Diffuse Retinal Nerve Fiber Layer Defects Identification and Quantification in Thickness Maps

Abstract

Purpose.: To report retinal nerve fiber layer (RNFL) defect identification and quantification in RNFL thickness maps according to the structural RNFL loss, and to evaluate diffuse RNFL defects.

Methods.: A total of 170 patients with glaucoma and 186 normal subjects were consecutively enrolled. We defined RNFL defects in an RNFL thickness map by the degree of RNFL loss. The reference level for RNFL defect determination was set as a 20% to 70% degree of RNFL loss with a 1% interval. To identify RNFL defects, each individual RNFL thickness map was compared to the normative database map by using MATLAB software, and the region below the reference level was detected. The area, volume, location, and angular width of each RNFL defect were measured. Diffuse RNFL defects were defined as having an angular width > 30°.

Results.: The optimal reference level for glaucomatous RNFL defects identification was 42% loss of RNFL. Retinal nerve fiber layer defects were identified in all (100%) of the 170 glaucoma patients and false-positive RNFL defects were detected in 16 (8.16%) cases among the 186 normal subjects. In all, 64.1% of glaucoma patients had diffuse RNFL defects, and 47.7% of diffuse RNFL defects were associated with mild glaucoma patients. The volume of diffuse RNFL defects was significantly associated with the severity of glaucomatous damage (P = 0.009). Diffuse RNFL defects were located closer to the center of the optic disc than localized RNFL defects (P < 0.001).

Conclusions.: Retinal nerve fiber layer thickness map analysis is an effective method for analyzing RNFL defects. Quantitative measurements (area, volume, location, and width) were useful to understanding diffuse RNFL defects.

Introduction

Glaucoma is a progressive optic neuropathy characterized by injury to the retinal ganglion cells, leading to loss of the retinal nerve fiber layer (RNFL). Retinal nerve fiber layer defect detection is important for diagnosing and managing glaucoma. Optical coherence tomography (OCT) allows objective and quantitative evaluation of the RNFL. ¹ Most studies have investigated the detection ability of OCT in glaucoma patients with definite localized RNFL defects identified on red-free fundus photography. ^2–5 However, only 20% of glaucoma patients have localized photographic RNFL defects. ⁶ In histologic studies, clinical detection of RNFL defects by red-free fundus photography is possible after a 50% loss of the RNFL. ⁷ We have observed that RNFL thinning measured by OCT does not appear as a defect in the red-free fundus photography. Moreover, diffuse RNFL defects are the initial structural abnormality of early glaucoma in 50% of cases. ^8,9 Despite the fact that many cases could have diffuse defects or invisible defects with red-free fundus photography, little is known about evaluating RNFL defects in such circumstances. 

In this study, we identified RNFL defects by the degree of RNFL loss, using RNFL thickness map of spectral-domain OCT (Cirrus HD-OCT; Carl Zeiss Meditec, Dublin, CA, USA) regardless of the presence of localized photographic RNFL defects, and we evaluated the effective degree of RNFL loss for the identification of glaucomatous RNFL defects. In addition, we quantitatively measured the area, volume, location, and width of RNFL defects in order to evaluate diffuse RNFL defects. 

Methods

Participants

A total of 170 glaucoma patients and 186 normal control subjects were consecutively enrolled. They had visited either the general health care clinic or the glaucoma clinic at Hanyang University Medical Center from September 2012 to January 2013. The study protocol was approved by the Institutional Review Board of Hanyang University Medical Center and adhered to the tenets of the Declaration of Helsinki. Informed consent was obtained from all participants before participation. 

All subjects underwent a comprehensive ophthalmic examination, which included a visual acuity test, applanation tonometry, anterior segment examination, refraction, fundus examination, pachymetry (SP-3000; Tomey, Nagoya, Japan), standard automated perimetry (Humphrey Field analyzer with SITA standard 30-2 test; Carl Zeiss Meditec), and RNFL imaging with a spectral-domain OCT (Cirrus HD-OCT). 

Inclusion criteria of normal subjects were best-corrected visual acuity of ≥20/30, normal anterior segment on slit-lamp examination, normal visual field, normal-appearing optic disc head, no RNFL defects, and no history of intraocular pressure > 21 mm Hg. Inclusion criteria of glaucoma subjects were the presence of RNFL defects on red-free photographs or glaucomatous appearance of the optic nerve head on color fundus photographs (neuroretinal rim notching or thinning, or optic disc hemorrhage) and the presence of visual field defects that corresponded to the RNFL defects or optic nerve head abnormalities. Visual field defects were defined as (1) a cluster of three or more nonedge contiguous points with probabilities of <5% on the pattern deviation plot, at least one of which was depressed below the 1% level, (2) glaucoma hemifield test results outside of normal limits, or (3) a pattern standard deviation (PSD) with a P value < 0.05 as confirmed by at least two reliable examinations. The severity of glaucomatous damage was classified into mild (mean deviation ≥ −6 dB) and moderate to advanced (mean deviation < −6 dB). The visual field tests were considered reliable on the basis of fixation losses and false-positive and false-negative results of 15% or less. 

The exclusion criteria consisted of best-corrected visual acuity worse than 20/30, spherical equivalent refractive errors of less than −6.0 diopters (D) or greater than +3.0 D, or the presence of any ophthalmic or neurologic disease known to affect RNFL thickness or visual function. 

Retinal Nerve Fiber Layer Image Acquisition

An Optic Disc Cube scan protocol was used to generate the RNFL thickness map with a Cirrus HD-OCT (software version 5.1; Carl Zeiss Meditec) with an 840-nm wavelength light source, a 5-μm axial image resolution, and a speed of 27,000 A-scans per second. This protocol consisted of 200 × 200 axial scans (pixels) on the 6 × 6-mm² optic disc region. The built-in analysis software automatically segmented the RNFL boundary and calculated the RNFL thickness. Retinal nerve fiber layer segmentation was checked for every OCT image. A built-in algorithm automatically detected the center of the optic disc, and its coordinates were displayed in the result report as the degree of movement from the center of RNFL thickness map (e.g., “Disc Center [−0.02, 0.04] mm”). To compare the same area from the center of the optic disc, the center of the optic disc was realigned to the center of the RNFL thickness map (equal to [0.00, 0.00] mm). All images had a signal strength of at least 7. Images with motion artifacts were rescanned at the same visit. 

Normative RNFL Thickness Map Database

A normative database was made up of another 261 eyes of healthy Korean subjects ranging in age from 18 to 81 years (mean, 56.4 ± 10.2 years). All individuals in the normative database had complete ophthalmic examinations and were selected by using the same criteria as the normal control group. The mean deviation of the visual field test was 0.27 ± 1.03 dB. The distribution of age was as follows: 44 eyes in 18- to 30-year-olds, 41 eyes in 31- to 40-year-olds, 45 eyes in 41- to 50-year-olds, 48 eyes in 51- to 60-year-olds, 41 eyes in 61- to 70-year-olds, and 42 eyes in >70-year-olds. Significant negative correlations were found between age and RNFL thickness (−0.26 μm/y; P = 0.014). To accurately differentiate age-related RNFL change from glaucomatous RNFL loss, the normative RNFL thickness map database was adjusted by age-related rate of change. 

Identification of RNFL Defects

Retinal nerve fiber layer thickness measurements of 200 × 200 pixels of the RNFL thickness map were extracted by a Cirrus HD-OCT Research Browser (Carl Zeiss Meditec). In the RNFL thickness map, we defined RNFL defects by the degree of RNFL loss. A 20% to 70% degree of RNFL loss with a 1% interval was set as the reference level for RNFL defect determination. To identify the boundary of RNFL defects, each individual RNFL thickness map was compared to the normative database map by using MATLAB software (The MathWorks, Inc., Natick, MA, USA; Fig. 1) and the region below the reference level (20%–70% loss of RNFL with 1% interval) was detected (Fig. 2). We excluded the eyes that moved more than 0.25 mm from the center of the RNFL thickness map, and analyzed within a 5.5-mm square. The optic disc area (range, 1.00–4.12 mm²) varies among subjects, so we removed the 2.5-mm-diameter circular area in the RNFL thickness map. 

Figure 1

 

Identification and measurement of the area, volume, center, and angular width of RNFL defects in thickness map. RNFL defects are detected by comparison (B) between normative (A) and individual (C) RNFL thickness maps. Through line-by-line analysis, the region ([B]: red zone; [C]: red transverse line indicated by red arrowheads) below the reference level ([B]: red line) is recognized as an RNFL defect ([C]: red dashed line). In this case, the reference level is set as 42% loss of RNFL, which is equal to 58% of normative data ([B]: 58% level of green line equal to red line). The area, volume, center, and angular width are automatically measured according to the identified RNFL defect boundary ([C]: text). Angular width was determined where the boundary of RNFL defect met the circle passing through the center of RNFL defect ([C]: green dot and arc). RNFL thickness maps were analyzed within a 5.5-mm square ([A, C]: inside black square, [B]: inside black dot line) and outside the 2.5-mm-diameter circular area ([A, C]: outside black circle).

Figure 2

 

Identified RNFL defects according to the percentage RNFL loss. A 20% to 70% degree of RNFL loss with a 1% interval was set as the reference level for RNFL defect determination. The higher the degree of RNFL loss set, the smaller the RNFL defect region (red line) detected.

Retinal Nerve Fiber Layer Defects Quantification

The quantitative characteristics (area, volume, location, and angular width) of RNFL defects were measured by a computer program written by using MATLAB software (Fig. 1C). 

The area of RNFL defects was calculated as follows:  

 

 

The volume of RNFL defects was calculated by the sum of each pixel's volume.  

 

 

To evaluate the location of RNFL defect, the center of RNFL defect was calculated.  

 

where R is the center of mass, M is the sum of the masses, m_i is the mass of particle, and r_i is the coordinates of particle.  

The center of the RNFL defect was described with polar coordinates, in which each point on a plane is determined by a distance from the center of the optic disc and an angle from a temporal equator. Angles were measured in a clockwise direction in right eyes and in a counterclockwise direction in left eyes, with the temporal equator set at 0°. The angular width of RNFL defects was determined where the boundary of RNFL defects met the circle passing through the center of RNFL defect. A diffuse RNFL defect was defined as having an angular width > 30°. ¹⁰ Multiple RNFL defects were separately measured. If multiple RNFL defects were present in superior or inferior quadrants, localized or diffuse RNFL defects were determined by the sum of the angular width within each quadrant. 

Statistical Analyses

Statistical analyses were performed by using SPSS software (version 18.0; SPSS, Inc., Chicago, IL, USA) and MedCalc (version 12.2.1; MedCalc Software, Ostend, Belgium). To evaluate the difference between the normal and glaucoma groups or between localized and diffuse RNFL defect groups, subject demographics and quantitative measurements of RNFL defect were compared by using an independent t-test and Pearson χ² test. To determine the appropriate reference level for glaucomatous RNFL defect identification, the glaucoma diagnostic ability of RNFL defect area and volume among several reference levels were compared by using the areas under the receiver operating characteristic curve (AUCs). Significant differences between AUCs were assessed by using the method described by DeLong et al. ¹¹ Univariate and multivariate logistic regressions were performed to evaluate the association of age, refractive error, signal strength, disc area, rim area, average RNFL thickness, and RNFL defect parameters (area, volume, angular location, and distance from optic disc center) with the severity of glaucomatous damage in diffuse RNFL defect. 

Results

Subject demographics are summarized in Table 1. There were significant differences in signal strength, mean deviation, PSD, rim area, and average RNFL thickness between the normal and glaucoma groups. Among 170 glaucoma patients, 31 (18.2%) had visible RNFL defects in red-free fundus photography. Retinal nerve fiber layer thickness map analysis, which detects at least 42% loss of RNFL, identified RNFL defects in all (100%) cases. Among 186 normal subjects, false-positive RNFL defects were detected in 16 (8.16%) cases in RNFL thickness map. The area, volume, and width of false-positive defects were significantly smaller than those of glaucomatous defects (all P < 0.001). False-positive defects were located in the superonasal (102.7°) or inferonasal (253.3°) quadrants, in contrast to glaucomatous defects, which were located in the superotemporal (71.8°) or inferotemporal (290.5°) quadrants (P < 0.001). False-positive defects were not associated with age, sex, intraocular pressure, signal strength, refractive error, mean deviation, PSD, disc area, rim area, or average RNFL thickness (P > 0.05). 

Table 1

 

Comparison of Various Characteristics Between Normal and Glaucoma Groups

	Normal	Glaucoma	P
N	186	170
General characteristics
Age, y	56.7 ± 11.2	57.3 ± 13.6	0.628*
Sex, male:female	100:86	100:71	0.430†
Intraocular pressure, mm Hg	15.2 ± 3.4	14.9 ± 3.2	0.214*
Signal strength	8.54 ± 0.93	7.70 ± 0.71	<0.001*
Refractive error, D	0.01 ± 1.27	−0.02 ± 2.47	0.947*
MD, dB	0.15 ± 1.19	−6.29 ± 7.20	<0.001*
MD ≥ −6:MD < −6	186:0	102:68	<0.001†
PSD, dB	1.42 ± 1.63	5.33 ± 4.67	<0.001*
Disc area, mm2	2.15 ± 0.51	2.09 ± 0.48	0.420*
Rim area, mm2	1.33 ± 0.33	0.97 ± 0.93	<0.001*
Average RNFL thickness, μm	103.6 ± 6.7	75.7 ± 11.8	<0.001*
Defects characteristics
Area, mm2	0.11 ± 0.34	3.48 ± 2.42	<0.001*
Volume, mm3	0.004 ± 0.014	0.199 ± 0.153	<0.001*
Width, °	4.5 ± 10.8	39.0 ± 28.8	<0.001*
Angular location, °
Superior	102.7 ± 5.2	71.8 ± 22.6	<0.001*
Inferior	253.3 ± 10.5	290.5 ± 19.6	<0.001*
Distance from disc center, mm	2.21 ± 0.5	2.14 ± 0.3	0.229*

P  values of 0.01 or less were considered statistically significant. MD, mean deviation.

*  Independent t-test.

†  χ² test.

The AUCs, sensitivity, and specificity according to the degree of RNFL loss are presented in Figure 3. At 42% loss of RNFL, the RNFL defect area (AUC, 0.978; sensitivity, 95.3%; specificity, 90.9%) and volume (AUC, 0.980; sensitivity, 95.9%; specificity, 91.9%) showed the highest glaucoma diagnostic performance. Average RNFL thickness (AUC, 0.953; sensitivity, 86.0%; specificity, 92.5%), which is analyzed by OCT's built-in algorithm on the 3.46-mm circle, showed the highest performance among circumpapillary RNFL parameters, followed by inferior (AUC, 0.933; sensitivity, 83.5%; specificity, 91.0%), and 7 o'clock (AUC, 0.927; sensitivity, 81.8%; specificity, 89.6%). The area and volume of RNFL defects had significantly greater AUCs than average RNFL thickness in the range of 35% to 46% loss of RNFL (all P < 0.01). Based on this result, the reference level for glaucomatous RNFL defect identification was set to 42% loss of RNFL in this study. 

Figure 3

 

The AUCs (A), sensitivity (B), and specificity (C) of area and volume of RNFL defects according to the degree of RNFL loss. At 42% loss of RNFL, glaucoma diagnostic performance was highest for area and volume measurements of RNFL defects. In the range of 35% to 46% loss of RNFL, area and volume of RNFL defect had significantly greater AUCs than average RNFL thickness (between vertical dot line, [A]).

Table 2 presents the distribution of the angular width and location of 307 RNFL defects identified on the RNFL thickness maps, when multiple RNFL defects were separately measured. Among 170 glaucoma patients, 62 (36.5%), 83 (48.8%), 22 (12.9%), 2 (1.2%), and 1 (0.6%) had single, double, triple, quadruple, and quintuple RNFL defects, respectively. The RNFL defects located most frequently at the inferotemporal area (42.0%), followed by the superotemporal area (32.5%), the superonasal area (11.0%), the inferonasal area (10.0%), and the temporal area (4.5%). The most common angular width of RNFL defect was 10° to <20° (28.6%), and RNFL defects with an angular width > 30° were 47.5%. 

Table 2

 

Distribution of Angular Width and Location of RNFL Defects Identified by RNFL Thickness Map

Width, °	Superotemporal, 45° to <90°	Superonasal, 90° to <135°	Nasal, 135° to <225°	Inferonasal, 225° to <270°	Inferotemporal, 270° to <315°	Temporal, 315° to <45°	Total
<10	5 (1.6)	1 (0.3)	0	0	7 (2.3)	1 (0.3)	14 (4.5)
10 to <20	28 (9.1)	7 (2.3)	0	9 (2.9)	36 (11.7)	8 (2.6)	88 (28.6)
20 to <30	17 (5.5)	9 (2.9)	0	9 (2.9)	21 (6.8)	4 (1.3)	60 (19.4)
30 to <40	11 (3.6)	6 (2.1)	0	5 (1.6)	13 (4.2)	1 (0.3)	36 (11.8)
40 to <50	7 (2.3)	2 (0.7)	0	4 (1.3)	10 (3.3)	0	23 (7.6)
50 to <60	4 (1.3)	2 (0.7)	0	2 (0.7)	8 (2.6)	0	16 (5.3)
60 to <70	5 (1.6)	3 (1.0)	0	1 (0.3)	10 (3.3)	0	19 (6.2)
70 to <80	3 (1.0)	1 (0.3)	0	1 (0.3)	8 (2.6)	0	13 (4.2)
80 to <90	5 (1.6)	0	0	0	7 (2.3)	0	12 (3.9)
≥90	15 (4.9)	2 (0.7)	0	0	9 (2.9)	0	26 (8.5)
Total	100 (32.5)	33 (11.0)	0	31 (10.0)	129 (42.0)	14 (4.5)	307 (100)

 Data shown as number of RNFL defects (%).

Defining diffuse RNFL defect as having a sum of angular width > 30° in superior or inferior quadrants (Figs. 4A, 4B, localized RNFL defect; Figs. 4C–E, diffuse RNFL defect), 109 (64.1%) glaucoma subjects had diffuse RNFL defects. The comparison between localized and diffuse RNFL defects is presented in Table 3. The area and volume (4.64 mm² and 0.28 mm³, respectively) of diffuse RNFL defects were significantly greater than those (1.45 mm² and 0.08 mm³, respectively) of localized RNFL defects (P < 0.001). Diffuse RNFL defects were located closer to the center of the optic disc than localized RNFL defects (P < 0.001). There was no significant difference in angular location between both groups. Diffuse RNFL defects showed a higher degree of myopia, more severe visual field defects, and a thinner average RNFL thickness than localized RNFL defects (all P < 0.001). The proportion of mild and moderate-to-advanced visual field defects was 50:11 (82.0%:18.0%) in the localized defect group and 52:57 (47.7%:52.3%) in the diffuse defect group (P < 0.001). When analyzed within mild glaucoma group, significant differences between localized and diffuse RNFL defect were found in area, volume, and distance location of RNFL defect, refractive error, and average RNFL thickness (all P < 0.01). 

Figure 4

 

(A) The visible localized RNFL defects (red arrowhead) in red-free RNFL photography could be detected in original (red arrowhead) and analyzed (red dashed line) RNFL thickness map. (B) The invisible localized RNFL defect in red-free RNFL photography might be missed in original RNFL thickness map, but could be detected in analyzed RNFL thickness map. (C) The visible diffuse RNFL defects in red-free RNFL photography could be detected by less fine striation of RNFL bundles (red curved arrow) in contrast to opposite side with healthy striation (green curved arrow). The boundaries of diffuse RNFL defects were difficult to determine in red-free RNFL photography. In the original RNFL thickness map, defects were determined subjectively by location of dark blue areas (red arrowhead). Analyzed RNFL thickness map could objectively delineate the boundaries of diffuse RNFL defects corresponding with visual field tests. (D) The invisible diffuse RNFL defects in red-free RNFL photography had the broadly diminished striations of RNFL bundles in both superior and inferior quadrants. It is difficult to determine the boundaries in the original RNFL thickness map owing to the broad ambiguous dark blue color patterns. Analyzed RNFL thickness map showed diffuse double RNFL defects with clear boundaries. The wider inferior RNFL defect corresponded to the more severe visual field defect in the superior quadrant. (E) Red-free RNFL photography and original RNFL thickness map showed double visible localized RNFL defects (red arrowhead). However, analyzed RNFL thickness map showed triple RNFL defects. The sum (>30°) of inferior RNFL defects and corresponding visual field defects was suitable for diffuse RNFL defects. In this case, the original RNFL thickness map missed RNFL defects proceeding beyond large vessels (asterisk).

Table 3

 

Comparison of Various Characteristics Between Localized and Diffuse RNFL Defects

	Localized RNFL Defects			Diffuse RNFL Defects			Comparisons (P Value)
	All, a	Mild, b, MD ≥ −6	Moderate-Advanced, c, MD < −6	All, d	Mild, e, MD ≥ −6	Moderate-Advanced, f, MD < −6	Between
							a–d	b–c	b–e	e–f
N	61	50	11	109	52	57
General characteristics
Age, y	55.4 ± 12.7	53.8 ± 13.0	62.5 ± 9.0	58.5 ± 13.9	55.7 ± 13.4	61.2 ± 14.1	0.145*	0.034†	0.475*	0.031*
Sex, male:female	35:26	32:18	3:8	65:44	35:17	30:27	0.901‡	0.041§	0.886‡	0.172‡
Intraocular pressure, mm Hg	15.0 ± 3.5	15.2 ± 3.7	13.9 ± 2.0	14.8 ± 3.1	14.9 ± 3.2	14.7 ± 2.9	0.681*	0.255†	0.626*	0.700*
Signal strength	7.79 ± 0.68	7.80 ± 0.63	7.72 ± 0.96	7.64 ± 0.75	7.78 ± 0.74	7.60 ± 0.94	0.647*	0.751†	0.841*	0.571*
Refractive error, D	0.69 ± 2.31	0.94 ± 11.42	−0.43 ± 1.49	−0.41 ± 2.67	−0.68 ± 2.47	−0.21 ± 2.99	<0.001*	0.567†	<0.001*	0.147*
MD, dB	−2.92 ± 3.98	−1.27 ± 1.87	−9.21 ± 3.65	−8.03 ± 7.84	−2.15 ± 2.55	−10.69 ± 7.01	<0.001*	<0.001†	0.067*	<0.001*
MD ≥ −6:MD < −6	50:11	-	-	52:57	-	-	<0.001‡	-	-	-
PSD, dB	2.84 ± 2.64	2.25 ± 1.71	5.07 ± 4.13	6.62 ± 4.95	3.27 ± 3.19	9.75 ± 4.29	<0.001*	0.016†	0.054*	<0.001*
Disc area, mm2	2.07 ± 0.46	2.06 ± 0.47	2.09 ± 0.45	2.10 ± 0.50	2.04 ± 0.44	2.15 ± 0.55	0.664*	0.948†	0.838*	0.237*
Rim area, mm2	1.05 ± 0.23	1.06 ± 0.24	1.02 ± 0.19	0.93 ± 1.16	1.11 ± 1.64	0.76 ± 0.28	0.428*	0.666†	0.785*	0.105*
Average RNFL thickness, μm	84.9 ± 6.7	85.5 ± 6.8	82.6 ± 6.2	70.4 ± 10.8	75.1 ± 9.8	66.2 ± 9.9	<0.001*	0.244†	<0.001*	<0.001*
Defects characteristic
Area, mm2	1.45 ± 0.93	1.44 ± 0.93	1.46 ± 0.98	4.64 ± 2.23	3.79 ± 1.79	5.42 ± 2.32	<0.001*	0.721†	<0.001*	<0.001*
Volume, mm3	0.080 ± 0.057	0.079 ± 0.056	0.083 ± 0.061	0.281 ± 0.143	0.219 ± 0.112	0.338 ± 0.147	<0.001*	0.866†	<0.001*	<0.001*
Width, °	17.9 ± 6.5	17.7 ± 6.0	19.7 ± 8.5	62.6 ± 25.4	42.8 ± 22.2	75.3 ± 26.0	<0.001*	0.302†	<0.001*	<0.001*
Angular location, °
Superior	70.9 ± 25.8	68.7 ± 21.3	75.8 ± 11.3	72.9 ± 18.5	71.8 ± 24.5	74.4 ± 23.2	0.602*	0.028†	0.180*	0.584*
Inferior	292.6 ± 22.6	292.0 ± 21.6	295.5 ± 18.6	288.3 ± 15.4	288.0 ± 19.9	288.7 ± 17.9	0.155*	0.237†	0.123*	0.956*
Distance from disc center, mm	2.32 ± 0.30	2.32 ± 0.28	2.32 ± 0.38	2.03 ± 0.27	2.12 ± 0.29	1.95 ± 0.26	<0.001*	0.924†	0.003*	0.004*

P  values of 0.01 or less were considered statistically significant.

*  Independent t-test.

†  Mann-Whitney U test.

‡  χ² test.

§  Fisher's exact test.

Within the diffuse RNFL defect group, moderate-to-advanced glaucoma had significantly greater area, volume, and width of RNFL defects than mild glaucoma (P < 0.001). The distance from the center of the optic disc to the center of the RNFL defect was shorter in moderate-to-advanced disease than in mild disease (P = 0.004). However, within the localized RNFL defect group, these parameters did not show significant differences between mild and moderate-to-advanced glaucoma. Univariate and multivariate logistic regression analyses are summarized in Table 4. The severity of glaucomatous damage in diffuse RNFL defects was associated with the rim area, the average RNFL thickness, and the area, volume, and width of the RNFL defect by univariate logistic regression analysis (P < 0.05). Among these factors, the volume of the RNFL defect was found to be significantly associated with the severity of glaucomatous damage in diffuse RNFL defects with multivariate logistic regression analysis (odds ratio: 4.843, 95% confidence interval: 1.536–11.327, P = 0.011). 

Table 4

 

Logistic Regression Analysis to Determine the Various Factors Associated With the Severity of Glaucomatous Damage in Diffuse RNFL Defects

	Univariate		Multivariate
	OR (95% CI)	P	OR (95% CI)	P
General characteristics
Age	1.013 (0.995–1.039)	0.156
Refractive error	1.094 (0.964–1.243)	0.115
Signal strength	0.949 (0.824–1.244)	0.239
Disc area	1.556 (0.713–3.394)	0.260
Rim area	0.117 (0.022–0.620)	0.003	1.013 (0.487–2.108)	0.972
Average RNFL thickness	0.909 (0.868–0.952)	<0.001	0.923 (0.842–1.013)	0.090
Defects characteristics
Area	1.476 (1.198–1.819)	<0.001	0.844 (0.363–2.142)	0.674
Volume	2.049 (1.445–2.905)	<0.001	4.843 (1.536–11.327)	0.011
Width	1.031 (1.012–1.049)	<0.001	1.001 (0.973–1.029)	0.928
Angular location
Superior	1.103 (0.978–1.215)	0.420
Inferior	0.973 (0.961–1.121)	0.225
Distance from disc center	0.994 (0.985–1.003)	0.191

P  values of 0.05 or less were considered statistically significant. OR, odds ratio; CI, confidence interval.

Discussion

How much RNFL loss is required for clinical detection of glaucomatous damage in human eyes? Quigley and Addicks ⁷ have reported that clinical detection of RNFL defects is possible after a loss of 50% of the neural tissue in primate eyes. However, RNFL thinning could be more extensive than it appears on fundus examination or red-free RNFL photography. ² Retinal nerve fiber layer defects affecting glaucoma development might begin at less than a 50% loss of RNFL. Optical coherence tomography can provide an answer with high-resolution, cross-sectional images that allow in vivo measurement of tissue thickness. ^12,13 In this study, we found that the optimal reference level for glaucomatous RNFL defect identification was 42% loss of RNFL, using spectral-domain OCT. The higher the degree of RNFL loss set as a reference level, the smaller the RNFL defect region detected (Fig. 2). From this, we can infer that red-free RNFL photography might miss or underestimate the actual extent of RNFL defects. 

Previous studies that assessed RNFL thickness maps have defined RNFL defects as dark blue or blue/black areas. ^14,15 These criteria were effective for finding visible localized RNFL defects with clear margins in red-free photography. However, only 18.2% of glaucoma subjects had photographically visible RNFL defects (Fig. 4A), and color interpretation might be subjective, leading to false-negative results (Figs. 4B, 4E) or difficulty in determining the boundaries of RNFL defects (Fig. 4D). Defining RNFL defects as structural RNFL loss could provide an objective approach for the interpretation of RNFL thickness maps. Retinal nerve fiber layer defects, defined by 42% loss of RNFL in this study, were identified in all glaucoma patients regardless of whether they were visible or invisible in red-free photography and whether they were localized or diffuse. 

Three-dimensional volumetric RNFL thickness map analysis showed better glaucoma diagnostic performance than circumpapillary RNFL thickness measurement. Better diagnostic ability can be explained by the selective analysis of RNFL damage. Because the damaged RNFL is mixed with the adjacent normal RNFL in clock-hour sector, dilution of glaucomatous damage is inevitable. Circumpapillary RNFL thickness may miss or underestimate RNFL defects, so it cannot exactly measure the damaged area (Fig. 5A). In addition, RNFL defect, which exists outside 3.46-mm-diameter scan circle, may be ignored (Fig. 5B). In this study, we selectively analyzed the damaged area defined by at least 42% loss from normative RNFL thickness database, such that the new parameters more closely reflect actual RNFL change. 

Figure 5

 

(A) Three-dimensional volumetric RNFL analysis detected diffuse and localized RNFL defects consistent with red-free RNFL photography (red arrowheads). Inferotemporal localized RNFL defect distributed from 312.2° to 327.7° (from 7 to 8 o'clock), and superotemporal diffuse RNFL defect distributed from 40.5° to 83.5° (from 10 to 12 o'clock). In conventional circumpapillary analysis, damaged RNFL is mixed with adjacent normal RNFL in clock-hour sectors, so its influence is diluted. Circumpapillary RNFL thickness missed localized RNFL defect and underestimated diffuse RNFL defect (black arrows; expected RNFL defect range). (B) Superotemporal localized RNFL defect existed outside the 3.46-mm-diameter scan circle. It leads to misidentification in circumpapillary RNFL thickness measurement. Inferotemporal diffuse RNFL defects (sum > 30°) were distributed from 259.8° to 316.3° including vessels between double defects. Fortunately, it nearly fits in the 6- to 7-o'clock sectors, so circumpapillary analysis successfully identified the RNFL defect range. This makes the difference in diagnostic performance between RNFL defect volume/area and circumpapillary RNFL thickness.

Although RNFL thickness maps were highly sensitive for identifying RNFL defects, the false-positive rate should be considered. Hwang et al. ¹⁵ have reported that 6.0% had false-positive RNFL color codes in the thickness map and these were associated with a higher degree of myopia. Kim et al. ¹⁶ have reported that 20.1% show false-positive RNFL color codes in the RNFL deviation map, and the axial length and disc area are significantly associated with an increased incidence of false positives. In this study, 8.16% among 186 normal subjects showed false-positive RNFL defects, but no factors were associated with false-positive detection. False-positive RNFL defects were nasally deviated compared to glaucomatous RNFL defects. This result is consistent with a previous study in which the highest frequency of false-positive defects is located at 5 and 6 o'clock in the inferior quadrant or 12 and 1 o'clock in the superior quadrant. ¹⁶ In clinical practice, RNFL defect interpretation combined with locational information could reduce false-positive errors. 

Retinal nerve fiber layer deviation map is another interpretation method of RNFL thickness map provided by the manufacturer. Abnormal RNFL measurements below the lower 95th or 99th percentile ranges at the 6 × 6-mm² parapapillary area were displayed with yellow or red code in the RNFL deviation map. In a previous study, we have found a topographic similarity of RNFL defect between RNFL thickness map and deviation map. ¹⁷ However, the RNFL defect area of RNFL thickness map is significantly greater than that of red-coded deviation map (<1% level), and smaller than that of yellow-coded deviation map (<5% level). ¹⁷ The RNFL deviation map might over- or underestimate RNFL defects, although it showed similar shape and location of RNFL defect, compared to RNFL thickness map analysis (Fig. 6). The discrepancy between the deviation map and the current study's method comes from the definition of RNFL defect. It is more appropriate to define RNFL defect by structural change than statistical percentile deviation, because RNFL defect results from progressive loss of retinal ganglion cell axons. 

Figure 6

 

(A) RNFL defect was represented with red dashed line in RNFL thickness map analysis. (B) Red dashed line was overlapped with the same patient's RNFL deviation map. It showed similar shape and location to the red- (<1% level) or yellow- (<5% level) coded region of RNFL deviation map. The red-coded area was smaller than the expected RNFL defect, and the yellow-coded area was greater than the expected RNFL defect.

The RNFL thickness map analysis of Cirrus HD-OCT may be a useful tool for evaluating diffuse RNFL defects. To date, several studies have focused on localized RNFL defects. ^2–5 Few studies ^18–20 have attempted a quantitative analysis to evaluate diffuse RNFL defects because it is difficult to clearly define their borders. These studies have used the semiquantitative grade scoring system, so that only limited information is available regarding the ability to identify and quantify diffuse RNFL defects. In this study, clear border determination by objective methods enabled researchers to analyze diffuse RNFL defects (Figs. 4C–E). The proportion of diffuse RNFL defects was 64.1% and of those, 47.7% had mild glaucoma. Half of mild glaucoma patients already had diffuse RNFL defects, which emphasizes the importance of early detection of diffuse RNFL defects. 

Compared with the localized RNFL defects, diffuse RNFL defects had larger areas and volumes. Within the diffuse RNFL defect group, moderate-to-advanced glaucoma had a significantly greater RNFL defect area, volume, and width than mild glaucoma. Quantitative measurements of RNFL defects might be useful to evaluate the severity of glaucomatous damage as well as to distinguish between localized and diffuse RNFL defects. The volume of RNFL defects was significantly associated with the severity of glaucomatous damage in diffuse RNFL defects. Jeoung et al. ¹⁸ have reported that circumpapillary RNFL thickness shows quantitative correlation with the degree of diffuse RNFL defects. The circumpapillary RNFL thickness is measured on 3.46-mm circle scan only in the z-axis, whereas diffuse RNFL defects are distributed throughout a broad area in the x-y axes. Therefore, three-dimensional volumetric analysis is a more suitable method for quantifying diffuse RNFL defects. 

The center of the RNFL defects is a useful reference point to evaluate the topographic characteristics of RNFL defects. Previous studies ^3–5 have used a 3.46-mm-diameter circle to measure angular location and width. However, if RNFL defects exist outside the 3.46-mm-diameter circle, it is impossible to measure angular information. ^14,15 The center of the RNFL defects, around which the distribution is balanced, can always be calculated by mathematical formula. The center of the defect could be an appropriate reference point for the defect location. In addition, the distance from the center of the optic disc to the center of the RNFL defects could be calculated. We found that diffuse RNFL defects were closer to the optic disc than localized RNFL defects, and within the diffuse RNFL defect group, moderate-to-advanced glaucoma was closer to the optic disc than mild glaucoma. If localized RNFL defects expand into diffuse RNFL defects as the disease advances, ¹⁰ it can be postulated that RNFL defects expand toward the optic disc. Close observation of RNFL change around the optic disc may be useful for early detection of RNFL defect progression in clinical practice. Monitoring the center of RNFL defects could provide an alternative approach to studying glaucoma progression. 

In this study, the volume of RNFL defect is defined as the volume of remaining RNFL in the region of RNFL defect. One should be careful in interpreting the volume of RNFL defect to avoid confusion with the volume of lost RNFL in the region of RNFL defect. The volume of RNFL defect does not necessarily represent the amount of glaucomatous damage. If RNFL defect progresses only by enlargement, with no further thinning of pre-existing RNFL defect, both RNFL defect area and volume will increase. If RNFL defect progresses only by thinning with no enlargement of its area, the RNFL defect area will not change and the RNFL defect volume will decrease. For example, within the localized RNFL defect group, the area of RNFL defect was not significantly different between mild and moderate-to-advanced glaucoma, so RNFL defect progression might depend on thinning rather than enlargement. Contrary to expectation, the volume of RNFL defect was not changed significantly between mild and moderate-to-advanced glaucoma. Because actual RNFL defect progresses simultaneously by both enlargement and thinning, a slight increase of RNFL defect area might affect the RNFL defect volume calculation. In further studies, a new parameter is needed to directly measure the amount of glaucomatous damage. 

Structural RNFL loss is known to precede functional visual field damage. ^21,22 Although this study enrolled 102 (60%) mild glaucoma and 68 (40%) moderate-to-advanced glaucoma patients, preperimetric glaucoma was not included. This study showed that, in mild glaucoma group, RNFL defects could be successfully identified by structural RNFL loss. Early detection of RNFL defects before functional visual field loss is expected to be possible, but this study is limited to prove this issue. Further study is required for the preperimetric subjects. All participants of this study were East Asian, so application to other ethnic groups is limited. 

In conclusion, RNFL thickness map analysis based on the degree of RNFL loss is an effective approach for analyzing RNFL defects. This technique can identify RNFL defects with clear borders, regardless of whether they are visible or invisible in red-free photography and whether they are localized or diffuse. Quantitative measurements (area, volume, location, and width) of RNFL defects were useful in understanding the diffuse RNFL defects. 

Supplementary Materials

pdf S1 

Acknowledgments

The authors alone are responsible for the content and writing of the paper. 

Disclosure: J.W. Shin, None; K.B. Uhm, None; M. Seong, None; Y.J. Kim, None