Measurement of Optic Disc Size and Rim Area with Spectral-Domain OCT and Scanning Laser Ophthalmoscopy

Abstract

Purpose.: To compare optic disc and neuroretinal rim area measurements from spectral-domain optical coherence tomography (SD-OCT) to those from confocal scanning laser ophthalmoscopy.

Methods.: Seventy-one eyes from 43 normal subjects or suspected/definite glaucoma patients were prospectively enrolled. All subjects had biometry with the IOLMaster and disc/retinal nerve fiber layer imaging with Cirrus SD-OCT (Optic Disc Cube 200×200) and Heidelberg Retina Tomograph (HRT). Uncorrected disc and rim areas and measurements corrected for eye magnification with Bennett's formula (AL-corrected measurements), along with 30° sectoral rim areas, vertical cup-to-disc ratio (VCDR), and cup volume, were compared between the two devices.

Results.: The median (range) axial length (AL) was 24.2 mm (22.4–27.7 mm). Mean keratometry-corrected HRT disc area measurements were larger than AL-corrected HRT and SD-OCT measurements (P < 0.001 for both) and the difference was a function of keratometry measurements (K-readings). The AL-corrected HRT disc area and uncorrected/corrected Cirrus disc areas were not significantly different (P > 0.481). HRT rim area was larger than Cirrus measurements (P < 0.001) and the difference decreased with decreasing rim area. HRT VCDR and cup volume were significantly smaller than Cirrus measurements (P < 0.001). The correlations for sectoral rim areas between the two devices were moderate at best (intraclass correlation coefficients = 0.12–0.65).

Conclusions.: HRT overestimated optic disc area as compared to SD-OCT. A portion of the difference in HRT and SD-OCT disc measurements is due to HRT's magnification correction algorithm. Rim area measurements from HRT were larger than from SD-OCT, likely a result of different definitions for the reference plane and differences in disc area measurements. Disc parameters from the two devices are not interchangeable.

Introduction

Measurement of the optic disc and neuroretinal rim is important for the diagnosis and management of glaucoma. ^1,2 It is increasingly recognized that the disc size is a major determinant of other disc parameters such as neuroretinal rim area or cup area or volume. ^3–6 Accurate measurement of the disc size in vivo is challenging. The main confounding factor for calculating the disc and neuroretinal rim areas is the magnification of the eye's optical system, which varies as a function of the dimensions of the eye. Biometric factors affecting the ocular magnification are refractive error, corneal curvature, anterior chamber depth, and most importantly, the axial length. ^7–10  

Optical coherence tomography (OCT) and scanning laser ophthalmoscopy have become important diagnostic tools for detection of glaucoma. ^11–14 The two devices use different approaches for detecting the disc margin or defining the reference plane, which forms the boundary between the cup and neuroretinal rim. High-resolution imaging of the peripapillary structures with spectral-domain OCT (SD-OCT) machines has shed light on the structures forming the disc border. From recent SD-OCT findings, the inner edge of the Bruch's membrane or, less commonly, the border tissue of Elschnig, are considered to constitute the neural canal opening, although the disc border defined clinically does not always match the location of the neural canal opening. ^15–17 The Heidelberg Retina Tomograph (HRT) and SD-OCT use different definitions for describing the rim–cup junction and handle the magnification of the eye's optical system in different ways. The HRT uses keratometry measurements to correct for eye's magnification. In contrast, there is no built-in system to correct for magnification with OCT devices in general. ^9,18,19 Although HRT and OCT measure similar parameters of the optic disc topography, there is evidence that the measurements may not be interchangeable. ^{2,4,18,20–28} A large amount of variation is observed in these studies with regard to estimated disc size with HRT, OCT, and fundus photographs. Moreover, different methodologies are used and most of the published studies do not correct for the eye's magnification. The purpose of this study was to compare the optic disc topographic measurements with Cirrus SD-OCT to those from HRT and to determine the sources of disagreement between the two devices. 

Methods

This prospective observational study was approved by the Institutional Review Board of the University of California Los Angeles. All participants signed an informed consent and the study protocol adhered to the tenets of the Declaration of Helsinki. Phakic eyes with reliable fields with mean deviation >−15.0 dB, spherical equivalent <8 diopters (D), and astigmatism <3 D, and no prior glaucoma surgery, were prospectively enrolled. Eyes with any evidence of other significant ocular diseases, such as retinopathies or neurologic problems or with peripapillary atrophy extending >1.7 mm (the radius of the Cirrus SD-OCT retinal nerve fiber layer measurement circle) beyond the border of the disc, were excluded. 

Eligible subjects underwent a comprehensive ophthalmic evaluation, including visual acuity measurement, autorefraction, slit-lamp examination, intraocular pressure measurement with Goldmann applanation tonometry, gonioscopy, and dilated fundus examination. Perimetry was performed with either the standard Swedish Interactive Thresholding Algorithm (SITA) or SITA short wavelength automated perimetry (SITA-SWAP) with the 24-2 pattern on the Humphrey Field Analyzer (Carl Zeiss Meditec, Dublin, CA). Refraction was recorded with an automated refractometer (Speedy-1; Nikon, Tokyo, Japan). Axial length was measured with an ocular biometry unit (IOLMaster; Carl Zeiss Meditec, Dublin, CA). Nonsimultaneous stereoscopic photographs of the optic disc were taken with a fundus camera (FF3; Carl Zeiss AG, Germany). Patients also underwent disc and retinal nerve fiber layer (RNFL) imaging with SD-OCT (Optic Disc Cube 200×200, Cirrus HD-OCT; Carl Zeiss Meditec) and HRT (software version 3.01; Heidelberg Engineering, Heidelberg, Jena, Germany). The optic disc photographs were separately reviewed by two experienced clinicians (K.N-M. and J.G.) and defined as normal, suspicious for glaucoma, or glaucomatous. In cases of disagreement, the two clinicians reviewed the disc photographs together again to reach agreement. 

Imaging with HRT was performed after entering each patient's keratometry measurements into the software. Three consecutive scans are obtained, aligned, and then averaged by HRT's software to create a single mean topography image for analysis. Images were included if the global pixel standard deviation was <50 μm. Optic disc contour was marked at the inner border of the scleral ring by an experienced clinician (S.M.) while viewing the stereoscopic optic disc photographs. Stereometric parameters including disc area, rim area, vertical cup-to-disc ratio, and cup volume were exported to a personal computer. Thirty-degree rim area sectors (centered on clock hours, right eye format) were also calculated and exported. 

Definitions

Primary open-angle glaucoma was defined as presence of glaucomatous optic nerve damage (i.e., vertical cup-to-disc ratio of >0.6, or cup-to-disc asymmetry >0.2, or the presence of focal thinning or notching) and an associated visual field defect, which correlated with the area of disc damage. A visual field defect was considered to be present if both of the following criteria were met: (1) Glaucoma Hemifield Test outside normal limits; and (2) four abnormal points with P < 0.05 on the pattern deviation plot, both confirmed at least once. These criteria have been shown to be highly specific and have demonstrated reasonable sensitivity for the detection of early glaucomatous visual field loss. ²⁹ Glaucoma suspect eyes were defined as those with evidence of glaucomatous optic neuropathy or suspicion thereof, from review of stereoscopic photographs along with normal or borderline visual fields (i.e., the visual field did not meet the criteria for presence of a defect as mentioned above). Normal subjects had open angles, no evidence of optic neuropathy or RNFL loss, and normal or borderline visual field as described above. 

Imaging and Measurements with Spectral-Domain OCT

The qualifying eye(s) of each participant was (were) dilated with 1% tropicamide and 2.5% neosynephrine before imaging. All SD-OCT scans were acquired with a Cirrus HD-OCT (software version 3.0.0.64) by using the Optic Disc Cube 200×200 protocol. The disc margin is determined automatically by the software without manual modification. Only good-quality scans, defined as scans with signal strength of at least 7 and no missing parts within the measurement circle and no motion artifacts, were used for analysis. The boundary of the optic disc (disc contour line on Cirrus) is defined as the inner termination of Bruch's membrane or Bruch's membrane opening (BMO). A plane located 200 μm above the level of the Bruch's membrane plane is defined as the reference plane or the plane separating the neuroretinal rim from the cup. Disc parameters including disc area, rim area, vertical cup-to-disc ratio (VCDR: ratio of a vertical line passing through the cup center and connecting the edges of the cup to the same vertical line extending to the disc margin), and cup volume were calculated with the Cirrus research browser version 5.0.0.326. To make these measurements, the Cirrus HD-OCT disc and RNFL algorithm automatically identifies the center of the optic disc and creates simulated radial B-scans every 2° (a total of 180 B-scans) around the disc perimeter. Thirty-degree rim area sectors centered on clock hours were then calculated as follows. The Cirrus research software provides the rim width for 2° pie-shaped disc segments. Each 2° rim sector was considered as a trapezoid, with the outer base formed by the disc edge and the inner edge by the cup–rim junction, and size of the base was estimated as disc perimeter divided by 180. The 2° sector areas were summed up over 15 segments to calculate the area for each 30° rim sector. The Cirrus SD-OCT rim sectors were created so as to correspond to the 30° HRT sectors in right eye format (i.e., centered on clock hours). 

Estimation of the Corrected Disc Size

Although based on recent findings by Reis et al., ³⁰ the clinical disc border and the neural canal opening may not coincide in an individual eye or in parts of the optic disc in the same eye and, for the purposes of this study, the clinical disc border was defined as the inner edge of what has been considered to be the scleral ring. 

For estimation of the corrected or “real” disc size, the automated SD-OCT disc area measurements were exported to a personal computer. The authors then used Bennett's formula to correct for the magnification of the eye's optical system.⁷ The relationship between the measured SD-OCT image and the actual size of the disc can be expressed as t = p * q * s, where t is the actual dimensions or size of an object, s is the SD-OCT measurement, p is the magnification factor related to the OCT's camera, and q is the magnification factor related to the eye. By using the default axial length (AL = 24.46 mm) and refraction (0 D) for a magnification of 1 with the Cirrus SD-OCT system (i.e., t = s), p can be calculated as 1/[0.01306 * (24.46 − 1.82)] = 3.382. The q factor based on AL would be 0.01306 * (AL − 1.82) according to Bennett's formula and therefore, disc area measurements obtained from the Cirrus SD-OCT's en face images should be corrected by the following fudge factor: 3.382² * 0.01306² * (AL −1.82)².¹⁸ 

 

So the final corrected OCT disc size will be:  

 

 

Registration of HRT Images to En Face Cirrus SD-OCT Images and Measurement of AL-Corrected HRT Disc Area

The current HRT software uses keratometry measurements (K-readings) for correcting for the magnification of the eye's optical system (K-corrected measurements). However, this has been demonstrated to be less than optimal. ⁸ Therefore, to make the comparison of the disc area measurements between the two devices more meaningful and to determine sources of discrepancy between the two devices, the authors also applied Bennett's formula to HRT disc area measurements (see below). 

The Cirrus en face images were exported to a personal computer in bmp format and opened with ImageJ (provided in the public domain by National Institutes of Health, Bethesda, MD, http://rsb.info.nih.gov/ij). The disc border is marked on such images on the basis of the location of the inner edge of Bruch's membrane. The measurement circle (3.46 mm in diameter in an emmetropic eye) is also shown on the en face Cirrus images. The Adobe Illustrator CS5 software (Adobe, San Jose, CA) was used to register HRT images (reflectance images as input images; resolution, 72 pixels per inch) to that of en face Cirrus images (base images; resolution, 96 pixels per inch). Transparency of HRT image was increased to from 50% to 60%, such that the details of underlying (base) image, such as disc border and retinal vasculature, could be easily observed. The two images were manually aligned (using rotation, translation, or magnification correction, if necessary) with retinal vessels as landmarks and saved as a single combined image file. Accuracy of the overlay images was verified by another observer (K.N-M.). The resulting image was opened in ImageJ software (version 1.43u). Measurement scale was set by using the 3.46-mm RNFL circle on the en face combined image. The HRT disc area (as marked by the contour line) and the area of the measurement circle were then computed on the en face images in pixels. After calculating disc area to measurement circle area ratio, the disc area could be subsequently measured with imageJ software as follows:  

 

The authors then applied Bennett's formula to correct for eye magnification as described above, using the same correction factor used for the Cirrus images. Examples of combined HRT/Cirrus overlay images are shown in Figure 1.  

Figure 1. 

 

Examples of HRT/Cirrus SD-OCT overlay images with the black line delineating disc border as defined by the Cirrus, and the green line representing HRT contour line. (A) An eye with very close agreement between HRT contour line and Cirrus disc border. (B) An eye with the Cirrus disc border located within the HRT contour line. (C) An eye with a larger Cirrus disc area compared to HRT disc border. (D) An example where the estimated Cirrus disc area, based on Bruch's membrane opening, is much larger than the clinically identified disc border on HRT (such eyes were excluded from analyses).

Measurement of Corrected Cirrus SD-OCT Rim Area

For calculating the corrected Cirrus global rim areas, the cup areas were first measured in a similar fashion as described above for the HRT disc areas by using imageJ software. The Cirrus global rim area measurements were then calculated by subtracting the cup area from the disc area and corrected with Bennett's formula. The rim area sectors were not corrected since the exact area for the sectors could not be calculated on en face images. 

Statistical Analysis

The P values for paired comparison of numerical (continuous) variables from HRT and OCT were computed by using the nonparametric Wilcoxon's signed rank test (Stata, version 11.0; Stata Corp., College Station, TX). Pearson's correlations and intraclass correlation coefficients (ICCs) were computed to quantify the association between topographic measurements of the two instruments correcting for correlation of the two eyes of the same patients. Univariate and multivariate regression analyses were performed to evaluate potential predictors influencing the difference between measurements from the two devices. The agreement of optic disc parameters obtained by the two instruments and corrected Cirrus SD-OCT measurements was evaluated with Bland-Altman plots to detect proportional bias (MedCalc version 11.6; MedCalc Software, Mariakerke, Belgium). Presence of proportional bias indicates that the methods do not agree equally through the range of measurements. P values less than 0.05 were considered significant. 

Results

Seventy-seven eyes (46 subjects) were available for this study. Three eyes were excluded secondary to blurred en face Cirrus SD-OCT images, making registration of HRT images unreliable. Three more eyes had large areas of peripapillary atrophy, where the Bruch's membrane was detected well outside the clinical disc border as determined by qualitative evaluation of the images by two of the authors (K.N-M. and H.H.). Therefore, data from 71 eyes (of 43 subjects) were used in this study, consisting of 13 normal eyes, 21 glaucoma suspect eyes, and 37 eyes with definite glaucoma. The demographic and biometric characteristics of the study sample are summarized in Table 1. The median (range) axial length was 24.2 (22.4–27.7) mm. The axial length and spherical equivalent for the study sample were skewed towards longer axial lengths and more myopic refractive errors, respectively (Fig. 2). 

Table 1. 

 

Demographic and Biometric Characteristics of the Enrolled Study Sample

Variable
No. of eyes (patients)	71 (43)
Age, y, mean ± SD	64.4 ± 7.9
Sex (F/M)	27/16
Visual acuity (LogMAR), mean ± SD	0.07 ± 0.09
Diagnosis (eyes)
Normal	13 (18.3%)
Glaucoma suspect	21 (29.6%)
Glaucoma	37 (52.1%)
Intraocular pressure, mm Hg, mean ± SD	13.6 ± 3.2
Axial length, mm, median (range)	24.2 (22.4–27.7)
Spherical equivalent, D, median, range	−1.2 ± 2.4
Mean deviation, dB, mean ± SD	−2.5 ± 3.4
Pattern standard deviation, dB, mean ± SD	3.5 ± 3.2

Figure 2. 

 

Distribution of the axial length (A) and refractive error (B) in 71 eyes of 43 patients.

The mean ± SD uncorrected Cirrus optic disc area was 1.86 ± 0.46 mm², which was not significantly different from corrected Cirrus disc area measurements (1.87 ± 0.46 mm², P = 0.655, Wilcoxon's signed rank test). The K-corrected HRT disc area was on average larger than the AL-corrected HRT disc area (2.02 ± 0.50 vs. 1.89 ± 0.49 mm², respectively; P < 0.001, Wilcoxon's signed rank test). The difference between the AL-corrected HRT and uncorrected or corrected Cirrus disc area measurements were not significant (P > 0.481). However, the distribution or spread of the differences between the HRT and Cirrus SD-OCT disc areas did not significantly decrease after correction of the HRT disc area for magnification with Bennett's method (P = 0.262, nonparametric resampling test). Table 2 compares the main uncorrected HRT and Cirrus parameters. Magnification correction could be performed only for the disc area with HRT images because of technical reasons. Table 3 demonstrates the pairwise ICCs between the K-corrected and AL-corrected HRT disc areas and uncorrected and corrected Cirrus disc areas (Figs. 3A, 3B). Bland-Altman plot showed a slight negative trend for the difference of K-corrected HRT disc area and corrected Cirrus disc area with increasing disc size, which was not statistically significant (β = −0.052, P = 0.454; Fig. 4A). The K-corrected HRT disc area measurements were consistently higher than corrected Cirrus measurements at all axial lengths (Fig. 5). The results of univariate regression analyses demonstrated that axial length (β = 0.063; P = 0.346) and disc area (β = 0.040; P = 0.454) did not have any influence on the difference between the K-corrected HRT and corrected Cirrus disc areas. The effect of spherical equivalent was of borderline significance (β = −0.025; P = 0.058), whereas the average keratometry measurement was significantly associated with the disc area difference (β = 0.044; P = 0.008). Results of multivariate analyses, including corrected Cirrus disc area, AL (or spherical equivalent), keratometry readings, age, or glaucoma severity, did not change the results (P = 0.007 vs. 0.014 for K-readings in analyses including AL vs. spherical equivalent, respectively). The results were fairly similar for the difference between K-corrected and AL-corrected HRT disc area (P = 0.087, 0.273, and 0.454 for AL, spherical equivalent, and disc area, respectively, and P = 0.005 for K-reading). The difference between uncorrected and corrected Cirrus disc areas did not vary as a function of corrected disc area, or K-readings, when adjusted for axial length (data not shown). 

Table 2. 

 

Comparison of Disc Parameters between HRT (Corrected with Keratometry Measurements) and Cirrus SD-OCT (Both Uncorrected and Corrected for Eye Magnification Using Bennett's Formula)

	K-Corrected HRT	Uncorrected Cirrus	P Value*	Corrected Cirrus	P Value†	P Value‡
Disc area, mm2, mean ± SD	2.02 ± 0.51	1.86 ± 0.46	<0.001	1.87 ± 0.46	<0.001	0.393
Rim area, mm2, mean ± SD	1.09 ± 0.37	0.88 ± 0.30	<0.001	0.94 ± 0.26	<0.001	<0.001
VCDR, mean ± SD	0.60 ± 0.21	0.68 ± 0.14	<0.001	N/A	N/A	N/A
Cup volume, mm3, mean ± SD	0.27 ± 0.28	0.42 ± 0.35	<0.001	N/A	N/A	N/A
1-o'clock rim area, mm2, mean ± SD	0.11 ± 0.04	0.10 ± 0.05	0.170	N/A	N/A	N/A
2-o'clock rim area, mm2, mean ± SD	0.10 ± 0.04	0.11 ± 0.06	0.909	N/A	N/A	N/A
3-o'clock rim area, mm2, mean ± SD	0.10 ± 0.04	0.11 ± 0.07	0.695	N/A	N/A	N/A
4-o'clock rim area, mm2, mean ± SD	0.11 ± 0.05	0.12 ± 0.07	0.526	N/A	N/A	N/A
5-o'clock rim area, mm2, mean ± SD	0.12 ± 0.05	0.11 ± 0.06	0.166	N/A	N/A	N/A
6-o'clock rim area, mm2, mean ± SD	0.11 ± 0.05	0.10 ± 0.05	0.045	N/A	N/A	N/A
7-o'clock rim area, mm2, mean ± SD	0.09 ± 0.05	0.07 ± 0.04	<0.001	N/A	N/A	N/A
8-o'clock rim area, mm2, mean ± SD	0.06 ± 0.03	0.06 ± 0.03	0.427	N/A	N/A	N/A
9-o'clock rim area, mm2, mean ± SD	0.05 ± 0.03	0.10 ± 0.07	<0.001	N/A	N/A	N/A
10-o'clock rim area, mm2, mean ± SD	0.06 ± 0.03	0.07 ± 0.07	0.120	N/A	N/A	N/A
11-o'clock rim area, mm2, mean ± SD	0.08 ± 0.04	0.09 ± 0.04	0.512	N/A	N/A	N/A
12-o'clock rim area, mm2, mean ± SD	0.11 ± 0.04	0.09 ± 0.04	0.003	N/A	N/A	N/A

*   P value for comparison of HRT and uncorrected Cirrus disc parameters.

†   P value for comparison of HRT and corrected Cirrus disc parameters.

‡   P value for comparison of uncorrected and corrected Cirrus disc parameters.

Table 3. 

 

Intraclass Correlation of Uncorrected and Corrected Disc Area Measurements with HRT and Cirrus SD-OCT Adjusted for the Correlation between the Two Eyes of the Same Subject

	AL-Corrected HRT	Uncorrected Cirrus	Corrected Cirrus
K-corrected HRT	0.830 (0.826–0.833)	0.757 (0.750–0.764)	0.807 (0.804–0.811)
AL-corrected HRT		0.785 (0.779–0.790)	0.887 (0.885–0.888)
Uncorrected SD-OCT			0.905 (0.904–0.906)

Figure 3. 

 

(A) Scatter plot for comparison of disc area corrected with keratometry versus axial length (Bennett's formula) as measured with HRT. (B) Scatter plot for comparison of disc area as measured with HRT and Cirrus SD-OCT, both corrected for eye's magnification according to Bennett's formula using axial length.

Figure 4. 

 

Bland-Altman plots for comparison of disc area (A) and rim area (B) measurements obtained by HRT and Cirrus SD-OCT (corrected for eye's magnification). The corresponding slopes and P values were −0.053 and 0.454 for disc area and 0.427 and <0.001 for rim area.

Figure 5. 

 

Scatter plot showing the correlation between the difference in the disc area as measured with HRT (corrected with K-readings) and Cirrus SD-OCT (corrected with Bennett's formula) and axial length.

The average ± SD uncorrected Cirrus global rim area was 0.88 ± 0.29 mm² and increased to 0.94 ± 0.26 mm² after correction for magnification (P = 0.003, paired t-test) (Table 2, Fig. 6). The K-corrected HRT rim area measurements were significantly higher (1.09 ± 0.37 mm²) than both uncorrected and corrected Cirrus rim areas (P < 0.001 for both, paired t-test). Although the rim area measurements from the two instruments were moderately correlated (ICC = 0.433 and 0.496 for correlations with uncorrected and corrected Cirrus rim area, 95% confidence interval [CI] = 0.391–0.476 and 0.464–0.528, respectively) (Fig. 7), the correlations were weaker than those of the disc area. The uncorrected and corrected Cirrus rim area measurements demonstrated a good correlation (ICC = 0.801, 95% CI = 0.796–0.805). Bland-Altman plots revealed that the difference in the rim area between the two devices (K-corrected HRT minus corrected Cirrus measurements) became larger as the average rim area increased (β = 0.43; P < 0.001) (Fig. 4B). On univariate and multivariate analyses, neither axial length nor the disc area had any influence on the difference between K-corrected HRT and corrected Cirrus rim areas (P > 0.24 for all). Table 4 demonstrates the correlation of 30° rim sector areas between Cirrus and HRT. The correlations were mostly fair except in the temporal (9-o'clock) sector, where the correlation was nonsignificant (ICC = 0.115, 95% CI = −0.120 to 0.351). 

Figure 6. 

 

The box plots demonstrate the median and 95% confidence intervals for the rim area as measured with HRT (corrected for K-readings) and Cirrus SD-OCT (uncorrected and corrected for eye's magnification with Bennett's formula).

Figure 7. 

 

Scatter plots comparing the rim area as measured by HRT to those from Cirrus SD-OCT, uncorrected (A) and corrected for eye's magnification (B).

Table 4. 

 

ICCs for Rim Sectors Measured with Cirrus SD-OCT (Uncorrected) versus HRT (Corrected with K-Readings)*

Rim Parameter	ICC	95% CI
1-o'clock rim area	0.474	0.439–0.509
2-o'clock rim area	0.473	0.438–0.509
3-o'clock rim area	0.367	0.311–0.424
4-o'clock rim area	0.430	0.388–0.474
5-o'clock rim area	0.498	0.467–0.53
6-o'clock rim area	0.659	0.646–0.673
7-o'clock rim area	0.593	0.574–0.614
8-o'clock rim area	0.572	0.55–0.594
9-o'clock rim area	0.115	−0.12–0.351
10-o'clock rim area	0.550	0.526–0.575
11-o'clock rim area	0.593	0.574–0.613
12-o'clock rim area	0.471	0.436–0.507

*  The P value for all ICCs was <0.001 for all comparison except for the 9-o'clock rim area (P > 0.05).

The results for comparison of HRT and Cirrus SD-OCT vertical cup-to-disc ratio and cup volume and the respective Bland-Altman regression coefficients are presented in Table 2 and Figure 8. Both VCDR and cup volume estimated by the two devices showed good correlations (ICC = 0.623, 95% CI = 0.603–0.643 and ICC = 0.765, 95% CI = 0.758–0.771, respectively; P < 0.001 for both). Overall, the HRT measurements for cup volume and VCDR tended to be smaller than the uncorrected Cirrus measurements. 

Figure 8. 

 

Bland-Altman plots for comparison of cup volume (A) and VCDR (B), as measured with HRT and Cirrus SD-OCT, show a negative trend (β = −0.24, P < 0.001) for cup volume and positive trend (β = 0.43, P < 0.001) for VCDR.

Discussion

The authors compared the disc and rim areas measured with HRT and Cirrus SD-OCT in a group of healthy and glaucomatous subjects. They found that the disc area was significantly larger when measured with HRT (corrected with K-readings) than with the uncorrected Cirrus measurements or the Cirrus disc area corrected for eye's magnification with Bennett's formula (estimated real disc size). This difference was not significantly related to axial length or disc area but was positively influenced by corneal power (K-readings). AL–corrected HRT disc measurements more closely approximated Cirrus measurements, suggesting that the disagreement between measurements from the two devices depended on the method used for correction of magnification and on the definition of the disc border. Uncorrected disc area measurements obtained from Cirrus SD-OCT were quite similar to corrected measurements. Likewise, the neuroretinal rim area was, on average, larger on K-corrected HRT measurements. However, the difference was positively correlated with the average rim area and was not influenced by the axial length or disc area. Conversely, the cup-related parameters (VCDR and cup volume) were larger with the Cirrus than K-corrected HRT measurements. The cup volume measurements were not corrected for possible axial magnification, since there is no evidence in the literature on whether this is necessary, and if so, on the best way to perform this task. 

Disc size is a major determinant of other disc parameters, and the size of both rim and cup parameters are highly dependent on the disc size. ^1,5,6 It has been suggested that the optic disc size needs be considered to optimize predictive performance of HRT parameters. The RNFL thickness, as measured by OCT and scanning laser polarimetry, also correlates with the disc size. ^27,31,32 The differences in disc measurements between any two devices can be attributed to two main factors: differences in methods used for detection of the disc border and differences in the magnification correction or lack thereof. The measured size of a feature in the fundus is dependent on the magnification of the camera (p factor) and magnification of the optical system of the eye (q factor). Various methods have been explored to correct for the q factor, based on the magnitude of ametropia, keratometry, or axial length. ^7,8,33 Garway-Heath et al. ⁸ have shown that Bennett's method, which uses the axial length as the main correction factor, is more accurate than methods using keratometry (HRT) or keratometry and ametropia (Littman's formula). The authors used Bennett's formula for estimating the “real” disc area and the corrected Cirrus rim area in this study. 

In contrast to Stratus OCT, the Cirrus SD-OCT determines the termination of Bruch's membrane as the edge of the disc (or neural canal opening), which should lead to more consistent and clinically more accurate measurements. ^17,18,28 The authors observed three eyes in their sample for which the disc border identified by the Cirrus algorithm was grossly inconsistent with the clinical disc border because of the temporal displacement of the edge of Bruch's membrane. In myopic eyes with varying amounts of disc tilt, the Bruch's membrane can end before what is perceived to be the clinical border of the disc on the temporal side, and in such eyes, the Bruch's membrane opening does not match the clinical disc margin. ¹⁶  

The K-corrected HRT disc area measurements were on average higher than corrected (and uncorrected) Cirrus measurements. The difference was not influenced by the disc size in this study but was rather a function of the corneal power (keratometric measurements) and tended to be smaller in eyes with larger axial length, although the trend was not statistically significant. Leung et al. ¹⁸ have used Bennett's formula to correct OCT disc measurements and found larger optic disc area measurement with Stratus OCT than with HRT in eyes with longer axial length and myopic refraction. 

AL–corrected HRT disc area measurements approximated those from uncorrected and corrected Cirrus SD-OCT in this study. This suggests that the more important reason for the discrepancy between HRT and Cirrus measurements is the lack of adequate correction for magnification by the HRT despite using keratometry for this task. It must be noted that determination of the disc border (contour line) with HRT is less than accurate, since a spline curve is fit to the disc border after an operator identifies 5 to 6 points on the disc border. The average corrected Cirrus disc area was slightly smaller than the average corrected HRT disc area, which is consistent with recent reports by Strouthidis et al. ¹⁶ in rhesus monkey eyes and Sharma and colleagues ¹⁷ and Manassakorn et al. ²⁸ in humans. More recently, Reis et al. ³⁰ have demonstrated that what is clinically considered to be the inner edge of the scleral ring does not always match the inner edge of the neural canal opening, which is to be considered the true border of the disc. They have shown that the latter is variably formed by the inner edge of BMO or the border tissue of Elschnig. The inner edge of Bruch's membrane was found to be clinically invisible and frequently was located within the clinically identified disc border. Cirrus SD-OCT defines the disc border as the inner edge of the BMO and therefore would be expected to be more consistent with the new definition of the disc border by Reis et al. ³⁰ Despite the lack of a large difference between the AL-corrected HRT and corrected Cirrus SD-OCT disc area measurements, significant variability between the two measurements persisted. The authors compared the variances of the difference in disc area measurements before and after magnification correction with a nonparametric resampling test (P = 0.262). A visual comparison of the distribution of the differences before and after correction of magnification also confirmed that although the spread of the distribution of the difference in disc area measurements improved after correction for magnification, using Bennett's method, significant variability (i.e., differences in individual eyes) persisted. The frequency and extent of possible errors in identification of the inner edge of BMO by Cirrus SD-OCT need to be further investigated. The authors found that correction of Cirrus measurements for the eye's magnification did not significantly change the average disc size in this study sample. The effect of magnification correction depends on the distribution of the axial length in the study sample, with the effect being more prominent if a larger number of eyes with longer or shorter axial length are included. A visual inspection of the individual Cirrus B-scan images (Disc Cube 200×200 algorithm of Cirrus HD-OCT) revealed that the algorithm detecting the edge Bruch's membrane opening sometimes failed to correctly identify it. Interestingly, the difference between AL-corrected HRT disc area and corrected Cirrus disc area was not a function of AL (P = 0.381, data not shown). 

Rim area measurements with the two devices followed a similar pattern, compared to the optic disc area measurements, with the caveat that a “real” rim area could not be defined since the rim area measurements highly depend on the definition for the reference plane. The difference in rim area measurements was a function of rim area, but was not affected by the axial length or disc size. Rim area measurements tended to be larger with the HRT. This is in contrast to previous studies that used older versions of OCT. ^18,34,35 The default reference plane in the Stratus OCT is determined by tracing a line with an anterior offset of 150 microns parallel to a line joining the ends of the RPE while for Cirrus SD-OCT, it is set at 200 microns above the Bruch's membrane edge. The reference plane for HRT is set 50 microns below and parallel to the peripapillary temporal retinal surface. The definition of the reference plane has a direct effect on the rim and cup area measurements. In this study, the difference in K-corrected HRT and corrected Cirrus rim areas increased with larger rim areas. This proportional bias has not been reported previously. ² This is, however, partly a mathematical issue. As the average rim area decreases and approaches the measurement floor, there is less room for a difference to exist. It is plausible to assume that the difference in rim area is partially due to imprecise magnification correction, as is the case with the disc area. One possible reason for the lower agreement of the rim areas from the two devices is the fact that Cirrus SD-OCT rim area measurements were calculated by subtracting the cup area from the disc area, making the Cirrus rim area variances (and SDs) slightly larger in some cases (Table 2). However, this does not appear to be the major reason for the fair correlation seen between rim areas from the two devices, since the SDs (and variances) were similar in most cases. There was also no evidence that the correlation between disc and rim areas from the two devices was different in control subjects/glaucoma suspects versus glaucoma patients (Figs. 3, 7) 

The agreement of cup-to-disc ratio (CDR) measurements between the earlier versions of HRT and OCT has been previously evaluated. ^23,26,36 Hoffman and coworkers ² have reported a fixed bias in CDR, with higher measurements with OCT than HRT. The authors found, on average, a larger VCDR with Cirrus than HRT with a proportional bias; that is, the difference between the two tended to intensify in eyes with larger VCDR. The same factors responsible for larger rim area measurements with HRT are likely at work here. 

The correlations for the rim sector areas between HRT and Cirrus were moderate at best, with ICCs ranging from 0.12 to 0.66. Specifically, the pairwise correlation was lowest in the temporal sector (ICC = 0.115, P > 0.05). In the current study, a large minority of the patients were myopic, with some demonstrating temporal peripapillary atrophy. In these cases, Bruch's membrane is more likely to be misidentified or actually end before the perceived temporal border of the disc. ¹⁶ This may explain larger rim area measurements with Cirrus than with HRT and the low correlation between the two methods in the temporal rim sector, despite the fact that three eyes with obvious temporal migration of Bruch's membrane were excluded. This finding might also be related to the slow sloping of the temporal rim area, especially in myopic eyes, which would potentially make rim area measurements larger and more variable. 

The limitations of this study need to be considered. Because of the fairly low resolution of the en face Cirrus images or HRT reflectance images (98 and 72 pixels per inch, respectively), the AL-corrected HRT disc measurements may have not been as accurate as desired. The axial length distribution in the patient sample was skewed towards longer values. This might have exaggerated the difference in the disc and rim area measurements between HRT and Cirrus SD-OCT. At the same time, it provided an opportunity to explore the effect the lack of magnification correction would have on disc parameters in eyes deviating from average values, which is a frequently encountered problem in tertiary care setting. Although Bennett's formula is the best available correction algorithm, it is not perfect, emphasizing the inherent difficulty in estimating in vivo disc dimensions. Finally, the sectoral neuroretinal rim areas were calculated by assuming that the optic disc was circular and that the larger sides of the trapezoids were parallel. These assumptions may not be true especially in eyes with small CDR or oval disc shape. The calculated global RA, based on the sum of all 180 two-degree sectors, was found to be larger than automated Cirrus measurements, especially in eyes with larger global rim area. This may have led to underestimation of the correlation coefficients for rim sectors. 

In summary, this study demonstrated that HRT and Cirrus SD-OCT disc or rim area measurements are not interchangeable. The K-corrected HRT measurements generally overestimated the disc and rim areas. The difference between HRT and Cirrus SD-OCT disc area measurements was mainly a function of anterior corneal curvature and likely a sign of incomplete correction for the eye's magnification by HRT's algorithm, while the difference in rim area was dependent on the magnitude of the rim area. Uncorrected Cirrus SD-OCT disc size measurements were fairly accurate and can be used clinically especially in eyes with average axial length. There was only moderate agreement in sectoral rim area measurements, VCDR, and cup volume between HRT and Cirrus SD-OCT. 

Acknowledgments

The authors thank Jeffrey Gornbein, DrPH, for assistance with statistical analyses. 

Footnotes

 This study was presented as a paper at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 1 to 5, 2011.

Footnotes

 Supported in part by an Early Career Clinician-Scientist Grant from the American Glaucoma Society (KN-M) and Research to Prevent Blindness.

Footnotes

 Disclosure: S. Moghimi, None; H. Hosseini, None; J. Riddle, None; G.Y. Lee, None; E. Bitrian, None; J. Giaconi, None; J. Caprioli, None; K. Nouri-Mahdavi, None