < 0.001
SE* (D) −2.45 ± 2.12 −2.69 ± 2.21 −9.67 ± 3.10 −9.80 ± 3.26 < 0.001
AL* (mm) 24.93 ± 1.11 25.19 ± 1.21 27.53 ± 1.28 27.74 ± 1.42 < 0.001
VF MD* (dB) −0.29 (−1.64 to 0.82) −2.93 (−9.29 to −0.92) −1.44 (−2.56 to −0.10) −6.04 (−13.87 to −2.55) < 0.001
VF PSD* (dB) 1.82 (1.54–2.13) 4.59 (2.39–10.36) 2.14 (1.61–2.94) 7.07 (3.33–13.27) < 0.001
RNFLT- global* 97.51 ± 10.32 72.63 ± 14.92 87.07 ± 13.02 63.43 ± 13.12 < 0.001
(microns)
Prior surgery
Cataract surgery, 0 6 (10.5) 2 (3.2) 10 (21.7) < 0.001
n (%)
Laser refractive 4 (6.6) 0 4 (6.5) 9 (19.6) < 0.001
surgery, n (%)
*Values are presented as mean ± SD or as median (interquartile range).
AL indicates axial length; D, diopter; dB, decibel; IOP, intraocular pressure; RNFLT, retinal nerve fiber layer thickness; SE, spherical equivalent; VF MD,
visual field mean deviation; VF PSD, visual field pattern standard deviation.

Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. www.glaucomajournal.com | 729
Poon et al J Glaucoma  Volume 32, Number 9, September 2023

TABLE 3. Frequency of Each Type of Artifact in the Study Population

Normal Non-high myopes with High myopes without High myopes with
(n = 61) glaucoma (n = 57) glaucoma (n = 62) glaucoma (n = 46) P
Software algorithm failure
Outer RNFL border 2 (3.3) 11 (19.3) 10 (16.1) 11 (23.9) 0.016
misidentification, n (%)
Inner RNFL border 0 0 1 (1.6) 2 (4.3) 0.185
misidentification, n (%)
No segmentation, n (%) 0 0 0 0 1.000
Retinal pathology-related artifact
PPA, n (%) 0 4 (7) 15 (24.2) 10 (21.7) < 0.001
PVD/ERM, n (%) 1 (1.6) 3 (5.3) 2 (3.2) 4 (8.7) 0.333
Peripapillary retinoschisis, 0 1 (1.8) 3 (4.8) 2 (4.3) 0.321
n (%)
Image acquisition artifact
Truncated image, n (%) 0 2 (3.5) 1 (1.6) 0 0.314
Low signal, n (%) 0 1 (1.8) 1 (1.6) 0 0.610
Motion artifact, n (%) 0 1 (1.8) 5 (8.1) 5 (10.9) 0.026
PPA indicates peripapillary atrophy; PVD/ERM, posterior vitreous detachment/epiretinal membrane; RNFL, retinal nerve fiber layer.

glaucoma patients, highly myopic patients, and highly which can make the border between the RNFL and the
myopic patients with glaucoma. Our study showed that RGC layer less distinguishable, resulting in outer RNFL
increased axial length is a predictive factor for the presence border misidentification. The association of glaucoma
of OCT artifacts and we also found that OCT artifacts were severity with the presence of OCT artifacts is also supported
present in over 50% of highly myopic patients, with the by the finding in previous studies.3,17 The loss of RNFL
prevalence increasing to over 60% if the highly myopic reflectivity in glaucomatous disease may in part be explained
patient also has glaucoma. by the changes in cytoskeletal components of the RGC
In our study, we found that outer RNFL border mis- axons and their supporting cells with the progression of the
identification was the most common type of OCT artifact in disease.18–20
non-high myopes and occurred predominantly in glaucoma In high myopes, outer RNFL border misidentification
patients (Table 3). On OCT, the adjacent layers of the was also common and occurred in 21 out of 108 (19.4%)
retina are visualized as different scattering intensities or scans of highly myopic patients. Although there have been
reflectances.12 Segmentation algorithms are designed to no prior studies that investigated the differences in RNFL
detect and delineate these intraretinal boundaries; however, reflectivity in myopic eyes, there was one study that dem-
reduced contrast between the adjacent retinal layers can onstrated changes in the cytoskeletal architecture and thin-
limit the ability of the algorithm to accurately detect the ning of the RGC axons in chicken eyes with induced
boundaries of interest.13 The RNFL can be observed as a myopia,21 and by inference, could also have an impact on
highly reflective layer of the retina with the internal limiting the reflectivity of the RNFL.18 In addition, myopia is
membrane as its inner border and the boundary between the known to cause axial elongation and deformation of the
RNFL and the more hyporeflective retinal ganglion cell eyeball, which has previously been well demonstrated on
(RGC) layer as its outer border. However, the reflectivity of
the RNFL decreases with the progression of glaucoma,14–16
TABLE 4. Univariable Regression Analysis for the Predictors of the
Presence of any Optical Coherence Tomography Artifacts
Variables Coefficient OR (95% CI) P
Age 0.017 1.017 (0.996–1.038) 0.115
IOP −0.082 0.921 (0.845–1.005) 0.065
Axial length 0.596 1.815 (1.479–2.227) < 0.001
VF PSD 0.177 1.194 (1.114–1.279) < 0.001
Scan Quality score (Q) −0.055 0.946 (0.876–1.023) 0.164
Retinal nerve fiber layer thickness
Global −0.045 0.947 (0.929–0.965) < 0.001
Superonasal sector 0.001 1.000 (0.996–1.004) 0.985
Superotemporal −0.013 0.987 (0.979–0.996) 0.003
sector
Temporal sector −0.005 0.995 (0.985–1.005) 0.347
Inferotemporal −0.018 0.982 (0.975–0.989) < 0.001
FIGURE 4. The prevalence of optical coherence tomography sector
(OCT) artifact in non-high myope controls, non-high myopes Inferonasal sector −0.002 0.998 (0.990–1.005) 0.543
with glaucoma, high myopes without glaucoma, and high Nasal sector 0.001 1.001 (0.998–1.005) 0.398
myopes with glaucoma *represents statistically significant
difference compared to non-high myope controls. **represent IOP indicates intraocular pressure; OR, odds ratio; VF PSD, visual field
statistically significant difference compared to non-high myopes pattern standard deviation.
with glaucoma).

730 | www.glaucomajournal.com Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc.
J Glaucoma  Volume 32, Number 9, September 2023 Types and Prevalence of Artifacts on OCT Scans

3-dimensional magnetic resonance imaging.22,23 Extreme regression analysis, we also found that increased glaucoma
elongation of the globe may result in an inadequately cap- severity and thinner RNFL thickness are associated with the
tured image due to exceeding the diopter compensation of presence of OCT artifacts, which is in agreement with pre-
the device or from insufficiently reflected image signals.24 vious studies.3,6,17 Although age was previously reported by
Distortion of the posterior pole may also result in a highly Li et al17 to be associated with the occurrence of OCT
curved peripapillary retinal image on OCT,6 and could artifacts, we did not find age to be a predictive factor in
hinder the segmentation algorithm from identifying the either the univariate or multivariate analysis. Older patients
outer border of the RNFL accurately. Furthermore, irreg- have a higher likelihood of having cataracts, dry eyes, and
ularity in the contour of the peripapillary region can lead to smaller pupils which may increase the difficulties of
a variable focusing effect and irregular light scattering obtaining a clear image on OCT; however, our study pop-
resulting in images with insufficient clarity and subsequently ulation is a relatively young population with a mean age of
leading to a higher likelihood of segmentation errors. 49 years, which is much younger than the study population
Retinal pathology-related artifact, which includes OCT in Li et al17’s study with a mean age of 72, and could par-
artifacts that occurred due to the presence of PPA, PVD/ tially explain why age was not found to be a contributory
ERM, and/or peripapillary retinoschisis, was the most factor for the presence of OCT artifacts in this study.
common category of artifact in our highly myopic patients. Although not necessarily resulting in segmentation
In contrast to the findings by Asrani et al2 who found that errors, temporal displacement of the RNFL, torsional
ERM was the most common ocular pathology resulting in changes in the RNFL topography, tilting of the optic disc,
OCT artifact, we found that PPA was the most frequent and difficulties in determining the center of a myopic disc
cause of retinal pathology-related artifacts (Table 3). Our with incorrect centration of the measurement circle can
study population, however, included mostly myopic result in a false classification by the machine of being out-
patients, whereas the refractive status of the patient pop- side of normal limits (Red disease), which can also confound
ulation by Asrani et al2 was not specified. PPA is very clinical interpretation.30 Furthermore, a magnification effect
common in highly myopic patients and can be present in as associated with myopia may result in a thinner measured
many as 79%–100% of highly myopic eyes.5,25 We regarded RNFL thickness,31 with some authors advocating adjust-
PPA as a cause of OCT artifact only if it was overlapped by ment for ocular magnification effect in patients with myopia
the OCT scan circle and found a significantly higher prev- when making glaucoma diagnosis.31
alence of PPA-associated artifact in high myopes (23.1%) Although studies in the past have demonstrated a good to
compared with non-high myopes (4%) (P < 0.001). This excellent glaucoma diagnostic capability of OCT RNFL
suggests that not only is PPA more common in high myopes thickness in high myopia,32–36 not all studies agreed on whether
but also PPAs larger than the 12 degrees (3.4–3.5 mm dia- this differs between high myopes or non-high myopes.
meter) scan circle are more common in high myopes. On Although some studies show that the diagnostic capability of
OCT, PPA can appear as the loss or disruption of the RPE OCT was significantly worse in highly myopic subjects,32 other
layer, RNFL plaques, retinal thinning, and abnormal studies show that there was no difference between highly
retinal sloping.26 Such loss of the normal retinal architecture myopic versus non-highly myopic subjects.33,34 Our study
or layers may subsequently result in the inability of found that despite a prevalence of 51.9% in high myopes and
the segmentation algorithm to accurately detect the 18.6% in non-high myopes for OCT artifacts, the AUCs for
RNFL borders.3,17,27 Although peripapillary retinoschisis is global RNFL thickness was over 0.9 in both groups of patients,
also considered a characteristic finding in highly myopic suggesting excellent glaucoma diagnostic capability for this
patients,4,5 we found the occurrence of retinoschisis to be OCT parameter. The high AUC value for the RNFL thickness
low and affected 5 out of 108 scans (4.6%) in highly myopic parameter in the highly myopic patients in this study could be
patients. the result of the analysis being performed between highly
Image acquisition artifacts, which included truncated myopic patients with glaucoma to highly myopic normal
image, low signal, and motion artifacts, are considered patients instead of being compared with normal patients
technician-dependent,2,3,27 and could be decreased with without high myopia and suggests the importance of using a
adequate technician training and experience.27 In this study, myopic normative database when assessing this group of
we found a relatively low prevalence of <1% for truncated patients. Biswas et al37 have also previously demonstrated the
image and low signal artifacts in both non-high myope and importance of incorporating a myopic normative database for
high myopes, which suggests that our technicians have been RNFL thickness analysis in highly myopic patients by showing
adequately trained to avoid these types of technician- a significantly higher diagnostic performance and specificity for
dependent artifacts; however, a significantly higher preva- glaucoma detection in this group of patients when compared
lence of 9.3% for motion artifact, compared with a 0.8% with the machine’s built-in normative database.
prevalence in non-high myopes, was still noted in high We also found that the AUC values did not differ
myopes (P = 0.004) and could reflect the difficulties in before and after the manual correction of segmentation
reaching an adequate focal plane, needing frequent adjust- errors [non-high myopes: AUC 0.915–0.913 (P = 0.955);
ments to obtain an OCT image in high myopes as their high myopes: AUC 0.906 to 0.917 (P = 0.806)]. Although it
eyeballs are more elongated than a typical normal eye.28,29 is important to correct ambiguous or erroneous segmenta-
Although often assumed to be associated with the tion and modify corresponding thickness values before
occurrence of OCT artifacts, none of the past studies have making clinical assessments, manual corrections are time-
clearly demonstrated the relationship between high myopia consuming and may not be feasible in busy clinical settings.
and OCT artifacts.2,6 The result of our multivariable The high AUCs before and after the correction of segmen-
regression analysis showed that the likelihood for the pres- tation artifacts suggest that in general, OCT RNFL thick-
ence of OCT artifact increased by 1.8 times for every 1 mm ness remains a useful tool for the detection of glaucoma
increase in axial length, which verifies myopia as a risk even without the manual correction of segmentation errors.
factor for having OCT artifacts. In our univariable However, it should be acknowledged that during manual

Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. www.glaucomajournal.com | 731
Poon et al J Glaucoma  Volume 32, Number 9, September 2023

correction of segmentation errors, it was found that 7 out of Our study has some limitations. Firstly, OCT machines
78 scans with OCT artifacts could not be re-segmented and from different manufacturers have their own segmentation
were not included in the AUC analysis for RNFL thickness, algorithm and may use different scanning protocols to
and the high AUC values shown in this study would thus generate RNFL thickness.50 Therefore the results presented
need to be interpreted with caution. Motion artifact was in this study may not be generalizable to OCT images
found to be the cause for all of the scans that could not be obtained from machines other than the one used in this
re-segmented. Therefore, it is important to repeat the RNFL study (Spectralis OCT, Heidelberg Engineering). Secondly,
scan when a motion artifact is present, as segmentation is patients with glaucoma in our study were, in general, older
often not feasible when such images are obtained. than those without (Table 2), and as older age has pre-
In addition, the high AUC for RNFL thickness in the viously been found to be associated with the presence of
high myopes shown in this study could have resulted from the artifacts,11,17 this could potentially contribute to a higher
lower-than-expected prevalence of PPA affecting the RNFL prevalence of OCT artifacts in glaucoma patients noted in
segmentation. PPA is very common in glaucoma and myopic this study. However, age was not identified as a contributing
patients, with its prevalence ranging from 68% to 95%38,39 factor to the presence of OCT artifacts in either the multi-
and 100%,5,40 respectively; however, in this study, we found a variate or univariate regression analysis in this study.
prevalence of PPA-related artifact of 7% in non-highly In conclusion, our study found a high prevalence of
myopic glaucoma patients, and 22%–24% in highly myopic over 50% for OCT artifacts in highly myopic patients, with
patients. Given the high prevalence of PPA in glaucoma and the prevalence increasing to over 60% if the highly myopic
myopic patients, this finding can be surprising, however; past patient also has glaucoma. The most common type of OCT
studies have found a mean maximum radial width for PPA artifact is outer RNFL border misidentification in non-high
(the distance from the optic disc border to the edge of PPA) of myopes and retinal pathology-related artifacts in high
0.24–0.37 mm,40–42 which suggests that in most patients with myopes. Physicians need to be aware of a higher likelihood
PPA, the PPA is not large enough to be intersected by the of OCT artifacts occurring particularly in those with a
3.45 mm diameter RNFL scan circle. However, adjustment of longer axial length, worse visual field, and thinner RNFL
the RNFL scan circle diameter to 4.1 and 4.7 mm using the thickness which can limit the usefulness of OCT in
Glaucoma Module Premium Edition software of the Spec- individual cases.
tralis OCT can be helpful in cases with larger PPAs that are
overlapped by the 3.45 mm scans.43 ACKNOWLEDGMENTS
Assessment of the neuroretinal rim thickness could be
an alternative to measuring peripapillary RNFL thickness The authors thank the Biostatistics Center of Kaohsiung
in myopic eyes. Previous studies have analyzed the Bruch’s Chang Gung Memorial Hospital for consultations for stat-
membrane opening—minimum rim width parameter and istical analysis.
reported comparable sensitivity and higher specificity for
glaucoma diagnosis in myopic eyes compared with the REFERENCES
RNFL thickness parameter.44,45 However, as the presence 1. Grewal DS, Tanna AP. Diagnosis of glaucoma and detection of
of gamma-PPA (peripapillary sclera without overlying glaucoma progression using spectral domain optical coherence
retinal layers, choroid, and Bruch’s membrane) is more tomography. Curr Opin Ophthalmol. 2013;24:150–161.
common in myopic eyes,46 and has previously been dem- 2. Asrani S, Essaid L, Alder BD, et al. Artifacts in spectral-
onstrated that around 30% of highly myopic eyes have domain optical coherence tomography measurements in glau-
indiscernible Bruch’s membrane opening in at least 1 coma. JAMA Ophthalmol. 2014;132:396–402.
meridian,47 it should be cautioned that the assessment of the 3. Liu Y, Simavli H, Que CJ, et al. Patient characteristics
neuroretinal rim may also be confounded by myopia. associated with artifacts in Spectralis optical coherence
tomography imaging of the retinal nerve fiber layer in
Macular ganglion cell analysis is another useful adjunct glaucoma. Am J Ophthalmol. 2015;159:565–576.
to peripapillary RNFL thickness measurement for evaluat- 4. Pan T, Su Y, Yuan ST, et al. Optic disc and peripapillary
ing patients with concomitant glaucoma and high myopia. changes by optic coherence tomography in high myopia. Int J
Previous investigations have shown that the macular gan- Ophthalmol. 2018;11:874–880.
glion cell complex analysis had comparable33,35,36 or 5. Shimada N, Ohno-Matsui K, Nishimuta A, et al. Peripapillary
better32,48 glaucoma diagnostic capability than the peri- changes detected by optical coherence tomography in eyes with
papillary RNFL measurements in highly myopic patients. high myopia. Ophthalmology. 2007;114:2070–2076.
Similar to the peripapillary RNFL thickness, the imple- 6. Suwan Y, Rettig S, Park SC, et al. Effects of circumpapillary
retinal nerve fiber layer segmentation error correction on
mentation of a myopic normative database has been
glaucoma diagnosis in myopic eyes. J Glaucoma. 2018;27:
suggested,36 which has been shown to improve the diag- 971–975.
nostic ability of the macular parameters.49 7. Xu L, Wang Y, Wang S, et al. High myopia and glaucoma
The high prevalence of OCT artifacts on RNFL scans susceptibility the Beijing Eye Study. Ophthalmology. 2007;114:
particularly in patients with high myopia should caution 216–220.
clinicians to carefully inspect RNFL segmentation and OCT 8. Marcus MW, de Vries MM, Junoy Montolio FG, et al. Myopia
images individually. Previously, Li et al,17 found that in all as a risk factor for open-angle glaucoma: a systematic review
patients with OCT artifacts, 23.9% of artifacts masked and meta-analysis. Ophthalmology. 2011;118:1989–1994.
actual RNFL thinning progression and led to a false-neg- 9. Bazvand F, Ghassemi F. Artifacts in macular optical coherence
ative diagnosis as no glaucoma progression, whereas 36.5% tomography. J Curr Ophthalmol. 2020;32:123–131.
10. Chen TC, Hoguet A, Junk AK, et al. Spectral-domain OCT:
of artifacts led to a false-positive interpretation of RNFL helping the clinician diagnose glaucoma: a report by the
thinning. These findings highlight the possibility of a false american academy of ophthalmology. Ophthalmology.
glaucoma diagnosis through an erroneous RNFL thickness 2018;125:1817–1827.
measurement and the importance of inspecting individual 11. Mansberger SL, Menda SA, Fortune BA, et al. Automated
OCT images for artifacts. segmentation errors when using optical coherence tomography

732 | www.glaucomajournal.com Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc.
J Glaucoma  Volume 32, Number 9, September 2023 Types and Prevalence of Artifacts on OCT Scans

to measure retinal nerve fiber layer thickness in glaucoma. Am J optical coherence tomography. Invest Ophthalmol Vis Sci.
Ophthalmol. 2017;174:1–8. 2010;51:4075–4083.
12. Chen TC. Spectral domain optical coherence tomography in 32. Shoji T, Nagaoka Y, Sato H, et al. Impact of high myopia on
glaucoma: qualitative and quantitative analysis of the optic the performance of SD-OCT parameters to detect glaucoma.
nerve head and retinal nerve fiber layer (an AOS thesis). Trans Graefes Arch Clin Exp Ophthalmol. 2012;250:1843–1849.
Am Ophthalmol Soc. 2009;107:254–281. 33. Kim NR, Lee ES, Seong GJ, et al. Comparing the ganglion cell
13. Kafieh R, Rabbani H, Kermani S. A review of algorithms for complex and retinal nerve fibre layer measurements by Fourier
segmentation of optical coherence tomography from retina. domain OCT to detect glaucoma in high myopia. Br J
J Med Signals Sens. 2013;3:45–60. Ophthalmol. 2011;95:1115–1121.
14. Gardiner SK, Demirel S, Reynaud J, et al. Changes in retinal 34. Choi YJ, Jeoung JW, Park KH, et al. Glaucoma detection
nerve fiber layer reflectance intensity as a predictor of ability of ganglion cell-inner plexiform layer thickness by
functional progression in glaucoma. Invest Ophthalmol Vis spectral-domain optical coherence tomography in high myopia.
Sci. 2016;57:1221–1227. Invest Ophthalmol Vis Sci. 2013;54:2296–2304.
15. Tan O, Liu L, You Q, et al. Focal loss analysis of nerve fiber 35. Akashi A, Kanamori A, Nakamura M, et al. The ability of
layer reflectance for glaucoma diagnosis. Transl Vis Sci macular parameters and circumpapillary retinal nerve fiber
Technol. 2021;10:9. layer by three SD-OCT instruments to diagnose highly myopic
16. Swanson WH, King BJ, Burns SA. Interpreting retinal nerve glaucoma. Invest Ophthalmol Vis Sci. 2013;54:6025–6032.
fiber layer reflectance defects based on presence of retinal nerve 36. Akashi A, Kanamori A, Ueda K, et al. The ability of SD-OCT
fiber bundles. Optom Vis Sci. 2021;98:531–541. to differentiate early glaucoma with high myopia from highly
17. Li A, Thompson AC, Asrani S. Impact of artifacts from optical myopic controls and nonhighly myopic controls. Invest
coherence tomography retinal nerve fiber layer and macula Ophthalmol Vis Sci. 2015;56:6573–6580.
scans on detection of glaucoma progression. Am J Ophthalmol. 37. Biswas S, Lin C, Leung CK. Evaluation of a myopic normative
2021;221:235–245. database for analysis of retinal nerve fiber layer thickness.
18. Huang XR, Knighton RW, Cavuoto LN. Microtubule JAMA Ophthalmol. 2016;134:1032–1039.
contribution to the reflectance of the retinal nerve fiber layer. 38. Jonas JB, Fernandez MC, Naumann GO. Glaucomatous
Invest Ophthalmol Vis Sci. 2006;47:5363–5367. parapapillary atrophy. Occurrence and correlations. Arch
19. Huang XR, Knighton RW, Spector YZ, et al. Cytoskeletal Ophthalmol. 1992;110:214–222.
alteration and change of retinal nerve fiber layer birefringence 39. Uchida H, Yamamoto T, Tomita G, et al. Peripapillary
in hypertensive retina. Curr Eye Res. 2017;42:936–947. atrophy in primary angle-closure glaucoma: a comparative
20. Richter-Landsberg C. The cytoskeleton in oligodendrocytes. study with primary open-angle glaucoma. Am J Ophthalmol.
Microtubule dynamics in health and disease. J Mol Neurosci. 1999;127:121–128.
2008;35:55–63. 40. Hill D, Heitmar R, Logan NS. Size and position of the optic
21. Swiatczak B, Feldkaemper M, Schraermeyer U, et al. Demyeli- disc crescent in a white European population with myopia.
nation and shrinkage of axons in the retinal nerve fiber layer in Ophthalmic Physiol Opt. 2022;42:1115–1123.
chickens developing deprivation myopia. Exp Eye Res. 41. Ahn JK, Kang JH, Park KH. Correlation between a disc
2019;188:107783. hemorrhage and peripapillary atrophy in glaucoma patients
22. Ohno-Matsui K. Proposed classification of posterior staph- with a unilateral disc hemorrhage. J Glaucoma. 2004;13:9–14.
ylomas based on analyses of eye shape by three-dimensional 42. Zhou LX, Shao L, Da Zhou W, et al. Beta zone parapapillary
magnetic resonance imaging and wide-field fundus imaging. atrophy in elderly Chinese. BMC Ophthalmol. 2022;22:431.
Ophthalmology. 2014;121:1798–1809. 43. Ghassibi MP, Chien JL, Patthanathamrongkasem T, et al.
23. Guo X, Xiao O, Chen Y, et al. Three-dimensional eye shape, Glaucoma diagnostic capability of circumpapillary retinal
myopic maculopathy, and visual acuity: The Zhongshan nerve fiber layer thickness in circle scans with different
Ophthalmic Center-Brien Holden Vision Institute High Myopia diameters. J Glaucoma. 2017;26:335–342.
Cohort Study. Ophthalmology. 2017;124:679–687. 44. Malik R, Belliveau AC, Sharpe GP, et al. Diagnostic accuracy
24. Ang M, Wong CW, Hoang QV, et al. Imaging in myopia: of optical coherence tomography and scanning laser tomog-
potential biomarkers, current challenges and future develop- raphy for identifying glaucoma in myopic eyes. Ophthalmology.
ments. Br J Ophthalmol. 2019;103:855–862. 2016;123:1181–1189.
25. Zemborain ZZ, Jarukasetphon R, Tsamis E, et al. Optical 45. Sastre-Ibanez M, Martinez-de-la-Casa JM, Rebolleda G, et al.
coherence tomography can be used to assess glaucomatous Utility of Bruch membrane opening-based optic nerve head
optic nerve damage in most eyes with high myopia. J Glaucoma. parameters in myopic subjects. Eur J Ophthalmol. 2018;28:
2020;29:833–845. 42–46.
26. Manjunath V, Shah H, Fujimoto JG, et al. Analysis of 46. Jonas JB, Jonas SB, Jonas RA, et al. Parapapillary atrophy:
peripapillary atrophy using spectral domain optical coherence histological gamma zone and delta zone. PLoS One. 2012;7:e47237.
tomography. Ophthalmology. 2011;118:531–536. 47. Zheng F, Wu Z, Leung CKS. Detection of Bruch’s membrane
27. Choi S, Jassim F, Tsikata E, et al. Artifact rates for 2D retinal opening in healthy individuals and glaucoma patients with and
nerve fiber layer thickness versus 3d retinal nerve fiber layer without high myopia. Ophthalmology. 2018;125:1537–1546.
volume. Transl Vis Sci Technol. 2020;9:12. 48. Shoji T, Sato H, Ishida M, et al. Assessment of glaucomatous
28. Meng W, Butterworth J, Malecaze F, et al. Axial length of changes in subjects with high myopia using spectral domain
myopia: a review of current research. Ophthalmologica. optical coherence tomography. Invest Ophthalmol Vis Sci.
2011;225:127–134. 2011;52:1098–1102.
29. Koomson NY, Kobia-Acquah E, Abdul-Kabir M, et al. 49. Seol BR, Kim DM, Park KH, et al. Assessment of optical
Relationship between peripheral refraction, axial lengths and coherence tomography color probability codes in myopic
parental myopia of young adult myopes. J Optom. 2021;15: glaucoma eyes after applying a myopic normative database.
122–128. Am J Ophthalmol. 2017;183:147–155.
30. Tan NYQ, Sng CCA, Ang M. Myopic optic disc changes and 50. Lee SY, Kwon HJ, Bae HW, et al. Frequency, type and cause
its role in glaucoma. Curr Opin Ophthalmol. 2019;30:89–96. of artifacts in swept-source and cirrus HD optical coherence
31. Kang SH, Hong SW, Im SK, et al. Effect of myopia on the tomography in cases of glaucoma and suspected glaucoma.
thickness of the retinal nerve fiber layer measured by Cirrus HD Curr Eye Res. 2016;41:957–964.

Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. www.glaucomajournal.com | 733