Three-Dimensional Imaging of the Macular Retinal Nerve Fiber Layer in Glaucoma with Spectral-Domain Optical Coherence Tomography

Abstract

Purpose.: To investigate the three-dimensional (3D), spectral-domain (SD) optical coherence tomography (3D,SD-OCT) imaging of the macular retinal nerve fiber layer (RNFL) in eyes with glaucoma.

Methods.: The study included 38 eyes of 38 patients with glaucoma and 38 normal eyes of 38 volunteers. With a 3D raster scan SD-OCT protocol, 512 × 128 axial scans were acquired over a 6-mm² area of the macula. Findings on 3D,SD-OCT images were compared with those on color and red-free fundus photographs and time-domain (TD) OCT.

Results.: Fourteen (30.4%) more RNFL defects were detected on 3D,SD-OCT images than on color fundus photographs. Of these 14, 12 were detected in 10 (90.9%) of 11 eyes with tessellated fundi (P < 0.0001). On 3D,SD-OCT images, complete loss of the RNFL reflectivity was seen in 63.0% of the RNFL defects and thinning of the RNFL in the rest. On TD-OCT cpRNFL analysis, RNFL defects that appeared on 3D,SD-OCT as a complete loss of RNFL reflectivity were detected more often (P = 0.012) than those that appeared as thinning of the RNFL. Inter-rater agreement was better for RNFL defects on 3D,SD-OCT (0.85) than for those on color (0.62–0.64) or red-free (0.68–0.70) fundus photographs. However, 3D,SD-OCT macular RNFL thickness measurements were substantially reproducible but not as reproducible as macular retinal thickness measurements, and neither was as sensitive as TD-OCT cpRNFL thickness measurements for detecting glaucoma.

Conclusions.: 3D,SD-OCT imaging of the macular RNFL is an effective means of detecting macular RNFL defects and their severity in eyes with glaucoma.

Retinal nerve fiber layer (RNFL) abnormalities precede the development of visual field defects in eyes with glaucoma. ^1,2 Thus, for early diagnosis of glaucoma, early recognition of RNFL defects is crucial. To date, RNFL photography has proved to be the most useful means of detecting RNFL defects. ^3–6 However, RNFL defects are visible on photographs only as areas of reduced optical reflectance compared with thicker areas of the RNFL, and it is sometimes difficult to delineate RNFL defects in color fundus photographs, especially in eyes with tessellated fundi or diffuse RNFL thinning. Furthermore, it is difficult to measure RNFL thickness on color and red-free fundus photographs. 

Optical coherence tomography (OCT), a noninvasive imaging technique used to obtain high-resolution cross-sectional images of the retina, permits analysis of structural abnormalities in the three anatomic locations that have abnormalities in eyes with glaucoma: the optic nerve head, circumpapillary (cp)RNFL, and macula. Many studies have demonstrated that, in glaucoma detection software based on OCT images obtained with time-domain (TD)-OCT equipment (Stratus OCT, Carl Zeiss Meditec, Inc., Dublin, CA), the thickness of the cpRNFL in an area 3.46 mm in diameter more accurately differentiates eyes with glaucoma from normal eyes than does the thickness of the macula. ^7–13 However, it is difficult to discern the shape of an RNFL defect on TD-OCT images, because image quality is limited by the acquisition of only six radial scans. 

With the newest technology, spectral-domain (SD) OCT, images are acquired 43 to 100 times faster than with TD-OCT, allowing for 3D imaging of macular structures ^14–16 that should provide visualization of RNFL defects. Commercially available SD-OCT instruments also offer higher axial resolution (5–7 μm, compared with 10 μm for Stratus OCT), which should improve the detection of areas where the RNFL in the macula has become thinner. ^17,18 However, the usefulness of SD-OCT imaging of the macular RNFL has not been fully explored. Thus, we performed a study to compare 3D,SD-OCT images with color and red-free fundus photographs in the detection of macular RNFL defects in eyes with known glaucoma. 

Methods

Seventy-six Japanese individuals who provided informed consent and met the described inclusion criteria were consecutively enrolled from the database of patients who were examined for glaucoma during the period from May 2007 to August 2007 in the Department of Ophthalmology, Kyoto University (Kyoto, Japan). All investigations adhered to the tenets of the Declaration of Helsinki, and the study was approved by the Institutional Review Board and Ethics Committee of Kyoto University Graduate School of Medicine. 

All participants had undergone complete ophthalmic examinations, including best-corrected visual acuity (BCVA) with the 5-m Landolt chart, refraction, slit lamp biomicroscopy, gonioscopy, Goldmann applanation tonometry, and dilated stereoscopic examination of the optic disc. Reliable standard automated perimetry (SAP) was performed with the 24-to 2 Swedish Interactive Threshold Algorithm (SITA; Humphrey Field Analyzer, HFA, Carl Zeiss Meditec) within 3 months of TD- and 3D,SD-OCT imaging for this study. 

To be included in the study, eyes had to have BCVA of 20/25 or better, spherical refractive error between +3.00 and −6.00 D, and open angles confirmed by gonioscopy. Eyes with any type of retinal abnormality, opaque media, or retinal laser procedure or surgery and individuals with neurologic disease, a history of diabetes, or a history of corticosteroid use were excluded. When both eyes met inclusion and exclusion criteria, one eye was randomly selected for inclusion in the study. 

A visual field was considered to be reliable when fixation losses were less than 20%, the false-positive rate was less than 15%, and the false-negative rate was less than 33%. A visual field defect due to glaucoma was defined as (1) glaucoma hemifield test (GHT) results outside normal limits; or (2) three characteristics with a P < 0.05 probability of being normal, including one characteristic with a <1% probability of being normal by pattern deviation; or (3) pattern standard deviation (PSD) with a less than <5% probability of being normal, as confirmed by two consecutive tests. 

Glaucomatous optic disc appearance was defined on stereoscopic examination as vertical cup-to-disc ratio of more than 0.7; signs of optic nerve changes such as diffuse or localized rim thinning, notching disc hemorrhage, or vertical cup-to-disc ratio more than 0.2 larger than the ratio in the fellow eye. 

Volunteers with normal results of ocular examinations, intraocular pressure less than 21 mm Hg, normal optic nerve head appearance, normal-appearing retinal nerve fiber layer, and normal results on automated static perimetry were also recruited. 

Optical Coherence Tomography

Optical coherence tomography imaging with TD-OCT (Stratus OCT; Carl Zeiss Meditec) and 3D,SD-OCT (3D OCT-1000; Topcon, Tokyo, Japan) was performed on the same day, after pupil dilation. 

3D Scan Protocols

A raster scan protocol was used with the SD-OCT equipment to analyze macular volume. The protocol involved acquiring 128 images with 512 axial scans (transverse pixels) × 1024 axial pixels over a 6× 6 × 2-mm volume centered on the fovea. The 3D OCT-1000 has an axial scanning speed of 27 kHz and acquires one set of raster scans in 2.6 seconds. Built-in software (ver. 2.10) calculates macular thickness as the distance between the vitreoretinal interface and the inner border of the retinal pigment epithelium, and it calculates macular RNFL thickness as the distance between the vitreoretinal interface and the outer border of the retinal nerve fiber layer (RNFL, Movie S1). For thickness analysis, a circular area 6 mm in diameter centered on the fovea was used. Only good-quality images (Q factor score higher than 60) were included. 

TD Scan Protocols

TD-OCT scan protocols included fast macular thickness (six radial line scans of 6-mm scan length, centered on the fovea) for measuring macular thickness and fast RNFL thickness (three consecutive circular scans [256 A-scans], each with a diameter of 3.46 mm, centered on the optic disc) for measuring cpRNFL thickness. Both macular and cpRNFL thicknesses were computed by the TD-OCT system (Stratus with software ver. 4.0; Carl Zeiss Meditec). Only good-quality images (image quality score >6) were included in the analyses. 

Subjective Assessment of RNFL Defects

Two glaucoma specialists with extensive experience with OCT images (NM, HN) assessed all images (color photographs, red-free photographs, and color-coded maps of macular RNFL thickness created by 3D,SD-OCT imaging) independently for every eye and recorded the number and locations of any visible RNFL defects. The assessors were masked to the clinical findings, including BCVA. When the first two assessors disagreed, a third assessor (AS) was invited to view the images, and the disagreement was resolved after discussion, so that the three assessors agreed on all the eyes. RNFL defects in color and red-free fundus photographs were defined as wedge-shaped areas with clear borders representing low reflectivity of the RNFL. RNFL defects in 3D,SD-OCT images were defined as wedge-shaped areas that appeared less thick than neighboring areas in color-coded maps of macular RNFL thickness. The RNFL defects observed in the 3D,SD-OCT scans were categorized as complete loss or thinning of RNFL reflectivity by the trained assessors, to estimate the effects of defect severity on agreement among modalities. We also counted the number of RNFL defects inside the central (macular) 6-mm² area on color and red-free fundus photographs. 

Reproducibility

The reproducibility of total retinal thickness and RNFL thickness measurements was assessed in 13 eyes of 13 subjects with glaucoma (mean age, 56.2 years; range, 36–79 years), a subset of the 38 glaucomatous eyes. To determine reproducibility, two trained examiners (YK, SM) obtained three series of 128 raster scans of each eye with the 3D,SD-OCT during the same session by. The first series was performed by examiner 1 (YK). The subject then sat back from the machine for at least 5 minutes before being repositioned. The second series of scans was then performed by examiner 2 (SM). After another rest period, the third series was performed by examiner 1. To assess intervisit reproducibility, an additional series of raster scans was obtained 1 week later by examiner 1. 

For the subgroup of the study participants, the mean total retinal thickness and the mean RNFL thickness measured with 3D,SD-OCT were calculated for each observer and each visit. The intraclass correlation coefficient (ICC) and coefficient of repeatability (CR) were calculated for each of the following: (1) intraobserver reproducibility, by comparing the results of two examinations performed on the same day by the same examiner; (2) interobserver reproducibility, by comparing the results of examinations performed on the same day by two different examiners; and (3) intervisit reproducibility, by comparing the results of two examinations performed 1 week apart by the same examiner. 

Statistical Analysis

The statistical significance of differences between the normal and glaucoma groups was evaluated with the Mann-Whitney U test for continuous data (age, refractive error, MD, and PSD). One-way analysis of variance (ANOVA) and Scheffé's multiple comparison were used to compare the mean number of RNFL defects per eye in color photographs, red-free photographs, and 3D macular RNFL maps. For categorical data, Fisher's exact test was used to compare differences between groups. Pearson's correlation coefficients and Cox and Snell's coefficient of determination were used to assess correlations between TD-OCT or 3D,SD-OCT measurements and mean deviations (MDs) in static automated perimetry. Receiver operating characteristic (ROC) curves were used to describe the sensitivity and specificity of each variable for distinguishing eyes with glaucoma from healthy eyes. The ROC curve shows the tradeoff between sensitivity and 1 − specificity. An area under the ROC curve (AROC) of 1.0 represents perfect discrimination, whereas an AROC of 0.5 represents chance discrimination. Commercially available software (MedCalc version 9.3.8.0, MedCalc Software, Mariakerke, Belgium) was used to draw and compare the ROC curves. Differences were significant at P < 0.05. 

Results

Characteristics of the 76 study participants (76 eyes; 38 with glaucoma and 38 normal eyes) are shown in Table 1. Although individuals in the two groups were similar in age and sex, the eyes in the two groups differed significantly in refractive error (P < 0.01) and in visual field test results (MD and PSD; P < 0.001). 

Table 1.

 

Characteristics of 38 Individuals (38 Eyes) with Glaucoma and 38 Individuals (38 Normal Eyes) without Glaucoma

Individuals (Eyes)	Normal (n = 38)	Glaucoma (n = 38)	P
Age, y
Mean ± 1 SD	59.2 ± 12.5	61.5 ± 10.3	0.58*
Range	36–77	33–77
Sex, female/male	17/21	19/19	0.65†
Refractive error, D	−0.31 ± 1.68	−1.67 ± 2.26	0.001*
MD, dB	−0.40 ± 1.00	−7.31 ± 6.16	<.0001*
PSD, dB	1.52 ± 0.24	8.75 ± 4.94	<.0001*

*  Mann-Whitney U-test.

†  Fisher's exact test.

RNFL Defects

At least one RNFL defect was observed in each of the 38 glaucomatous eyes. Various numbers of RNFL defects were detected by various imaging techniques in the maculas of the 38 eyes with glaucoma, but no defects were seen in any images of any of the 38 normal eyes. 

Both the 3D volume images of the macular RNFL and color-coded maps of macular RNFL thickness clearly showed RNFL defects in the macular regions that corresponded well with the RNFL defects seen in color fundus and red-free photographs at various stages of glaucoma (Figs. 1 2 3–4). 

Figure 1.

 

Imaging results in the left eye of a 51-year-old woman with early normal-tension glaucoma. Intraocular pressure (IOP) was 15 to 17 mm Hg without glaucoma medication, and MD on static automated perimetry was −5.15 dB. (A) Color fundus photograph shows marked thinning of the inferior neuroretinal rim, with bayonet-like vessel kinking and undermining of the cup border that appears to correspond with an RNFL defect in the inferior–temporal quadrant but the margins are unclear. (B) Red-free fundus photograph shows an RNFL defect in the inferior–temporal region. (C) Stratus TD-OCT fast macular thickness map. (D) Stratus OCT cpRNFL thickness profile and significance maps for quadrant and clock hours show RNFL atrophy inferiorly and temporally (red segment). (E) Visual field pattern deviation map. (F) Macular retinal thickness map generated by 3D,SD-OCT software. (G) 3D,SD-OCT volume image shows the RNFL defect. (H) Macular RNFL thickness map generated by 3D,SD-OCT software clearly shows an RNFL defect that corresponds to the RNFL defect in the fundus photographs. (I) Vertical cross-section along the green line indicated in (H) shows that the highly reflective layer representing the RNFL is completely absent in the inferior hemisphere.

Figure 2.

 

Imaging results in the left eye of a 56-year-old man with early normal-tension glaucoma. Intraocular pressure (IOP) was 10 to 12 mm Hg with one glaucoma medication, and MD on static automated perimetry was −2.07 dB. (A, B) Color and red-free fundus photographs show narrow and broad RNFL defects in the superior and inferior hemispheres, respectively. (C) TD-OCT Stratus fast macular thickness map. It is difficult to recognize RNFL defects, especially the superior narrow RNFL defect, on this image alone. (D) Stratus OCT cpRNFL thickness profile and significance maps detect cpRNFL atrophy corresponding to the broad RNFL defect in the inferior hemisphere, but no cpRNFL atrophy corresponding to the narrow RNFL defect in the superior hemisphere. (E) Visual field pattern deviation map shows a visual field defect corresponding to the broad RNFL defect but no significant visual field defect corresponding to the narrow RNFL defect. (F) Macular retinal thickness map generated by 3D,SD-OCT software. (G) 3D,SD-OCT volume image depicts the narrow and broad RNFL defects. (H) Macular RNFL thickness map generated by 3D,SD-OCT software depicts the RNFL defects that correspond to the narrow and broad RNFL defects in the fundus photographs. (I) Vertical cross-section along the yellow dashed line in (H) shows complete loss of the highly reflective layer representing the RNFL inferiorly, in the region corresponding to the broad RNFL defect, but incomplete loss of RNFL reflectivity in the superior region corresponding to the narrow RNFL defect. Arrowheads indicate cross-sectional images of both the RNFL defects.

Figure 3.

 

Imaging results in the left eye of a 62-year-old woman with preperimetric normal-tension glaucoma. Intraocular pressure (IOP) was 15 to 17 mm Hg without glaucoma medication, and MD on static automated perimetry was −1.98 dB. (A, B) Color and red-free fundus photographs show a preperimetric RNFL defect in the inferior hemisphere. Color fundus photograph shows an inferotemporal neuroretinal rim notch that appears to correspond to the RNFL. (C) TD Stratus OCT fast macular thickness map. It is difficult to recognize RNFL defects on this image alone. (D) Stratus OCT cpRNFL thickness profile and significance maps do not show the RNFL defect in the inferior hemisphere. (E) Visual field pattern deviation map. (F) Macular retinal thickness map generated by 3D,SD-OCT software. (G) 3D,SD-OCT volume image clearly shows the RNFL defect, which extends to the temporal raphe. (H) Macular RNFL thickness map generated by 3D,SD-OCT software clearly shows the RNFL defect that corresponds to the RNFL defect in the fundus photographs. (I) Vertical cross-section along the yellow dashed line in (H) shows a thin, highly reflective line (arrowhead) representing the RNFL remaining in the region that corresponds to the RNFL defect in fundus photographs.

Figure 4.

 

Imaging results in the right eye of a 56-year-old man with advanced primary open-angle glaucoma. Intraocular pressure (IOP) was 16 to 18 mm Hg with two glaucoma medications and MD on static automated perimetry was −13.08 dB. (A) Color fundus photograph shows extensive optic disc cupping with marked thinning of the neuroretinal rim and bayonet-like vessel kinking and undermining of the cup border, but the borders of the RNFL defect are unclear. (B) A red-free fundus photograph shows the RNFL to be partially preserved in the papillomacular region but does not clearly show whether the RNFL is preserved superotemporally or inferotemporally. (C) TD Stratus OCT fast macular thickness map. It is difficult to recognize preserved RNFL on this image alone. (D) Stratus OCT cpRNFL thickness profile and significance maps. (E) Visual field 10-2 pattern deviation. (F) Macular retinal thickness map generated by 3D,SD-OCT software. (G) 3D,SD-OCT volume image from the nasal side clearly shows preservation of the RNFL in superotemporal and papillomacular regions but total loss of the RNFL inferotemporally. (H) Macular RNFL thickness map generated by 3D,SD-OCT software clearly shows where the RNFL is preserved, and this area of preservation corresponded with the area of visual field threshold preservation on standard automated perimetry. (I) Vertical cross-section along the yellow dashed line indicated in (H) shows preservation of the RNFL (very thin, highly reflective line, arrowheads).

Table 2 shows the number of defects detected in images from the 38 eyes with glaucoma. Thirty-four RNFL defects were detected in color fundus photographs and 32 (94.1%) of the 34 appeared in the 6-mm² scan area on 3D,SD-OCT images of these eyes. In red-free photographs, 51 RNFL defects were detected, 45 (88.2%) of which appeared in the 6-mm² scan area on 3D,SD-OCT. Forty-six RNFL defects were detected in 3D macular RNFL maps (6-mm² scan area), including all 32 defects that were detected in color fundus photographs and an additional 14 (30.4% more) that were detected only on 3D macular RNFL maps and not on color fundus photographs. Of the 45 RNFL defects that were detected on red-free photographs, 43 were also detected on 3D macular RNFL maps, whereas 2 that were detected on red-free photographs were not detected on 3D macular RNFL maps, and 3 that were detected on 3D macular RNFL maps were not detected on red-free photographs. Thus, 3D macular RNFL maps and red-free photographs revealed more RNFL defects than did color fundus photographs, but the difference was not significantly different (P = 0.086). Among the 14 RNFL defects that were detected on 3D macular RNFL maps but not on color fundus photographs, 12 (85.7%; P < 0.0001) were detected in eyes with tessellated fundi. In 11 eyes with tessellated fundi, 3D macular RNFL maps, and red-free photographs showed significantly more RNFL defects than did color fundus photographs (P = 0.006 and 0.023, respectively, compared with 50° color fundus photographs). 

Table 2.

 

Numbers of RNFL Defects Detected by Photography versus 3D,SD-OCT

	All (n = 38)		Tesselated Fundus (n = 11)
	50°	6 mm2
Color	34	32	6 (50°)
Red-free	51	45	16 (50°)*
3D,SD-OCT	—	46	18†

  P-values versus 50° color fundus photographs, by one-way ANOVA and Scheffé's multiple comparison.

*   P = 0.023.

†   P = 0.006.

Detection of RNFL Defects with Complete Loss versus Thinning

Serial vertical cross-sectional images generated from 3D,SD-OCT scan data revealed that some RNFL defects involved complete loss of the highly reflective line representing the RNFL (Figs. 1, 2, 4) whereas others involved thinning of the RNFL layer without complete loss of the line (Figs. 2, 3). 

Table 3 shows how many of the RNFL defects seen on 3D,SD-OCT as complete loss or just thinning of the reflective line were detected as a cpRNFL defect on TD-OCT, on color fundus photographs, and on red-free photographs. Of the 46 RNFL defects that were detected on 3D macular RNFL maps, complete loss of the RNFL was seen in 29 (63.0%) eyes, and these defects were more frequently (P = 0.012) detected on TD-OCT cpRNFL thickness profiles and significance maps than were RNFL defects that showed thinning of the RNFL layer (17 eyes, 37.0%). When only the 32 RNFL defects in color fundus photograph that appeared in the 6-mm² 3D,SD-OCT scan areas were considered, complete loss of the highly reflective line representing the RNFL on 3D,SD-OCT images was seen in 20 (69.0%) of the 29 cases with complete loss and in 12 (70.6%) of the 17 cases without complete loss (nonsignificant; P = 0.99). When only the 43 RNFL defects in the red-free fundus photograph that appeared in the 6-mm² 3D,SD-OCT scan areas were considered, complete loss of the highly reflective line representing the RNFL on 3D,SD-OCT images was seen in 27 (93.1%) of the 29 cases with complete loss and in 16 (94.1%) of the 17 cases without complete loss (nonsignificant; P = 0.99). 

Table 3.

 

Relationship between Cross-Sectional Features of RNFL Defects on 3D,SD-OCT and Detection of the RNFL Defects by TD-OCT, Color Photographs, and Red-Free Photographs in 38 Eyes with Glaucoma

Cross-Sectional Features (i.e. Complete Loss vs. Thinning) and Number of RNFL Defects from 3D,SD-OCT	Number of RNFL Defects Detected in a 6-mm2 Area
	TD-OCT cpRNFL (n = 46)	Color Photographs (n = 32)	Red-Free Detectable on 3D,SD-OCT (n = 43)
Complete loss 29/46 (63.0%)	24 TD-OCT defects vs. 29 SD-OCT defects (82.8%)*	20 color photographic defects vs. 29 SD-OCT defects (69.0%)†	27 red-free photographic defects vs. 29 SD-OCT defects (93.1%)†
Thinning only 17/46 (37.0%)	8 TD-OCT defects vs. 17 SD-OCT defects (47.1%)	12 color photographic defects vs. 17 SD-OCT defects (70.6%)	16 red-free photographic defects vs. 17 SD-OCT defects (94.1%)

 Complete loss, RNFL defects that include complete loss of the reflective line representing the RNFL on 3D,SD-OCT images; thinning only, RNFL defects that involve thinning of the RNFL layer without complete loss of the highly reflective line representing the RNFL on 3D,SD-OCT images.

*   P = 0.012 (Fisher's exact test) compared with thinning only.

†   P > 0.05 (Fisher's exact test) compared with thinning only.

The Cohen's κ (index of intercoder reliability) for the two assessors for RNFL defects was 0.62 for color fundus photographs, 0.64 for the 6-mm² macular area in color fundus photographs, 0.70 for red-free photographs, 0.68 for the 6 mm² macular area in red-free photographs, and 0.85 for color-coded maps of macular RNFL thickness. 

Usefulness of Results for Glaucoma Diagnosis

Table 4 shows significantly lower mean cpRNFL thickness measured by TD-OCT, macular thickness measured by TD-OCT and 3D,SD-OCT, and macular RNFL thickness measured by 3D,SD-OCT in eyes with glaucoma compared with normal eyes in our study (P < 0.0001 for all values). When each average thickness of the eyes with glaucoma was divided by that of normal eyes, the percentage difference was greatest for macular RNFL thickness measured by SD-OCT (59.2%), followed by cpRNFL measured by TD-OCT (70.8%). The reduction in macular thickness was similar when measured by TD-OCT (91.9%) and SD-OCT (91.7%). Linear regression analysis among the eyes with glaucoma showed significant (P < 0.001) correlation of all thickness measurements with mean deviation (MD) values on automated static perimetry (Table 4). 

Table 4.

 

Diagnostic Value of RNFL Thickness Measurements in Eyes with Glaucoma

Measurement/Imaging Method	Normal (n = 38)	Glaucoma (n = 38)	P *	AROC (95% CI)	r/R 2/P *
cpRNFL thickness/TD-OCT	106 ± 8.85	75.1 ± 17.4	<0.0001	0.918 (0.832–0.968)	0.87/0.61/<0.001
Macular thickness/TD-OCT	248 ± 11.9	228 ± 19.7	<0.0001	0.815 (0.709–0.895)	0.63/0.40/<0.001
Macular thickness/SD-OCT	265 ± 9.74	243 ± 17.1	<0.0001	0.877 (0.781–0.941)	0.62/0.39/<0.001
Macular RNFL thickness/SD-OCT	26.2 ± 3.86	15.7 ± 6.7	<0.0001	0.896 (0.805–0.954)	0.73/0.54/<0.001

 Data are the mean ± SD or statistical significance (Mann-Whitney U-test). 95% CI, 95% confidence interval.

*  Pearson's correlation coefficients and Cox and Snell's coefficient of determination.

The AROC was greatest for mean cpRNFL thickness measured by TD-OCT (0.918), followed by mean macular RNFL thickness measured by 3D-OCT (0.896), macular thickness measured by SD-OCT (0.877), and macular thickness measured by TD-OCT (0.815) (Table 4). Statistically significant differences were detected only between cpRNFL thickness and macular thickness measured by TD-OCT (P < 0.05) and between macular thickness measured by SD-OCT and TD-OCT (P = 0.05). 

Reproducibility and Repeatability

Table 5 shows the ICC and CR for measurements of retinal thickness and macular RNFL thickness with 3D,SD-OCT in eyes with glaucoma. Retinal thickness measurements had higher ICCs (range, 0.995–0.998) than did RNFL measurements (range, 0.835–0.878), indicating that the former has the more reliable reproducibility. Retinal thickness measurements had lower CR (range, 2.57–3.64) than did RNFL measurements (range, 7.51–8.37), indicating that the former has the higher repeatability. 

Table 5.

 

ICCs and CR of Macular Measurements with 3D,SD-OCT in Eyes with Glaucoma

	Retinal Thickness (n = 13)	RNFL (n = 13)
Intraobserver ICC	0.998	0.875
Interobserver ICC	0.998	0.835
Intervisit ICC	0.995	0.878
Intraobserver CR	2.72	7.51
Interobserver CR	2.57	8.37
Intervisit CR	3.64	7.98

Discussion

In this study, we assessed the potential usefulness of 3D,SD-OCT imaging of the macular RNFL for detecting RNFL defects in the diagnosis of glaucoma. A large number of studies have been performed to examine TD-OCT measurement of cpRNFL thickness as a tool for glaucoma diagnosis, but TD-OCT is too slow for effective 3D imaging, and even the fast macular thickness mode on the Stratus (TD)-OCT provided too few sampling points to clearly visualize the shape of RNFL defects. Our study showed, however, that high-speed 3D raster scanning using 3D,SD-OCT provided excellent images of the macular RNFL, with several advantages compared with TD-OCT and fundus photography. 

First, we found that on 3D,SD-OCT images of macular RNFL, RNFL defects in the macular region could be identified more effectively than on color fundus photographs, especially in eyes with tessellated fundi and as effectively as on red-free fundus photographs. The identification of RNFL defects on color fundus photographs has been one of the major diagnostic indicators of glaucoma. However, RNFL defects can be difficult or impossible to detect on photographs of tessellated fundi and may be difficult to detect in photographs of eyes with diffuse RNFL thinning. In contrast, we clearly saw macular RNFL defects in the 3D volume images and thickness maps of macular RNFLs in our study, even in images from eyes with tessellated fundi and diffuse RNFL thinning. This advantage of macular RNFL imaging using SD-OCT appears to result from the difference in imaging principals between OCT and photography. On OCT images, the RNFL appears as a line of high reflectivity, distinct from the reflectivity representing other retinal and subretinal tissues, whereas on photographs, light reflected from the RNFL is mixed with light reflected from other retinal and subretinal tissues, and so RNFL defects, which appear in color fundus photographs as localized reductions in light reflected from the RNFL, may be masked by reflections from deeper tissues such as the choroid and retinal pigment epithelium. Red-free processing can decrease the reflections from these deeper tissues. 

Because the Stratus OCT (TD-OCT) software (Carl Zeiss Meditec, Inc.) does not allow automated segmentation of macular RNFL thickness, researchers developed a new segmentation algorithm for RNFL segmentation on Stratus OCT images. ^19,20 The maps of macular RNFL thickness in these studies did not clearly depict macular RNFL defects, probably because of the conventional scanning protocol they used—six radial linear scans with 6-mm scan length, centered at the fovea—which limits the number of sampling lines (the number of sampling lines of each scan is 256) and leaves large areas unsampled. This limitation in scanning density stems from the slow scanning speed (400 A-scans/s) of the Stratus OCT. In this study, we used 65,536 A-scans for 3D imaging of the macular RNFL, approximately 85 times the six radial linear scans in TD-OCT. In addition, our 3D,SD-OCT protocol allowed even distribution of a much higher density of A-scans over the macula. 

The 3D,SD-OCT scans in our study showed the severity of RNFL atrophy, whereas the severity of RNFL atrophy cannot be determined from color or red-free fundus photographs. Our serial vertical B-scan images and color-coded macular RNFL thickness maps clearly showed whether each RNFL defect in color or red-free fundus photographs involved complete loss or just thinning of the RNFL. Thus, the 3D macular RNFL imaging using SD-OCT has advantages over fundus photography for RNFL imaging. 

Although RNFL thinning was more severe than total retinal thinning in the eyes with glaucoma in our study, mean macular RNFL thickness was not a significantly better indicator of glaucoma than mean macular thickness measured with SD-OCT. This result is consistent with studies by Ishikawa et al. ¹⁹ and Leung et al. ²¹ The outer retinal layers are hardly affected in glaucoma, whereas the ganglion cell layer (GCL) and inner plexiform layer (IPL) of the retina are affected in glaucoma as well as the RNFL. ^20,22 The ganglion cell layer (GCL) in the macula, which has between two and seven layers of ganglion cell bodies, may especially be affected. ²³ Recent studies showed that the mean thickness of the innermost three or four retinal layers, which include the RNFL, GCL, and IPL, is better than the mean macular thickness and comparable to mean cpRNFL thickness in discriminating whether glaucoma is present. ^19,20 In addition, it was shown that measurements of the RNFL layer have relatively low repeatability compared with measurements of combined inner retinal layers and of the total retina, which is in agreement with our results. ²⁰  

Many studies have shown that Stratus (TD) OCT measurements of mean cpRNFL thickness are the best means of discriminating between glaucomatous and normal eyes, ^7–13 but Stratus OCT cpRNFL thickness profiles and significance maps are not sensitive indicators of RNFL defects. ^24,25 In our study, we found that localized RNFL defects were more often detected on Stratus OCT cpRNFL thickness profiles and significance maps when macular RNFL defects included complete loss of RNFL reflectivity. In contrast, when macular RNFL defects did not include complete loss of RNFL reflectivity (only RNFL thinning), half or more of the RNFL defects were not detected on Stratus OCT cpRNFL thickness profiles and significance maps. These findings suggest that conventional cpRNFL thickness analyses may fail to reveal less severe RNFL atrophy, which could help to explain why RNFL defects are not sensitively detected by Stratus OCT measurements of cpRNFL thickness. 

In the present study, the correlation with MD of macular RNFL thickness measured using SD-OCT as a visual function was higher than that of other macular parameters. However, the correlation of all the macular parameters with MD was lower than that of cpRNFL from TD-OCT. The difference in the coverage area between the two scanning patterns might be responsible for the highest correlation of the cpRNFL parameter with the MD parameter; a macular 6-mm² scan only covers the area corresponding to HFA 10-2, whereas cpRNFL thickness globally reflects the changes from the whole retinal area. We used the MD parameter from HFA 24-2, which covers changes from areas much larger than the macula. 

There are some limitations to the usefulness of 3D,SD-OCT imaging in detecting macular RNFL defects for early diagnosis of glaucoma. One is that the scanning area is limited to 6 mm², and among our study patients, 5.9% and 11.8% of RNFL defects on color and red-free fundus photographs, respectively, were not within the macular scan area. It may be desirable to try to widen the scanning area for detection of RNFL defects. Another limitation is that the axial resolution of commercially available SD-OCT systems, which is 5 to 6 μm, may not be adequate for automated segmentation of thin RNFLs in the vicinity of the central fovea and temporal raphe, even in healthy eyes, although it is unknown whether examining the RNFL in these regions is important. It is also uncertain whether the axial resolution of 5 to 6 μm is adequate for automated segmentation of atrophic RNFLs in eyes with glaucoma. The axial resolution on 3D,SD-OCT may account, at least in part, for the lower reproducibility of macular RNFL thickness measurements compared with total macular retinal thickness measurements. Despite this limitation, however, it was possible to effectively identify RNFL defects from color-coded 3D,SD-OCT maps of RNFL thickness. 

In our subjects, normal eyes and those with glaucoma differed significantly in refractive error. This difference in ocular magnification may not have an influence on the qualitative assessment of RNFL defects on 3D,SD-OCT images but may be a factor on the macular RNFL and macular thickness estimates in the normal versus glaucomatous eyes. In future work, the effects of ocular magnification correction on these macular thickness parameters should be investigated. 

In conclusion, our study showed that 3D,SD-OCT imaging of macular RNFL has advantages and shows promise for early detection of macular RNFL defects due to glaucoma. The current minor limitations of this approach to early glaucoma diagnosis could be overcome by advances in SD-OCT technology to allow a wider scanning area and higher axial resolution. 

Supplementary Materials

Movie S1 - 19.5 MB (.avi) 

Footnotes

 Supported in part by Grant-in-Aid for Scientific Research 20592038 from the Japan Society for the Promotion of Science (JSPS) and by Topcon, Inc. (Tokyo, Japan).

Footnotes

 Disclosure: A. Sakamoto, Topcon, Inc. (F); M. Hangai, Topcon, Inc. (F, C), Nidek Co., Ltd. (C); M. Nukada, Topcon, Inc. (F); H. Nakanishi, Topcon, Inc. (F); S. Mori, Topcon, Inc. (F); Y. Kotera, Topcon, Inc. (F); R. Inoue, Topcon, Inc. (F); N. Yoshimura, Topcon, Inc. (F, C), Nidek Co., Ltd. (C)