Free
Glaucoma  |   September 2013
Cluster Analyses of Grid-Pattern Display in Macular Parameters Using Optical Coherence Tomography for Glaucoma Diagnosis
Author Notes
  • Division of Ophthalmology, Department of Surgery, Kobe University Graduate School of Medicine, Kobe, Japan 
  • Correspondence: Akiyasu Kanamori, Division of Ophthalmology, Department of Surgery, Kobe University Graduate School of Medicine, 7-5-1, Kusunoki-cho, Chuo-ku, Kobe, Japan, 650-0017; kanaaki@med.kobe-u.ac.jp
Investigative Ophthalmology & Visual Science September 2013, Vol.54, 6401-6408. doi:10.1167/iovs.13-12805
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Akiyasu Kanamori, Maiko Naka, Azusa Akashi, Masashi Fujihara, Yuko Yamada, Makoto Nakamura; Cluster Analyses of Grid-Pattern Display in Macular Parameters Using Optical Coherence Tomography for Glaucoma Diagnosis. Invest. Ophthalmol. Vis. Sci. 2013;54(9):6401-6408. doi: 10.1167/iovs.13-12805.

      Download citation file:


      © 2016 Association for Research in Vision and Ophthalmology.

      ×
  • Supplements
Abstract

Purpose.: Using spectral-domain optical coherence tomography (SD-OCT), we assessed the ability of cluster analyses, based on the grid-pattern of macular parameters, to detect glaucoma.

Methods.: We enrolled 75 normal eyes, 64 early glaucomatous eyes (EG), and 40 preperimetric glaucomatous eyes (PPG). Each participant was imaged using 3-dimensional optical coherence tomography (3D-OCT) to examine the macular retinal nerve fiber layer (mRNFL) and the thickness of the ganglion cell layer, together with the inner plexiform layer (GCL/IPL). Diagnostic criteria based on the clustering of abnormal grids from the mRNFL and GCL/IPL measurements were applied. The sensitivity and specificity of glaucoma detection were compared between the cluster criteria (CC) and the average thickness criteria (ATC) of total and hemiretinal sectors, and the cut-off criteria were determined using receiver operating characteristic (ROC) curve analyses from our normal controls.

Results.: The specificity values of CC and ATC from mRNFL measurements were 97% and 100%, respectively. The sensitivity of CC was 94% for EG and 68% for PPG. The sensitivity of ATC was 81% for EG and 38% for PPG. The specificity values of CC and ATC from GCL/IPL measurements were 96% and 100%, respectively. The sensitivity values of CC and ATC were 92% for EG and 63% for PPG. The sensitivity of ATC was 84% for EG and 25% for PPG. When compared to ATC and ROC-based cut-off criteria, CC showed a higher diagnostic capability.

Conclusions.: Judging abnormality based on a clustering of abnormal grids from macular OCT parameters may be a reliable approach for diagnosing early glaucoma. ( http://www.umin.ac.jp/ctr/index/htm9 number, UMIN000006900.)

Introduction
Glaucoma is an optic neuropathy that is characterized by a specific and progressive injury to the retinal nerve fiber layer (RNFL) and retinal ganglion cell (RGC). 1 Approximately 50% of these cells are located within 4.5 mm of the foveal center. 2 Histologic studies of human eyes of patients with glaucoma and of primate eyes with experimental glaucoma indicate that considerable loss of macular RGCs has occurred when detectable visual field (VF) defects appear. 3 5 These studies suggest that an assessment of macular structure could improve the detection of early glaucoma. 
Structural evaluation of RGCs using optical coherence tomography (OCT) has an important role in the diagnosis and management of glaucomatous patients. In addition to the measurement of circumpapillary RNFL thickness (cpRNFL), which has been a standard biomarker since the development of OCT, spectral-domain OCT (SD-OCT) instruments have enabled the segmentation between the RNFL and the ganglion cell layer (GCL) in the macular area. 6 8 Thus, the macular RNFL (mRNFL) thickness, and the total GCL and inner plexiform layer (GCL/IPL) thickness may be measured. The most typical glaucomatous VF defect spreads in an arcuate pattern or as an altitudinal nasal step due to the pattern of trajectory of the retinal nerve fiber bundles. 9,10 Macular GCL thickness showed significant spatial relationships with cpRNFL thickness in glaucomatous patients evaluated by SD-OCT. 11 Some reports using receiver operating characteristic (ROC) curve analyses have demonstrated that these parameters have a significant ability to detect early glaucoma. 12,13  
Glaucoma typically begins as localized changes; thus, averaged measurements inevitably are insensitive for detecting these changes. Although the average thickness has the advantage of being a global parameter for the detection of structural abnormalities, it is likely to be less sensitive for detecting RGC loss limited to a specific area. In general, the evaluation of narrower regions targeted to the damaged area would be ideal for detecting local damage. However, predefinition of analyzing area using the diagnosing instrument is difficult because the region of RGC or RNFL loss differs between individuals with glaucoma. 
Three-dimensional OCT (3D-OCT; Topcon, Inc., Tokyo, Japan) creates a 10 × 10 grid map of mRNFL and GCL/IPL thickness of the macular region. Commercially available 3D-OCT software displays a deviation map that reflects a significant reduction of the total macular retinal thickness, the superior hemiretinal thickness, or the inferior hemiretinal thickness compared to a built-in normal database, with a probability of less than 1%. Because the data, divided into 10 × 10 grids, can be exported using software provided by Topcon, Inc., the user also can evaluate the thickness deviation of a specific grid or clustered grids-of-interest. The purpose of this study was to determine whether clustering of abnormally color-coded grids using a standardized criterion discriminates early glaucomatous (EG) eyes or preperimetric eyes effectively from healthy eyes when they are compared using commercially available deviation maps of mean thicknesses of macula parameters. 
Materials and Methods
Japanese subjects were recruited at the Kobe University Hospital (Kobe, Japan) for this observational cross-sectional study. The institutional review board of Kobe University approved the study protocol, which adhered to the tenets of the Declaration of Helsinki. Written, informed consent was obtained from each subject after the study protocol was explained. 
All subjects received a full ocular examination. The Humphrey Field Analyzer 30-2 SITA standard program (HFA; Carl Zeiss Meditec, Inc., Dublin, CA) was used to perform VF tests. Subjects with best-corrected visual acuity of 20/40 or better, spherical refraction of greater than −6.0 diopters (D), cylinder correction within ±3.0 D, and gonioscopically open angles were included. Axial length was acquired using an IOL Master (Carl Zeiss Meditec, Inc.). No subject had undergone ocular surgeries. VF tests and measurements obtained using the 3D-OCT instrument (Topcon, Inc.) were performed within a period of 6 months. 
Glaucomatous optic neuropathy (GON) is defined as consisting of neuroretinal rim damage, increased cup-to-disc ratio, rim thinning, and notches with or without RNFL defects. The glaucomatous VF defect was defined based on liberal criteria: two or more contiguous points with a pattern deviation sensitivity loss of P < 0.01, three or more contiguous points with sensitivity loss of P < 0.05 not crossing the horizontal meridian line, or a 10-dB difference across the nasal horizontal midline at two or more adjacent locations, and an abnormal result on the glaucoma hemifield test. 14 Furthermore, early glaucoma was defined by GON with VF loss based on the classification of Anderson and Patella when the mean deviation (MD) was >−6 dB. Participants who exhibited the same criteria for optic nerve head change without VF loss criteria were defined as having preperimetric glaucoma by 2 independent blinded glaucomatous specialists (AA and AK) at the time of OCT imaging. In cases of disagreement, a third specialist (MN) reviewed and assessed the stereo photographs (nonmyd WX3D; Kowa, Inc., Tokyo, Japan) Self-reported healthy subjects of at least 20 years of age also were invited to participate in this study. The exclusion criteria for normal eyes were as follows: intraocular pressure > 21 mm Hg, unreliable HFA results (fixation loss, false-positive or false-negative > 33%), abnormal findings in HFA tests suggestive of glaucoma as mentioned above, any abnormal VF loss due to vitreoretinal diseases, and optic nerve or RNFL abnormalities unrelated to glaucomatous optic neuropathy. Only 1 randomly selected eye was enrolled. 
3D-OCT Measurements
The 3D–OCT-2000 (Topcon, Inc.) has an axial scanning speed of 27,000 A-scans per second at an axial resolution of 5 to 6 μm (software version 8.00; Topcon, Inc.). The magnification effect in each eye was corrected based on the formula (modified Littman's method) provided by the manufacturer, which was based on the refraction, corneal radius, and axial length, to obtain more accurate scan sizes during measurements. Images with a quality factor > 60 were used for analyses. 
Raster scanning of a 7 × 7 mm area centered on the fovea at a scan density of 512 (vertical) × 128 (horizontal) scans was performed using 3D-OCT. A 6 × 6 mm area centered on the fovea was measured using software embedded in the 3D-OCT instrument (Topcon, Inc.). The data were divided automatically into 10 × 10 grids, and were exported using software provided by Topcon, Inc. The total thickness, and the superior and inferior hemiretina thicknesses of mRNFL, GCL/IPL were calculated. 
Statistical Analyses
All numerical data showed normal distributions, confirmed using the Kolmogorov-Smirnov test. The built-in viewer presented color-coded displays of the total, superior, and inferior hemiretina thicknesses of the mRNFL, GCL/IPL, and GCC (ganglion cell complex, mRNFL + GCL/IPL). If the average thickness varied by more than 1% of the level of the normal data distribution, red color was displayed. When an eye showed at least one of these three areas in red, that eye was defined as abnormal. The judgment based on these parameters was termed the average thickness criteria (ATC). 
The internal software also generated deviation plots by color-coding 10 × 10 grids as clear color (within the 95% normal limit), yellow (outside of the 95% normal limit), or red (outside the 99% normal limit). In addition to the average thickness analyses based on the total or hemiretinal deviation maps mentioned above, cluster criteria (CC) analogous to the liberal criteria of the Humphrey VF were conducted as follows: If an eye showed at least four contiguous, horizontally located grids and an additional two grids adjacent to these grids that were displayed in red in mRNFL measurements of the same hemiretina, the eye was considered to be abnormal (Fig. 1). For GCL/IPL measurements, if an eye showed at least three contiguous grids in red, the eye was considered to be abnormal. Other criteria were tested, but the present CC had the best performance (data not shown). 
Figure 1
 
Image of a right eye with early glaucoma. (A) A fundus photograph shows an enlargement of disc cupping and an RNFL defect in the inferotemporal quadrant. (B) A pattern deviation map of the VF test shows glaucomatous VF loss. (C) A printout of the macular analysis using 3D-OCT. The built-in viewer shows macular RNFL thickness, GCL/IPL thickness, and RNFL+ GCL+ IPL thickness. Upper panel: a pseudo-colored map of the measured thickness. Center panel: Each grid in the 10 × 10 grid was color-coded with no color (within the normal limit), yellow (outside of the 95% normal limit), or red (outside of the 99% normal limit). Lower panel: the thicknesses of the total and superior and inferior hemiretinal sectors. Each thickness was color-coded in green or yellow based on the software's normal built-in dataset. Arrows indicate the contiguous grids that defined the eye as abnormal.
Figure 1
 
Image of a right eye with early glaucoma. (A) A fundus photograph shows an enlargement of disc cupping and an RNFL defect in the inferotemporal quadrant. (B) A pattern deviation map of the VF test shows glaucomatous VF loss. (C) A printout of the macular analysis using 3D-OCT. The built-in viewer shows macular RNFL thickness, GCL/IPL thickness, and RNFL+ GCL+ IPL thickness. Upper panel: a pseudo-colored map of the measured thickness. Center panel: Each grid in the 10 × 10 grid was color-coded with no color (within the normal limit), yellow (outside of the 95% normal limit), or red (outside of the 99% normal limit). Lower panel: the thicknesses of the total and superior and inferior hemiretinal sectors. Each thickness was color-coded in green or yellow based on the software's normal built-in dataset. Arrows indicate the contiguous grids that defined the eye as abnormal.
The ROC curve analyses were performed to evaluate the diagnostic ability of the mRNFL and GCL/IPL measurements to discriminate preperimetric or early-glaucoma eyes from our control cohort. The area under the ROC curve (AUC) was calculated for each parameter. ROC curves were adjusted for differences in age using covariate-adjusted ROC curves, as demonstrated by Pepe. 15 A bootstrap resampling procedure was used (N = 1000 resamples). We set a third criterion for glaucoma diagnosis from the OCT parameters by defining eyes as being abnormal when the total or hemiretinal OCT measurements were thinner than a specific value set (= an ROC curve-based cut-off criterion). 
The sensitivity and specificity values based on the ATC, the CC, and the ROC curve-based cut-off criteria were calculated, and compared using McNemar tests. 
Statistical analyses were performed using software programs (Stata version 12.0; StataCorp, College Station, TX; and Medcalc version 11.6.1.0; Medcalc Software, Mariakerke, Belgium). A P value of <0.05 was considered to be statistically significant. 
Results
We studied 75 normal eyes, 64 EG eyes, and 40 preperimetric glaucomatous (PPG) eyes. Table 1 shows the demographics and ocular characteristics of the subjects. There were no significant differences in refraction or axial length between normal eyes and PPG eyes, or normal eyes and EG eyes. 
Table 1
 
Characteristics of the Studied Eyes (Mean ± SD)
Table 1
 
Characteristics of the Studied Eyes (Mean ± SD)
Normal Eyes, n = 75 EG, n = 64 PPG, n = 40 P Value
Normal vs. EG Normal vs. PPG
Age, y 45.5 ± 12.8 49.5 ± 10.9 49.9 ± 11.7 0.045* 0.043*
Sex, % female 58.8 51.6 55 0.148† 0.619†
Refraction, D −2.16 ± 2.36 −2.76 ± 2.3 −2.64 ± 2.1 0.14* 0.30*
Axial length, mm 24.7 ± 1.26 25.1 ± 1.23 24.8 ± 1.02 0.11* 0.88*
Mean deviation, dB −0.08 ± 1.69 −3.12 ± 2.11 −0.82 ± 3.69 <0.001* 0.15*
Initially, the ability to detect EG eyes was evaluated. We evaluated the specificity–sensitivity of the built-in software that was based on the ATC. Total and hemiretinal sectors in mRNFL and GCL/IPL measurements in all normal eyes were not below 1% normal database of the built-in software. None of normal eyes was judged to be abnormal from mRNFL and GCL/IPL measurements. In contrast, 52 EG eyes were defined as abnormal in mRNFL measurements. Also, 41 EG eyes were defined as abnormal in GCL/IPL measurements. Hence, the sensitivity of mRNFL and GCL/IPL measurements to diagnose early glaucoma based on a normal database and the ATC was 81.3% (52/64) and 64.1% (41/64), respectively, with 100% specificity. Table 2 summarizes these results. 
Table 2
 
Sensitivities and Specificities (%) for the Parameters Based on the Three Classification Criteria
Table 2
 
Sensitivities and Specificities (%) for the Parameters Based on the Three Classification Criteria
mRNFL Measurements GCL/IPL Measurements
The Cluster Criteria The Average Criteria The ROC Curve Criteria The Cluster Criteria The Average Criteria The ROC Curve Criteria
Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity
Early glaucomatous eyes, n = 64 93.8 97.4 81.3 100 50 97.4 92.2 96 64 100 25 96
Preperimetric glaucomatous eyes, n = 40 67.5 97.4 37.5 100 25 97.4 62.5 96 25 100 30 96
We then evaluated the specificity–sensitivity of the CC based on the 10 × 10 grid map. Two normal eyes were defined as abnormal in this cohort, whereas 60 glaucomatous eyes were defined as abnormal based on the CC from mRNFL measurements. Thus, the mRNFL CC showed 93.8% (60/64) sensitivity and 97.3% specificity (73/75). Similarly, 59 eyes with early glaucoma and three normal eyes were defined as abnormal based on the CC from GCL/IPL measurements, which resulted in 92.2% (59/64) sensitivity and 96% specificity (72/75). A Venn diagram indicated that agreement in glaucoma diagnosis between the mRFNL and GCL/IPL based on CC was 87.5% ( = 56/64, Fig. 2). Only one EG eye was not detected by either the mRNFL or GCL/IPL CC. 
Figure 2
 
Venn diagrams illustrating the percentage of eyes that were judged as abnormal based on the grid-pattern analyses of the mRNFL measurements (solid lines) and the GCL/IPL measurements (broken lines) obtained using 3D-OCT for EG eyes (A) and PPG eyes (B).
Figure 2
 
Venn diagrams illustrating the percentage of eyes that were judged as abnormal based on the grid-pattern analyses of the mRNFL measurements (solid lines) and the GCL/IPL measurements (broken lines) obtained using 3D-OCT for EG eyes (A) and PPG eyes (B).
To evaluate the diagnostic capability of the mRNFL and GCL/IPL compared to independent normal controls, the age-adjusted AUC for the average thicknesses of these two parameters was calculated using ROC analyses (Fig. 3). As shown in Table 3, the total mRNFL thickness showed the highest AUC of the three mRNFL parameters. Furthermore, the inferior hemiretinal GCL/IPL thickness yielded the highest AUC of the three GCL/IPL parameters. The sensitivities of the mRNFL and GCL/IPL parameters are shown in Table 3, when the specificity was set to 97.4% or 96% (the same values as used for the CC). 
Figure 3
 
ROC curves of the total (blue lines), superior (red lines), and inferior (green lines) sectors of the mRNFL thickness (A) and the GCL/IPL thickness (B) measured using 3D-OCT for discriminating EG eyes. ROC curves of the total, superior, and inferior sectors of the mRNFL thickness (C) and the GCL/IPL thickness (D) for discriminating PPG eyes also are shown.
Figure 3
 
ROC curves of the total (blue lines), superior (red lines), and inferior (green lines) sectors of the mRNFL thickness (A) and the GCL/IPL thickness (B) measured using 3D-OCT for discriminating EG eyes. ROC curves of the total, superior, and inferior sectors of the mRNFL thickness (C) and the GCL/IPL thickness (D) for discriminating PPG eyes also are shown.
Table 3
 
The ROC Analyses and Sensitivity in the Total, Superior, and Inferior Hemiretinal Thickness
Table 3
 
The ROC Analyses and Sensitivity in the Total, Superior, and Inferior Hemiretinal Thickness
Early Glaucomatous Eyes Preperimetric Glaucomatous Eyes
AUC Standard Error Sensitivity AUC Standard Error Sensitivity
mRNFL
 Total 0.949 0.172 70.3%* 0.853 0.043 20%*
 Superior 0.846 0.332 37.5%* 0.751 0.049 17.5%*
 Inferior 0.926 0.228 67.2%* 0.863 0.042 25%*
GCL/IPL
 Total 0.895 0.278 42.2%† 0.748 0.065 17.5%†
 Superior 0.819 0.370 20.3%† 0.668 0.064 12.5%†
 Inferior 0.912 0.245 42.2%† 0.812 0.063 30%†
The McNemar test was performed to analyze consistencies and to compare the diagnostic capabilities of the three criteria. As shown in Table 4, the CC revealed higher detection capabilities from mRNFL measurements compared to the ATC (P = 0.002) and total mRNFL thickness using the ROC-based cut-off criteria (P < 0.0001). Similarly, the CC showed higher detection ability from GCL/IPL compared to the ATC (P < 0.0001) and inferior hemiretinal GCL/IPL thickness using the ROC curve-based cut-off criteria (P < 0.0001). 
Table 4
 
Contingency Table for the Comparison of the Three Criteria in Terms of Early Glaucoma by McNemar Test
Table 4
 
Contingency Table for the Comparison of the Three Criteria in Terms of Early Glaucoma by McNemar Test
mRNFL Measurements GCL/IPL Measurements
The Cluster Criteria P The Cluster Criteria P
Normal Abnormal Total Normal Abnormal Total
The average thickness criteria
 Normal 77 10 87 0.002 77 21 98 <0.0001
 Abnormal 0 52 52 0 41 41
 Total 77 62 139 77 62 139
The ROC curve criteria
 Normal 77 28 105 <0.0001 77 43 120 <0.0001
 Abnormal 0 34 34 0 19 19
 Total 77 62 139 77 62 139
The detection capability for PPG eyes was evaluated next. As expected, less-PPG eyes were identified as abnormal based on the ATC compared to EG eyes. The sensitivity levels of the mRNFL and GCL/IPL measurements based on the ATC were 37.5% and 25%, respectively. 
We defined 24 and 25 PPG eyes as abnormal based on the CC from mRNFL and GCL/IPL measurements, respectively. Thus, mRNFL and GCL/IPL CC showed 67.5% (27/40) sensitivity and 62.5% specificity (25/40). A Venn diagram showed 50% (20/40) agreement between mRNFL and GCL/IPL CC. A total of 11 PPG eyes could not be detected by either mRNFL or GCL/IPL CC. 
The age-adjusted AUC of average mRNFL and GCL/IPL thickness discriminating PPG eyes is shown in Table 3. The inferior hemiretinal thickness exhibited the highest AUC among mRNFL parameters or GCL/IPL parameters. However, the sensitivity levels of the inferior mRNFL thickness and the GCL/IPL thickness were only 25% and 30%, respectively, when the specificity was set to 97.4% or 96%. 
As shown in Table 5, the McNemar test revealed that the CC in mRNFL and GCL/IPL measurements showed higher discriminating capabilities for preperimetric glaucoma compared to the ATC and the ROC curve-based cut-off criteria. 
Table 5
 
Contingency Table for the Comparison of the Three Criteria in Terms of Preperimetric Glaucoma by McNemar Test
Table 5
 
Contingency Table for the Comparison of the Three Criteria in Terms of Preperimetric Glaucoma by McNemar Test
mRNFL Measurements GCL/IPL Measurements
The Cluster Criteria P The Cluster Criteria P
Normal Abnormal Total Normal Abnormal Total
The average criteria
 Normal 86 14 100 0.0001 87 18 105 0.0001
 Abnormal 0 15 15 0 10 10
 Total 86 29 115 87 28 115
The ROC curve criteria
 Normal 86 17 103 <0.0001 87 13 100 <0.0001
 Abnormal 0 12 12 0 15 15
 Total 86 29 115 87 28 115
Discussion
Our study evaluated whether the cluster analyses of macular OCT parameters, analogous to the VF diagnostic criteria, would strengthen the ability of these parameters to discriminate correctly early and preperimetric glaucoma from normal control eyes. For this purpose, we compared three sets of criteria: CC that defined eyes as abnormal when several contiguous grids within a hemiretina from either mRNFL or GCL/IPL showed reduced thickness that deviated from the built-in normal database, with a probability of less than 1% (color coded red); ATC that defined eyes as abnormal when the thickness of the total retina, superior hemiretina, or inferior hemiretina was reduced compared to the built-in normal database; and ROC-based cut-off criteria that defined eyes as abnormal when the mRNFL or GCL/IPL was thinner than the value determined from the ROC analyses of our normal control cohort. 
With regard to the CC, clustering grids were created based essentially on the trajectories of the retinal nerve fiber bundles, because it is well-known that the EG eye shows a cluster of damaged RGCs. Clustering of test points also is known to enhance the sensitivity and specificity (to reduce the signal-to-noise ratio) in standard automated perimetry and multifocal visual evoked potential. 16,17 In contrast, the ATC are analogous to the glaucoma hemifield test, a global index in the VF, in which the average sensitivity obtained from many test points is compared to a normal database. Although easy to understand, averaging thickness underestimates local structural damage because areas with a normal or less-affected mRNFL and RGC population may be included. Indeed, an arbitrary cut-off criterion derived from ROC analyses showed very poor sensitivity reflecting a significant overlap of the OCT parameters between controls and glaucomatous eyes. This study demonstrated that in our cohort, the ability of CC to discriminate early and PPG eyes from healthy eyes was superior to the ATC. Theoretically, the measurements of mRNFL (axon) and GCL/IPL (cell body) evaluate the same target of glaucomatous structural loss. The CC showed good consistency between mRNFL and GCL/IPL damage in eyes with early glaucoma, as shown in Figure 2. Hence, we believe that the CC based on the color-coded grid-pattern optimally reflected the anatomic loss in early glaucoma. Taken together, an assessment of localized regions may offer greater ability to detect structural damage than average thickness measurements. 
Evidence is accumulating that measurements of the inner retinal layers in the macular region may provide additional parameters for the detection of glaucoma. 12,18 21 Previous studies have shown that GCC and circumpapillary RNFL thickness exhibit similar diagnostic performance for the detection of early glaucoma. 13,19,22 The recent versions of Cirrus and 3D-OCT enable the separation of RNFL from GCL at the macula. 6 8 We have reported that the ability to discriminate glaucoma based on mRNFL thickness using 3D-OCT was superior to that of Cirrus. 13 Although the cause of this superior performance remains unclear, this discrepancy may be due to the scanning area (Cirrus has an oval area of 14.13 mm2 and 3D-OCT has a square area of 36 mm2). The 3D-OCT instrument was capable of evaluating a greater area of the macula than the Cirrus device. A study using fundus photography demonstrated that RNFL defects occurred preferentially in the 7 and 11 o'clock sectors during early glaucoma, but that these defects also appeared in the 6 and 12 o'clock sectors in some cases. 23 This observation indicated that local RNFL thinning may occur at variable locations; therefore, macular parameters may miss the structural damage that converges into the superior or inferior pole of the optic disc. 23 Such eyes cannot be detected using the grid-pattern analyses of 3D-OCT, even if the instrument covers a wider area of the macula. However, the opposite case also may be true, as reported recently by Garvin et al. 11 Their map indicated that axons that originated from the parafoveal area entered the temporal side only at the optic nerve head. Macular assessment may be advantageous in analyzing these types of eyes. Hence, an assessment of a wider area of the macula may offer more chance to detect structural damage during early glaucoma. 
Our study has some limitations. First, our study design was a case-controlled study that included patients with well-established glaucoma and a separate group of normal subjects as hospital based-controls, and could substantially overestimate the diagnostic performance. 24,25 Also, when the sample is small, the bootstrap method provides less coverage than its theoretic coverage. 26 The ROC curve-based cut-off criteria using the bootstrap method might deliver optimistic results. However, the conclusion that the CC had superiority to the other criteria was solid if the weakness of the bootstrap method was taken into account. Second, an anatomic mismatch may cause errors in aggregating a given cell location among eyes in the normative data. Ocular rotation could be considered as a factor of individual anatomic mismatch. However, a recent study found that the reproducibility of macular measurements in 3D-OCT was not statistically different in normal and glaucomatous eyes when the ocular rotation was corrected. 27 The study concluded that macular measurements are affected minimally by the ocular rotation. Third, the 10-2 standard automated perimetry test was not used to evaluate the glaucomatous patients. Only 12 test points in our VF test corresponded with macular OCT measurements. We included some EG patients with VF defects outside the central 10°. These protocols may lead to a large variation in functional reduction in the central 10° VF. A recent study that enrolled eyes only with VF defects inside the central 10° demonstrated that arcuate mRNFL defects showed an apparent spatial continuum. 28 However, the purpose of this study was to demonstrate the detection ability of the CC for early glaucoma and preperimetric glaucoma diagnosed using standard definition in glaucoma. Moreover, the macular analyses of 3D-OCT measurements did not detect any RGC loss in glaucomatous patients whose VF defects were located outside the central 10°. In these cases, only mRNFL abnormalities could be detected using OCT measurements. 
In conclusion, we introduced the CC for macular OCT measurements to detect early structural damage in glaucoma. This method adapted anatomic change in glaucoma and yielded a higher ability to detect early glaucoma compared to conventional averaging thickness of global or divided sectors. This method will be a useful and convenient tool, and enhance the ability for discriminating early glaucoma, particularly preperimetric glaucoma from healthy subjects. 
Acknowledgments
Supported by JSPS KAKENHI Grant Number 25462715 (AK) from the Japanese Government, the Suda Memorial Foundation (AK), the Mishima Memorial Foundation (AK), and the Santan Pharmaceutical Founder Commemoration Ophthalmic Research Fund (AK). 
Disclosure: A. Kanamori, None; M. Naka, None; A. Akashi, None; M. Fujihara, None; Y. Yamada, None; M. Nakamura, None 
References
Weinreb RN Khaw PT. Primary open-angle glaucoma. Lancet . 2004; 363: 1711–1720. [CrossRef] [PubMed]
Curcio CA Allen KA. Topography of ganglion cells in human retina. J Comp Neurol . 1990; 300: 5–25. [CrossRef] [PubMed]
Quigley HA Dunkelberger GR Green WR. Retinal ganglion cell atrophy correlated with automated perimetry in human eyes with glaucoma. Am J Ophthalmol . 1989; 107: 453–464. [CrossRef] [PubMed]
Kerrigan-Baumrind LA Quigley HA Pease ME Kerrigan DF Mitchell RS. Number of ganglion cells in glaucoma eyes compared to threshold visual field tests in the same persons. Invest Ophthalmol Vis Sci . 2000; 41: 741–748. [PubMed]
Harwerth RS Carter-Dawson L Shen F Smith ELIII, Crawford ML. Ganglion cell losses underlying visual field defects from experimental glaucoma. Invest Ophthalmol Vis Sci . 1999; 40: 2242–2250. [PubMed]
Mwanza JC Durbin MK Budenz DL Profile and predictors of normal ganglion cell-inner plexiform layer thickness measured with frequency-domain optical coherence tomography. Invest Ophthalmol Vis Sci . 2011; 52: 7872–7879. [CrossRef] [PubMed]
Mwanza JC Oakley JD Budenz DL Chang RT Knight OJ Feuer WJ. 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: 8323–8329. [CrossRef] [PubMed]
Kotera Y Hangai M Hirose F Mori S Yoshimura N. Three-dimensional imaging of macular inner structures in glaucoma by using spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci . 2011; 52: 1412–1421. [CrossRef] [PubMed]
Fitzgibbon T Taylor SF. Retinotopy of the human retinal nerve fibre layer and optic nerve head. J Comp Neurol . 1996; 375: 238–251. [CrossRef] [PubMed]
Jansonius NM Nevalainen J Selig B A mathematical description of nerve fiber bundle trajectories and their variability in the human retina. Vision Res . 2009; 49: 2157–2163. [CrossRef] [PubMed]
Garvin MK Abramoff MD Lee K Niemeijer M Sonka M Kwon YH. 2-D pattern of nerve fiber bundles in glaucoma emerging from spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci . 2012; 53: 483–489. [CrossRef] [PubMed]
Mwanza JC Durbin MK Budenz DL Glaucoma diagnostic accuracy of ganglion cell-inner plexiform layer thickness: comparison with nerve fiber layer and optic nerve head. Ophthalmology . 2012; 119: 1151–1158. [CrossRef] [PubMed]
Akashi A Kanamori A Nakamura M Fujihara M Yamada Y Negi A. Comparative assessment for the ability of Cirrus, RTVue and 3D OCT to diagnose glaucoma. Invest Ophthalmol Vis Sci . 2013; 54: 4478–4484. [CrossRef] [PubMed]
Anderson DR Patella VM. Automated Static Perimetry. 2nd ed. St. Louis, MO: The C. V. Mosby Co.; 1999.
Pepe MS. The Statistical Evaluation of Medical Tests for Classification and Prediction . New York, NY: Oxford University Press; 2004; xvi, 302.
Nakamura M Ishikawa K Nagai T Negi A. Receiver-operating characteristic analysis of multifocal VEPs to diagnose and quantify glaucomatous functional damage. Doc Ophthalmol . 2011; 123: 93–108. [CrossRef] [PubMed]
Thonginnetra O Greenstein VC Chu D Liebmann JM Ritch R Hood DC. Normal versus high tension glaucoma: a comparison of functional and structural defects. J Glaucoma . 2010; 19: 151–157. [CrossRef] [PubMed]
Tan O Chopra V Lu AT Detection of macular ganglion cell loss in glaucoma by Fourier-domain optical coherence tomography. Ophthalmology . 2009; 116: 2305–2314. [CrossRef] [PubMed]
Schulze A Lamparter J Pfeiffer N Berisha F Schmidtmann I Hoffmann EM. Diagnostic ability of retinal ganglion cell complex, retinal nerve fiber layer, and optic nerve head measurements by Fourier-domain optical coherence tomography. Graefes Arch Clin Exp Ophthalmol . 2011; 249: 1039–1045. [CrossRef] [PubMed]
Garas A Vargha P Hollo G. Diagnostic accuracy of nerve fibre layer, macular thickness and optic disc measurements made with the RTVue-100 optical coherence tomograph to detect glaucoma. Eye (Lond) . 2011; 25: 57–65. [CrossRef] [PubMed]
Moreno PA Konno B Lima VC Spectral-domain optical coherence tomography for early glaucoma assessment: analysis of macular ganglion cell complex versus peripapillary retinal nerve fiber layer. Can J Ophthalmol . 2011; 46: 543–547. [CrossRef] [PubMed]
Rao HL Babu JG Addepalli UK Senthil S Garudadri CS. Retinal nerve fiber layer and macular inner retina measurements by spectral domain optical coherence tomograph in Indian eyes with early glaucoma. Eye (Lond) . 2012; 26: 133–139. [CrossRef] [PubMed]
Lee EJ Kim TW Weinreb RN Park KH Kim SH Kim DM. Trend-based analysis of retinal nerve fiber layer thickness measured by optical coherence tomography in eyes with localized nerve fiber layer defects. Invest Ophthalmol Vis Sci . 2011; 52: 1138–1144. [CrossRef] [PubMed]
Medeiros FA Ng D Zangwill LM Sample PA Bowd C Weinreb RN. The effects of study design and spectrum bias on the evaluation of diagnostic accuracy of confocal scanning laser ophthalmoscopy in glaucoma. Invest Ophthalmol Vis Sci . 2007; 48: 214–222. [CrossRef] [PubMed]
Rao HL Kumbar T Addepalli UK Effect of spectrum bias on the diagnostic accuracy of spectral-domain optical coherence tomography in glaucoma. Invest Ophthalmol Vis Sci . 2012; 53: 1058–1065. [CrossRef] [PubMed]
Sun J Chernick MR LaBudde RA. A bootstrap test for comparing two variances: simulation of size and power in small samples. J Biopharm Stat . 2011; 21: 1079–1093. [CrossRef] [PubMed]
Hirasawa H Araie M Tomidokoro A Reproducibility of thickness measurements of macular inner retinal layers using SD-OCT with or without correction of ocular rotation. Invest Ophthalmol Vis Sci . 2013; 54: 2562–2570. [CrossRef] [PubMed]
Hood DC Raza AS de Moraes CG Initial arcuate defects within the central 10 degrees in glaucoma. Invest Ophthalmol Vis Sci . 2011; 52: 940–946. [CrossRef] [PubMed]
Figure 1
 
Image of a right eye with early glaucoma. (A) A fundus photograph shows an enlargement of disc cupping and an RNFL defect in the inferotemporal quadrant. (B) A pattern deviation map of the VF test shows glaucomatous VF loss. (C) A printout of the macular analysis using 3D-OCT. The built-in viewer shows macular RNFL thickness, GCL/IPL thickness, and RNFL+ GCL+ IPL thickness. Upper panel: a pseudo-colored map of the measured thickness. Center panel: Each grid in the 10 × 10 grid was color-coded with no color (within the normal limit), yellow (outside of the 95% normal limit), or red (outside of the 99% normal limit). Lower panel: the thicknesses of the total and superior and inferior hemiretinal sectors. Each thickness was color-coded in green or yellow based on the software's normal built-in dataset. Arrows indicate the contiguous grids that defined the eye as abnormal.
Figure 1
 
Image of a right eye with early glaucoma. (A) A fundus photograph shows an enlargement of disc cupping and an RNFL defect in the inferotemporal quadrant. (B) A pattern deviation map of the VF test shows glaucomatous VF loss. (C) A printout of the macular analysis using 3D-OCT. The built-in viewer shows macular RNFL thickness, GCL/IPL thickness, and RNFL+ GCL+ IPL thickness. Upper panel: a pseudo-colored map of the measured thickness. Center panel: Each grid in the 10 × 10 grid was color-coded with no color (within the normal limit), yellow (outside of the 95% normal limit), or red (outside of the 99% normal limit). Lower panel: the thicknesses of the total and superior and inferior hemiretinal sectors. Each thickness was color-coded in green or yellow based on the software's normal built-in dataset. Arrows indicate the contiguous grids that defined the eye as abnormal.
Figure 2
 
Venn diagrams illustrating the percentage of eyes that were judged as abnormal based on the grid-pattern analyses of the mRNFL measurements (solid lines) and the GCL/IPL measurements (broken lines) obtained using 3D-OCT for EG eyes (A) and PPG eyes (B).
Figure 2
 
Venn diagrams illustrating the percentage of eyes that were judged as abnormal based on the grid-pattern analyses of the mRNFL measurements (solid lines) and the GCL/IPL measurements (broken lines) obtained using 3D-OCT for EG eyes (A) and PPG eyes (B).
Figure 3
 
ROC curves of the total (blue lines), superior (red lines), and inferior (green lines) sectors of the mRNFL thickness (A) and the GCL/IPL thickness (B) measured using 3D-OCT for discriminating EG eyes. ROC curves of the total, superior, and inferior sectors of the mRNFL thickness (C) and the GCL/IPL thickness (D) for discriminating PPG eyes also are shown.
Figure 3
 
ROC curves of the total (blue lines), superior (red lines), and inferior (green lines) sectors of the mRNFL thickness (A) and the GCL/IPL thickness (B) measured using 3D-OCT for discriminating EG eyes. ROC curves of the total, superior, and inferior sectors of the mRNFL thickness (C) and the GCL/IPL thickness (D) for discriminating PPG eyes also are shown.
Table 1
 
Characteristics of the Studied Eyes (Mean ± SD)
Table 1
 
Characteristics of the Studied Eyes (Mean ± SD)
Normal Eyes, n = 75 EG, n = 64 PPG, n = 40 P Value
Normal vs. EG Normal vs. PPG
Age, y 45.5 ± 12.8 49.5 ± 10.9 49.9 ± 11.7 0.045* 0.043*
Sex, % female 58.8 51.6 55 0.148† 0.619†
Refraction, D −2.16 ± 2.36 −2.76 ± 2.3 −2.64 ± 2.1 0.14* 0.30*
Axial length, mm 24.7 ± 1.26 25.1 ± 1.23 24.8 ± 1.02 0.11* 0.88*
Mean deviation, dB −0.08 ± 1.69 −3.12 ± 2.11 −0.82 ± 3.69 <0.001* 0.15*
Table 2
 
Sensitivities and Specificities (%) for the Parameters Based on the Three Classification Criteria
Table 2
 
Sensitivities and Specificities (%) for the Parameters Based on the Three Classification Criteria
mRNFL Measurements GCL/IPL Measurements
The Cluster Criteria The Average Criteria The ROC Curve Criteria The Cluster Criteria The Average Criteria The ROC Curve Criteria
Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity
Early glaucomatous eyes, n = 64 93.8 97.4 81.3 100 50 97.4 92.2 96 64 100 25 96
Preperimetric glaucomatous eyes, n = 40 67.5 97.4 37.5 100 25 97.4 62.5 96 25 100 30 96
Table 3
 
The ROC Analyses and Sensitivity in the Total, Superior, and Inferior Hemiretinal Thickness
Table 3
 
The ROC Analyses and Sensitivity in the Total, Superior, and Inferior Hemiretinal Thickness
Early Glaucomatous Eyes Preperimetric Glaucomatous Eyes
AUC Standard Error Sensitivity AUC Standard Error Sensitivity
mRNFL
 Total 0.949 0.172 70.3%* 0.853 0.043 20%*
 Superior 0.846 0.332 37.5%* 0.751 0.049 17.5%*
 Inferior 0.926 0.228 67.2%* 0.863 0.042 25%*
GCL/IPL
 Total 0.895 0.278 42.2%† 0.748 0.065 17.5%†
 Superior 0.819 0.370 20.3%† 0.668 0.064 12.5%†
 Inferior 0.912 0.245 42.2%† 0.812 0.063 30%†
Table 4
 
Contingency Table for the Comparison of the Three Criteria in Terms of Early Glaucoma by McNemar Test
Table 4
 
Contingency Table for the Comparison of the Three Criteria in Terms of Early Glaucoma by McNemar Test
mRNFL Measurements GCL/IPL Measurements
The Cluster Criteria P The Cluster Criteria P
Normal Abnormal Total Normal Abnormal Total
The average thickness criteria
 Normal 77 10 87 0.002 77 21 98 <0.0001
 Abnormal 0 52 52 0 41 41
 Total 77 62 139 77 62 139
The ROC curve criteria
 Normal 77 28 105 <0.0001 77 43 120 <0.0001
 Abnormal 0 34 34 0 19 19
 Total 77 62 139 77 62 139
Table 5
 
Contingency Table for the Comparison of the Three Criteria in Terms of Preperimetric Glaucoma by McNemar Test
Table 5
 
Contingency Table for the Comparison of the Three Criteria in Terms of Preperimetric Glaucoma by McNemar Test
mRNFL Measurements GCL/IPL Measurements
The Cluster Criteria P The Cluster Criteria P
Normal Abnormal Total Normal Abnormal Total
The average criteria
 Normal 86 14 100 0.0001 87 18 105 0.0001
 Abnormal 0 15 15 0 10 10
 Total 86 29 115 87 28 115
The ROC curve criteria
 Normal 86 17 103 <0.0001 87 13 100 <0.0001
 Abnormal 0 12 12 0 15 15
 Total 86 29 115 87 28 115
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×