- Email: [email protected]
Automated Detection of Hemifield Difference across Horizontal Raphe on Ganglion Cell–Inner Plexiform Layer Thickness Map
Automated Detection of Hemifield Difference across Horizontal Raphe on Ganglion CelleInner Plexiform Layer Thickness Map Young Kook Kim, MD,1 Byeong Wook Yoo, BE,2 Hee Chan Kim, PhD,3 Ki Ho Park, MD, PhD1 Purpose: A MATLAB-based (The MathWorks, Inc, Natick, MA) computer program (the ganglion cell-inner plexiform layer [GCIPL] hemifield test) for automated detection of GCIPL thickness difference across the horizontal raphe was developed, and its glaucoma diagnostic performance was assessed. Design: Cross-sectional study. Participants: A total of 65 eyes of normal, healthy subjects along with 162 eyes of patients with glaucoma (79 preperimetric and 83 early perimetric). Methods: Cirrus high-definition optical coherence tomography (HD-OCT) (Carl Zeiss Meditec, Dublin, CA) was used to scan all of the subjects’ macular and optic discs. A positive (i.e., “outside normal limits”) GCIPL hemifield test result was declared if the following 3 conditions were all met: (1) The reference line (a horizontal line dividing the superior and inferior hemifields) is continuously detected for longer than one-half of the distance from the temporal inner elliptical annulus to the outer elliptical annulus; (2) the average GCIPL thickness difference within 10 pixels of the reference line, both above and below, is 5 mm; and (3) the average RGB color ranges of the 10 pixels above and below the reference line display blue in 1 hemifield and red/yellow/white in the other hemifield. Main Outcome Measures: Comparison of diagnostic ability using the areas under the receiver operating characteristic curves (AUCs). Results: A positive GCIPL hemifield test result was observed more frequently in the glaucomatous eyes (74/79 preperimetric, 78/83 early perimetric) than in the normal eyes (1/65). In the preperimetric group, the AUC of the GCIPL hemifield test (0.967; sensitivity 94.94%, specificity 98.46%) was greater than that of the minimum GCIPL thickness (0.933), the inferotemporal GCIPL thickness (0.907), and the average GCIPL thickness (0.899) (P ¼ 0.09, 0.06, and 0.03, respectively). In the early perimetric group, the AUC of the GCIPL hemifield test (0.962; sensitivity 93.98%, specificity 96.46%) was greater than that of the inferotemporal GCIPL thickness (0.938), the minimum GCIPL thickness (0.919), and the average GCIPL thickness (0.912) (P ¼ 0.38, 0.17, and 0.11, respectively). Conclusions: For discrimination of early glaucomatous structural loss, most notably in preperimetric glaucoma cases, identification of the GCIPL thickness difference across the horizontal raphe was effective. Ophthalmology 2015;122:2252-2260 ª 2015 by the American Academy of Ophthalmology.
Glaucomatous structural and functional changes require timely evaluation and detection to prevent possible blindness. Because glaucoma primarily affects the retinal ganglion cells and their axons, measurement of the thickness of ganglion cell layers or retinal nerve fiber layers (RNFLs) has proved an effective glaucoma detection strategy.1e5 Ganglion cell analysis (GCA) is a new Cirrus highdefinition optical coherence tomography (HD-OCT) (Carl Zeiss Meditec, Dublin, CA) algorithm developed for macular ganglion cell-inner plexiform layer (GCIPL) thickness measurement.6 The ability of the macular GCIPL parameters to discriminate between normal eyes and eyes with glaucoma was reported to be comparable to the circumpapillary RNFL and optic nerve head parameters.5,7,8 However, GCIPL thickness, because of the large value range overlap between normal eyes and eyes with early glaucoma,8e13 is
2252
2015 by the American Academy of Ophthalmology Published by Elsevier Inc.
known to be less sensitive and less specific than optic disc or circumpapillary RNFL thickness.10e12,14 For well-known anatomic reasons, glaucomatous visual field loss most commonly is not symmetric across the horizontal meridian.15e17 Thus, hemifield comparative techniques, for example, the glaucoma hemifield test, often have been used for the detection of glaucomatous change.16e18 Likewise, structural glaucomatous loss is often asymmetric, especially in the early stages of glaucoma.9,19e21 Correspondingly, a useful approach is to observe, on a Cirrus HD-OCT GCIPL thickness map, the step-like configuration of the GCIPL thickness near the horizontal raphe. However, as a clinical glaucoma diagnostics strategy, this has generated little interest, probably because of the absence of objective configuration evaluation standards.
http://dx.doi.org/10.1016/j.ophtha.2015.07.013 ISSN 0161-6420/15
Kim et al
GCIPL Hemifield Difference in Early Glaucoma
This study was undertaken to develop a new algorithm for automated evaluation of a single macular GCIPL thickness map and detection of hemifield difference across the horizontal raphe for identification of early glaucoma.
Methods This study was approved by the Seoul National University Hospital Institutional Review Board and adhered to the tenets of the Declaration of Helsinki.
Subjects All of the study subjects were examined between December 2012 and January 2014 at the Seoul National University Hospital Glaucoma Clinic in Seoul, Korea. Eligible participants were consecutively enrolled on the basis of a retrospective medical record review. All underwent a complete ophthalmic examination, including visual acuity assessment, refraction, slit-lamp biomicroscopy, gonioscopy, Goldmann applanation tonometry (HaagStreit, Koniz, Switzerland), and dilated stereoscopic examination of the optic disc. They also underwent central corneal thickness measurement (Orbscan 73 II, Bausch & Lomb Surgical, Rochester, NY), axial length measurement (IOLMaster ver. 5, Carl Zeiss Meditec, Dublin, CA), red-free RNFL photography, optic nerve head and macular imaging by Cirrus HD-OCT (Carl Zeiss Meditec), and a central 30-2 threshold test of the Humphrey Visual Field (HFA II; Humphrey Instruments Inc, Dublin, CA). For inclusion, subjects were required to have a best-corrected visual acuity 20/40 in the Snellen equivalent, spherical refraction >6 diopters (D) and <3 D, a normal open anterior chamber angle, and reliable visual field tests. Individuals were excluded from further analysis on the basis of the following criteria: (1) the existence of a secondary cause of glaucomatous optic neuropathy; (2) a history of intraocular surgery (except cataract surgery) or retinal laser photocoagulation; and (3) any neurologic and systemic diseases that could affect retina and visual field results. One eye was randomly selected if both eyes were found to be eligible. Glaucomatous eyes were defined by the presence of characteristic optic disc (localized or diffuse neuroretinal rim thinning) on stereo disc photograph or the presence of RNFL defect on redfree fundus imaging, regardless of the presence or absence of glaucomatous visual field defects. Eyes with glaucomatous visual field defects were defined as (1) a cluster of 3 points with probabilities <5% in at least 1 hemifield on the pattern deviation map, including at least 1 point with a probability <1% or a cluster of 2 points with a probability <1%; (2) glaucomatous hemifield test results outside of the normal limits; or (3) a pattern standard deviation beyond 95% of the normal limits, as confirmed by at least 2 reliable examinations (false-positive/negatives <15%, fixation losses <15%). The healthy controls had an intraocular pressure (IOP) 21 mm Hg, no IOP elevation history, no glaucomatous optic disc appearance, no RNFL defects, and normal visual field test results. On the basis of visual field test results, patients with glaucoma were divided into 2 groups: preperimetric glaucoma (normal visual field) and early perimetric glaucoma (visual field mean deviation [MD] >6 decibels [dB]). The appearance of the optic disc on stereoscopic photographs and that of RNFL on red-free imaging were evaluated by 2 glaucoma specialists (Y.K.K. and K.H.P.) who were masked to all other information on the eyes. Discrepancies between the 2 observers’ findings were resolved by consensus.
Cirrus High-Definition Optical Coherence Tomography Optic Disc Cube Optic disc (Optic Disc Cube 200200 protocol) and macular scans (Macular Cube 512128 protocol) using Cirrus HD-OCT instrument software (version 6.0) for RNFL and GCIPL thickness measurements, respectively, were carried out. Poor-quality images showing eye motion, blinking artifacts, or poor centration were discarded by the examiner, and those with a signal strength <6 were excluded from the study. The circumpapillary RNFL thicknesses were measured overall, in each of the 4 quadrants, and in each of the 12 clock-hour sectors. The average, minimum, and 6 sectoral (superotemporal, superior, superonasal, inferonasal, inferior, inferotemporal) GCIPL thicknesses in an elliptical annulus were measured in the macular cube scan mode.5,22
Ganglion Cell-Inner Plexiform Layer Hemifield Test The GCIPL thickness maps were processed using the GCIPL Hemifield Test (Medical Electronics Lab, Seoul National University, Seoul, Korea), a customized software in MATLAB (2013a version, The MathWorks, Inc, Natick, MA). The GCIPL Hemifield Test automatically extracted, from the GCIPL thickness map, a 32bit color-scale image of an elliptical annulus of 2.0 mm vertical outer radius and 2.4 mm horizontal outer radius. Then, automated image processing for line detection was conducted. The GCIPL Hemifield Test uses the Hough transform algorithm to detect lines on a GCIPL thickness map. Hough transform is a widely used technique in the image processing field for detection of lines according to their parametric representation. In general, to implement the Hough transform, the threshold is selected heuristically. Setting the threshold too low or too high can give rise to false positivity or mis-detection, respectively, for a given shape. According to the threshold, it was determined whether the small gaps in line segments were automatically filled or not. Subsequently, the end points of the line segments corresponding to the peaks in the Hough transform were found. Subsequent image processing for detection of color values was conducted only in cases in which horizontal reference lines longer than one-half the distance from the temporal inner elliptical annulus to the outer elliptical annulus were successfully detected. The red, green, and blue (RGB) color values of the pixels above and below the detected line dividing the superior and inferior hemifields were discriminated. They were converted to GCIPL thicknesses on the basis of the GCIPL thickness map’s reference color bar. Ultimately, a positive (i.e., “outside normal limits”) GCIPL Hemifield Test result was declared if the following 3 conditions were all met: (1) The reference line (a horizontal line dividing the superior and inferior hemifields) is continuously detected for longer than one half of the distance from the temporal inner elliptical annulus to the outer elliptical annulus; (2) the average GCIPL thickness difference within 10 pixels of the reference line, both above and below, is 5 mm; and (3) the average RGB color ranges of the 10 pixels above and below the reference line display blue in 1 hemifield and red/yellow/white in the other hemifield (Figs 1 and 2).
Optimal Cutoff Values for Ganglion Cell-Inner Plexiform Layer Hemifield Test The optimal cutoff values for classifying cases as positive or negative were confirmed by using the areas under the receiver operating characteristic curves (AUCs) as the following process. First, the probable diagnostic criteria for discriminating early glaucoma was postulated that calculation of the average GCIPL thickness difference is within approximately 10 pixels both above and below the reference line and the baseline of the difference is 7 mm. Thus, the
2253
Ophthalmology Volume 122, Number 11, November 2015
Figure 1. Schematic representation of ganglion cell-inner plexiform layer (GCIPL) hemifield test for automated detection of hemifield difference across horizontal raphe on Cirrus high-definition optical coherence tomography (HD-OCT; Carl Zeiss Meditec, Dublin, CA) GCIPL thickness map. A, The GCIPL hemifield test automatically extracted, from a GCIPL thickness map, a 32-bit color-scale image of an elliptical annulus of 2.0 mm vertical outer radius and 2.4 mm horizontal radius. B, C, The reference line (red dashed line), running from the temporal inner elliptical annulus to the outer elliptical annulus and dividing the superior and inferior hemifields, is detected using a computer program. C, The average color of the RGB values in pixels included in 10 pixels both up and down (black dashed line), as set on the basis of the reference line, is calculated. Those values were automatically converted to GCIPL thicknesses on the basis of the GCIPL thickness map’s reference color bar. D, The Glaucoma Hemifield Test on a right eye showed a positive result, because the following 3 conditions all met together: (1) The reference line (blue line; a horizontal line dividing the superior and inferior hemifields) is continuously detected for more than one-half of the distance from the temporal inner elliptical annulus to the outer elliptical annulus; (2) the average GCIPL thickness difference within 10 pixels of the reference line, both above and below, is 5 mm; and (3) the average RGB color ranges of the 10 pixels above and below the reference line display blue in 1 hemifield and red/yellow/white in the other hemifield. OD ¼ right eye; OS ¼ left eye.
2254
Kim et al
GCIPL Hemifield Difference in Early Glaucoma
Figure 2. Examples of circumpapillary retinal nerve fiber layer (RNFL) and macular ganglion cell analysis (GCA) maps as measured by high-definition optical coherence tomography (HD-OCT) for a healthy eye (first row), a preperimetric glaucoma eye (second row), and an early-perimetric glaucoma eye (third row). The temporal and nasal borders of the RNFL defects are indicated by the black arrows on the RNFL thickness map or deviation map. In the preperimetric glaucoma eye, although there was no defect on the RNFL deviation map, a hemifield difference across the horizontal raphe (indicated by the blue arrow) was detected by the ganglion cell-inner plexiform layer (GCIPL) hemifield test. A hemifield difference across the horizontal raphe (indicated by the blue arrow) also was detected in the early perimetric glaucoma eye. OD ¼ right eye; OS ¼ left eye.
5 candidate values (4, 7, 10, 13, and 16 pixels) were taken to determine the optimal cutoff values for the up-down ranges of pixel numbers to be analyzed. Likewise, 5 candidate values (3, 5, 7, 9, and 11 mm) were chosen to obtain the optimal cutoff values for the minimum up-down difference of GCIPL thickness to be applied in diagnosing glaucoma. Then, the AUCs, calculated by using 25 matched pairs of cutoff scores, were compared among a total of 91 subjects (30 normal, 31 preperimetric, and 30 early-perimetric glaucoma) randomly sampled from previously aggregated clinical data and who were not associated with the current study. The high-ranked AUC values of the paired cutoff scores are listed in Table 1. Through this process, finally, the optimal cutoff values of the GCIPL Hemifield Test were determined as an up-down range of 10 pixels, and the difference of the minimum GCIPL thickness as 5 mm. On the basis of the selected optimal cutoff values, the GCIPL
Hemifield Test AUC was 0.967 (sensitivity 96.77%, specificity 96.67%) for discriminating glaucomatous changes between normal and preperimetric cases and 0.950 (sensitivity 93.33%, specificity 96.67%) for discriminating glaucomatous changes between normal and early perimetric cases.
Statistical Analysis Between the healthy and glaucoma groups, age and spherical equivalent of refractive error differences were obtained using analysis of variance tests; sex differences were computed using chisquare tests. The glaucoma diagnostic abilities were compared on the basis of computed AUC values. The sensitivities and specificities were calculated according to the optimal cutoff point, which was set as the maximum of the Youden index (obtained as J¼max
Table 1. Discriminating Ability of Ganglion Cell-Inner Plexiform Layer Hemifield Test in Different Matched Pairs of Cutoff Values (95% Confidence Interval) Cutoff Values* I (Pixels) 10 10 7 13 7
Preperimetric Glaucoma
II (mm) 5 7 5 5 7
AUCy 0.967 0.934 0.918 0.934 0.902
(0.89e1.00) (0.84e0.98) (0.82e0.97) (0.84e0.98) (0.80e0.96)
Sensitivity (%) 96.77 93.55 93.55 93.55 90.32
(83.3e99.9) (78.6e99.2) (78.6e99.2) (78.6e99.2) (74.2e98.0)
Early Perimetric Glaucoma AUCy
Specificity (%) 96.67 93.33 90.00 93.33 90.00
(82.8e99.9) (77.9e99.2) (73.5e97.9) (77.9e99.2) (73.5e97.9)
0.950 0.933 0.917 0.883 0.900
(0.86e0.99) (0.84e0.98) (0.82e0.97) (0.77e0.95) (0.80e0.96)
Sensitivity (%) 93.33 93.33 93.33 83.33 90.00
(77.9e99.2) (77.9e99.2) (77.9e99.2) (65.3e94.4) (73.9e97.9)
Specificity (%) 96.67 93.33 90.00 93.33 90.00
(82.8e99.9) (77.9e99.2) (73.5e97.9) (77.9e99.2) (73.9e97.9)
AUC ¼ area under the receiver operating characteristic curve. I, The cutoff values for up-down ranges of pixel numbers to be analyzed. II, The cutoff values for minimum up-down difference of GCIPL thickness to be applied in diagnosing glaucoma. *The paired cutoff values showing the high-ranked AUC values. y Evaluated in a total of 91 subjects (30 normal, 31 preperimetric, and 30 early perimetric glaucoma) randomly sampled from previously aggregated clinical data who are not associated with the current study.
2255
Ophthalmology Volume 122, Number 11, November 2015 Table 2. Clinical Characteristics of Study Participants
Age (yrs, average SD) Sex (M/F) Spherical equivalent (D, average SD) IOP (mmHg) CCT (mm, average SD) Axial length (mm, average SD) VF MD (dB, average SD) VF PSD (dB, average SD) OCT signal strength (average SD)
Healthy (n [ 65)
Preperimetric Glaucoma (n [ 79)
Early Perimetric Glaucoma (n [ 83)
P Value
53.59.3 33/32 0.762.36 13.52.5 545.435.3 24.71.98 0.051.27 1.790.55 7.90.7
54.611.8 39/40 0.922.24 13.32.1 546.133.9 24.31.72 0.721.92 1.880.75 8.00.6
57.511.6 40/43 0.792.15 12.72.2 536.932.1 24.81.80 2.391.65 5.193.37 7.80.8
0.072* 0.757y 0.897* 0.076* 0.161* 0.192* <0.001* <0.001* 0.200*
CCT ¼ central corneal thickness; D ¼ diopters; dB ¼ decibels; F ¼ female; IOP ¼ intraocular pressure; M ¼ male; MD ¼ mean deviation; OCT ¼ optical coherence tomography; PSD ¼ pattern standard deviation; SD ¼ standard deviation; VF ¼ visual field. *Analysis of variance test. y Chi-square test.
[sensitivity þ specificity 1]). Statistical analyses were performed using SPSS software (version 19, SPSS, Inc, Chicago, IL). The AUC comparisons were achieved with statistical software (MedCalc14.12; MedCalc Software, Mariakerke, Belgium). P values less than 0.05 were considered statistically significant. The data ranges were recorded as mean standard deviations.
Results Subjects A total of 239 eyes, representing 239 subjects, were initially included in the study. Of these, 12 patients were excluded: (1) Six subjects had poor-quality stereoscopic disc photography or red-free RNFL photography, and (2) 6 subjects had poor-quality scanned spectral-domain optical coherence tomography (signal strength <6). A total of 227 remaining eyes representing 162 patients with glaucoma and 65 healthy subjects were further examined in the ensuing analysis. The sample distribution was as follows: 79 preperimetric glaucoma eyes, 83 early perimetric glaucoma eyes, and 65 healthy eyes. The study population’s clinical characteristics are shown in Table 2. The age of those with preperimetric glaucoma ranged from 42 to 82 years (54.611.8 years, n ¼ 79). The spherical equivalent of the refractive error was 0.922.24, and the MD ranged from 1.34 to 2.87 dB (0.721.92 dB). Among the early perimetric glaucoma eye group (n ¼ 83), the age range on the first visit was 57.511.6 years, the mean spherical equivalent of the refractive error was 0.792.15 D, and the MD ranged from 4.56 to 1.54 (2.391.65). A total of 69 of the 83 eyes (83.1%) showed a baseline IOP 21 mmHg, and 14 of the 83 eyes (16.9%) showed a baseline IOP >21 mmHg. In the group of healthy subjects (n ¼ 65), the age range was 53.59.3 years. The spherical equivalent of the refractive error was 0.762.36, and the MD ranged from 0.85 to 2.02 dB (0.051.27 dB). The differences in sex, age, refractive error, central corneal thickness, and axial length among the study groups were not statistically significant.
Cirrus High-Definition Optical Coherence Tomography Index Table 3 shows that in the normal eyes, the average GCIPL and circumpapillary RNFL thicknesses (81.27.5 mm and 93.79.1 mm, respectively) were significantly greater than in eyes with
2256
preperimetric glaucoma (78.95.5 mm and 88.212.7 mm, respectively) or early perimetric glaucoma (72.96.5 mm and 73.811.6 mm, respectively) (all P < 0.001). The absolute value of GCIPL thickness differences between superotemporal and inferotemporal sectors was significantly lower in the normal eyes (0.80.3 mm) than in the eyes with preperimetric glaucoma (8.64.2 mm) or early perimetric glaucoma (12.24.8 mm) (P < 0.001). The minimum GCIPL thickness and inferotemporal GCIPL thickness were significantly greater in the normal eyes (77.910.3 mm and 80.49.7 mm, respectively) than in the eyes with preperimetric glaucoma (67.37.3 mm and 70.65.3 mm, respectively) or early perimetric glaucoma (60.98.4 mm and 64.75.7 mm, respectively) (all P < 0.001).
Results of Ganglion Cell-Inner Plexiform Layer Hemifield Test Ganglion cell-inner plexiform layer Hemifield Test positivity, interpreted as “outside normal limits,” was shown in 93.7% (74/79) of the preperimetric glaucoma eyes and in 94.0% (78/83) of the early perimetric glaucoma eyes. By contrast, only 1 of 65 healthy eyes (1.5%) showed positivity.
Diagnostic Ability of Ganglion Cell-Inner Plexiform Layer Hemifield Test The GCIPL Hemifield Test AUC for discrimination of glaucomatous changes between normal and preperimetric cases (0.967; sensitivity 94.94%, specificity 98.46%) was significantly higher than that of the minimum GCIPL thickness (0.933; sensitivity 92.43%, specificity 85.15%), the GCIPL thickness difference between superotemporal and inferotemporal sectors (0.911; sensitivity 87.34%, specificity 84.62%), the inferotemporal GCIPL thickness (0.907; sensitivity 90.14%, specificity 88.74%), and the average GCIPL thickness (0.899; sensitivity 79.75%, specificity 84.53%) (P ¼ 0.09, 0.03, 0.06, and 0.03, respectively). In the early perimetric glaucoma group, the AUC of the GCIPL Hemifield Test (0.962; sensitivity 93.98%, specificity 96.46%) was greater than that of the inferotemporal GCIPL thickness (0.938; sensitivity 92.77%, specificity 89.23%), the minimum GCIPL thickness (0.919; sensitivity 92.77%, specificity 86.13%), the GCIPL thickness difference between the superotemporal and inferotemporal sectors (0.916; sensitivity 82.34%, specificity 84.62%), and the average GCIPL thickness (0.912; sensitivity 86.75%, specificity 86.15%) (P ¼ 0.38, 0.17, 0.06, and 0.11, respectively) (Tables 4, 5, and Fig 3).
Kim et al
GCIPL Hemifield Difference in Early Glaucoma
Table 3. Comparison of Optical Coherence Tomography Parameters as Determined by Cirrus High-Definition Optical Coherence Tomography
GCIPL Hemifield Test positivity* (%) cpRNFL thickness (mm) Average Superior Nasal Inferior Temporal GCIPL thickness (mm) Average Minimum Differencey Superotemporal Superior Superonasal Inferonasal Inferior Inferotemporal
Healthy (n [ 65)
Preperimetric Glaucoma (n [ 79)
Early Perimetric Glaucoma (n [ 83)
P Valuez
1.5%
93.7%
94.0%
<0.001
93.79.1 119.910.4 64.78.2 118.69.9 71.86.7
88.212.7 107.716.3 71.211.5 99.713.9 74.110.9
73.811.6 93.211.5 57.49.9 85.311.5 60.29.3
<0.001 <0.001 <0.001 <0.001 <0.001
81.27.5 77.910.3 0.80.3 79.67.5 83.36.4 84.96.5 81.07.1 78.28.3 80.49.7
78.95.5 67.37.3 8.64.2 79.26.8 82.87.2 84.45.3 79.98.0 76.57.9 70.65.3
72.96.5 60.98.4 12.24.8 76.96.8 76.06.4 77.86.9 74.76.1 67.37.1 64.75.7
<0.001 <0.001 <0.001 0.037 <0.001 <0.001 <0.001 <0.001 <0.001
cpRNFL ¼ circumpapillary retinal nerve fiber layer; GCIPL ¼ ganglion cell-inner plexiform layer. Data are presented as mean standard deviation. *Interpreted as “outside normal limits.” y The absolute value of GCIPL thickness difference between superotemporal and inferotemporal sectors. z Analysis of variance test.
Discussion
difference across the horizontal raphe on GCIPL thickness maps. Although several maps have been used in single GCA interpretation, a more condensed and objective evaluation, which can be especially helpful for less-experienced clinicians, often is preferred. The GCIPL Hemifield Test satisfies
In this study, a good early glaucoma detection ability was demonstrated for the GCIPL Hemifield Test. This study has investigated the glaucoma diagnostic utility of hemifield
Table 4. Discriminating Ability of Ganglion Cell-Inner Plexiform Layer Hemifield Test and Cirrus High-Definition Optical Coherence Tomography Parameters for Glaucoma Detection (95% Confidence Interval) Preperimetric Glaucoma
Early Perimetric Glaucoma
Parameters
Sensitivity (%)
Specificity (%)
þLR
LR
Sensitivity (%)
Specificity (%)
þLR
LR
GCIPL hemifield test cpRNFL thickness Average Superior Nasal Inferior Temporal GCIPL thickness Average Minimum Difference* Superotemporal Superior Superonasal Inferonasal Inferior Inferotemporal
94.94 (87.5e98.6)
98.46 (91.7e100.0)
61.65
0.05
93.98 (86.5e98.0)
96.46 (90.7e99.2)
26.55
0.06
81.01 69.31 43.05 74.19 49.41
(70.6e89.0) (55.7e82.8) (33.7e62.2) (62.0e86.1) (37.5e69.8)
80.81 77.22 77.26 84.18 75.22
(70.5e90.1) (64.5e88.1) (64.5e90.0) (78.3e94.7) (61.3e84.5)
4.22 3.04 1.89 4.69 1.99
0.23 0.40 0.74 0.31 0.67
90.36 63.33 47.96 70.27 53.04
(81.9e95.7) (52.6e77.1) (35.2e68.6) (59.3e82.7) (39.6e72.2)
81.54 79.04 76.57 87.36 78.45
(70.0e90.1) (63.5e90.2) (64.4e88.9) (79.2e95.8) (64.0e88.3)
4.89 3.02 2.05 5.56 2.46
0.12 0.46 0.68 0.34 0.60
79.75 92.43 87.34 83.11 75.32 68.73 77.44 88.55 90.14
(69.2e88.0) (84.2e97.5) (78.0e93.8) (74.5e91.2) (64.5e85.9) (57.5e78.3) (62.7e84.6) (78.7e94.4) (81.6e95.4)
84.53 85.15 84.62 82.36 83.77 82.68 79.29 78.41 88.74
(74.3e93.5) (79.1e95.6) (73.5e92.4) (71.2e89.7) (73.3e89.1) (74.6e91.2) (69.3e88.6) (63.4e85.9) (78.0e95.7)
5.16 6.22 5.68 4.71 4.64 3.97 3.74 4.10 8.01
0.24 0.09 0.15 0.21 0.29 0.38 0.28 0.15 0.11
86.75 92.77 82.34 85.31 83.75 82.12 81.73 89.54 92.77
(77.5e93.2) (84.9e97.3) (74.7e91.4) (76.2e95.7) (74.5e90.2) (69.5e90.3) (70.1e88.9) (81.2e93.9) (84.9e97.3)
86.15 86.13 84.62 82.36 84.87 82.31 79.15 80.29 89.23
(75.3e93.5) (73.4e92.2) (73.5e92.4) (71.0e92.1) (73.7e92.1) (73.5e91.6) (69.0e88.1) (68.7e88.5) (79.1e95.6)
6.26 6.69 5.35 4.84 5.54 4.64 3.92 4.54 8.61
0.15 0.08 0.21 0.18 0.19 0.22 0.23 0.13 0.08
cpRNFL ¼ circumpapillary retinal nerve fiber layer; GCIPL ¼ ganglion cell-inner plexiform layer; þLR ¼ positive likelihood ratio; LR ¼ negative likelihood ratio. *The absolute value of GCIPL thickness difference between superotemporal and inferotemporal sectors.
2257
Ophthalmology Volume 122, Number 11, November 2015 Table 5. Areas Under the Receiver Operating Characteristic Curve of Ganglion Cell-Inner Plexiform Layer Hemifield Test and Cirrus High-Definition Optical Coherence Tomography Parameters (95% Confidence Interval) Preperimetric Glaucoma Parameters
AUC
GCIPL hemifield test cpRNFL thickness Average Superior Nasal Inferior Temporal GCIPL thickness Average Minimum Differencey Superotemporal Superior Superonasal Inferonasal Inferior Inferotemporal
0.967 (0.92e0.99)
Early Perimetric Glaucoma P*
AUC
P*
0.962 (0.92e0.99)
0.880 0.856 0.712 0.878 0.719
(0.83e0.94) (0.79e0.90) (0.64e0.76) (0.81e0.93) (0.64e0.77)
0.01 0.01 <0.001 0.01 0.01
0.907 0.872 0.713 0.898 0.735
(0.85e0.95) (0.79e0.93) (0.64e0.77) (0.81e0.94) (0.68e0.81)
0.07 0.03 <0.001 0.01 0.01
0.899 0.933 0.911 0.830 0.857 0.812 0.806 0.886 0.907
(0.84e0.94) (0.88e0.97) (0.85e0.95) (0.76e0.88) (0.78e0.91) (0.74e0.87) (0.73e0.87) (0.81e0.94) (0.85e0.95)
0.03 0.09 0.03 0.01 0.01 0.01 0.01 0.02 0.06
0.912 0.919 0.916 0.859 0.850 0.828 0.822 0.902 0.938
(0.86e0.95) (0.86e0.96) (0.86e0.96) (0.79e0.90) (0.78e0.93) (0.77e0.89) (0.75e0.89) (0.82e0.94) (0.89e0.97)
0.11 0.17 0.06 0.01 0.01 0.01 0.01 0.05 0.38
AUC ¼ area under the curve; cpRNFL ¼ circumpapillary retinal nerve fiber layer; GCIPL ¼ ganglion cell-inner plexiform layer. *Comparing between the AUC of the GCIPL Hemifield Test and each parameter. y The absolute value of GCIPL thickness difference between superotemporal and inferotemporal sectors.
those criteria: A brief GCIPL thickness evaluation is printed on field charts in plain text to distinguish glaucomatous structural loss on GCIPL thickness maps. In a previous study investigating the factors associated with the presence or absence of abnormal GCA findings,23 the GCA maps showed abnormal findings in healthy eyes, especially those with higher myopic refractive error. Therefore, high myopic eyes showing less than 6 D
spherical refraction were excluded from the present study. Another relevant previous finding was that for large angular distances between the fovea and the RNFL defect, GCA maps showed no abnormalities.23 In the present study, in 5 of 79 preperimetric and 5 of 83 early glaucoma eyes, the GCIPL Hemifield Test’s hemifield difference across the horizontal raphe was not identified. On retrospective review of the relevant RNFL photographs, the angular distance
Figure 3. The areas under the receiver operating characteristic curves (AUCs) in discriminating healthy eyes from (A) preperimetric glaucoma eyes and (B) early perimetric glaucoma eyes. P1 indicates the comparison of the AUC value between the ganglion cell-inner plexiform layer (GCIPL) Hemifield Test and the GCIPL thickness of inferotemporal sector; P2 indicates the comparison of the AUC value between the GCIPL Hemifield Test and the minimum GCIPL thickness; P3 indicates the comparison of the AUC value between the GCIPL Hemifield Test and the average GCIPL thickness; P4 indicates the comparison of the AUC value between the GCIPL Hemifield Test and the average circumpapillary retinal nerve fiber layer (RNFL) thickness; P5 indicates the comparison of the AUC value between the GCIPL Hemifield Test and the GCIPL thickness difference between superotemporal and IT sectors. IT ¼ inferotemporal; ST ¼ superotemporal.
2258
Kim et al
GCIPL Hemifield Difference in Early Glaucoma
between the fovea and the RNFL defect was greater in subjects who did not show any hemifield difference across the horizontal raphe. In cases in which no visual field defects are apparent in eyes with glaucoma, there might be significant macular ganglion cell layer and RNFL thinning.24 This indicates that even if abnormal RNFL and ganglion cell layer thinning are manifest in the area near the fovea, detectable visual field defects at the corresponding test points will not necessarily be evident.25 Therefore, macular structural parameters can be used as markers of subclinical (preperimetric) abnormalities of the ganglion cell layer and RNFL near the fovea.25 This is particularly important, because such subclinical abnormalities can, in time, morph into clinical (perimetric) abnormalities. Earlier studies have shown that macular retinal layer thickness asymmetry indices, especially those for the ganglion cell layer, are potential indicators of glaucomatous retinal damage.9 Furthermore, the overlap in values for normal and glaucomatous eyes, on the ganglion cell layer asymmetry index, has been minimal. In the current study, the AUC results of the GCIPL Hemifield Test were greater than those of the other GCIPL parameters in eyes with both preperimetric and early perimetric glaucoma. An earlier study showed that in eyes with preperimetric glaucoma, although both the RNFL and the macular ganglion cell layer were thinned, the ganglion cell layer was substantially thinner; also, the macular ganglion cell layer thinning relative to that of the RNFL was both more localized and abrupt.9 Of note, although preperimetric glaucoma with localized ganglion cell layer thinning does not affect average macular ganglion cell layer thickness, it might cause ganglion cell layer asymmetry and thus might indicate a hemifield difference across the horizontal raphe. For that, in this study, the GCIPL Hemifield Test’s higher diagnostic ability relative to average GCILP thickness, for preperimetric glaucoma, seems to be associated with macular glaucomatous anatomic changes. In this study, the values of AUCs for the GCIPL Hemifield Test were not much different between preperimetric glaucoma (0.967) and early perimetric glaucoma (0.962). These results suggest that preperimetric glaucoma and early perimetric glaucoma form a continuum and are not separate entities, and the former can progress to the latter over time; the structural differences between these 2 disease entities are minimal, the main difference being the absence or presence, respectively, of visual field defect. Furthermore, these facts might indicate that the GCIPL Hemifield Test is not suitable for discriminating between preperimetric and early perimetric glaucoma. To clarify the glaucoma diagnostic ability and the natural time course of GCIPL hemifield asymmetry, further examination of additional patients with ocular hypertension, suspected glaucoma, and moderate-to-severe stages of glaucoma will be necessary. In this study, all manual procedures were eliminated, in that hemifield difference was detected by an automated customized MATLAB-based program. Of note, the line found by the program’s “line-detecting function” was used as the reference line, thereby enabling more accurate diagnosis.
Study Limitations The GCIPL Hemifield Test program is intended for use in glaucoma diagnostics and it is ineffective for other types of diagnosis, including differential diagnosis. Because a considerable variety of other diseases also might result in significant hemifield differences, a positive GCIPL Hemifield Test result does not necessarily indicate GCIPL loss due to glaucoma. Certainly, the GCIPL Hemifield Test will have to be developed for better efficacy in the assessment of the presence or absence of glaucomatous structural loss and the diagnosis of glaucoma. Furthermore, because we studied a group of mostly normal-tension glaucomatous eyes (83.1% of patients with glaucoma had a baseline IOP 21 mmHg) and early-stage glaucomatous eyes (preperimetric and early perimetric), our results might not be applicable to other populations, such as patients with primary open-angle glaucoma or those with advanced glaucoma. Another limitation is that the RGB color values of the pixels were converted to GCIPL thicknesses on the basis of the GCIPL thickness map’s reference color bar. Thus, some thickness bias might have been incurred. In conclusion, automated detection of hemifield difference across the horizontal raphe on a GCIPL thickness map, providing easily interpretable test results in a short, plaintext display, was proved to be useful for discrimination of early glaucomatous structural loss, most notably in preperimetric glaucoma cases. In the future, this type of GCA results presentation scheme will be more frequently used once expert systems are applied to the interpretation of GCA data.
References 1. Zeimer R, Asrani S, Zou S, et al. Quantitative detection of glaucomatous damage at the posterior pole by retinal thickness mapping. A pilot study. Ophthalmology 1998;105:224–31. 2. Brusini P, Tosoni C, Miani F. Quantitative mapping of the retinal thickness at the posterior pole in chronic open angle glaucoma. Acta Ophthalmol Scand Suppl 2000:42–4. 3. Nouri-Mahdavi K, Nowroozizadeh S, Nassiri N, et al. Macular ganglion cell/inner plexiform layer measurements by spectral domain optical coherence tomography for detection of early glaucoma and comparison to retinal nerve fiber layer measurements. Am J Ophthalmol 2013;156:1297–307. 4. Jeoung JW, Choi YJ, Park KH, Kim DM. Macular ganglion cell imaging study: glaucoma diagnostic accuracy of spectraldomain optical coherence tomography. Invest Ophthalmol Vis Sci 2013;54:4422–9. 5. Mwanza JC, Oakley JD, Budenz DL, et al. Macular ganglion cell-inner plexiform layer: automated detection and thickness reproducibility with spectral domain-optical coherence tomography in glaucoma. Invest Ophthalmol Vis Sci 2011;52: 8323–9. 6. Tan O, Chopra V, Lu AT, et al. Detection of macular ganglion cell loss in glaucoma by Fourier-domain optical coherence tomography. Ophthalmology 2009;116:2305–14. 7. Shin HY, Park HY, Jung KI, et al. Glaucoma diagnostic ability of ganglion cell-inner plexiform layer thickness differs according to the location of visual field loss. Ophthalmology 2014;121:93–9.
2259
Ophthalmology Volume 122, Number 11, November 2015 8. Mwanza JC, Durbin MK, Budenz DL, et al. Glaucoma diagnostic accuracy of ganglion cell-inner plexiform layer thickness: comparison with nerve fiber layer and optic nerve head. Ophthalmology 2012;119:1151–8. 9. Yamada H, Hangai M, Nakano N, et al. Asymmetry analysis of macular inner retinal layers for glaucoma diagnosis. Am J Ophthalmol 2014;158:1318–29. 10. Lisboa R, Paranhos A Jr, Weinreb RN, et al. Comparison of different spectral domain OCT scanning protocols for diagnosing preperimetric glaucoma. Invest Ophthalmol Vis Sci 2013;54:3417–25. 11. Morooka S, Hangai M, Nukada M, et al. Wide 3-dimensional macular ganglion cell complex imaging with spectral-domain optical coherence tomography in glaucoma. Invest Ophthalmol Vis Sci 2012;53:4805–12. 12. Kotera Y, Hangai M, Hirose F, et al. Three-dimensional imaging of macular inner structures in glaucoma by using spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci 2011;52:1412–21. 13. Seong M, Sung KR, Choi EH, et al. Macular and peripapillary retinal nerve fiber layer measurements by spectral domain optical coherence tomography in normal-tension glaucoma. Invest Ophthalmol Vis Sci 2010;51:1446–52. 14. Arintawati P, Sone T, Akita T, et al. The applicability of ganglion cell complex parameters determined from SD-OCT images to detect glaucomatous eyes. J Glaucoma 2013;22:713–8. 15. Katz J, Sommer A. Asymmetry and variation in the normal hill of vision. Arch Ophthalmol 1986;104:65–8. 16. Duggan C, Sommer A, Auer C, Burkhard K. Automated differential threshold perimetry for detecting glaucomatous visual field loss. Am J Ophthalmol 1985;100:420–3.
17. Asman P, Heijl A. Glaucoma Hemifield Test. Automated visual field evaluation. Arch Ophthalmol 1992;110:812–9. 18. Sommer A, Enger C, Witt K. Screening for glaucomatous visual field loss with automated threshold perimetry. Am J Ophthalmol 1987;103:681–4. 19. Um TW, Sung KR, Wollstein G, et al. Asymmetry in hemifield macular thickness as an early indicator of glaucomatous change. Invest Ophthalmol Vis Sci 2012;53:1139–44. 20. Seo JH, Kim TW, Weinreb RN, et al. Detection of localized retinal nerve fiber layer defects with posterior pole asymmetry analysis of spectral domain optical coherence tomography. Invest Ophthalmol Vis Sci 2012;53: 4347–53. 21. Asrani S, Rosdahl JA, Allingham RR. Novel software strategy for glaucoma diagnosis: asymmetry analysis of retinal thickness. Arch Ophthalmol 2011;129:1205–11. 22. Mwanza JC, Durbin MK, Budenz DL, et al. Profile and predictors of normal ganglion cell-inner plexiform layer thickness measured with frequency-domain optical coherence tomography. Invest Ophthalmol Vis Sci 2011;52: 7872–9. 23. Hwang YH, Jeong YC, Kim HK, Sohn YH. Macular ganglion cell analysis for early detection of glaucoma. Ophthalmology 2014;121:1508–15. 24. Nakano N, Hangai M, Nakanishi H, et al. Macular ganglion cell layer imaging in preperimetric glaucoma with speckle noise-reduced spectral domain optical coherence tomography. Ophthalmology 2011;118:2414–26. 25. Kimura Y, Hangai M, Matsumoto A, et al. Macular structure parameters as an automated indicator of paracentral scotoma in early glaucoma. Am J Ophthalmol 2013;156:907–17.
Footnotes and Financial Disclosures Originally received: March 2, 2015. Final revision: June 30, 2015. Accepted: July 10, 2015. Available online: August 13, 2015.
Author Contributions:
Manuscript no. 2015-335.
1
Department of Ophthalmology, Seoul National University Hospital, Seoul National University, College of Medicine, Seoul, Korea.
2
Bioengineering Major, Graduate School, Seoul National University, Seoul, Korea.
3
Department of Biomedical Engineering, College of Medicine and Institute of Medical and Biological Engineering, Medical Research Center, Seoul National University, Seoul, Korea. Financial Disclosure(s): The author(s) have no proprietary or commercial interest in any materials discussed in this article.
2260
Conception and design: Y.K. Kim, Yoo, H.C. Kim, Park Data collection: Y.K. Kim, Yoo, H.C. Kim, Park Analysis and interpretation: Y.K. Kim, Yoo, H.C. Kim, Park Obtained funding: Not applicable Overall responsibility: Y.K. Kim, Park Abbreviations and Acronyms: AUC ¼ area under the receiver operating characteristic curve; D ¼ diopter; dB ¼ decibels; GCA ¼ ganglion cell analysis; GCIPL ¼ ganglion cellinner plexiform layer; HD-OCT ¼ high-definition optical coherence tomography; IOP ¼ intraocular pressure; MD ¼ mean deviation; RGB ¼ red, green, and blue; RNFL ¼ retinal nerve fiber layer. Correspondence: Ki Ho Park, MD, PhD, Department of Ophthalmology, Seoul National University Hospital, Seoul National University College of Medicine, 101 Daehak-ro, Chongno-gu, Seoul 110-744, Republic of Korea. E-mail: [email protected].
Glaucomatous structural and functional changes require timely evaluation and detection to prevent possible blindness. Because glaucoma primarily affects the retinal ganglion cells and their axons, measurement of the thickness of ganglion cell layers or retinal nerve fiber layers (RNFLs) has proved an effective glaucoma detection strategy.1e5 Ganglion cell analysis (GCA) is a new Cirrus highdefinition optical coherence tomography (HD-OCT) (Carl Zeiss Meditec, Dublin, CA) algorithm developed for macular ganglion cell-inner plexiform layer (GCIPL) thickness measurement.6 The ability of the macular GCIPL parameters to discriminate between normal eyes and eyes with glaucoma was reported to be comparable to the circumpapillary RNFL and optic nerve head parameters.5,7,8 However, GCIPL thickness, because of the large value range overlap between normal eyes and eyes with early glaucoma,8e13 is
2252
2015 by the American Academy of Ophthalmology Published by Elsevier Inc.
known to be less sensitive and less specific than optic disc or circumpapillary RNFL thickness.10e12,14 For well-known anatomic reasons, glaucomatous visual field loss most commonly is not symmetric across the horizontal meridian.15e17 Thus, hemifield comparative techniques, for example, the glaucoma hemifield test, often have been used for the detection of glaucomatous change.16e18 Likewise, structural glaucomatous loss is often asymmetric, especially in the early stages of glaucoma.9,19e21 Correspondingly, a useful approach is to observe, on a Cirrus HD-OCT GCIPL thickness map, the step-like configuration of the GCIPL thickness near the horizontal raphe. However, as a clinical glaucoma diagnostics strategy, this has generated little interest, probably because of the absence of objective configuration evaluation standards.
http://dx.doi.org/10.1016/j.ophtha.2015.07.013 ISSN 0161-6420/15
Kim et al
GCIPL Hemifield Difference in Early Glaucoma
This study was undertaken to develop a new algorithm for automated evaluation of a single macular GCIPL thickness map and detection of hemifield difference across the horizontal raphe for identification of early glaucoma.
Methods This study was approved by the Seoul National University Hospital Institutional Review Board and adhered to the tenets of the Declaration of Helsinki.
Subjects All of the study subjects were examined between December 2012 and January 2014 at the Seoul National University Hospital Glaucoma Clinic in Seoul, Korea. Eligible participants were consecutively enrolled on the basis of a retrospective medical record review. All underwent a complete ophthalmic examination, including visual acuity assessment, refraction, slit-lamp biomicroscopy, gonioscopy, Goldmann applanation tonometry (HaagStreit, Koniz, Switzerland), and dilated stereoscopic examination of the optic disc. They also underwent central corneal thickness measurement (Orbscan 73 II, Bausch & Lomb Surgical, Rochester, NY), axial length measurement (IOLMaster ver. 5, Carl Zeiss Meditec, Dublin, CA), red-free RNFL photography, optic nerve head and macular imaging by Cirrus HD-OCT (Carl Zeiss Meditec), and a central 30-2 threshold test of the Humphrey Visual Field (HFA II; Humphrey Instruments Inc, Dublin, CA). For inclusion, subjects were required to have a best-corrected visual acuity 20/40 in the Snellen equivalent, spherical refraction >6 diopters (D) and <3 D, a normal open anterior chamber angle, and reliable visual field tests. Individuals were excluded from further analysis on the basis of the following criteria: (1) the existence of a secondary cause of glaucomatous optic neuropathy; (2) a history of intraocular surgery (except cataract surgery) or retinal laser photocoagulation; and (3) any neurologic and systemic diseases that could affect retina and visual field results. One eye was randomly selected if both eyes were found to be eligible. Glaucomatous eyes were defined by the presence of characteristic optic disc (localized or diffuse neuroretinal rim thinning) on stereo disc photograph or the presence of RNFL defect on redfree fundus imaging, regardless of the presence or absence of glaucomatous visual field defects. Eyes with glaucomatous visual field defects were defined as (1) a cluster of 3 points with probabilities <5% in at least 1 hemifield on the pattern deviation map, including at least 1 point with a probability <1% or a cluster of 2 points with a probability <1%; (2) glaucomatous hemifield test results outside of the normal limits; or (3) a pattern standard deviation beyond 95% of the normal limits, as confirmed by at least 2 reliable examinations (false-positive/negatives <15%, fixation losses <15%). The healthy controls had an intraocular pressure (IOP) 21 mm Hg, no IOP elevation history, no glaucomatous optic disc appearance, no RNFL defects, and normal visual field test results. On the basis of visual field test results, patients with glaucoma were divided into 2 groups: preperimetric glaucoma (normal visual field) and early perimetric glaucoma (visual field mean deviation [MD] >6 decibels [dB]). The appearance of the optic disc on stereoscopic photographs and that of RNFL on red-free imaging were evaluated by 2 glaucoma specialists (Y.K.K. and K.H.P.) who were masked to all other information on the eyes. Discrepancies between the 2 observers’ findings were resolved by consensus.
Cirrus High-Definition Optical Coherence Tomography Optic Disc Cube Optic disc (Optic Disc Cube 200200 protocol) and macular scans (Macular Cube 512128 protocol) using Cirrus HD-OCT instrument software (version 6.0) for RNFL and GCIPL thickness measurements, respectively, were carried out. Poor-quality images showing eye motion, blinking artifacts, or poor centration were discarded by the examiner, and those with a signal strength <6 were excluded from the study. The circumpapillary RNFL thicknesses were measured overall, in each of the 4 quadrants, and in each of the 12 clock-hour sectors. The average, minimum, and 6 sectoral (superotemporal, superior, superonasal, inferonasal, inferior, inferotemporal) GCIPL thicknesses in an elliptical annulus were measured in the macular cube scan mode.5,22
Ganglion Cell-Inner Plexiform Layer Hemifield Test The GCIPL thickness maps were processed using the GCIPL Hemifield Test (Medical Electronics Lab, Seoul National University, Seoul, Korea), a customized software in MATLAB (2013a version, The MathWorks, Inc, Natick, MA). The GCIPL Hemifield Test automatically extracted, from the GCIPL thickness map, a 32bit color-scale image of an elliptical annulus of 2.0 mm vertical outer radius and 2.4 mm horizontal outer radius. Then, automated image processing for line detection was conducted. The GCIPL Hemifield Test uses the Hough transform algorithm to detect lines on a GCIPL thickness map. Hough transform is a widely used technique in the image processing field for detection of lines according to their parametric representation. In general, to implement the Hough transform, the threshold is selected heuristically. Setting the threshold too low or too high can give rise to false positivity or mis-detection, respectively, for a given shape. According to the threshold, it was determined whether the small gaps in line segments were automatically filled or not. Subsequently, the end points of the line segments corresponding to the peaks in the Hough transform were found. Subsequent image processing for detection of color values was conducted only in cases in which horizontal reference lines longer than one-half the distance from the temporal inner elliptical annulus to the outer elliptical annulus were successfully detected. The red, green, and blue (RGB) color values of the pixels above and below the detected line dividing the superior and inferior hemifields were discriminated. They were converted to GCIPL thicknesses on the basis of the GCIPL thickness map’s reference color bar. Ultimately, a positive (i.e., “outside normal limits”) GCIPL Hemifield Test result was declared if the following 3 conditions were all met: (1) The reference line (a horizontal line dividing the superior and inferior hemifields) is continuously detected for longer than one half of the distance from the temporal inner elliptical annulus to the outer elliptical annulus; (2) the average GCIPL thickness difference within 10 pixels of the reference line, both above and below, is 5 mm; and (3) the average RGB color ranges of the 10 pixels above and below the reference line display blue in 1 hemifield and red/yellow/white in the other hemifield (Figs 1 and 2).
Optimal Cutoff Values for Ganglion Cell-Inner Plexiform Layer Hemifield Test The optimal cutoff values for classifying cases as positive or negative were confirmed by using the areas under the receiver operating characteristic curves (AUCs) as the following process. First, the probable diagnostic criteria for discriminating early glaucoma was postulated that calculation of the average GCIPL thickness difference is within approximately 10 pixels both above and below the reference line and the baseline of the difference is 7 mm. Thus, the
2253
Ophthalmology Volume 122, Number 11, November 2015
Figure 1. Schematic representation of ganglion cell-inner plexiform layer (GCIPL) hemifield test for automated detection of hemifield difference across horizontal raphe on Cirrus high-definition optical coherence tomography (HD-OCT; Carl Zeiss Meditec, Dublin, CA) GCIPL thickness map. A, The GCIPL hemifield test automatically extracted, from a GCIPL thickness map, a 32-bit color-scale image of an elliptical annulus of 2.0 mm vertical outer radius and 2.4 mm horizontal radius. B, C, The reference line (red dashed line), running from the temporal inner elliptical annulus to the outer elliptical annulus and dividing the superior and inferior hemifields, is detected using a computer program. C, The average color of the RGB values in pixels included in 10 pixels both up and down (black dashed line), as set on the basis of the reference line, is calculated. Those values were automatically converted to GCIPL thicknesses on the basis of the GCIPL thickness map’s reference color bar. D, The Glaucoma Hemifield Test on a right eye showed a positive result, because the following 3 conditions all met together: (1) The reference line (blue line; a horizontal line dividing the superior and inferior hemifields) is continuously detected for more than one-half of the distance from the temporal inner elliptical annulus to the outer elliptical annulus; (2) the average GCIPL thickness difference within 10 pixels of the reference line, both above and below, is 5 mm; and (3) the average RGB color ranges of the 10 pixels above and below the reference line display blue in 1 hemifield and red/yellow/white in the other hemifield. OD ¼ right eye; OS ¼ left eye.
2254
Kim et al
GCIPL Hemifield Difference in Early Glaucoma
Figure 2. Examples of circumpapillary retinal nerve fiber layer (RNFL) and macular ganglion cell analysis (GCA) maps as measured by high-definition optical coherence tomography (HD-OCT) for a healthy eye (first row), a preperimetric glaucoma eye (second row), and an early-perimetric glaucoma eye (third row). The temporal and nasal borders of the RNFL defects are indicated by the black arrows on the RNFL thickness map or deviation map. In the preperimetric glaucoma eye, although there was no defect on the RNFL deviation map, a hemifield difference across the horizontal raphe (indicated by the blue arrow) was detected by the ganglion cell-inner plexiform layer (GCIPL) hemifield test. A hemifield difference across the horizontal raphe (indicated by the blue arrow) also was detected in the early perimetric glaucoma eye. OD ¼ right eye; OS ¼ left eye.
5 candidate values (4, 7, 10, 13, and 16 pixels) were taken to determine the optimal cutoff values for the up-down ranges of pixel numbers to be analyzed. Likewise, 5 candidate values (3, 5, 7, 9, and 11 mm) were chosen to obtain the optimal cutoff values for the minimum up-down difference of GCIPL thickness to be applied in diagnosing glaucoma. Then, the AUCs, calculated by using 25 matched pairs of cutoff scores, were compared among a total of 91 subjects (30 normal, 31 preperimetric, and 30 early-perimetric glaucoma) randomly sampled from previously aggregated clinical data and who were not associated with the current study. The high-ranked AUC values of the paired cutoff scores are listed in Table 1. Through this process, finally, the optimal cutoff values of the GCIPL Hemifield Test were determined as an up-down range of 10 pixels, and the difference of the minimum GCIPL thickness as 5 mm. On the basis of the selected optimal cutoff values, the GCIPL
Hemifield Test AUC was 0.967 (sensitivity 96.77%, specificity 96.67%) for discriminating glaucomatous changes between normal and preperimetric cases and 0.950 (sensitivity 93.33%, specificity 96.67%) for discriminating glaucomatous changes between normal and early perimetric cases.
Statistical Analysis Between the healthy and glaucoma groups, age and spherical equivalent of refractive error differences were obtained using analysis of variance tests; sex differences were computed using chisquare tests. The glaucoma diagnostic abilities were compared on the basis of computed AUC values. The sensitivities and specificities were calculated according to the optimal cutoff point, which was set as the maximum of the Youden index (obtained as J¼max
Table 1. Discriminating Ability of Ganglion Cell-Inner Plexiform Layer Hemifield Test in Different Matched Pairs of Cutoff Values (95% Confidence Interval) Cutoff Values* I (Pixels) 10 10 7 13 7
Preperimetric Glaucoma
II (mm) 5 7 5 5 7
AUCy 0.967 0.934 0.918 0.934 0.902
(0.89e1.00) (0.84e0.98) (0.82e0.97) (0.84e0.98) (0.80e0.96)
Sensitivity (%) 96.77 93.55 93.55 93.55 90.32
(83.3e99.9) (78.6e99.2) (78.6e99.2) (78.6e99.2) (74.2e98.0)
Early Perimetric Glaucoma AUCy
Specificity (%) 96.67 93.33 90.00 93.33 90.00
(82.8e99.9) (77.9e99.2) (73.5e97.9) (77.9e99.2) (73.5e97.9)
0.950 0.933 0.917 0.883 0.900
(0.86e0.99) (0.84e0.98) (0.82e0.97) (0.77e0.95) (0.80e0.96)
Sensitivity (%) 93.33 93.33 93.33 83.33 90.00
(77.9e99.2) (77.9e99.2) (77.9e99.2) (65.3e94.4) (73.9e97.9)
Specificity (%) 96.67 93.33 90.00 93.33 90.00
(82.8e99.9) (77.9e99.2) (73.5e97.9) (77.9e99.2) (73.9e97.9)
AUC ¼ area under the receiver operating characteristic curve. I, The cutoff values for up-down ranges of pixel numbers to be analyzed. II, The cutoff values for minimum up-down difference of GCIPL thickness to be applied in diagnosing glaucoma. *The paired cutoff values showing the high-ranked AUC values. y Evaluated in a total of 91 subjects (30 normal, 31 preperimetric, and 30 early perimetric glaucoma) randomly sampled from previously aggregated clinical data who are not associated with the current study.
2255
Ophthalmology Volume 122, Number 11, November 2015 Table 2. Clinical Characteristics of Study Participants
Age (yrs, average SD) Sex (M/F) Spherical equivalent (D, average SD) IOP (mmHg) CCT (mm, average SD) Axial length (mm, average SD) VF MD (dB, average SD) VF PSD (dB, average SD) OCT signal strength (average SD)
Healthy (n [ 65)
Preperimetric Glaucoma (n [ 79)
Early Perimetric Glaucoma (n [ 83)
P Value
53.59.3 33/32 0.762.36 13.52.5 545.435.3 24.71.98 0.051.27 1.790.55 7.90.7
54.611.8 39/40 0.922.24 13.32.1 546.133.9 24.31.72 0.721.92 1.880.75 8.00.6
57.511.6 40/43 0.792.15 12.72.2 536.932.1 24.81.80 2.391.65 5.193.37 7.80.8
0.072* 0.757y 0.897* 0.076* 0.161* 0.192* <0.001* <0.001* 0.200*
CCT ¼ central corneal thickness; D ¼ diopters; dB ¼ decibels; F ¼ female; IOP ¼ intraocular pressure; M ¼ male; MD ¼ mean deviation; OCT ¼ optical coherence tomography; PSD ¼ pattern standard deviation; SD ¼ standard deviation; VF ¼ visual field. *Analysis of variance test. y Chi-square test.
[sensitivity þ specificity 1]). Statistical analyses were performed using SPSS software (version 19, SPSS, Inc, Chicago, IL). The AUC comparisons were achieved with statistical software (MedCalc14.12; MedCalc Software, Mariakerke, Belgium). P values less than 0.05 were considered statistically significant. The data ranges were recorded as mean standard deviations.
Results Subjects A total of 239 eyes, representing 239 subjects, were initially included in the study. Of these, 12 patients were excluded: (1) Six subjects had poor-quality stereoscopic disc photography or red-free RNFL photography, and (2) 6 subjects had poor-quality scanned spectral-domain optical coherence tomography (signal strength <6). A total of 227 remaining eyes representing 162 patients with glaucoma and 65 healthy subjects were further examined in the ensuing analysis. The sample distribution was as follows: 79 preperimetric glaucoma eyes, 83 early perimetric glaucoma eyes, and 65 healthy eyes. The study population’s clinical characteristics are shown in Table 2. The age of those with preperimetric glaucoma ranged from 42 to 82 years (54.611.8 years, n ¼ 79). The spherical equivalent of the refractive error was 0.922.24, and the MD ranged from 1.34 to 2.87 dB (0.721.92 dB). Among the early perimetric glaucoma eye group (n ¼ 83), the age range on the first visit was 57.511.6 years, the mean spherical equivalent of the refractive error was 0.792.15 D, and the MD ranged from 4.56 to 1.54 (2.391.65). A total of 69 of the 83 eyes (83.1%) showed a baseline IOP 21 mmHg, and 14 of the 83 eyes (16.9%) showed a baseline IOP >21 mmHg. In the group of healthy subjects (n ¼ 65), the age range was 53.59.3 years. The spherical equivalent of the refractive error was 0.762.36, and the MD ranged from 0.85 to 2.02 dB (0.051.27 dB). The differences in sex, age, refractive error, central corneal thickness, and axial length among the study groups were not statistically significant.
Cirrus High-Definition Optical Coherence Tomography Index Table 3 shows that in the normal eyes, the average GCIPL and circumpapillary RNFL thicknesses (81.27.5 mm and 93.79.1 mm, respectively) were significantly greater than in eyes with
2256
preperimetric glaucoma (78.95.5 mm and 88.212.7 mm, respectively) or early perimetric glaucoma (72.96.5 mm and 73.811.6 mm, respectively) (all P < 0.001). The absolute value of GCIPL thickness differences between superotemporal and inferotemporal sectors was significantly lower in the normal eyes (0.80.3 mm) than in the eyes with preperimetric glaucoma (8.64.2 mm) or early perimetric glaucoma (12.24.8 mm) (P < 0.001). The minimum GCIPL thickness and inferotemporal GCIPL thickness were significantly greater in the normal eyes (77.910.3 mm and 80.49.7 mm, respectively) than in the eyes with preperimetric glaucoma (67.37.3 mm and 70.65.3 mm, respectively) or early perimetric glaucoma (60.98.4 mm and 64.75.7 mm, respectively) (all P < 0.001).
Results of Ganglion Cell-Inner Plexiform Layer Hemifield Test Ganglion cell-inner plexiform layer Hemifield Test positivity, interpreted as “outside normal limits,” was shown in 93.7% (74/79) of the preperimetric glaucoma eyes and in 94.0% (78/83) of the early perimetric glaucoma eyes. By contrast, only 1 of 65 healthy eyes (1.5%) showed positivity.
Diagnostic Ability of Ganglion Cell-Inner Plexiform Layer Hemifield Test The GCIPL Hemifield Test AUC for discrimination of glaucomatous changes between normal and preperimetric cases (0.967; sensitivity 94.94%, specificity 98.46%) was significantly higher than that of the minimum GCIPL thickness (0.933; sensitivity 92.43%, specificity 85.15%), the GCIPL thickness difference between superotemporal and inferotemporal sectors (0.911; sensitivity 87.34%, specificity 84.62%), the inferotemporal GCIPL thickness (0.907; sensitivity 90.14%, specificity 88.74%), and the average GCIPL thickness (0.899; sensitivity 79.75%, specificity 84.53%) (P ¼ 0.09, 0.03, 0.06, and 0.03, respectively). In the early perimetric glaucoma group, the AUC of the GCIPL Hemifield Test (0.962; sensitivity 93.98%, specificity 96.46%) was greater than that of the inferotemporal GCIPL thickness (0.938; sensitivity 92.77%, specificity 89.23%), the minimum GCIPL thickness (0.919; sensitivity 92.77%, specificity 86.13%), the GCIPL thickness difference between the superotemporal and inferotemporal sectors (0.916; sensitivity 82.34%, specificity 84.62%), and the average GCIPL thickness (0.912; sensitivity 86.75%, specificity 86.15%) (P ¼ 0.38, 0.17, 0.06, and 0.11, respectively) (Tables 4, 5, and Fig 3).
Kim et al
GCIPL Hemifield Difference in Early Glaucoma
Table 3. Comparison of Optical Coherence Tomography Parameters as Determined by Cirrus High-Definition Optical Coherence Tomography
GCIPL Hemifield Test positivity* (%) cpRNFL thickness (mm) Average Superior Nasal Inferior Temporal GCIPL thickness (mm) Average Minimum Differencey Superotemporal Superior Superonasal Inferonasal Inferior Inferotemporal
Healthy (n [ 65)
Preperimetric Glaucoma (n [ 79)
Early Perimetric Glaucoma (n [ 83)
P Valuez
1.5%
93.7%
94.0%
<0.001
93.79.1 119.910.4 64.78.2 118.69.9 71.86.7
88.212.7 107.716.3 71.211.5 99.713.9 74.110.9
73.811.6 93.211.5 57.49.9 85.311.5 60.29.3
<0.001 <0.001 <0.001 <0.001 <0.001
81.27.5 77.910.3 0.80.3 79.67.5 83.36.4 84.96.5 81.07.1 78.28.3 80.49.7
78.95.5 67.37.3 8.64.2 79.26.8 82.87.2 84.45.3 79.98.0 76.57.9 70.65.3
72.96.5 60.98.4 12.24.8 76.96.8 76.06.4 77.86.9 74.76.1 67.37.1 64.75.7
<0.001 <0.001 <0.001 0.037 <0.001 <0.001 <0.001 <0.001 <0.001
cpRNFL ¼ circumpapillary retinal nerve fiber layer; GCIPL ¼ ganglion cell-inner plexiform layer. Data are presented as mean standard deviation. *Interpreted as “outside normal limits.” y The absolute value of GCIPL thickness difference between superotemporal and inferotemporal sectors. z Analysis of variance test.
Discussion
difference across the horizontal raphe on GCIPL thickness maps. Although several maps have been used in single GCA interpretation, a more condensed and objective evaluation, which can be especially helpful for less-experienced clinicians, often is preferred. The GCIPL Hemifield Test satisfies
In this study, a good early glaucoma detection ability was demonstrated for the GCIPL Hemifield Test. This study has investigated the glaucoma diagnostic utility of hemifield
Table 4. Discriminating Ability of Ganglion Cell-Inner Plexiform Layer Hemifield Test and Cirrus High-Definition Optical Coherence Tomography Parameters for Glaucoma Detection (95% Confidence Interval) Preperimetric Glaucoma
Early Perimetric Glaucoma
Parameters
Sensitivity (%)
Specificity (%)
þLR
LR
Sensitivity (%)
Specificity (%)
þLR
LR
GCIPL hemifield test cpRNFL thickness Average Superior Nasal Inferior Temporal GCIPL thickness Average Minimum Difference* Superotemporal Superior Superonasal Inferonasal Inferior Inferotemporal
94.94 (87.5e98.6)
98.46 (91.7e100.0)
61.65
0.05
93.98 (86.5e98.0)
96.46 (90.7e99.2)
26.55
0.06
81.01 69.31 43.05 74.19 49.41
(70.6e89.0) (55.7e82.8) (33.7e62.2) (62.0e86.1) (37.5e69.8)
80.81 77.22 77.26 84.18 75.22
(70.5e90.1) (64.5e88.1) (64.5e90.0) (78.3e94.7) (61.3e84.5)
4.22 3.04 1.89 4.69 1.99
0.23 0.40 0.74 0.31 0.67
90.36 63.33 47.96 70.27 53.04
(81.9e95.7) (52.6e77.1) (35.2e68.6) (59.3e82.7) (39.6e72.2)
81.54 79.04 76.57 87.36 78.45
(70.0e90.1) (63.5e90.2) (64.4e88.9) (79.2e95.8) (64.0e88.3)
4.89 3.02 2.05 5.56 2.46
0.12 0.46 0.68 0.34 0.60
79.75 92.43 87.34 83.11 75.32 68.73 77.44 88.55 90.14
(69.2e88.0) (84.2e97.5) (78.0e93.8) (74.5e91.2) (64.5e85.9) (57.5e78.3) (62.7e84.6) (78.7e94.4) (81.6e95.4)
84.53 85.15 84.62 82.36 83.77 82.68 79.29 78.41 88.74
(74.3e93.5) (79.1e95.6) (73.5e92.4) (71.2e89.7) (73.3e89.1) (74.6e91.2) (69.3e88.6) (63.4e85.9) (78.0e95.7)
5.16 6.22 5.68 4.71 4.64 3.97 3.74 4.10 8.01
0.24 0.09 0.15 0.21 0.29 0.38 0.28 0.15 0.11
86.75 92.77 82.34 85.31 83.75 82.12 81.73 89.54 92.77
(77.5e93.2) (84.9e97.3) (74.7e91.4) (76.2e95.7) (74.5e90.2) (69.5e90.3) (70.1e88.9) (81.2e93.9) (84.9e97.3)
86.15 86.13 84.62 82.36 84.87 82.31 79.15 80.29 89.23
(75.3e93.5) (73.4e92.2) (73.5e92.4) (71.0e92.1) (73.7e92.1) (73.5e91.6) (69.0e88.1) (68.7e88.5) (79.1e95.6)
6.26 6.69 5.35 4.84 5.54 4.64 3.92 4.54 8.61
0.15 0.08 0.21 0.18 0.19 0.22 0.23 0.13 0.08
cpRNFL ¼ circumpapillary retinal nerve fiber layer; GCIPL ¼ ganglion cell-inner plexiform layer; þLR ¼ positive likelihood ratio; LR ¼ negative likelihood ratio. *The absolute value of GCIPL thickness difference between superotemporal and inferotemporal sectors.
2257
Ophthalmology Volume 122, Number 11, November 2015 Table 5. Areas Under the Receiver Operating Characteristic Curve of Ganglion Cell-Inner Plexiform Layer Hemifield Test and Cirrus High-Definition Optical Coherence Tomography Parameters (95% Confidence Interval) Preperimetric Glaucoma Parameters
AUC
GCIPL hemifield test cpRNFL thickness Average Superior Nasal Inferior Temporal GCIPL thickness Average Minimum Differencey Superotemporal Superior Superonasal Inferonasal Inferior Inferotemporal
0.967 (0.92e0.99)
Early Perimetric Glaucoma P*
AUC
P*
0.962 (0.92e0.99)
0.880 0.856 0.712 0.878 0.719
(0.83e0.94) (0.79e0.90) (0.64e0.76) (0.81e0.93) (0.64e0.77)
0.01 0.01 <0.001 0.01 0.01
0.907 0.872 0.713 0.898 0.735
(0.85e0.95) (0.79e0.93) (0.64e0.77) (0.81e0.94) (0.68e0.81)
0.07 0.03 <0.001 0.01 0.01
0.899 0.933 0.911 0.830 0.857 0.812 0.806 0.886 0.907
(0.84e0.94) (0.88e0.97) (0.85e0.95) (0.76e0.88) (0.78e0.91) (0.74e0.87) (0.73e0.87) (0.81e0.94) (0.85e0.95)
0.03 0.09 0.03 0.01 0.01 0.01 0.01 0.02 0.06
0.912 0.919 0.916 0.859 0.850 0.828 0.822 0.902 0.938
(0.86e0.95) (0.86e0.96) (0.86e0.96) (0.79e0.90) (0.78e0.93) (0.77e0.89) (0.75e0.89) (0.82e0.94) (0.89e0.97)
0.11 0.17 0.06 0.01 0.01 0.01 0.01 0.05 0.38
AUC ¼ area under the curve; cpRNFL ¼ circumpapillary retinal nerve fiber layer; GCIPL ¼ ganglion cell-inner plexiform layer. *Comparing between the AUC of the GCIPL Hemifield Test and each parameter. y The absolute value of GCIPL thickness difference between superotemporal and inferotemporal sectors.
those criteria: A brief GCIPL thickness evaluation is printed on field charts in plain text to distinguish glaucomatous structural loss on GCIPL thickness maps. In a previous study investigating the factors associated with the presence or absence of abnormal GCA findings,23 the GCA maps showed abnormal findings in healthy eyes, especially those with higher myopic refractive error. Therefore, high myopic eyes showing less than 6 D
spherical refraction were excluded from the present study. Another relevant previous finding was that for large angular distances between the fovea and the RNFL defect, GCA maps showed no abnormalities.23 In the present study, in 5 of 79 preperimetric and 5 of 83 early glaucoma eyes, the GCIPL Hemifield Test’s hemifield difference across the horizontal raphe was not identified. On retrospective review of the relevant RNFL photographs, the angular distance
Figure 3. The areas under the receiver operating characteristic curves (AUCs) in discriminating healthy eyes from (A) preperimetric glaucoma eyes and (B) early perimetric glaucoma eyes. P1 indicates the comparison of the AUC value between the ganglion cell-inner plexiform layer (GCIPL) Hemifield Test and the GCIPL thickness of inferotemporal sector; P2 indicates the comparison of the AUC value between the GCIPL Hemifield Test and the minimum GCIPL thickness; P3 indicates the comparison of the AUC value between the GCIPL Hemifield Test and the average GCIPL thickness; P4 indicates the comparison of the AUC value between the GCIPL Hemifield Test and the average circumpapillary retinal nerve fiber layer (RNFL) thickness; P5 indicates the comparison of the AUC value between the GCIPL Hemifield Test and the GCIPL thickness difference between superotemporal and IT sectors. IT ¼ inferotemporal; ST ¼ superotemporal.
2258
Kim et al
GCIPL Hemifield Difference in Early Glaucoma
between the fovea and the RNFL defect was greater in subjects who did not show any hemifield difference across the horizontal raphe. In cases in which no visual field defects are apparent in eyes with glaucoma, there might be significant macular ganglion cell layer and RNFL thinning.24 This indicates that even if abnormal RNFL and ganglion cell layer thinning are manifest in the area near the fovea, detectable visual field defects at the corresponding test points will not necessarily be evident.25 Therefore, macular structural parameters can be used as markers of subclinical (preperimetric) abnormalities of the ganglion cell layer and RNFL near the fovea.25 This is particularly important, because such subclinical abnormalities can, in time, morph into clinical (perimetric) abnormalities. Earlier studies have shown that macular retinal layer thickness asymmetry indices, especially those for the ganglion cell layer, are potential indicators of glaucomatous retinal damage.9 Furthermore, the overlap in values for normal and glaucomatous eyes, on the ganglion cell layer asymmetry index, has been minimal. In the current study, the AUC results of the GCIPL Hemifield Test were greater than those of the other GCIPL parameters in eyes with both preperimetric and early perimetric glaucoma. An earlier study showed that in eyes with preperimetric glaucoma, although both the RNFL and the macular ganglion cell layer were thinned, the ganglion cell layer was substantially thinner; also, the macular ganglion cell layer thinning relative to that of the RNFL was both more localized and abrupt.9 Of note, although preperimetric glaucoma with localized ganglion cell layer thinning does not affect average macular ganglion cell layer thickness, it might cause ganglion cell layer asymmetry and thus might indicate a hemifield difference across the horizontal raphe. For that, in this study, the GCIPL Hemifield Test’s higher diagnostic ability relative to average GCILP thickness, for preperimetric glaucoma, seems to be associated with macular glaucomatous anatomic changes. In this study, the values of AUCs for the GCIPL Hemifield Test were not much different between preperimetric glaucoma (0.967) and early perimetric glaucoma (0.962). These results suggest that preperimetric glaucoma and early perimetric glaucoma form a continuum and are not separate entities, and the former can progress to the latter over time; the structural differences between these 2 disease entities are minimal, the main difference being the absence or presence, respectively, of visual field defect. Furthermore, these facts might indicate that the GCIPL Hemifield Test is not suitable for discriminating between preperimetric and early perimetric glaucoma. To clarify the glaucoma diagnostic ability and the natural time course of GCIPL hemifield asymmetry, further examination of additional patients with ocular hypertension, suspected glaucoma, and moderate-to-severe stages of glaucoma will be necessary. In this study, all manual procedures were eliminated, in that hemifield difference was detected by an automated customized MATLAB-based program. Of note, the line found by the program’s “line-detecting function” was used as the reference line, thereby enabling more accurate diagnosis.
Study Limitations The GCIPL Hemifield Test program is intended for use in glaucoma diagnostics and it is ineffective for other types of diagnosis, including differential diagnosis. Because a considerable variety of other diseases also might result in significant hemifield differences, a positive GCIPL Hemifield Test result does not necessarily indicate GCIPL loss due to glaucoma. Certainly, the GCIPL Hemifield Test will have to be developed for better efficacy in the assessment of the presence or absence of glaucomatous structural loss and the diagnosis of glaucoma. Furthermore, because we studied a group of mostly normal-tension glaucomatous eyes (83.1% of patients with glaucoma had a baseline IOP 21 mmHg) and early-stage glaucomatous eyes (preperimetric and early perimetric), our results might not be applicable to other populations, such as patients with primary open-angle glaucoma or those with advanced glaucoma. Another limitation is that the RGB color values of the pixels were converted to GCIPL thicknesses on the basis of the GCIPL thickness map’s reference color bar. Thus, some thickness bias might have been incurred. In conclusion, automated detection of hemifield difference across the horizontal raphe on a GCIPL thickness map, providing easily interpretable test results in a short, plaintext display, was proved to be useful for discrimination of early glaucomatous structural loss, most notably in preperimetric glaucoma cases. In the future, this type of GCA results presentation scheme will be more frequently used once expert systems are applied to the interpretation of GCA data.
References 1. Zeimer R, Asrani S, Zou S, et al. Quantitative detection of glaucomatous damage at the posterior pole by retinal thickness mapping. A pilot study. Ophthalmology 1998;105:224–31. 2. Brusini P, Tosoni C, Miani F. Quantitative mapping of the retinal thickness at the posterior pole in chronic open angle glaucoma. Acta Ophthalmol Scand Suppl 2000:42–4. 3. Nouri-Mahdavi K, Nowroozizadeh S, Nassiri N, et al. Macular ganglion cell/inner plexiform layer measurements by spectral domain optical coherence tomography for detection of early glaucoma and comparison to retinal nerve fiber layer measurements. Am J Ophthalmol 2013;156:1297–307. 4. Jeoung JW, Choi YJ, Park KH, Kim DM. Macular ganglion cell imaging study: glaucoma diagnostic accuracy of spectraldomain optical coherence tomography. Invest Ophthalmol Vis Sci 2013;54:4422–9. 5. Mwanza JC, Oakley JD, Budenz DL, et al. Macular ganglion cell-inner plexiform layer: automated detection and thickness reproducibility with spectral domain-optical coherence tomography in glaucoma. Invest Ophthalmol Vis Sci 2011;52: 8323–9. 6. Tan O, Chopra V, Lu AT, et al. Detection of macular ganglion cell loss in glaucoma by Fourier-domain optical coherence tomography. Ophthalmology 2009;116:2305–14. 7. Shin HY, Park HY, Jung KI, et al. Glaucoma diagnostic ability of ganglion cell-inner plexiform layer thickness differs according to the location of visual field loss. Ophthalmology 2014;121:93–9.
2259
Ophthalmology Volume 122, Number 11, November 2015 8. Mwanza JC, Durbin MK, Budenz DL, et al. Glaucoma diagnostic accuracy of ganglion cell-inner plexiform layer thickness: comparison with nerve fiber layer and optic nerve head. Ophthalmology 2012;119:1151–8. 9. Yamada H, Hangai M, Nakano N, et al. Asymmetry analysis of macular inner retinal layers for glaucoma diagnosis. Am J Ophthalmol 2014;158:1318–29. 10. Lisboa R, Paranhos A Jr, Weinreb RN, et al. Comparison of different spectral domain OCT scanning protocols for diagnosing preperimetric glaucoma. Invest Ophthalmol Vis Sci 2013;54:3417–25. 11. Morooka S, Hangai M, Nukada M, et al. Wide 3-dimensional macular ganglion cell complex imaging with spectral-domain optical coherence tomography in glaucoma. Invest Ophthalmol Vis Sci 2012;53:4805–12. 12. Kotera Y, Hangai M, Hirose F, et al. Three-dimensional imaging of macular inner structures in glaucoma by using spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci 2011;52:1412–21. 13. Seong M, Sung KR, Choi EH, et al. Macular and peripapillary retinal nerve fiber layer measurements by spectral domain optical coherence tomography in normal-tension glaucoma. Invest Ophthalmol Vis Sci 2010;51:1446–52. 14. Arintawati P, Sone T, Akita T, et al. The applicability of ganglion cell complex parameters determined from SD-OCT images to detect glaucomatous eyes. J Glaucoma 2013;22:713–8. 15. Katz J, Sommer A. Asymmetry and variation in the normal hill of vision. Arch Ophthalmol 1986;104:65–8. 16. Duggan C, Sommer A, Auer C, Burkhard K. Automated differential threshold perimetry for detecting glaucomatous visual field loss. Am J Ophthalmol 1985;100:420–3.
17. Asman P, Heijl A. Glaucoma Hemifield Test. Automated visual field evaluation. Arch Ophthalmol 1992;110:812–9. 18. Sommer A, Enger C, Witt K. Screening for glaucomatous visual field loss with automated threshold perimetry. Am J Ophthalmol 1987;103:681–4. 19. Um TW, Sung KR, Wollstein G, et al. Asymmetry in hemifield macular thickness as an early indicator of glaucomatous change. Invest Ophthalmol Vis Sci 2012;53:1139–44. 20. Seo JH, Kim TW, Weinreb RN, et al. Detection of localized retinal nerve fiber layer defects with posterior pole asymmetry analysis of spectral domain optical coherence tomography. Invest Ophthalmol Vis Sci 2012;53: 4347–53. 21. Asrani S, Rosdahl JA, Allingham RR. Novel software strategy for glaucoma diagnosis: asymmetry analysis of retinal thickness. Arch Ophthalmol 2011;129:1205–11. 22. Mwanza JC, Durbin MK, Budenz DL, et al. Profile and predictors of normal ganglion cell-inner plexiform layer thickness measured with frequency-domain optical coherence tomography. Invest Ophthalmol Vis Sci 2011;52: 7872–9. 23. Hwang YH, Jeong YC, Kim HK, Sohn YH. Macular ganglion cell analysis for early detection of glaucoma. Ophthalmology 2014;121:1508–15. 24. Nakano N, Hangai M, Nakanishi H, et al. Macular ganglion cell layer imaging in preperimetric glaucoma with speckle noise-reduced spectral domain optical coherence tomography. Ophthalmology 2011;118:2414–26. 25. Kimura Y, Hangai M, Matsumoto A, et al. Macular structure parameters as an automated indicator of paracentral scotoma in early glaucoma. Am J Ophthalmol 2013;156:907–17.
Footnotes and Financial Disclosures Originally received: March 2, 2015. Final revision: June 30, 2015. Accepted: July 10, 2015. Available online: August 13, 2015.
Author Contributions:
Manuscript no. 2015-335.
1
Department of Ophthalmology, Seoul National University Hospital, Seoul National University, College of Medicine, Seoul, Korea.
2
Bioengineering Major, Graduate School, Seoul National University, Seoul, Korea.
3
Department of Biomedical Engineering, College of Medicine and Institute of Medical and Biological Engineering, Medical Research Center, Seoul National University, Seoul, Korea. Financial Disclosure(s): The author(s) have no proprietary or commercial interest in any materials discussed in this article.
2260
Conception and design: Y.K. Kim, Yoo, H.C. Kim, Park Data collection: Y.K. Kim, Yoo, H.C. Kim, Park Analysis and interpretation: Y.K. Kim, Yoo, H.C. Kim, Park Obtained funding: Not applicable Overall responsibility: Y.K. Kim, Park Abbreviations and Acronyms: AUC ¼ area under the receiver operating characteristic curve; D ¼ diopter; dB ¼ decibels; GCA ¼ ganglion cell analysis; GCIPL ¼ ganglion cellinner plexiform layer; HD-OCT ¼ high-definition optical coherence tomography; IOP ¼ intraocular pressure; MD ¼ mean deviation; RGB ¼ red, green, and blue; RNFL ¼ retinal nerve fiber layer. Correspondence: Ki Ho Park, MD, PhD, Department of Ophthalmology, Seoul National University Hospital, Seoul National University College of Medicine, 101 Daehak-ro, Chongno-gu, Seoul 110-744, Republic of Korea. E-mail: [email protected].