Development of Topographic Scoring

System for Identifying Glaucoma in Myopic

Eyes
A Spectral-Domain OCT Study
Sung Uk Baek, MD,1,2,* Ko Eun Kim, MD, PhD,3,* Young Kook Kim, MD,1,2 Ki Ho Park, MD, PhD,1,2
Jin Wook Jeoung, MD, PhD1,2

Purpose: To develop a new scoring system that uses topographic diagnostic signs of spectral-domain (SD)
OCT to enhance glaucoma diagnostic performance for myopic eyes and to validate the system’s diagnostic
ability.
Design: Cross-sectional study.
Participants: A total of 517 patients (517 eyes; spherical equivalent [SE] <e1.0 diopters [D] or axial length
>24.0 mm), including 175 highly myopic eyes (SE <e6.0 D or axial length >26.0 mm), were recruited and divided
into 2, training (241 eyes) and validation (276 eyes) test sets.
Methods: Retinal nerve fiber layer (RNFL) and ganglion celleinner plexiform layer (GCIPL) topographic signs
were selected based on the morphologic patterns of RNFL (size, shape, location, and agreement between
deviation and thickness maps) and GCIPL (size, shape, location, color tone, agreement between maps, and step
sign) abnormalities indicative of higher likelihood of myopic glaucoma on deviation and thickness maps. The
diagnostic score was compiled according to the sensitivity, specificity, and positive likelihood ratio (PLR) of each
diagnostic sign using the training set. The area under the receiver operating characteristic curve (AUC) was
plotted and compared between the OCT-provided parameters and the scoring system in the validation set.
Main Outcome Measures: The diagnostic performance of a new scoring system as validated by AUC.
Results: Among all of the RNFL and GCIPL parameters, the presence of temporal hemifield asymmetry on
the GCIPL thickness map (PLR, 5.98) showed the highest diagnostic ability, followed by location of the RNFL
defect (PLR, 5.79) and color tone of the GCIPL defect (PLR, 5.04). The AUC of the topographic scoring system in
myopic eyes was 0.979, which was significantly higher than those of the inferior (0.895; P < 0.001) and average
(0.894; P < 0.001) RNFL thickness parameters. For highly myopic eyes, the scoring system (AUC, 0.983) also
showed a higher diagnostic performance than that of the RNFL and GCIPL thickness parameters (all P < 0.001).
Conclusions: Our scoring system including OCT topographic parameters demonstrated to be beneficial for
clinicians to differentiate real glaucomatous damage from myopic healthy eyes. Our results support the value of
using multitopographic OCT parameters for detecting glaucoma in myopic eyes. Ophthalmology 2018;-
:1e10 ª 2018 by the American Academy of Ophthalmology

Supplemental material available at www.aaojournal.org.

The approximate worldwide prevalence of myopia has been
reported to be 30% for the general population1,2 and 74% to
84% among people 5 to 18 years of age in East Asia.3,4
These figures are especially important in light of several
population-based studies’ findings of significant associa-
tions between myopia and glaucoma.5e7 In myopic eyes,
evaluation of glaucomatous structural change in the optic
nerve head can be challenging because of considerable
differences in morphologic characteristics such as tilted
disc, peripapillary atrophy, larger diameters of optic disc,
and larger cup-to-disc ratio.8e10
Previous studies have suggested useful methods for dif-
ferentiation of myopic glaucomatous eyes from myopic
healthy eyes using spectral-domain (SD) OCT.11e13
Spectral-domain OCT provides various peripapillary
retinal nerve fiber layer (RNFL) and macular ganglion
celleinner plexiform layer (GCIPL) parameters including
thickness, color-coded diagnostic classification, and patterns
of glaucomatous damage, enabling objective and quantita-
tive evaluation of myopic eyes.14e16 Still, a major impedi-
ment of SD OCT in diagnosing myopic glaucomatous eyes
includes erroneous measurements or high false-positive

ª 2018 by the American Academy of Ophthalmology https://doi.org/10.1016/j.ophtha.2018.05.002 1

Published by Elsevier Inc. ISSN 0161-6420/18
 Ophthalmology Volume -, Number -, Month 2018
rates of diagnostic classification.11,17,18 Despite introduction
of multiple OCT thickness parameters or diagnostic signs
for enhancement of glaucoma diagnostic ability, the sig-
nificance and the diagnostic weight of each sign or param-
eter is not understood fully in myopic eyes having
distinctive structural characteristics. Therefore, to use these
parameters optimally for diagnosis of myopic glaucoma,
there is a need to understand the diagnostic implication of
each SD OCT parameter. Moreover, as one parameter alone
cannot be used in actual clinical settings, the effective
combination of multiple parameters could provide more
comprehensive and useful information to clinicians.
In this regard, in this study, using OCT thickness and
deviation maps for detection of myopic glaucomatous eyes,
we evaluated the quantitative diagnostic impacts of
commonly used topographic signs. Based on those param-
eters, we formulated a new scoring system to enhance
glaucoma diagnostic performance for myopic eyes and
validated the system’s diagnostic ability via comparison
with other OCT parameters.
Methods
This investigation was based on the Macular Ganglion Cell
Imaging Study, an ongoing prospective study of patients with
glaucoma and healthy individuals at the Glaucoma Clinic of Seoul
National University Hospital, Seoul, Korea. Eyes were chosen
from a database of glaucoma patients and healthy individuals. This
study adhered to the tenets of the Declaration of Helsinki and was
approved by the institutional review board of Seoul National
University Hospital. Informed consent to participate was obtained
from all participants.
Participants
All participants underwent complete ophthalmologic examinations,
including intraocular pressure measurement by Goldmann appla-
nation tonometry, best-corrected visual acuity assessment, refrac-
tive value assessment with an autorefractor (KR-890; Topcon
Corporation, Tokyo, Japan), central corneal thickness measurement
(Pocket II Pachymeter Echograph; Quantel Medical, Clermont-
Ferrand, France), axial length measurement (Axis II PR; Quantel
Medical, Inc., Bozeman, MT), slit-lamp examination, gonioscopy,
and dilated fundus examination. After maximum pupil dilation, all
participants underwent red-free RNFL photography (Vx-10; Kowa
Optimed, Inc., Tokyo, Japan), stereo optic disc photography, and
SD OCT (Cirrus-HD software version 6.0; Carl Zeiss Meditec,
Inc., Dublin, CA). They underwent standard automated perimetry
using the Swedish interactive threshold algorithm with the 30-2
standard program (Humphrey Field Analyzer II; Carl Zeiss
Meditec, Inc.).
All participants had myopic eyes defined as a spherical equiv-
alent (SE) of less than e1.0 diopter (D) or axial length of more than
24.0 mm. In addition, highly myopic eyes were defined as those
with SE of less than e6.0 D or axial length of more than 26.0 mm.
They had to meet the following criteria for inclusion:
best-corrected visual acuity of 20/40 or better and a normal anterior
chamber and open angle on slit-lamp and gonioscopic examina-
tions. Patients with a history of ocular surgery other than uncom-
plicated cataract surgery; dense nuclear sclerosis representing
myopic shift; any history of retinal pathologic characteristics,
diabetes mellitus, or nonglaucomatous optic neuropathy; or
apparent pathologic myopic changes were excluded. For
2
participants with a history of cataract surgery, preoperative SE and
axial length were adopted as the eligibility criteria.
Glaucomatous eyes were defined as those showing glaucoma-
tous optic disc neuropathy (e.g., notching, neuroretinal rim thin-
ning, RNFL defects, or a combination thereof) and corresponding
glaucomatous visual field defects, as confirmed by at least 2
consecutive visual field examinations. The glaucomatous visual
field defects were defined as a cluster of 3 points or more with
P < 0.05 on the pattern deviation map in at least 1 hemifield,
including 1 point or more with P < 0.01; a pattern standard de-
viation of P < 0.05; or glaucoma hemifield test results outside the
normal limits. Only patients with reliable visual field test results,
defined as fixation loss of less than 20%, false-positive errors of
less than 15%, and false-negative errors of less than 15%, were
included. In cases in which both eyes of a patient were eligible for
inclusion, one eye was chosen randomly. The criteria to be eligible
for the healthy eyes were a normal anterior segment on slit-lamp
examination, intraocular pressure between 10 and 21 mmHg, no
RNFL defects in red-free RNFL photographs, and no glaucoma-
tous visual field defects.
Determination of Training and Validation Sets
We divided the participants into 2 sets, a training set and a vali-
dation set. Using the training set, we designed a new scoring
system based on topographic diagnostic parameters of SD OCT.
Next, using the validation set, we confirmed the diagnostic ability
of the scoring system via comparison with other OCT parameters.
Computerized randomization (Statistical Package for the Social
Sciences version 21.0 for Windows; IBM Corp., Armonk, NY)
selected 241 eyes (46.6%) of the study population as the training
set. The devised topographic scoring system then was applied to
the remaining 276 eyes (53.4%) of the patients in the validation set.
Color Disc Photography and Red-Free Retinal
Nerve Fiber Layer Photography
Color disc photographs and red-free RNFL photographs were
evaluated independently by 2 of the authors (K.E.K. and J.W.J.)
who were masked to the patients’ clinical information including
OCT results. The presence of glaucomatous optic disc change was
determined by a consensus agreement between the 2 observers. In
cases where the 2 observers disagreed, those participants were
classified as ambiguous and were excluded from further analysis.
Spectral-Domain OCT Measurements
Macular and optic disc scans were acquired through a dilated pupil.
All of the SD OCT images were obtained by 1 experienced tech-
nician. Only qualified images with a signal strength of 8 or more
(maximum of 10) without segmentation failure, motion artifacts, or
decentration of the measurement circle were included.
One macular scan focused on the fovea (200
200 cube pro-
tocol) and 1 peripapillary scan focused on the optic disc (200
200
cube protocol) were acquired for every participant. The details of
macular and optic disc cube scan of SD OCT have been reported
previously.19,20 Briefly, an optic disc cube obtained from 3-
dimensional data that cover a 6-mm2 area is centered on the
optic disc. The ganglion cell analysis algorithm detects and
measures macular GCIPL thickness within an annulus with inner
vertical and horizontal diameters of 1 and 1.2 mm, respectively,
and outer vertical and horizontal diameters of 4 and 4.8 mm,
respectively. Using these automatically measured thickness data,
SD OCT provides a thickness deviation map of RNFL and
GCIPL. An RNFL thickness deviation map, in comparison with
age-matched normative data, uses yellow and red colors to
 Baek et al 
Scoring System for Diagnosing Myopic Glaucoma

Table 1. Glaucoma Diagnostic Signs on OCT Retinal Nerve Fiber Layer and Ganglion CelleInner Plexiform Layer Maps

Glaucoma Diagnostic Topographic Signs Definition

RNFL defect based on OCT RNFL analysis
Size >20 superpixels (on OCT RNFL deviation map)
Shape Wedge shape (on OCT RNFL deviation map)
Location Superotemporal or inferotemporal area (on OCT RNFL deviation map)
Agreement with RNFL thickness map Decrease in RNFL thickness in the corresponding area of the OCT RNFL defect on OCT RNFL
deviation map
GCIPL defect based on OCT GCIPL analysis
Size >10 superpixels (on OCT GCIPL deviation map)
Shape Arcuate or circular shape (on OCT GCIPL deviation map)
Location Predominantly located in the temporal part (on OCT GCIPL deviation map)
Color tone Yellow or red, dense, monotonous, uniformly color-coded area (on OCT GCIPL deviation map)
Agreement with GCIPL thickness map Decrease in GCIPL thickness in the corresponding area of the OCT GCIPL defect on OCT GCIPL
deviation map
Step sign on GCIPL thickness map Step sign at temporal raphe (hemifield asymmetry)

GCIPL ¼ ganglion celleinner plexiform layer; RNFL ¼ retinal nerve fiber layer.

indicate superpixels in which RNFL thickness falls in less than 5%
and less than 1% of normal distribution percentiles, respectively.
Uncolored areas demonstrate RNFL thickness within the normal
range. Likewise, the ganglion cell analysis algorithm generates a
GCIPL deviation map, on which areas appear as yellow or red to
represent GCIPL thickness less than the lower 5% or 1% of
normative data, respectively, and uncolored areas indicating
normal GCIPL thickness.
Formulation and Evaluation of Topographic
Scoring System Based on Retinal Nerve Fiber
Layer and Ganglion CelleInner Plexiform Layer
Maps
Previous studies have demonstrated typical glaucomatous patterns
of OCT RNFL and GCIPL damage.11,17,18,21e24 We selected
diagnostic signs of OCT from the morphologic patterns of RNFL
and GCIPL abnormalities on deviation and thickness maps indi-
cating glaucomatous damage. We then evaluated the diagnostic
signs of RNFL (size, shape, location, agreement between maps)
and GCIPL (size, shape, location, color tone, agreement between
maps, step sign) defects on such maps. The specific characteristics
of the topographic glaucoma diagnostic signs for the RNFL and
GCIPL analyses were as follows (Table 1): for the OCT RNFL
map, (1) size more than 20 superpixels, (2) wedge shaped, (3)
located in the superotemporal or inferotemporal area, and (4)
presence of agreement between RNFL deviation and thickness
maps; and for the OCT GCIPL map, (1) size more than 10
superpixels, (2) arcuate to crescent, circular shape, (3) located
predominantly in the temporal area, (4) more dense, monotonous,
uniformly yellow or red color-coded area, (5) presence of agree-
ment between GCIPL deviation map and thickness map, and (6)
step sign on GCIPL thickness map.
Two observers (K.E.K. and J.W.J.) individually evaluated the
patterns of RNFL and GCIPL defects shown on the deviation maps
in consideration of the aforementioned variables in an independent
and masked manner. The diagnostic score was constructed based
on the sensitivity, specificity, and positive likelihood ratio (PLR) of
each diagnostic sign using the training set. The diagnostic score
then was weighted by the PLR rounded to the nearest integer. The
total score was calculated for each patient by summing the

Table 2. Baseline Characteristics of Participants

Training Set Validation Set

Myopic Eyes Myopic Glaucomatous Eyes Myopic Eyes Myopic Glaucomatous Eyes
Characteristic (n ¼ 107) (n ¼ 134) P Value (n ¼ 112) (n ¼ 164) P Value
Age (yrs) 48.911.4 51.311.2 0.103* 48.011.9 50.311.4 0.116*
Male 64 (59.8) 76 (56.7) 0.724y 63 (56.3) 92 (56.4) 0.986y
Axial length (mm) 25.81.1 26.11.1 0.074* 25.91.2 26.11.2 0.181*
SE (D) e5.572.45 e6.262.41 0.090* e5.303.25 e6.023.13 0.103*
Average RNFLT (mm) 92.19.9 74.88.9 <0.001* 91.49.9 74.59.3 <0.001*
Average GCIPLT (mm) 77.57.9 68.67.0 <0.001* 75.66.0 68.77.6 <0.001*
Minimum GCIPLT (mm) 72.412.4 59.79.8 <0.001* 72.78.5 59.011.2 <0.001*
SAP C30-2 MD (dB) e0.721.52 e3.863.90 <0.001* e0.671.47 e3.653.73 <0.001*
SAP C30-2 PSD (dB) 1.840.94 6.464.50 <0.001* 2.000.99 5.904.49 <0.001*

D ¼ diopter; GCIPLT ¼ ganglion celleinner plexiform layer thickness; MD ¼ mean deviation; PSD ¼ pattern standard deviation; RNFLT ¼ retinal nerve
fiber layer thickness; SAP ¼ standard automated perimetry; SE ¼ spherical equivalent.
Data are mean  standard deviation or no. (%) unless otherwise indicated.
*Student t test.
y
Chi-square test.

3
 Ophthalmology Volume -, Number -, Month 2018
Table 3. Topographic Scoring System Based on the Positive
Likelihood Ratio of Each Diagnostic Sign Using Training Set
Positive Likelihood
Ratio* Scorey
RNFL defect based on OCT RNFL analysis
Size 2.43 2 for the Social Sciences version 21.0 for Windows (IBM Corp.,
Shape 2.76 3 Armonk, NY). Statistical significance was defined as P < 0.05.
Location 5.79 6
Agreement with RNFL thickness map 1.87 2
GCIPL defect based on OCT GCIPL analysis
Size 1.94 2
Shape 1.86 2
Location 4.78 5
Color tone 5.04 5
Agreement with GCIPL thickness map 1.99 2
Step sign on GCIPL thickness map 5.98 6 ization selected a training set (107 myopic healthy eyes and 134
GCIPL ¼ ganglion celleinner plexiform layer; RNFL ¼ retinal nerve fiber
layer.
*Positive likelihood ratio ¼ sensitivity / (100 e specificity).
y
Scores were assigned according to the positive likelihood ratio values
rounded off to the nearest whole number for convenience.
weighted scores of 10 diagnostic signs when topographic findings
were present. It arithmetically ranged from a minimum of 0 to a
maximum of 35 points. Of the maximum score of 35 points, 13
were allocated to 4 RNFL signs and 22 to 6 GCIPL signs. The
scoring system was validated by 2 observers (K.E.K. and S.U.B.)
using the independent validation set.
Statistical Analysis
The baseline characteristics of the myopic nonglaucomatous and
glaucomatous eyes in the training and validation sets were
compared using the Student t test and chi-square test for continuous
and categorical variables, respectively. Sensitivity, specificity, and
PLR were evaluated for each of the diagnostic variables using the
training set to determine the scoring system.
To assign different weights of points to each topographic
sign, the PLR of each sign was derived from the training set. The
PLR was calculated by the following formula: sensitivity / (100 e
specificity). Ultimately, the PLR of each sign was converted to the
topographic scoring system by rounding it off to the nearest whole
number. Then, the total score was obtained by summing the con-
verted points of the respective signs. The cutoff points were those
with the best balance between sensitivity and specificity.
The calculated intraobserver reproducibility of the evaluation of
topographic diagnostic signs determining glaucomatous defect for
each rater (S.U.B., K.E.K., J.W.J.) was based on first- and second-
round scoring using a scoring system with a 2-week scoring
interval. A 2-way random effects model was applied to determine
reliability based on the intraclass correlation coefficient (ICC) by
pooling together the myopic healthy eyes and glaucomatous eyes.
An ICC with a 2-way random effects model also was applied to
determine the interobserver reproducibility for the estimate of
agreement among the 3 raters. An ICC of 0.41 to 0.60 was
considered fair agreement, an ICC of 0.61 to 0.80 was considered
moderate agreement, and an ICC of more than 0.80 was considered
excellent agreement.25
The area under the receiver operating characteristic curve
(AUC) with the corresponding 95% confidence interval (CI) was
calculated to determine the diagnostic ability of each parameter to
discriminate glaucomatous eyes from healthy ones: an AUC of 1.0
represents perfect discrimination, and one of 0.5 represents chance
4
discrimination. DeLong’s test was used to test for the statistical
significance of the diagnostic performance difference (represented
as AUC) between any 2 parameters.26 The AUCs of different
variables were compared using MedCalc software version 12.0
(MedCalc Statistical Software, Marakierke, Belgium). Other
statistical analyses were performed using the Statistical Package
Results
This study included 219 myopic healthy eyes and 298 myopic
glaucomatous eyes, including 67 healthy eyes and 108 glaucom-
atous eyes with a high degree of myopia. Computerized random-
myopic glaucomatous eyes) and a validation set (112 healthy
myopic eyes and 164 myopic glaucomatous eyes).
The baseline characteristics are summarized in Table 2.
Moreover, in our validation set, a total of 175 highly myopic
eyes were included, with the mean age of 49.111.4 years,
mean axial length of 26.80.79 mm, and mean SE of
e7.272.71 D, respectively. There were no significant
differences in mean age between the myopic healthy eyes and
myopic glaucomatous eyes. Neither gender, nor axial length, nor
SE showed any differences between the groups in either the
training or validation set (all P > 0.05). In a separate analysis of
highly myopic eyes, no significant difference was found between
healthy eyes and glaucomatous eyes with regard to age
(47.111.2 years vs. 50.211.4 years; P ¼ 0.100), axial length
(26.80.7 mm vs. 26.80.9 mm; P ¼ 0.960), or SE
(e7.272.36 D vs. e7.262.92 D; P ¼ 0.981). However, as
expected, the myopic glaucomatous eyes demonstrated thinner
average RNFL and GCIPL thicknesses as well as thinner
minimum GCIPL thickness and showed worse visual field
results than the myopic healthy eyes in both sets (all P < 0.001;
Table 2).
Diagnostic Ability of Each Retinal Nerve Fiber
Layer or Ganglion CelleInner Plexiform Layer
Topographic Parameter for Discrimination of
Myopic Glaucomatous Eyes from Myopic
Healthy Eyes
To investigate the discriminatory ability of each topographic sign,
its PLR was evaluated using the training set (Table 3). Among all
of the RNFL and GCIPL parameters, the presence of step sign
(temporal hemifield asymmetry) on a GCIPL thickness map
(PLR, 5.98) was given the highest value for detection of myopic
glaucoma on OCT deviation and thickness maps. The location of
RNFL defect, with a PLR of 5.79, was ranked second. Among
the RNFL topographic parameters, the shape of the RNFL defect
received the second highest PLR of 2.76, followed by size of the
RNFL defect (PLR, 2.43) and agreement between RNFL
deviation and thickness maps (PLR, 1.87). Among the GCIPL
topographic parameters, color tone of GCIPL defect was
assigned a PLR of 5.04, followed by location (PLR, 4.78),
agreement between GCIPL deviation and thickness maps (PLR,
1.99), size (PLR, 1.94), and shape (PLR, 1.86). In the training
set, the cutoff point of our scoring system suggestive of a high
likelihood of myopic glaucoma was 23 (sensitivity, 92.7%;
specificity, 97.6%; AUC, 0.977). Figure 1 shows 2 cases of
myopic glaucoma and myopic healthy eyes, respectively, from
the validation set.
Baek et al 
Scoring System for Diagnosing Myopic Glaucoma

5
 Ophthalmology Volume -, Number -, Month 2018

Assessment of Intraobserver and Interobserver
Reproducibility of Evaluation of Topographic
Diagnostic Signs for Determination of
Glaucomatous Damage in Myopic Eyes
Overall, the intraobserver and interobserver reproducibilities for
the RNFL (ICC, 0.865e0.986) and GCIPL (ICC, 0.863e0.990)
showed excellent agreements (all P < 0.001; Table S1, available at
www.aaojournal.org). Despite high intraobserver reproducibility
for the RNFL signs, the size of the RNFL defect (ICC, 0.969)
showed the lowest reproducibility, and the location of the RNFL
defect (ICC, 0.986) showed the highest. For the GCIPL signs,
the color tone of the GCIPL defect (ICC, 0.952) showed the
lowest reproducibility, and the step sign (ICC, 0.990) showed the
highest. For the interobserver variance, the agreement between
the RNFL deviation map and the thickness map (ICC, 0.865)
showed the lowest value, and the location of the RNFL defect
(ICC, 0.974) showed the highest for the RNFL signs. For the
GCIPL signs, the shape of the GCIPL defect (ICC, 0.863)
presented the lowest interobserver reproducibility, and the step
sign (ICC, 0.968) showed the highest.
Validation of Diagnostic Performance of
Topographic Scoring System
To validate the diagnostic performance of the topographic scoring
system, its AUC was compared with those of commonly used RNFL
(average and 4 quadrants) and GCIPL (average, minimum, and 6
sectors) thickness parameters using the validation set (Table 4).
Among the single OCT parameters, those that were remarkable for
discriminating normal from glaucomatous eyes were inferior
RNFL thickness (AUC, 0.895; 95% CI, 0.846e0.932), average
RNFL thickness (AUC, 0.894; 95% CI, 0.846e0.932),
inferotemporal GCIPL thickness (AUC, 0.865; 95% CI,
0.794e0.919), minimum GCIPL thickness (AUC, 0.840; 95% CI,
0.765e0.898), and inferior GCIPL thickness (AUC, 0.792; 95%
CI, 0.711e0.858). The topographic scoring system showed an
AUC of 0.979 for all myopic eyes, which was significantly higher
than those of the inferior and average RNFL thickness parameters
(all P < 0.001) and inferotemporal, minimum, and inferior GCIPL
thickness parameters (all P < 0.001; Fig 2A). Even for highly
myopic eyes (67 healthy and 108 glaucomatous eyes), the scoring
system (AUC, 0.983; 95% CI, 0.951e0.997) showed a
significantly higher diagnostic performance than that of the RNFL
or GCIPL thickness parameters (all P < 0.001; Fig 2B).
Discussion
In this study, we formulated a topographic scoring system
for typical morphologic changes that determine myopic
glaucoma. Our scoring system consists of important
topographic diagnostic signs from RNFL and GCIPL maps
whose quantitative diagnostic values were confirmed
through the training set of myopic eyes. After confirmation,
the score was assigned to each topographic sign based on its
degree of significance in detecting glaucomatous damage in
myopic eyes. In this way, we could determine how impor-
tant a role each diagnostic sign plays in distinguishing
myopic glaucoma. We also confirmed that the topographic
scoring system, compared with other RNFL and GCIPL
parameters in SD OCT, has a better diagnostic ability for
discriminating between healthy and glaucomatous eyes in
myopia. Further, a classification algorithm based on our
scoring system would be useful to OCT interpretation for
glaucoma diagnostic purposes.
There have been many studies on SD OCT-based diag-
nosis of glaucoma in myopic eyes. The corresponding re-
ports have introduced several diagnostic OCT signsdsuch
as size, shape, location, color tone, and temporal hemifield
asymmetry of OCT RNFL and GCIPL defectsdfor
discrimination of myopic glaucomatous eyes from normal
eyes.11,12,21e23 In clinical settings, however, clinicians
diagnose glaucoma based not on a single diagnostic sign on
an OCT printout; rather, they confirm the diagnosis by
compiling and combining information gathered from mul-
tiple OCT results. Accordingly, we created and evaluated
our new scoring system incorporating topographic diag-
nostic signs found on RNFL and GCIPL deviation and
thickness maps.
In the current study, the presence of temporal hemifield
asymmetry on the OCT GCIPL thickness map was given the
highest PLR for differentiating myopic glaucoma from
myopic healthy eyes. It is well recognized that glaucoma-
tous RNFL defects often occur at the superotemporal or
inferotemporal parapapillary region, or both.27,28 Further-
more, previous studies29e31 have supposed that the retinal
ganglion cells in the superior or inferior temporal macula
project their axons in an arcuate pattern toward the super-
otemporal or inferotemporal portions of the optic nerve
head. Another study, this one performed by our group,
showed that the AUC of a positive step sign (0.938) was
greater than that of either inferotemporal GCIPL thickness
(0.814) or minimum GCIPL thickness (0.818).13 By
extension of these spatial and morphologic considerations,
it is also possible to conclude that characteristic
glaucomatous changes are located predominantly in the
temporal macular area along the horizontal raphe, which
usually is in the same hemifield as the corresponding
RNFL defect.

Figure 1. Myopic glaucomatous and healthy eyes from the validation set. A, Myopic glaucomatous eye of a 39-year-old woman with a spherical equivalent of e5.00
diopters (D). An inferior localized retinal nerve fiber layer (RNFL) defect shown by red-free RNFL photograph (arrowhead) was detected on RNFL and ganglion
celleinner plexiform layer (GCIPL) maps of spectral-domain (SD) OCT. In detail, her left eye had a wedge-shaped (3 points) inferotemporal (6 points) RNFL defect of
more than 20 superpixels (2 points) on the OCT RNFL deviation map. The RNFL defect on the deviation map corresponded well with that on the thickness map (2
points). As for the GCIPL maps, an arcuate (2 points) defect of a size of more than 10 superpixels (2 points) on the OCT GCIPL deviation map was located mainly in
the temporal parafoveal area (5 points) showing a dense and monotonous red color (5 points). The GCIPL thickness map showed a defect corresponding to that on the
GCIPL deviation map (2 points) and a temporal step sign respecting the temporal horizontal raphe (6 points). As a result, the patient had a total score of 35 points (13
points from RNFL the analysis and 22 points from the GCIPL analysis), which suggests a high probability of having glaucoma. B, Myopic healthy eye of a 34-year-old
woman with a spherical equivalent of e4.50 D. The red-free RNFL photograph showed no definite structural damage. However, the RNFL and GCIPL deviation maps
with their abnormal color codes could be misinterpreted as indicating glaucomatous damage. The score according to the topographic scoring system was 11 points,
which suggests a low probability of having glaucoma.

6
 Baek et al 
Scoring System for Diagnosing Myopic Glaucoma

Table 4. Glaucoma Diagnostic Ability of Topographic Scoring System, Retinal Nerve Fiber Layer Thickness, and Ganglion CelleInner
Plexiform Layer Thickness

Area under the Receiver Operating Characteristic Curve (95% Confidence Interval)
All Myopic Eyes (Healthy, n ¼ 112; Highly Myopic Eyes (Healthy, n ¼ 67;
Parameters Glaucomatous, n ¼ 163) P Value* Glaucomatous, n ¼ 108) P Value*
Topographic scoring system 0.979 (0.954e0.993) N/A 0.983 (0.951e0.997) N/A
RNFL thickness
Average 0.894 (0.846e0.932) <0.001 0.895 (0.834e0.940) <0.001
Superior 0.780 (0.719e0.832) <0.001 0.753 (0.675e0.820) <0.001
Nasal 0.611 (0.543e0.676) <0.001 0.589 (0.505e0.669) <0.001
Inferior 0.895 (0.846e0.932) <0.001 0.902 (0.842e0.945) <0.001
Temporal 0.730 (0.666e0.787) <0.001 0.792 (0.718e0.855) <0.001
GCIPL thickness
Average 0.773 (0.691e0.842) <0.001 0.782 (0.693e0.855) <0.001
Minimum 0.840 (0.765e0.898) <0.001 0.850 (0.769e0.911) <0.001
Superotemporal 0.699 (0.612e0.777) <0.001 0.715 (0.621e0.797) <0.001
Superior 0.591 (0.503e0.676) <0.001 0.602 (0.505e0.694) <0.001
Superonasal 0.602 (0.513e0.685) <0.001 0.619 (0.522e0.709) <0.001
Inferonasal 0.676 (0.588e0.756) <0.001 0.693 (0.598e0.778) <0.001
Inferior 0.792 (0.711e0.858) <0.001 0.805 (0.719e0.874) <0.001
Inferotemporal 0.865 (0.794e0.919) <0.001 0.889 (0.815e0.941) <0.001

GCIPL ¼ ganglion celleinner plexiform layer; N/A ¼ not applicable; RNFL ¼ retinal nerve fiber layer.
*Comparison between topographic scoring system and thickness parameters.

Initially, the 10 (4 RNFL and 6 GCIPL) diagnostic signs,
which can be recognized easily from the OCT printout
without further manipulation, were determined based on
previous investigation of glaucomatous RNFL and GCIPL
defect pattern using SD OCT.11,17,18,21e24,31 Then, we
evaluated the diagnostic implication of each diagnostic
parameter with our training set of myopic eyes. Despite
thorough analysis, a considerable point of our scoring sys-
tem is that the elements of GCIPL analysis in the total scores
are more predominant than those of RNFL analysis. Spe-
cifically, 2 additional factors (color tone and step sign on
GCIPL thickness map) were included in GCIPL signs
compared with RNFL components. As previously
mentioned, the step sign on the GCIPL thickness map has

Figure 2. Graphs comparing the diagnostic performance between the topographic scoring system and global and sectoral retinal nerve fiber layer (RNFL) and
ganglion celleinner plexiform layer (GCIPL) thickness parameters in (A) all myopia and (B) high myopia using the validation set. The areas under the receiver
operating characteristic curve of the topographic scoring system were 0.979 in the all myopia group and 0.983 in the high myopia group, which were significantly
higher than those of the RNFL and GCIPL thickness parameters (all P < 0.001). Ave_RNFLT ¼ average RNFL thickness; Inf_GCIPLT ¼ inferior GCIPL
thickness; Inf_RNFLT ¼ inferior RNFL thickness; IT_GCIPLT ¼ inferotemporal GCIPL thickness; Min_GCIPLT ¼ minimum GCIPL thickness.

7
 Ophthalmology Volume -, Number -, Month 2018
shown highly sensitive glaucoma diagnostic power in our
study as well as in previous studies.13,23 At the same time,
for the scoring system to have both high sensitivity and high
specificity, we should be aware of false-positive signs.
Therefore, for differentiation of true and false-positive
diagnostic signs, color tone of GCIPL defect, reported in a
previous study,18 also was tested with our training set and
was proven to be highly important (PLR, 5.04). A score
weighted toward GCIPL signs can cause a false-positive
diagnostic error in patients with nonglaucomatous macular
abnormality, but the high diagnostic ability of our scoring
system implied that effective combination of RNFL and
GCIPL signs with different diagnostic capabilities could
overcome this limit.
Previous studies have shown the SD OCT device’s
RNFL thickness parameter AUC ranges from 0.60 to 0.98,
depending on the specific parameters and patient charac-
teristics.21,32e37 In those reports, inferior RNFL thickness
often demonstrated the best accuracy for discriminating
normal eyes from eyes with glaucoma.21,32,33 In the present
study, the AUC of our topographic scoring system was
0.979, which was significantly higher than that of inferior
RNFL thickness parameters (AUC, 0.895; P < 0.001;
Table 4; Fig 2A). Moreover, the results of the present study
correspond well with those of an earlier study reporting on a
scoring system (0e5) developed to analyze the RNFL
thickness deviation map by taking into consideration
defect size, shape, depth, location, and distance from the
disc margin (a map score of 5 points attained a
significantly higher sensitivity compared with clock-hour
or average RNFL thickness categorical classification by
SD OCT).21
The initial purpose of this study was to create a scoring
system to facilitate glaucoma diagnosis in myopic eyes.
However, after confirming from additional analyses
(Table S2, available at www.aaojournal.org) the high
diagnostic ability of the scoring system even for nonmyopic
eyes, the authors realized that a new scoring system could
be comparable with other commonly used RNFL and
GCIPL thickness parameters in eyes with various degrees of
refractive error and also could be applicable to a wider
spectrum of patients. Also, the current manual system could
have problems with regard to subjectivity, reproducibility,
and discrepancy in observers’ clinical experience. To
overcome these obstacles and to shorten the score
calculation time, the authors believe that an automated
version is required in the near future. Further
complementary investigations aiming toward the creation of
an automated version of the topographic scoring system will
be helpful to the goal of improved clinical efficiency.
Several limitations of our study need to be considered.
First, all of the enrolled participants are Korean. Because SD
OCT uses a built-in internal normative database for its an-
alyses, detection of the RNFL and GCIPL thinning might
have been affected by the participants’ ethnicity. Second,
the present study included patients with early-to-moderate
glaucoma, which might have influenced the diagnostic
performance of our scoring system. Some topographic signs
(e.g., temporal hemifield asymmetry, shape of RNFL or
GCIPL defect) can have higher diagnostic values at earlier
8
disease stages than at more advanced ones. Thus, the
topographic scoring system might not be as effective for
patients with advanced glaucoma. However, we should note
that we included patients with early-to-moderate glaucoma
because advanced patients can be detected even with basic
clinical examinations. Third, cataract itself can cause a
myopic shift in some patients; therefore, including such
cases can be problematic for representing true myopic eyes.
Therefore, we rechecked the data of 5 patients from the
training set and that of 7 patients from the validation set who
had undergone simple cataract surgery. They all showed a
moderate degree of myopia by SE and an axial length 25.1
mm or more before surgery. Still, participants with acquired
myopic shift resulting from cataract could have been
included in the study, regardless of cataract surgery. Fourth,
the most reasonable comparison of glaucoma diagnostic
ability would be that between the scoring system and the
thickness parameters of certain RNFL or GCIPL defect
areas. Unfortunately, no system determining RNFL or
GCIPL thickness within a specific area by Cirrus OCT is
available. Thus, we compared the topographic scoring sys-
tem with the currently available global and sectoral thick-
ness parameters that can be identified easily from OCT
printouts. Finally, our scoring system presented merely
validates OCT results; it offers neither any consideration of
other structural measurements (e.g., red-free RNFL photo-
graphs and optic nerve head photographs) nor functional
assessment. That is, our topographic scoring system is a
supplemental method that can be used in combination with
other clinical parameters.
In conclusion, our scoring system incorporating OCT
topographic parameters was demonstrated to be beneficial
for clinicians’ differentiation of real glaucomatous damage
from myopic healthy eyes. Our results suggest that the OCT
topographic scoring system can be an effective adjunctive
diagnostic tool for detection of myopic glaucoma.
References
1. Sperduto RD, Seigel D, Roberts J, Rowland M. Prevalence of
myopia in the United States. Arch Ophthalmol. 1983;101(3):
405e407.
2. Wang Q, Klein B, Klein R, Moss SE. Refractive status in the
Beaver Dam Eye Study. Invest Ophthalmol Vis Sci.
1994;35(13):4344e4347.
3. He M, Zeng J, Liu Y, et al. Refractive error and visual
impairment in urban children in southern China. Invest Oph-
thalmol Vis Sci. 2004;45(3):793e799.
4. He M, Huang W, Zheng Y, et al. Refractive error and visual
impairment in school children in rural southern China.
Ophthalmology. 2007;114(2):374e382.
5. Kuzin AA, Varma R, Reddy HS, et al. Ocular biometry and
open-angle glaucoma: the Los Angeles Latino Eye Study.
Ophthalmology. 2010;117(9):1713e1719.
6. Mitchell P, Hourihan F, Sandbach J, Wang JJ. The relationship
between glaucoma and myopia: the Blue Mountains Eye
Study. Ophthalmology. 1999;106(10):2010e2015.
7. Perera SA, Wong TY, Tay W-T, et al. Refractive error, axial
dimensions, and primary open-angle glaucoma: the Singapore
Malay Eye Study. Arch Ophthalmol. 2010;128(7):900e905.
 Baek et al 
Scoring System for Diagnosing Myopic Glaucoma
8. Aref AA, Sayyad FE, Mwanza JC, et al. Diagnostic specific-
ities of retinal nerve fiber layer, optic nerve head, and macular
ganglion cell-inner plexiform layer measurements in myopic
eyes. J Glaucoma. 2014;23(8):487e493.
9. Jonas JB, Dichtl A. Optic disc morphology in myopic primary
open-angle glaucoma. Graefes Arch Clin Exp Ophthalmol.
1997;235(10):627e633.
10. Jonas JB, Gusek GC, Naumann GO. Optic disk morphometry
in high myopia. Graefes Arch Clin Exp Ophthalmol.
1988;226(6):587e590.
11. Leung CK-S, Yu M, Weinreb RN, et al. Retinal nerve fiber
layer imaging with spectral-domain optical coherence tomog-
raphy: interpreting the RNFL maps in healthy myopic eyes.
Invest Ophthalmol Vis Sci. 2012;53(11):7194e7200.
12. Shoji T, Sato H, Ishida M, et al. Assessment of glaucomatous
changes in subjects with high myopia using spectral domain
optical coherence tomography. Invest Ophthalmol Vis Sci.
2011;52(2):1098e1102.
13. Kim YK, Yoo BW, Jeoung JW, et al. Glaucoma-diagnostic
ability of ganglion cell-inner plexiform layer thickness dif-
ference across temporal raphe in highly myopic eyes. Invest
Ophthalmol Vis Sci. 2016;57(14):5856e5863.
14. Park H-YL, Shin H-Y, Yoon J-Y, et al. Intereye comparison of
Cirrus OCT in early glaucoma diagnosis and detecting
photographic retinal nerve fiber layer abnormalities: intereye
comparison in glaucoma diagnosis. Invest Ophthalmol Vis Sci.
2015;56(3):1733e1742.
15. Hwang YH, Jeong YC, Kim HK, Sohn YH. Macular ganglion
cell analysis for early detection of glaucoma. Ophthalmology.
2014;121(8):1508e1515.
16. Lee JE, Sung KR, Park JM, et al. Optic disc and peripapillary
retinal nerve fiber layer characteristics associated with glau-
comatous optic disc in young myopia. Graefes Arch Clin Exp
Ophthalmol. 2017;255(3):591e598.
17. Kim NR, Lim H, Kim JH, et al. Factors associated with false
positives in retinal nerve fiber layer color codes from spectral-
domain optical coherence tomography. Ophthalmology.
2011;118(9):1774e1781.
18. Kim KE, Jeoung JW, Park KH, et al. Diagnostic classification
of macular ganglion cell and retinal nerve fiber layer analysis:
differentiation of false-positives from glaucoma. Ophthal-
mology. 2015;122(3):502e510.
19. Leite MT, Rao HL, Weinreb RN, et al. Agreement among
spectral-domain optical coherence tomography instruments for
assessing retinal nerve fiber layer thickness. Am J Ophthal-
mology. 2011;151(1):85e92.e1.
20. Mwanza J-C, Oakley JD, Budenz DL, et al. Ability of cirrus
HD-OCT optic nerve head parameters to discriminate normal
from glaucomatous eyes. Ophthalmology. 2011;118(2):
241e248.e1.
21. Leung CK, Lam S, Weinreb RN, et al. Retinal nerve fiber layer
imaging with spectral-domain optical coherence tomography:
analysis of the retinal nerve fiber layer map for glaucoma
detection. Ophthalmology. 2010;117(9):1684e1691.
22. Leung CK, Choi N, Weinreb RN, et al. Retinal nerve fiber
layer imaging with spectral-domain optical coherence
Footnotes and Financial Disclosures
Originally received: January 11, 2018.
Final revision: April 27, 2018.
Accepted: May 1, 2018.
Available online: ---. Manuscript no. 2018-97.
1
Department of Ophthalmology, Seoul National University College of
Medicine, Seoul, Korea.
tomography: pattern of RNFL defects in glaucoma. Ophthal-
mology. 2010;117(12):2337e2344.
23. Kim YK, Yoo BW, Kim HC, Park KH. Automated detection
of hemifield difference across horizontal raphe on ganglion
celleinner plexiform layer thickness map. Ophthalmology.
2015;122(11):2252e2260.
24. Kim KE, Park KH, Yoo BW, et al. Topographic localization of
macular retinal ganglion cell loss associated with localized
peripapillary retinal nerve fiber layer defect topographic rela-
tionship between GCIPL and RNFL. Invest Ophthalmol Vis
Sci. 2014;55(6):3501e3508.
25. Landis JR, Koch GG. The measurement of observer agreement
for categorical data. Biometrics. 1977;33(1):159e174.
26. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the
areas under two or more correlated receiver operating char-
acteristic curves: a nonparametric approach. Biometrics.
1988;44(3):837e845.
27. Jonas JB, Dichtl A. Evaluation of the retinal nerve fiber layer.
Surv Ophthalmol. 1996;40(5):369e378.
28. Quigley HA, Vitale S. Models of open-angle glaucoma prev-
alence and incidence in the United States. Invest Ophthalmol
Vis Sci. 1997;38(1):83e91.
29. Garway-Heath DF, Poinoosawmy D, Fitzke FW,
Hitchings RA. Mapping the visual field to the optic disc in
normal tension glaucoma eyes. Ophthalmology. 2000;107(10):
1809e1815.
30. Hood DC, Raza AS, de Moraes CGV, et al. The nature of macular
damage in glaucoma as revealed by averaging optical coherence
tomography data. Trans Vis Sci Technol. 2012;1(1):1e15.
31. Hood DC, Raza AS, de Moraes CGV, et al. Glaucomatous
damage of the macula. Prog Retinal Eye Res. 2013;32:1e21.
32. Medeiros FA, Zangwill LM, Bowd C, Weinreb RN. Com-
parison of the GDx VCC scanning laser polarimeter, HRT II
confocal scanning laser ophthalmoscope, and Stratus OCT
optical coherence tomograph for the detection of glaucoma.
Arch Ophthalmol. 2004;122(6):827e837.
33. Jeoung JW, Park KH. Comparison of Cirrus OCT and Stratus
OCT on the ability to detect localized retinal nerve fiber layer
defects in preperimetric glaucoma. Invest Ophthalmol Vis Sci.
2010;51(2):938e945.
34. Mwanza J-C, Durbin MK, Budenz DL, et al. Glaucoma
diagnostic accuracy of ganglion celleinner plexiform layer
thickness: comparison with nerve fiber layer and optic nerve
head. Ophthalmology. 2012;119(6):1151e1158.
35. Takayama K, Hangai M, Durbin M, et al. A novel method to
detect local ganglion cell loss in early glaucoma using spectral-
domain optical coherence tomography detection of local RGC
loss in glaucoma. Invest Ophthalmol Vis Sci. 2012;53(11):
6904e6913.
36. Leite MT, Rao HL, Zangwill LM, et al. Comparison of the
diagnostic accuracies of the Spectralis, Cirrus, and RTVue
optical coherence tomography devices in glaucoma. Ophthal-
mology. 2011;118(7):1334e1339.
37. Oddone F, Centofanti M, Tanga L, et al. Influence of disc size on
optic nerve head versus retinal nerve fiber layer assessment for
diagnosing glaucoma. Ophthalmology. 2011;118(7):1340e1347.
2
Department of Ophthalmology, Seoul National University Hospital,
Seoul, Korea.
3
Department of Ophthalmology, Nowon Eulji Medical Center, Eulji Uni-
versity, Seoul, Korea.
*Both authors contributed equally as first authors.
9
 Ophthalmology Volume -, Number -, Month 2018
Financial Disclosure(s):
The author(s) have no proprietary or commercial interest in any materials
discussed in this article.
HUMAN SUBJECTS: This study included human subjects or tissues. No
animals were used in this study. Study protocol was approved by the
institutional review board of the Seoul National University Hospital.
Informed consent was obtained from all human subjects. This research
adhered to the tenets of the Declaration of Helsinki.
Author Contributions:
Conception and design: Baek, K.E.Kim, Y.K.Kim, Park, Jeoung
Analysis and interpretation: Baek, K.E.Kim, Y.K.Kim, Park, Jeoung
Data collection: Baek, K.E.Kim, Y.K.Kim, Park, Jeoung
Obtained funding: none
10
Overall responsibility: Baek, K.E.Kim, Jeoung
Abbreviations and Acronyms:
AUC ¼ area under the receiver operating characteristic curve;
CI ¼ confidence interval; D ¼ diopter; GCIPL ¼ ganglion celleinner
plexiform layer; ICC ¼ intraclass correlation coefficient; PLR ¼ positive
likelihood ratio; SD ¼ spectral-domain; RNFL ¼ retinal nerve fiber layer;
SE ¼ spherical equivalent.
Correspondence:
Jin Wook Jeoung, MD, PhD, Department of Ophthalmology, Seoul Na-
tional University Hospital, Seoul National University College of Medicine,
101 Daehak-ro, Jongno-gu, Seoul 03080, Korea. E-mail: neuroprotect@
gmail.com.