Study Dataset

The study was approved by the University of North Carolina at Chapel Hill Institutional Review Board, adhered to the tenets of the Declaration of Helsinki, and conformed to Health Insurance Portability and Accountability Act regulations. The OCT RNFL data used in this study are those of 282 healthy subjects from the Cirrus HD-OCT normative database, 86 glaucomatous patients (46 mild, 23 moderate, 17 severe by Hodapp-Anderson-Parrish criteria ; visual field mean deviation: −7.4 ± 6.3 dB) from the Ganglion Cell Analysis FDA registration (GCA) study, and 49 nonglaucomatous myopic subjects (mean spherical equivalent: −4.6 ± 2.1 dB) from another earlier study.

Data Handling

Temporal-superior-nasal-inferior-temporal (TSNIT) profiles were extracted and used to generate average, quadrant, and clock-hour RNFL thicknesses. For each subject, the prototype Cirrus PanoMap feature and other custom analysis tools were used to register the macular OCT fundus image to the peripapillary ocular fundus image. An illustrative example of the PanoMap export image is shown in Figure 1 . The fovea-disc angle was determined using the PanoMap feature as follows: the affine transform parameters used by the algorithm were used to calculate the relative positions of the optic disc and macular centers for fovea-disc angle calculation. This was also done externally in Matlab using the disc positions and the affine transform parameters outputted from the registration exerciser.

Illustrative step-by-step example of the PanoMap export image for registration of the macular optical coherence tomography fundus image to the peripapillary ocular fundus image.

The fovea-disc angle, determined by the horizontal meridian through the center of the ONH and the axis passing through the fovea and the ONH center ( Figure 2 , Left) was used to correct the start of the TSNIT profile to a common landmark ( Figure 2 , Right). One set of nonparametric normative limits (1%, 5%, and 95%) was generated for both the standard and the corrected data from all of the normal eyes. Then halves of the normal eyes (first half, second half, and a randomly selected half) were selected to generate a second set of 5% limits. The remaining halves of the eyes from each set were used to test the specificities. The myopic subjects were also used to separately test the 95% specificity on the full set of normative limits.

Optical coherence tomography fundus image showing the standard horizontal axis (dotted line) through the center of the optic disc and fovea-disc angle formed by the horizontal axis and the line joining the center of the disc and the fovea (Left), and the same fundus image when the fovea-disc axis is taken as landmark (Right).

The locations of the RNFL TSNIT profile peaks and the blood vessel bundle peaks for the normative database and nonglaucomatous eyes were also determined using custom code. The RNFL peak index was found by running a localized peak finder algorithm on the RNFL TSNIT profile for each subject and then prioritizing the peaks that were closest to the canonical average RNFL profile from averaging the 282 normative database RNFL TSNIT profiles together (custom MATLAB code). This process was not performed in glaucomatous eyes because it was not likely to work very well owing to loss of thickness in the peaks. Blood vessel peaks were found by deriving a blood vessel map, extracting the TSNIT profile of the blood vessel map around the same RNFL TSNIT circle, and running a similar peak finder to the RNFL peak.

Statistical Analysis

All statistical analyses in this investigation were performed in MATLAB (R2012B). Paired t tests were used to compare standard and corrected RNFL data after landmarking to fovea-BMO. Receiver operating characteristics (ROC) analyses, including area under the curve (AUC) values, were calculated for each sector. Pairwise comparisons between standard and corrected AUCs were computed using the DeLong method. The correlation between the locations of the RNFL peaks and both the fovea-disc angle and the blood vessel peaks were compared by calculating the Pearson correlation coefficient. Statistical significance for tests was set at P < .05.

Effect of Fovea-Disc Angle Axis Correction on Retinal Nerve Fiber Layer Thickness

The fovea-disc angles across the study populations were 7.4 ± 4.2 degrees for all 3 study groups combined, 7.4 ± 4.1 degrees in the 282 subjects from the Cirrus normative database, 7.6 ± 4.2 degrees for the glaucomatous subjects, and 7.1 ± 4.3 degrees in the nonglaucomatous myopic subjects (all P > .05). Figure 3 shows the fovea-disc angle of 6 different eyes, as indication of fovea-disc variability among individuals. The 1%, 5%, and 95% normative limits for the standard and fovea-disc-corrected RNFL TSNIT profiles were calculated from the full normative database set (n = 282) and are illustrated in Figure 4 at a reference age of 46.7 years (the mean age of the normative database population). The mean TSNIT in each case is also shown for illustration purposes. Overall, the net effect of the correction on the limits was a “bulk” shift in the size of the mean fovea-disc angle (7.4 degrees) to the right. The same analysis performed in nonglaucomatous myopic subjects and in subjects with glaucoma yielded the same trends as in the normative database set; individual examples of RNFL TSNIT profile against the age-corrected normative limits in these 2 groups of subjects are shown in Supplemental Figure 1 (supplemental material available at AJO.com ).

Fovea-disc angle variability across 6 different eyes.

Normative 1%, 5%, and 95% limits for the standard (Top) and fovea-disc-corrected (Bottom) retinal nerve fiber layer temporal-superior-nasal-inferior-temporal (TSNIT) profiles calculated from the full normative database set (n = 282).

The mean thicknesses of the 17 RNFL parameters in the 282 subjects from the normative database prior to and following fovea-disc correction are presented in Figure 5 . No changes in the average RNFL values were observed after correction, the correction being merely a circular shift. Small but significant changes were observed in the RNFL quadrant thicknesses, whereas larger significant changes occurred in clock-hour thicknesses (all P < .05). Similar trends were observed in the standard deviation and coefficient of variation values ( Supplemental Figure 2 , available at AJO.com ). However, there were no discernible spatial patterns in the direction of the change in the various RNFL parameters, as some increased while others decreased.

Histogram bar representation of global and quadrant (Top), and clock-hour (Bottom) retinal nerve fiber layer thickness before (blue) and after (red) adjusting to fovea-disc angle.

Effect of Fovea-Disc Angle Axis Correction on False-Positive Rates and Glaucoma Diagnostic Accuracy

The false-positive rates using the 95% limits derived from the normative database subjects (all 282, first half, second half, or random half) and the nonglaucomatous myopic subjects performed with and without fovea-disc correction failed to show a consistent pattern as to which method has a better specificity ( Table 1 ). The pre- and post-fovea-disc correction specificity of RNFL thickness remained the same (4.96%) for all 17 RNFL parameters in the 282 subjects from the normative database. In the first half, the specificity remained unchanged for global RNFL and 2 clock hours following fovea-disc correction; it improved in 3 quadrants and 5 clock hours but worsened in 1 quadrant and 5 other clock hours. In the second half, the specificity worsened in 3 quadrants and 5 clock hours, improved in 1 quadrant and 4 clock hours after correction, and did not change in 3 clock hours. When fovea-disc correction was applied to the random half of the normative database subjects, the specificity remained the same in 3 clock hours, decreased in 2 quadrants and 5 clock hours, and increased in 2 other quadrants and 4 clock hours. For the group of nonglaucomatous myopic subjects, the specificity did not change for global RNFL, in 1 quadrant and 7 clock hours, it decreased in 2 quadrants and 2 clock hours, and it increased in 1 quadrant and 3 clock hours.

The results of the pairwise comparison of AUC values of the 17 RNFL measured before and after fovea-disc correction are presented in Table 2 . The AUCs of 6 RNFL parameters (global, superior and inferior quadrants, and clock hours 9, 1, and 5) showed no significant differences (all P > .05) following the correction. The temporal and nasal quadrants significantly improved their AUCs following fovea-disc correction, but the AUCs were only fair (0.70–0.79). Among the remaining 9 clock hours, the AUC increased in 4 but decreased in 5 of them, without consistent pattern with regard to direction of the change.

Table 2

Regular and Fovea-Disc-Adjusted Retinal Nerve Fiber Later Thickness in the Normative Database

RNFL Parameter	Regular AUC	Corrected AUC	P
Global	0.9618	0.9618	1.00
Temporal quadrant	0.7346	0.7583	<.01
Superior quadrant	0.9576	0.9579	.90
Nasal quadrant	0.6928	0.7181	<.01
Inferior quadrant	0.9630	0.9594	.14
Clock hour 9	0.5880	0.5689	.18
Clock hour 10	0.7825	0.7300	<.01
Clock hour 11	0.9091	0.8718	<.01
Clock hour 12	0.8780	0.9146	<.01
Clock hour 1	0.8800	0.8787	.90
Clock hour 2	0.7498	0.7901	<.01
Clock hour 3	0.5280	0.5658	<.01
Clock hour 4	0.6328	0.5728	<.01
Clock hour 5	0.8428	0.8291	.13
Clock hour 6	0.9178	0.8946	<.01
Clock hour 7	0.9507	0.9675	<.01
Clock hour 8	0.7266	0.8466	<.01

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. A Comparative Study of Rebound Tonometry With Tonopen and Goldmann Applanation Tonometry Following Vitreoretinal Surgery
2. Microvascular Abnormalities on Optical Coherence Tomography Angiography in Macular Edema Associated With Branch Retinal Vein Occlusion
3. Evaluation of Objective Vitritis Grading Method Using Optical Coherence Tomography: Influence of Phakic Status and Previous Vitrectomy
4. Validation of Concentric Rings Method as a Topographic Measure of Retinal Nonperfusion in Ultra-widefield Fluorescein Angiography

Stay updated, free articles. Join our Telegram channel

Join