Retinal nerve fiber layer (RNFL) defects on red-free RNFL photography compared with the fast RNFL scan mode of Stratus optical coherence tomography (OCT) and the optic disc cube mode of Cirrus OCT (Stratus and Cirrus are time-domain and spectral-domain OCT, respectively; Carl Zeiss Meditec, Dublin, California). A, Clock-hour locations on red-free RNFL photography of the left eye. A circle is placed around the optic nerve head, in which the location and diameter correspond as closely as possible to the circle displayed in the video mode of the OCT. Nine o’clock represents the temporal-most clock hour for both eyes. Angle width was determined by drawing a line from the center of the disc to the 2 points where the RNFL defect and disc converged. B, The RNFL normative database map of Stratus OCT. In Stratus OCT and Cirrus OCT, yellow (bottom fifth percentile) or red (bottom first percentile) in a clock hour that exactly corresponds to the clock hour of the RNFL defect on the red-free RNFL photograph is regarded as an indication of a localized RNFL defect. Yellow or red in the quadrant that includes the clock hour of the RNFL defect on red-free RNFL photography is also considered an indication of a localized RNFL defect. C, The RNFL normative database map from Cirrus OCT. D, Deviation from the normative database map of Cirrus OCT. The yellow and red wedge-shaped figures across the 3.46-mm scan circle appear in the clock hour that exactly corresponds to the clock hour of the RNFL defect on the red-free RNFL photograph, which is a positive indication of an RNFL defect.
Frequency distribution of localized retinal nerve fiber layer (RNFL) defects in terms of clock-hour locations on RNFL photography.
Scatterplot showing the relationship of angle width in red-free retinal nerve fiber layer (RNFL) photography and Cirrus deviation maps (spectral-domain optical coherence tomography [OCT]; Carl Zeiss Meditec, Dublin, California). There was a strong correlation in angle width (r = 0.88, P < .001).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kim NR, Lee ES, Seong GJ, Choi EH, Hong S, Kim CY. Spectral-Domain Optical Coherence Tomography for Detection of Localized Retinal Nerve Fiber Layer Defects in Patients With Open-Angle Glaucoma. Arch Ophthalmol. 2010;128(9):1121–1128. doi:10.1001/archophthalmol.2010.204
To evaluate and compare time-domain (Stratus) and spectral-domain (Cirrus; both Carl Zeiss Meditec, Dublin, California) optical coherence tomography (OCT) for the detection of localized retinal nerve fiber layer (RNFL) defects in patients with open-angle glaucoma.
Patients with localized RNFL defects and age-matched normal control participants were consecutively enrolled from July 1 to December 31, 2008. Sixty-six eyes from 66 patients and 66 eyes from 66 normal controls were imaged with Stratus OCT (fast RNFL scan mode) and Cirrus OCT (optic disc cube mode). The ability to detect the RNFL defect by using quadrant clock-hour maps from both OCTs and a Cirrus OCT deviation map were compared with red-free RNFL photography, which is the criterion standard for visualizing RNFL defects.
The Cirrus OCT deviation map exhibited significantly higher overall sensitivity (92.42%) in detecting the RNFL defects compared with the other maps, which were derived from a 3.46-mm-diameter peripapillary cross-sectional RNFL scan of both OCTs (P < .001). The Cirrus OCT quadrant map had a higher specificity; however, it was not statistically significant (P = .07). Compared with the other maps, the Cirrus OCT derivation map had the lowest cutoff angle for the width (10.69°) of the RNFL defect.
The deviation map from Cirrus OCT was more sensitive in detecting RNFL defects than the clock-hour and quadrant maps derived from cross-sectional peripapillary RNFL measurements by Stratus and Cirrus OCTs. The ability to detect localized RNFL defects on clock-hour or quadrant RNFL maps did not significantly differ between Stratus OCT and Cirrus OCT.
Glaucoma is a chronic optic neuropathy characterized by selective and progressive loss of ganglion cells, leading to loss of the retinal nerve fiber layer (RNFL).1-3 Imaging of the RNFL can be used to diagnose glaucomatous RNFL atrophy and can serve as a qualitative or semiquantitative reference in determining the extent of RNFL damage.3-5 Loss of RNFL precedes measurable optic nerve head and visual field damage.4,6,7 Although imaging methods provide objective information for comparison, the interpretation of images remains subjective, and there is significant variation in their assessment even among experienced professionals.8
Optical coherence tomography (OCT) was originally developed to assess tissue thickness in vivo. It has several advantages: it provides high-resolution cross-sectional imaging of the eye, it can evaluate and quantify the cross-sectional peripapillary RNFL thickness, and it can detect glaucoma in its early stages.9,10 Recently, OCT technology has developed considerably; it now incorporates spectral-domain imaging, which offers significant advantages over the traditional time-domain OCT techniques.11 These advantages include a faster scanning speed and higher-resolution imaging of RNFL.12,13 Previous studies reported that spectral-domain OCT is more reproducible14 and offers lower measurement variability15 and better scan quality16 than time-domain OCT. However, the discriminating ability between time-domain OCT and spectral-domain OCT for glaucoma has been reported to be comparable.14-16 Therefore, we sought to determine whether the technical advantages of spectral-domain OCT could improve the discriminating ability of OCT in detecting glaucomatous RNFL defects.
The purpose of this study was to evaluate and compare Stratus time-domain OCT (Carl Zeiss Meditec, Dublin, California) and Cirrus spectral-domain OCT (Carl Zeiss Meditec) in regard to their capability for detecting localized RNFL defects in patients with open-angle glaucoma.
Patients with localized RNFL defects and normal control participants from the Glaucoma-Cataract Clinic of Severance Hospital, Seoul, Korea, were consecutively enrolled in this study from July 1 to December 31, 2008. This study was approved by the institutional review board of Severance Hospital, Yonsei University, and performed in adherence with the Declaration of Helsinki. Informed consent was obtained from all participants. Eyes with localized RNFL defects were included if the defects were detected by clinical examination, disc imaging, or RNFL photography. The control group comprised healthy volunteers recruited from the staff, nurses, spouses or friends of patients, and patients referred for a routine visual acuity examination who had no ocular diseases.
All participants underwent a full ophthalmologic examination that included measurement of visual acuity, refraction measurement, intraocular pressure measurements using Goldmann applanation tonometry, slitlamp examination, gonioscopy, fundus examination with a 90-diopter lens, and standard automated perimetry (Humphrey field analyzer II with Swedish Interactive Thresholding Algorithm; Carl Zeiss Meditec). After pupillary dilation to a minimum diameter of 5 mm, each eye was imaged by color disc photography, RNFL photography, Stratus time-domain OCT, and Cirrus spectral-domain OCT. Each of these imaging methods was performed on each participant within a 1-month period.
Localized RNFL defects were considered to be present in the RNFL images if their width at a 1-disc-diameter distance from the edge of the disc was larger than a major retinal vessel and if they diverged in an arcuate or wedge shape reaching the edge of the disc.17 Only patients with an open anterior chamber angle, a best-corrected visual acuity of 20/40 or better, and clear ocular media were selected for participation in the study. Patients with a history of intraocular surgery, diabetes, and neurologic disorders were excluded.
Normal eyes were defined as eyes from patients who had no family history of glaucoma in a first-degree relative, no history or evidence of intraocular surgery, and no retinal pathological features. Normal eyes had to have an intraocular pressure of 21 mm Hg or lower, normal-appearing RNFL, normal-appearing optic nerve heads, and normal results of visual field tests.
The RNFL photographs were obtained with the Heidelberg Retina Angiograph 1 (Heidelberg Engineering, Heidelberg, Germany). This is a modern confocal laser scanning system used for retinal angiography that can generate digital, high-resolution, red-free fundus reflectance images.18,19 The RNFL photographs were independently and randomly evaluated by 2 of us (N.R.K and C.Y.K.) who were unaware of the identity of and clinical information for the corresponding patients.
The size of the RNFL photography image was adjusted to the red-free fundus image of the OCTs. Image tilting was corrected on the basis of the course of retinal vessels. An imaginary circle with a clock-hour indication was drawn around the optic nerve head on the RNFL photograph. The location and diameter of this circle corresponded as closely as possible to the circle displayed in the red-free video mode of the Stratus and Cirrus OCT, which centered on the optic disc and had a 3.46-mm diameter (Figure 1A). Nine o’clock represents the temporal-most clock hour for both eyes. The location of a localized RNFL defect was assigned a clock hour based on its corresponding location within the circle. In the cases of broad RNFL defects spanning 2 clock hours, the observers assigned only 1 clock hour based on the larger portion. If there were 2 or more RNFL defects in 2 eyes, the dominant RNFL defect was selected for data analysis. The width of each RNFL defect was measured (with Image J [http://rsbweb.nih.gov/ij/]) as the angle between the 2 lines from the center of the disc to the points at which the RNFL defect and circle converged.
The fast RNFL scan protocol was used to measure the peripapillary RNFL on the Stratus OCT (software version 4.0). Circular scans of 3.46-mm diameter centered on the optic disc were obtained by means of 256 A-scans acquired during 1.92 seconds. Scans with signal strengths of less than 6 were excluded from the analysis.
Spectral-domain OCT imaging was performed with the Cirrus OCT (software version 3.0), and the optic disc cube protocol was used for acquisition and analysis. This protocol generates a cube of data through a 6-mm square grid by acquiring a series of 200 horizontal scan lines each composed of 200 A-scans. Then a calculation circle 3.46 mm in diameter, consisting of 256 A-scans, was automatically positioned around the optic disc. All obtained images had signal strengths of at least 6.
The results from the comparison of RNFL thickness with normative data were indicated with a stoplight color scheme for both Stratus OCT and Cirrus OCT. The RNFL thicknesses in the normal range are represented by green backgrounds, those that are abnormal at the 5% level are represented by yellow backgrounds, and those that are abnormal at the 1% level are represented by red backgrounds.20
A yellow or red display in a clock-hour map from Stratus and Cirrus OCTs was defined as detection of an RNFL defect if it matched the clock-hour location of the RNFL defects observed in the RNFL photographs (Figure 1B). A yellow or red display in a quadrant map was defined as detection of an RNFL defect when the clock-hour location of the RNFL defects in the RNFL photographs was in the range of that on the OCT quadrant map. For example, 10:30 to 1:30 o’clock, 1:30 to 4:30 o’clock, 4:30 to 7:30 o’clock, and 7:30 to 10:30 o’clock for the right eye represented the superior, nasal, inferior, and temporal quadrants, respectively (Figure 1C).
“Deviation from normal maps” is provided in the optic disc cube mode of Cirrus OCT (Figure 1D). Noncolored areas indicate a normal RNFL thickness. Areas with RNFL thicknesses that occurred in less than 5% of age-matched controls are colored in yellow, and areas with RNFL thicknesses that occurred in less than 1% of age-matched controls are colored in red. Detection of RNFL defects on the Cirrus deviation map is considered to have occurred when a wedge-shaped color pattern across a 3.46-mm calculation circle is represented by yellow or red. The width of the wedge-shaped color pattern was measured by determining the angle between the 2 lines from the center of the disc to the points at which the color pattern and the 3.46-mm calculation circle converged.
One eye from each subject was randomly selected for analysis. A paired t test was used to evaluate the differences in the RNFL measurements generated by Stratus OCT and Cirrus OCT. The area under the receiver operating characteristic curve was analyzed to assess the ability of RNFL thickness measures to detect localized RNFL defects. The overall sensitivity and specificity were tested on the basis of the comparison of the measurements with a normative database built into the OCT equipment. The sensitivity and specificity of the OCT maps were compared by χ2 test, and multiple comparisons were compared with Bonferroni correction. We divided the eyes into 2 groups: those with and those without RNFL defects, according to the RNFL internal database. We used the unpaired t test to compare these 2 groups with respect to the widths of the localized RNFL defects observed in RNFL photography. We calculated the optimal cutoff point for the angle width of localized RNFL defects that could be used to distinguish between the presence and absence of an RNFL defect on OCT. The lowest limit for the angle of RNFL defects to be detected as abnormal in the 5 OCT maps was defined as the value that provided the best pair of sensitivity/specificity values based on receiver operating characteristic curve analysis.21 To investigate whether the ability to detect localized RNFL defects is affected by the width of localized RNFL defects, as determined by the RNFL photography, the Mantel-Haenszel χ2 test was applied to the data set. The correlation between the width of RNFL defects on RNFL photography and the width of the color pattern on the Cirrus OCT deviation map was analyzed to determine whether the photography and topographic information from the Cirrus deviation map were consistent with one another.
Statistical analysis was performed with SPSS for Windows (version 12.0.0; SPSS Inc, Chicago, Illinois). P < .05 was accepted to be statistically significant.
During the enrollment period, localized RNFL defects were observed in 341 eyes of patients with open-angle glaucoma. Of these, 166 eyes of 121 participants were examined with both Stratus OCT and Cirrus OCT, and 1 eye was selected from each participant. The quality of RNFL photography images was assessed, and the following eyes were excluded from the study: 13 eyes because of general RNFL atrophy despite definitive localized RNFL defects, 17 eyes because of ambiguous RNFL defect borders, and 9 eyes because of an additional diffuse or ambiguous defect located within the same hemifield as the localized RNFL defect. In addition, 16 eyes were excluded because of poor OCT images (signal strength <6, 10 eyes), epiretinal membrane (1 eye), and unreliable visual fields (5 eyes). Fifty-five eyes had single RNFL defects and 11 had 2 RNFL defects. Localized RNFL defects were most frequently located at 7 o’clock. The next most frequent location was 11 o’clock (Figure 2).
Table 1 summarizes the demographic characteristics of the participants. Sixty-six eyes from 66 participants and 66 eyes from 66 normal controls were analyzed. All the participants were Asian. There were 24 eyes with primary open-angle glaucoma and 42 with normal-tension glaucoma. Glaucomatous eyes exhibited early to moderate visual field damage with a mean deviation of −4.66 (3.98) dB (mean [SD]). No differences in the signal strength were observed between the scans acquired with the 2 devices (average signal strengths for Stratus OCT and Cirrus OCT were 8.20 [1.50] and 8.26 [1.04], respectively; t test, P = .71). Table 2 shows the RNFL thickness values along the whole calculation circle and by quadrants and clock hours for eyes with RNFL defects and the normal controls. The average RNFL thicknesses of eyes with localized RNFL defects, as measured by Stratus OCT and Cirrus OCT, were 86.11 (11.08) μm and 78.95 (8.74) μm, respectively (P < .001).
Table 3 shows the areas under the receiver operating characteristic curve with 95% confidence intervals. The discriminating ability of peripapillary RNFL sectoral measurements was similar for Stratus OCT and Cirrus OCT. The best-performing variable to discriminate between normal eyes and eyes with localized RNFL defects was the inferior RNFL thickness on Cirrus OCT (area under the curve, 0.94). We measured the angle of the wedge-shaped color pattern on the Cirrus deviation map, which demonstrated an almost equivalent discriminating ability (area under the curve, 0.91).
The overall sensitivities of the OCT maps for detecting localized RNFL defects that were identified on the basis of the RNFL normative database significantly differed from one another (P < .001). Although the clock-hour and quadrant criteria for Cirrus OCT showed slightly higher sensitivities than those for Stratus OCT, these differences in sensitivity were not statistically significant. The deviation map from Cirrus OCT showed superior sensitivity (92.4%) compared with the other maps (all P < .05). The overall specificities of the OCT maps were not significantly different from one another (P = .07). The quadrant map of Cirrus OCT exhibited a higher specificity for detection of RNFL defects (95.4%). However, there were no statistically significant differences in the specificities between Stratus OCT and Cirrus OCT (P > .05) (Table 4).
The mean angle of the RNFL defects on photography was significantly smaller in eyes without defects, as determined by using the normative databases, than in eyes with defects (all P < .05). The mean (SD) angle of RNFL defects detected by Cirrus OCT deviation maps was the smallest (21.07° [11.11°]) among the 5 maps, and the mean (SD) angle of RNFL defects that were overlooked by the Cirrus OCT deviation map was also the smallest (9.70° [3.73°]) (Table 5).
The optimal cutoff points for the angle width of RNFL defects were greater than 10.69° on the deviation map from Cirrus, greater than 21.13° on the quadrant map from Cirrus, greater than 13.74° on the clock-hour map from Cirrus, greater than 16.74° on the quadrant map from Stratus, and greater than 15.38° on the clock-hour map from Stratus. The Cirrus OCT deviation map had the smallest cutoff angle, 10.69°, which had an 83.6% sensitivity (95% confidence interval, 71.9%-91.8%) and 80.0% specificity (28.4%-99.4%) (Table 6).
Table 7 summarizes the sensitivities of the different cutoff points for the angle width of RNFL defects on OCT. The sensitivity of the Cirrus deviation map to detect the localized RNFL defect with angle widths less than 10° was 73%. This sensitivity was higher than those of the other maps. We examined whether a trend toward greater angle widths corresponded to a more sensitive detection of RNFL defects with the OCT normative database. The sensitivity of OCT to detect RNFL defects was closely related to the angle widths of the defects on RNFL photography (all P ≤ .05).
The angle width of RNFL defects on the Cirrus deviation map strongly correlated with that from RNFL photography (r = 0.88, P < .001) (Figure 3).
A previous study demonstrated that RNFL photography has a sensitivity of approximately 80% with a specificity of 94% for glaucomatous eyes with nerve fiber defects.22 Precise visualization of the RNFL requires a skilled imager, dilation, relatively clear media, and high contrast, which makes examination difficult in patients with cataract and lightly pigmented fundi. The use of OCT may be helpful in diagnosing RNFL defects, especially for physicians who are not as experienced in assessing RNFL on the basis of RNFL photography. Optical coherence tomography may be especially beneficial in cases of myopia with retinal pigment epithelial atrophy and among white patients for whom the quality of RNFL photography is sometimes insufficient. In addition, quantitative analysis is possible with OCT. Recently, a new generation of OCT devices relying on spectral-domain technology was introduced by different manufacturers. These instruments were reported to be able to detect localized RNFL defects in agreement with masked assessment of stereophotographs.23 Spectral-domain technology is capable of a faster scanning speed12 and offers a higher resolution than time-domain OCT.13 The fast scanning speed and the algorithm for automated placement of the scan circle are expected to reduce RNFL measurement variability. In addition, the higher axial resolution may improve the delineation accuracy of the RNFL, which would subsequently improve the diagnostic performance for glaucoma detection.
This study was designed to assess the validity of OCT maps compared with established methods for evaluating localized RNFL defects by means of RNFL photography. In this study, the peripapillary sectoral RNFL measurements and the clock-hour and quadrant map color scales based on normative databases from the optic disc cube scan of Cirrus OCT were not notably better than the corresponding profiles and maps from the fast RNFL scan mode of the conventional Stratus OCT in detecting localized RNFL defects, despite its higher resolution and more accurate method of data acquisition. Clock-hour and quadrant maps for both OCTs from 2 different generations are commonly based on the measurements acquired from cross-sectional 3.46-mm-diameter circumpapillary RNFL scans consisting of 256 A-scans. These cross-sectional scans could not demonstrate improved diagnostic advantages over those of spectral-domain OCT. Previous studies with spectral-domain OCT reported a similar discriminating ability for time-domain OCT and spectral-domain OCT; however, those studies compared the average, quadrant, and clock-hour peripapillary RNFL thickness measurements acquired from a cross-sectional scan circle of 3.46 mm in diameter.14-16
Spectral-domain OCT gathers a large amount of information with minimal sampling error, whereas time-domain OCT acquires RNFL thickness profiles predominantly by using circumpapillary scans. Much denser sampling and imaging of a broader area of spectral-domain OCT provides new analysis of RNFL. The maps created by “deviation from normal maps” from Cirrus OCT are derived from superpixel average thickness measurements of a 200 × 200 area, and they report the results of a statistical comparison against the normal thickness range for each superpixel overlaid on the OCT fundus image. Similar to the Zeiss GDx (Carl Zeiss Meditec Inc), these maps apply the yellow and red colors (not the green) of the age-matched normative data to superpixels in which average thickness falls in the yellow and red normal distribution percentiles. This analysis is a distinctive method of spectral-domain OCT used to detect glaucoma, which was not possible with time-domain OCT. In this study, the Cirrus deviation map derived from regional RNFL measurements exhibited a significantly higher sensitivity to detect localized RNFL defects than the other maps. In addition, the abnormal RNFL thickness in the Cirrus OCT deviation map is morphologically similar to that observed in RNFL photography; therefore, it is easier for clinicians to identify abnormal RNFL thickness that is presented in a quadrant or clock-hour map in a 3.46-mm-diameter circle.
The present study has several limitations that must be considered. We did not account for demographic and clinical factors, such as age, severity of visual field deficit, central corneal thickness, and refractive error. These variables have not been shown to affect the width of the RNFL defect to date. Although the red-free fundus image in the analysis printout shows the position of the scan circle, the scan circle captured in the red-free image may not reflect the exact location registered during OCT imaging.24 Our analysis is based on the assumption that RNFL photography is the criterion standard for detecting localized RNFL defects; however, the ideal method for detecting RNFL defects has not yet been established.
In conclusion, there was no difference between Cirrus OCT and Stratus OCT in terms of their capability of detecting localized RNFL defects with the clock-hour and quadrant maps. However, the Cirrus OCT deviation map was more sensitive than other maps generated from cross-sectional RNFL measurements. To determine the performance of Cirrus OCT for detecting localized RNFL defects, a longitudinal study is necessary.
Correspondence: Chan Yun Kim, MD, PhD, Institute of Vision Research, Department of Ophthalmology, Yonsei University College of Medicine, 134 Shinchon-dong, Seodaemun-gu, Seoul 120-752, Korea (email@example.com).
Submitted for Publication: October 5, 2009; final revision received December 16, 2009; accepted January 7, 2010.
Financial Disclosure: None reported.