July 2019
Volume 60, Issue 8
Open Access
Glaucoma  |   July 2019
En Face Slab Images Visualize Nerve Fibers With Residual Visual Sensitivity in Significantly Thinned Macular Areas of Advanced Glaucomatous Eyes
Author Affiliations & Notes
  • Mari Sakamoto
    Kobe University Graduate School of Medicine, Department of Surgery, Division of Ophthalmology, Kobe, Japan
  • Sotaro Mori
    Kobe University Graduate School of Medicine, Department of Surgery, Division of Ophthalmology, Kobe, Japan
  • Kaori Ueda
    Kobe University Graduate School of Medicine, Department of Surgery, Division of Ophthalmology, Kobe, Japan
  • Takuji Kurimoto
    Kobe University Graduate School of Medicine, Department of Surgery, Division of Ophthalmology, Kobe, Japan
  • Sentaro Kusuhara
    Kobe University Graduate School of Medicine, Department of Surgery, Division of Ophthalmology, Kobe, Japan
  • Yuko Yamada-Nakanishi
    Kobe University Graduate School of Medicine, Department of Surgery, Division of Ophthalmology, Kobe, Japan
  • Makoto Nakamura
    Kobe University Graduate School of Medicine, Department of Surgery, Division of Ophthalmology, Kobe, Japan
  • Correspondence: Makoto Nakamura, Kobe University Graduate School of Medicine, Department of Surgery, Division of Ophthalmology, 7-5-2 Kusunoki-cho, Chuo-ku, Kobe 650-0017, Japan; [email protected]
Investigative Ophthalmology & Visual Science July 2019, Vol.60, 2811-2821. doi:https://doi.org/10.1167/iovs.18-25910
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Mari Sakamoto, Sotaro Mori, Kaori Ueda, Takuji Kurimoto, Sentaro Kusuhara, Yuko Yamada-Nakanishi, Makoto Nakamura; En Face Slab Images Visualize Nerve Fibers With Residual Visual Sensitivity in Significantly Thinned Macular Areas of Advanced Glaucomatous Eyes. Invest. Ophthalmol. Vis. Sci. 2019;60(8):2811-2821. https://doi.org/10.1167/iovs.18-25910.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: The present study aimed to assess the ability of en face slab images of maculae to detect residual nerve fiber bundles in eyes with advanced glaucoma.

Methods: This study included 36 eyes with diffuse thinning of the ganglion cell and inner plexiform layers (GCL+IPL). Inclusion criterion was GCL+IPL thickness of <1% of the normative database level as detected using optical coherence tomography (OCT). En face slab images (50-μm thickness) were reconstructed from the macular cube scan data using advanced visualization settings. Thereafter, we assessed the agreement of the locations between hyperreflective nerve fiber bundles and normal points in the Humphrey visual field test 10-2 pattern deviation (PD) plots. Additionally, total deviation (TD) corresponding to hyperreflective and hyporeflective areas was compared.

Results: Hyperreflective areas were detected in 31 out of 36 eyes; all 31 eyes exhibited at least one normal PD point despite the substantial GCL+IPL thinning in the macular region. Two eyes with abnormalities in all PD points showed no hyperreflective area. The remaining three eyes had normal PD points despite the lack of high reflectivity areas in the slab images. Therefore, 91.7% of eyes showed agreement between en face slab images and PD plots. Moreover, hyperreflective areas demonstrated significantly better TD than hyporeflective areas.

Conclusions: En face slab images of maculae were able to reveal the residual nerve fiber bundles in the eyes with advanced glaucoma despite the GCL+IPL thickness in the maculae being diffusely and substantially reduced.

Analysis of the thickness of the inner retinal layers of maculae using optical coherence tomography (OCT) has been widely used to diagnose and monitor glaucoma. However, OCT thickness analysis often fails to capture the structural changes during advanced stages of glaucoma due to the floor effect, where the thickness of inner retinal layers approaches baseline level.13 Typically, the structural changes progress more rapidly during early stages of glaucoma while the functional changes still progress during the advanced stages, where the structural change reaches a floor as aforementioned, indicating the structure–function dissociation at the two extremes of the stage.13 
Hood et al.4 reported that the detection of the local glaucomatous changes using the average reflectance intensity on en face slab images is better compared with the use of the traditional OCT retinal nerve fiber layer (RNFL) thickness analysis. In contrast to the traditional thickness analysis, wherein thickness of the retinal layers of interest varies depending on the severity of disease, aging, and axial length elongation, slab analysis evaluates the retinal tissues with a fixed arbitrary thickness that is sandwiched between a reference landmark, such as inner limiting membrane (ILM)/vitreous border, and a specific line that is below a predefined fixed distance from the landmark (Fig. 1). In the en face slab images obtained using this approach, RNFL is considered to contribute to high reflectance, irrespective of the thickness of each retinal layer. However, other studies4,5 have focused on the usefulness of RNFL reflectivity in en face slab images for early detection of glaucoma. 
Figure 1
 
Representative en face slab image of the macula in a normal eye. (A) A pseudocolor en face slab image overlaid on a fundus photograph. Areas with high reflectivity are expressed by warm color, whereas those with low reflectivity are indicated in cold color. (B) A pseudocolor horizontal cross-section image at the blue line in (A). (C) A pseudocolor vertical cross-section image at the purple line in (A). (D, E) Grayscale images of (B, C), respectively. Yellow solid lines indicate the surface of inner limiting membrane (ILM). Dotted yellow lines are 50 μm beneath and parallel to the ILM. The pseudocolor en face slab image in (A) reflects the average reflectance intensity of the “slab” sandwiched between the ILM and border 50 μm beneath and parallel to the ILM (black arrows in D, E).
Figure 1
 
Representative en face slab image of the macula in a normal eye. (A) A pseudocolor en face slab image overlaid on a fundus photograph. Areas with high reflectivity are expressed by warm color, whereas those with low reflectivity are indicated in cold color. (B) A pseudocolor horizontal cross-section image at the blue line in (A). (C) A pseudocolor vertical cross-section image at the purple line in (A). (D, E) Grayscale images of (B, C), respectively. Yellow solid lines indicate the surface of inner limiting membrane (ILM). Dotted yellow lines are 50 μm beneath and parallel to the ILM. The pseudocolor en face slab image in (A) reflects the average reflectance intensity of the “slab” sandwiched between the ILM and border 50 μm beneath and parallel to the ILM (black arrows in D, E).
The present study aimed to assess the ability of en face slab images of maculae to reveal residual nerve fiber bundles in advanced glaucomatous eyes, wherein the retinal thickness was substantially reduced. 
Methods
Participants
This is a substudy of a larger investigation titled “Retinal Imaging Using Optical Coherence Tomography in Several Optic Neuropathies,” registered on the University Hospital Medical Information Network-Clinical Trial Registry (UMIN000006900). The study protocol was approved by the Institutional Review Board of Kobe University (approval no. 1362) and was conducted in accordance with the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants. 
In the present study, we reviewed the medical charts of patients with all types of glaucoma who visited the Kobe University Hospital between 2015 and 2017, for whom OCT images of their maculae were obtained and visual field tests measured on the Humphrey Field Analyzer Swedish Interactive Thresholding Algorithm standard 10-2 program (HVF; Carl Zeiss Meditec, Inc., Dublin, CA, USA) were performed within a period of 3 months of the two-test interval. Patients with ocular or systemic diseases other than glaucoma that are known to affect the retina and optic nerve were excluded. Moreover, patients with an axial length >27 mm were excluded. 
All the participants underwent a comprehensive ophthalmic evaluation, which included refraction, best-corrected Landolt decimal visual acuity, slit-lamp biomicroscopy, gonioscopy, Goldmann applanation tonometry, dilated funduscopy, and stereophotography. Axial length was measured using the IOL Master (Carl Zeiss Meditec AG, Jena, Germany). The results of HVF were considered reliable when the fixation loss was <20% and the false-positive and false-negative error rates were <25%. 
We aimed to evaluate the usefulness of en face slab images of the maculae for the detection of residual RNFL reflectivity in advanced glaucomatous eyes, wherein the thickness analysis would not be effective in capturing the glaucomatous structural change owing to the floor effect and relatively thin RNFL at the temporal disc region with high variability.2 Therefore, we included only the eyes with total cupping of the optic disc and a ganglion cell analysis (GCA) using Cirrus HD-OCT model 4000 (software version 6.1.0.96; Carl Zeiss Meditec AG) demonstrating that all sectors in the map were <1% of the normative database levels and could be displayed in red (Fig. 2), as described in detail below. A total of 36 eyes fulfilled all the criteria above. 
Figure 2
 
The structure–function relationships in four representative cases (cases 3, 5, 7, and 16 in Figs. 5, 6). Each raw image shows five concepts for the same eye: an optical coherence tomography ganglion cell layer (GCL) + inner plexiform layer (IPL) thickness map, an average GCL+IPL sector map, an average retinal nerve fiber layer (RNFL) sector map, an en face slab image of macula, and a Humphrey visual field (HVF) 10-2 grayscale (flipped upside down). Diffuse and substantial GCL+IPL thinning indicated by the GCL+IPL thickness maps (A, F, K, P) that are almost uniformly in blue and GCL+IPL sector maps in red, with essentially similar GCL+IPL thickness values among sectors (B, G, L, Q). In comparison, the average RNFL thickness across sectors is inconsistent. Some are two or more times thicker than others (C, H, M, R). The average RNFL thickness sector maps are not color coded according to the deviation due to a lack of the normative data. En face slab images show bundle-shaped high-reflective areas, colored in orange (D, I, N, S), that approximately correspond to the thicker RNFL sectors and test points with a normal sensitivity on the HVF (E, J, O, T).
Figure 2
 
The structure–function relationships in four representative cases (cases 3, 5, 7, and 16 in Figs. 5, 6). Each raw image shows five concepts for the same eye: an optical coherence tomography ganglion cell layer (GCL) + inner plexiform layer (IPL) thickness map, an average GCL+IPL sector map, an average retinal nerve fiber layer (RNFL) sector map, an en face slab image of macula, and a Humphrey visual field (HVF) 10-2 grayscale (flipped upside down). Diffuse and substantial GCL+IPL thinning indicated by the GCL+IPL thickness maps (A, F, K, P) that are almost uniformly in blue and GCL+IPL sector maps in red, with essentially similar GCL+IPL thickness values among sectors (B, G, L, Q). In comparison, the average RNFL thickness across sectors is inconsistent. Some are two or more times thicker than others (C, H, M, R). The average RNFL thickness sector maps are not color coded according to the deviation due to a lack of the normative data. En face slab images show bundle-shaped high-reflective areas, colored in orange (D, I, N, S), that approximately correspond to the thicker RNFL sectors and test points with a normal sensitivity on the HVF (E, J, O, T).
OCT Measurement and En Face Slab Image Acquisition
Experienced technicians conducted the OCT examinations. Images of each macula were obtained by using the Cirrus HD-OCT system employing the 200 × 200 macular cube protocol (200 horizontal B-scans comprising 200 A-scans per B-scan over 1024 samplings within a cube measuring 6 × 6 × 2 mm). Images with signal strength of <6 were excluded. The sum of the thickness of ganglion cell layer (GCL) and inner plexiform layer (IPL) was automatically presented using GCA settings in the onboard software. Briefly, the GCA algorithm identified the outer boundaries of the RNFL and IPL, and the tissue sandwiched between them was presented as the combined thickness of GCL+IPL (Figs. 2A, 2F, 2K, 2P). The sectoral thicknesses of GCL+IPL were measured within an elliptical annulus surrounding the fovea. In the sector map of the GCA report, the macular area was divided into six sectors, and the average GCL+IPL thickness of each sector was determined. This thickness was compared to age-matched normative data and color coded according to the deviation from the normative database (Figs. 2B, 2G, 2L, 2Q). The Cirrus HD-OCT model 4000 does not provide macular RNFL thickness analysis; however, Zeiss provided us with a software package that extracts average macular RNFL thickness at six sectors corresponding to the GCL+IPL thickness sector map (Figs. 2C, 2H, 2M, 2R). 
The detailed methodology for the acquisition of en face slab images using the Cirrus HD-OCT system has been reported elsewhere46; however, briefly, our en face slab images were reconstructed from the 200 × 200 macular cube scan data using advanced visualization settings (Fig. 1). Choosing the “ILM” tab from the dropdown menu, the software automatically creates a 6 × 6-mm en face slab image, which comprises retinal layers from the ILM to a specified depth (band areas sandwiched between the yellow solid lines and yellow dotted lines in Figs. 1D and 1E). The default setting for the slab thickness is 20 μm. The en face slab image represents an average signal intensity value for each A-scan location through the selected depth of the slab. 
The images were displayed in pseudocolors (0–255) based on the average reflectance intensity of the slab; areas with high reflectance intensity were indicated in warm colors, whereas those with low reflectance intensity were indicated in cold colors (Figs. 2D, 2I, 2N, 2S). In the OCT images, the membranes and nerve fibers are shown to be hyperreflective, whereas cell bodies are hyporeflective; therefore, the RNFL contributes more than the ganglion cell bodies to the en face intensity image produced by a fixed thickness slab. 
First, we attempted to determine the optimal slab thickness for analyzing the RNFL reflectance intensity by creating en face slab images of various thicknesses from the ILM using macular cube scan data from healthy and glaucomatous individuals (Fig. 3). In the en face slab images of healthy individuals, the average reflectance intensity initially increased with an increase in the slab thickness from ILM. However, on approaching approximately 50 μm thickness, the average reflectance intensity reached a plateau. On approaching approximately 100 μm thickness, the reflectance intensity started to decrease and increased again on approaching the most hyperreflective retinal pigment epithelium at approximately 250-μm thickness. Slabs with 50-μm thickness from the ILM included almost the entire RNFL in healthy eyes (Fig. 3D). On the other hand, RNFL was observed to be <30 μm thick in advanced glaucomatous eyes (Figs. 2C, 2H, 2M, 2R; Table); thus a part of the GCL beneath the RNFL may be included in the 50-μm-thick slab in those eyes (Fig. 3F). However, as GCL is relatively hyporeflective to RNFL (Fig. 3), the average reflectance intensity does not considerably differ at approximately 50 to 80 μm thickness. Therefore, we set the slab thickness to 50 ± 5 μm in the present study. For the same reason, Hood et al.4 set the slab thickness to 52 μm using swept-source OCT, although they also mentioned that reflectance intensity in en face OCT images should be analyzed complementarily, for it sacrifices the information about the thickness of RNFL with a fixed thickness slab analysis.4 
Figure 3
 
Representative pseudocolor (0–255) en face slab images in a series of different settings of slab thickness. (A) A normal eye. (B) An eye with advanced glaucoma. Areas with high reflectance intensity are indicated in warm colors, whereas those with low reflectance intensity are indicated in cold colors. The average reflectance intensity initially increases with an increase in slab thickness from the inner limiting membrane (ILM); on approaching approximately 50-μm thickness, this intensity reaches a plateau. On approaching approximately 100-μm thickness, the intensity starts decreasing, and again increases on approaching the most hyperreflective retinal pigment epithelium at approximately 250 μm. (C, E) Pseudocolor horizontal cross-sectional images of the macular region through the fovea in (A, B), respectively. (D, F) Grayscale images of (C, E), respectively. Yellow solid lines indicate the surface of the ILM. Dotted yellow lines are 50 μm beneath and parallel to the ILM. Black arrows indicate a highly reflective RNFL within the 50-μm-thick retinal slabs. It is noteworthy that RNFL is the major component showing high reflectivity in these 50-μm-thick slabs.
Figure 3
 
Representative pseudocolor (0–255) en face slab images in a series of different settings of slab thickness. (A) A normal eye. (B) An eye with advanced glaucoma. Areas with high reflectance intensity are indicated in warm colors, whereas those with low reflectance intensity are indicated in cold colors. The average reflectance intensity initially increases with an increase in slab thickness from the inner limiting membrane (ILM); on approaching approximately 50-μm thickness, this intensity reaches a plateau. On approaching approximately 100-μm thickness, the intensity starts decreasing, and again increases on approaching the most hyperreflective retinal pigment epithelium at approximately 250 μm. (C, E) Pseudocolor horizontal cross-sectional images of the macular region through the fovea in (A, B), respectively. (D, F) Grayscale images of (C, E), respectively. Yellow solid lines indicate the surface of the ILM. Dotted yellow lines are 50 μm beneath and parallel to the ILM. Black arrows indicate a highly reflective RNFL within the 50-μm-thick retinal slabs. It is noteworthy that RNFL is the major component showing high reflectivity in these 50-μm-thick slabs.
Table
 
Demographic and Ophthalmic Data of the Participants
Table
 
Demographic and Ophthalmic Data of the Participants
Agreement Between En Face Slab Images and the HVF Pattern Deviation Plots
Before we compared the 6 × 6-mm en face slab images of maculae and HVF 10-2 pattern deviation (PD) plots, the pseudocolor en face slab images (8-bit, 0–255 colors) were converted to a color-adjusted black and white version using Photoshop (version 11.4; Adobe Systems Co. Ltd, Tokyo, Japan). With the use of the “black and white” color adjusting tool, color images were converted to a black and white version with default color balance of red 40%, yellow 60%, green 40%, cyan 40%, blue 20%, and magenta 80%. After raising only the red color balance to the highest level (300%), areas with high reflectance intensity (red component-rich areas in pseudocolor image) were displayed in white, those with low reflectance intensity in gray (red component-poor areas) and black (no color component areas) (Fig. 4D). Test points of the HVF 10-2 PD plots were adjusted by considering the retinal ganglion cell (RGC) displacement according to previous studies2,7 (Fig. 4B). These PD plots were flipped upside down and superimposed on the en face slab images (Figs. 4E, 4F), thereby enabling comparison of the en face slab color-adjusted black and white images with the PD plots in a black circle zone, which encompassed all 68 test points in PD plots. 
Figure 4
 
Representative case 3 shows the correspondence of en face slab images with Humphrey visual field test (HVF) 10-2 pattern deviation (PD) plots. (A) A HVF 10-2 PD plot with original test point locations. (B) A HVF 10-2 PD plot with the test locations being adjusted after considering the retinal ganglion cell displacement. Open circles correspond to dots with a normal sensitivity in (A), whereas closed circles correspond to dashed squares with 1% level of significance below the normal sensitivity in (A). (C) A pseudocolor en face slab image of the same eye. (D) A color-adjusted black and white en face slab image converted from (C). (E) A composite image of (A) (flipped upside down) and (D). (F) A composite image of (B) (flipped upside down) and (D). Black circles in (B, E, F) indicate an analyzed zone wherein all 68 test points in HVF 10-2 are located.
Figure 4
 
Representative case 3 shows the correspondence of en face slab images with Humphrey visual field test (HVF) 10-2 pattern deviation (PD) plots. (A) A HVF 10-2 PD plot with original test point locations. (B) A HVF 10-2 PD plot with the test locations being adjusted after considering the retinal ganglion cell displacement. Open circles correspond to dots with a normal sensitivity in (A), whereas closed circles correspond to dashed squares with 1% level of significance below the normal sensitivity in (A). (C) A pseudocolor en face slab image of the same eye. (D) A color-adjusted black and white en face slab image converted from (C). (E) A composite image of (A) (flipped upside down) and (D). (F) A composite image of (B) (flipped upside down) and (D). Black circles in (B, E, F) indicate an analyzed zone wherein all 68 test points in HVF 10-2 are located.
First, we analyzed the relationship between structure and function in a qualitative manner. When both the white area with bundle-shaped appearance in the color-adjusted black and white image and points with normal sensitivity in the PD plots existed, slab and HVF judgments were flagged as “agree.” In addition, when the gray area in the slab image exhibited no normal point in the PD plots, judgments were flagged as “agree.” However, when the gray area in the slab images exhibited at least one normal test point in the PD plots, two-test judgment was flagged as “disagree.” These decisions were independently made by two examiners (M.S. and M.N.). 
Furthermore, we calculated the average total deviation (TD) of the test points located in the white and gray areas, by first anti-logging the TD values, which were then averaged, and relogging them.1,8 Paired t-test was used for the comparison. All statistical analyses were performed using Med Calc (version 18.2.1; MedCalc Software, Mariakerte, Belgium) with type I error for significance set at P < 0.05. 
Results
The Table shows demographic and ophthalmic data pertaining to the 36 glaucomatous eyes included in the present study. Figure 2 presents representative cases of glaucomatous eyes and shows that the en face slab images of advanced glaucomatous eyes seem to contain hyperreflective areas with nerve fiber bundle-like appearance in the macular region, where the GCL+IPL thickness at all sectors was reduced to <1% of the normative database levels despite the presence of normal points of HVF 10-2 PD plots. Remarkably, the GCL+IPL thickness map showed almost uniform thinning of these layers (Figs. 2A, 2F, 2K, 2P). Although the average RNFL thickness at each sector was reduced to <30 μm, in some sectors, RNFL was approximately twice as thick as the others in all four cases (Figs. 2C, 2H, 2M, 2R). The en face slab images demonstrated hyperreflective areas (Figs. 2D, 2I, 2N, 2S), which seemed to approximately correspond to the normal visual field points (Figs. 2E, 2J, 2O, 2T) and thicker RNFL sectors. 
The structure–function correlation did not appear to be influenced by the HVF test locations, after considering the RGC displacement. Figure 4 depicts a representative case where the correspondence of the hyperreflective areas was expressed as either a pseudocolor image (Fig. 4C) or a color-adjusted black and white image (Fig. 4D); HVF 10-2 PD plots were with either naïve test point locations (Fig. 4A) or adjusted locations after considering the RGC displacement (Fig. 4B). After flipping the HVF test points upside down and superimposing them onto the en face slab images, there was a fine spatial agreement between the hyperreflective areas and normal HVF test points (Figs. 4E, 4F), regardless of the HVF test point adjustment. 
Figure 5 depicts a series of en face slab images of all the included eyes. High-reflective areas with bundle-shaped appearance were observed in these images, except for cases 11, 34, and 36. Figure 6 demonstrates the HVF PD plots without and with RGC displacement adjustments (Figs. 6A, 6B, respectively), which were flipped upside down and superimposed on the color-adjusted black and white en face slab images of all 36 eyes. Among the glaucomatous eyes with significantly low macular GCL+IPL thickness (below 1% level of the lower limit of normal), 34 glaucomatous eyes had at least one normal test point in HVF10-2 PD plots. In addition, the en face slab images exhibited high-reflective areas with bundle-shaped appearance in 31 out of 34 eyes. In two eyes, all visual field test points were abnormal and no high-reflective areas were observed in the slab images (Figs. 5, 6; cases 11 and 36). With regard to the structural and functional agreement, 33 eyes were judged to “agree” with respect to the presence or absence of high reflectivity in the en face slab images and presence or absence of normal test points in the HVF10-2 PD plots. The remaining three eyes exhibited normal HVF test points despite a lack of high reflectivity areas in the analyzed circle zones of the slab images (Figs. 5, 6; cases 1, 32, and 34; red checks inside the images), thereby indicating a disagreement between the two tests. Therefore, 91.7% of the eyes showed agreement between the en face slab images and HVF test results. Two independent examiners (M.S. and M.N.) made completely identical judgments regarding the agreement between structure–function outcomes. 
Figure 5
 
Raw data of pseudocolor en face slab images of maculae of all 36 eyes. Areas with high reflectance intensity are indicated in warm colors, whereas those with low reflectance intensity are indicated in cold colors.
Figure 5
 
Raw data of pseudocolor en face slab images of maculae of all 36 eyes. Areas with high reflectance intensity are indicated in warm colors, whereas those with low reflectance intensity are indicated in cold colors.
Figure 6
 
Comparison between the en face slab images and Humphrey visual field test (HVF) 10-2 pattern deviation (PD) plots in 36 eyes. (A) Color-adjusted black and white en face slab images superimposed on the HVF PD plots (flipped upside down) using the original HVF test locations. (B) Images after the HVF test locations were adjusted for retinal ganglion cell displacement. Red checks are marked on three images (1, 32, and 34) that do not exhibit high reflectivity within the analyzed circle zone in en face slab images, despite the presence of normal visual field test points, indicating structure–function disagreement in these three eyes. Notably, in eyes numbered 11 and 36, all visual field test points are abnormal and no high reflectivity area is observed in the en face slab images, indicating agreement between the HVF and slab images. In the remaining 31 eyes, the two tests agree on the presence of functional areas within a circle of 10° angle radius. It is noteworthy that the adjustment of the HVF test locations for RGC displacement showed minor impact on the structure–function correlation.
Figure 6
 
Comparison between the en face slab images and Humphrey visual field test (HVF) 10-2 pattern deviation (PD) plots in 36 eyes. (A) Color-adjusted black and white en face slab images superimposed on the HVF PD plots (flipped upside down) using the original HVF test locations. (B) Images after the HVF test locations were adjusted for retinal ganglion cell displacement. Red checks are marked on three images (1, 32, and 34) that do not exhibit high reflectivity within the analyzed circle zone in en face slab images, despite the presence of normal visual field test points, indicating structure–function disagreement in these three eyes. Notably, in eyes numbered 11 and 36, all visual field test points are abnormal and no high reflectivity area is observed in the en face slab images, indicating agreement between the HVF and slab images. In the remaining 31 eyes, the two tests agree on the presence of functional areas within a circle of 10° angle radius. It is noteworthy that the adjustment of the HVF test locations for RGC displacement showed minor impact on the structure–function correlation.
The paired t-tests revealed that the average sensitivity of the TD of HVF 10-2 in white and gray areas of color-adjusted black and white en face slab images was −7.50 ± 5.01 and −22.61 ± 7.11 decibels (dB), respectively (P < 0.0001), when the HVF test points were unadjusted. TD in the former and latter were −7.46 ± 4.98 and −22.85 ± 6.98 dB, respectively (P < 0.0001), when the HVF test points were adjusted after considering the RGC displacement, indicating little impact of test location adjustments on the structure–function correlation. 
Figure 7 shows a representative case (case 16 in Figs. 5, 6) that indicates that the hyperreflective area on the en face slab image contains thicker RNFL than the hyporeflective area. Note that the retinal thickness of the slab analyses is maintained constant between the hyporeflective (Figs. 7C, 7D) and hyperreflective areas (Figs. 7E, 7F). 
Figure 7
 
En face slab image and B-scans of representative case 16. (A) A vertical section at the most nasal side of the scan cube. (B) An en face slab image. The two blue lines that are drawn at the same distance from the horizontal line via fovea represent the positions of horizontal sections shown in (CF). The upper line traverses a hyporeflective area, whereas the lower line passes through a hyperreflective area. (C) Grayscale and (D) pseudocolor OCT images of the horizontal section correspond to the upper blue line in (B). (E, F) Images of the section corresponding to the lower blue line in (B). Yellow solid lines in (C, E) indicate an inner limiting membrane (ILM), and dotted yellow lines are 50 μm beneath these ILMs. The area between these solid and dotted lines corresponds to the slab section being analyzed. Notably, the hyperreflective retinal nerve fibers, as shown in orange, are observed within the 50-μm-thick slab (F) that traverses the hyperreflective area on the en face slab image, whereas such fibers are not observed within the slab (D) that traverses the hyporeflective area on the en face slab image.
Figure 7
 
En face slab image and B-scans of representative case 16. (A) A vertical section at the most nasal side of the scan cube. (B) An en face slab image. The two blue lines that are drawn at the same distance from the horizontal line via fovea represent the positions of horizontal sections shown in (CF). The upper line traverses a hyporeflective area, whereas the lower line passes through a hyperreflective area. (C) Grayscale and (D) pseudocolor OCT images of the horizontal section correspond to the upper blue line in (B). (E, F) Images of the section corresponding to the lower blue line in (B). Yellow solid lines in (C, E) indicate an inner limiting membrane (ILM), and dotted yellow lines are 50 μm beneath these ILMs. The area between these solid and dotted lines corresponds to the slab section being analyzed. Notably, the hyperreflective retinal nerve fibers, as shown in orange, are observed within the 50-μm-thick slab (F) that traverses the hyperreflective area on the en face slab image, whereas such fibers are not observed within the slab (D) that traverses the hyporeflective area on the en face slab image.
Considering that segmentation error may have underestimated the GCL+IPL thickness, we examined the autosegmentation status in all the eyes in this study. Figure 8 depicts four representative cases, where autosegmentations of the border between RNFL and GCL (purple lines) and that between IPL and inner nuclear layer (yellow lines) were appropriately delineated (Figs. 8B, 8D, 8F, 8H). 
Figure 8
 
Confirmation on the absence of segmentation error in the automatic ganglion cell analysis in four representative cases that were randomly sampled. (A, C, E, G) These are the pseudocolor images before autosegmentation, whereas (B, D, F, H) are grayscale images of (A, C, E, G), respectively, with autosegmentation in the ganglion cell analysis. Purple lines indicate a border between the retinal nerve fiber layer and ganglion cell layer. Yellow lines indicate a border between the inner plexiform and inner nuclear layers.
Figure 8
 
Confirmation on the absence of segmentation error in the automatic ganglion cell analysis in four representative cases that were randomly sampled. (A, C, E, G) These are the pseudocolor images before autosegmentation, whereas (B, D, F, H) are grayscale images of (A, C, E, G), respectively, with autosegmentation in the ganglion cell analysis. Purple lines indicate a border between the retinal nerve fiber layer and ganglion cell layer. Yellow lines indicate a border between the inner plexiform and inner nuclear layers.
Figure 9 shows a representative case where the manually measured GCL+IPL thickness at one point in each sector, expressed by yellow open circles (Fig. 9A), was remarkably similar to the automatically calculated average GCL+IPL thickness at the corresponding sectors. Moreover, the average differences between manually measured GCL+IPL thickness and automatically calculated average GCL+IPL thickness in each sector were −1.6 ± 6.2 (superonasal), 0.1 ± 5.2 (superior), 0.7 ± 6.6 (superotemporal), 1.5 ± 4.1 (inferotemporal), 0.7 ± 5.4 (inferior), and 0.2 ± 7.0 μm (inferonasal). The average of the manually measured GCL+IPL thickness of all eyes was 53.2 ± 3.4 μm, which was similar to that of automatic calculation based on the GCA analysis (52.8 ± 3.4 μm; Table). Indeed, no statistical difference was observed between the two values (paired t-test, P = 0.62). These lines of evidence indicate that the reduced GCL+IPL thickness was not owing to segmentation error but attributable to a true reduction of these layers in these eyes. 
Figure 9
 
Comparison of the ganglion cell layer (GCL) and inner plexiform layer (IPL) thicknesses measured by an automatic segmentation versus manual segmentation in a representative case. (A) A fundus photograph showing an ellipse in red, within which the ganglion cell analysis automatically calculated the GCL+IPL thickness; the points wherein these thicknesses were manually measured are indicated as yellow open circles. Values displayed along with the yellow open circles indicate average values of GCL+IPL thickness that were manually measured in horizontal and vertical sections at each point. (B) The GCL+IPL average sector map automatically measured by ganglion cell analysis. These values are averaged measurements of GCL+IPL thickness in each sector. (C) Vertical and (D) horizontal B-scans, respectively, at the upper middle point in (A). Yellow dashes on the blue (C) and pink (D) lines, corresponding to A-scans at the upper middle point in (A), are the manually measured GCL+IPL thicknesses. The number displayed above the dashes is the GCL+IPL thickness. Notably, the manually measured values in each sector were similar to the automeasured average values in the corresponding sector.
Figure 9
 
Comparison of the ganglion cell layer (GCL) and inner plexiform layer (IPL) thicknesses measured by an automatic segmentation versus manual segmentation in a representative case. (A) A fundus photograph showing an ellipse in red, within which the ganglion cell analysis automatically calculated the GCL+IPL thickness; the points wherein these thicknesses were manually measured are indicated as yellow open circles. Values displayed along with the yellow open circles indicate average values of GCL+IPL thickness that were manually measured in horizontal and vertical sections at each point. (B) The GCL+IPL average sector map automatically measured by ganglion cell analysis. These values are averaged measurements of GCL+IPL thickness in each sector. (C) Vertical and (D) horizontal B-scans, respectively, at the upper middle point in (A). Yellow dashes on the blue (C) and pink (D) lines, corresponding to A-scans at the upper middle point in (A), are the manually measured GCL+IPL thicknesses. The number displayed above the dashes is the GCL+IPL thickness. Notably, the manually measured values in each sector were similar to the automeasured average values in the corresponding sector.
Figure 10 depicts four representative cases that further demonstrated that the pixel-based GCL+IPL thickness, which was manually measured, was similarly reduced in the hyperreflective and hyporeflective areas in the en face slab images. In Figure 10, the intersections of a purple line and two blue parallel lines with the same distance from the horizontal midline of the image indicate the B-scan locations of the GCL+IPL thickness measurement, while the yellow dashes indicate the measured range of the GCL+IPL thickness. It is noteworthy that the hyperreflective areas contain the remaining RNFL (white layers observed at the left side adjacent to the yellow dashes in Figs. 10B, 10D, 10F, 10H), indicating that hyperreflective area with bundle-shaped appearance on en face slab image corresponds to the residual RNFL, probably with a preserved visual sensitivity in advanced glaucomatous eyes despite the GCL+IPL thickness being substantially reduced. 
Figure 10
 
Four representative cases of manually measured pixel-based GCL+IPL thicknesses in a set of B-scans of the same eyes [(A, B) case 3; (C, D) case 5; (E, F) case 7; (G, H) case 16, respectively]. (A, C, E, G) Grayscale en face slab images with two blue lines traversing both hyper- and hyporeflective nerve fiber bundles. (B, D, F, H) The vertical sections at purple lines in (A, C, E, G), respectively. The GCL+IPL thickness was manually measured at the intersection points of the blue and purple lines in (A, C, E, G), and the values are indicated in yellow dashes in (B, D, F, H), respectively. As the two blue lines are parallel and at the same distance from the horizontal midline of the image, the eccentricity of the two measured points in each image is the same. Notably, the hyperreflective area on the en face slab image contains hyperreflective RNFL in the B-scans, whereas the hyporeflective area does not. However, the pixel-based thickness of the GCL+IPL on the same B-scans is remarkably similar in the same eyes.
Figure 10
 
Four representative cases of manually measured pixel-based GCL+IPL thicknesses in a set of B-scans of the same eyes [(A, B) case 3; (C, D) case 5; (E, F) case 7; (G, H) case 16, respectively]. (A, C, E, G) Grayscale en face slab images with two blue lines traversing both hyper- and hyporeflective nerve fiber bundles. (B, D, F, H) The vertical sections at purple lines in (A, C, E, G), respectively. The GCL+IPL thickness was manually measured at the intersection points of the blue and purple lines in (A, C, E, G), and the values are indicated in yellow dashes in (B, D, F, H), respectively. As the two blue lines are parallel and at the same distance from the horizontal midline of the image, the eccentricity of the two measured points in each image is the same. Notably, the hyperreflective area on the en face slab image contains hyperreflective RNFL in the B-scans, whereas the hyporeflective area does not. However, the pixel-based thickness of the GCL+IPL on the same B-scans is remarkably similar in the same eyes.
Discussion
In the present study, en face slab images of maculae demonstrated hyperreflective areas with bundle-shaped appearance that exhibited significantly higher TD than the corresponding hyporeflective areas in eyes with advanced glaucoma, despite the overall reduction of GCL+IPL thickness. 
Hood et al.4 analyzed 52-μm en face slab images using a RNFL thickness map derived from swept-source OCT and spatially compared them with peripapillary images obtained from an adaptive optics scanning laser ophthalmoscope (AO-SLO) of six glaucomatous eyes with deep defects near fixation, as seen on the HVF 10-2 tests. These images demonstrated that en face OCT images were able to show the details of the small regions of preserved and/or missing RNFL bundles as observed in AO-SLO images, which traditional RNFL thickness analysis failed to detect.4 The en face slab image mode has already been incorporated as “Hood's report” in a commercially available swept-source OCT to assess the structural damage in glaucomatous eyes. We used the spectral-domain Cirrus HD-OCT system to compare the en face slab images with GCL+IPL thickness analysis of the macular region. 
Hood et al.4 discussed that one of the reasons for the superiority of en face slab image compared with RNFL thickness analysis was an erroneous autosegmentation in the thickness analysis. Segmentation error is less of a concern in the case of en face slab image analysis, considering that the most identifiable ILM is used as a reference line. On the other hand, they pointed out the limitation of the en face image with fixed thickness. Considering that the RNFL thickness varies depending on the location or individuals, and that the abnormal regions of RNFL may vary in reflectance intensity and thickness, there is no single fixed slab thickness that can cover all the glaucomatous changes in RNFL. In addition, the en face slab images can overlook information about RNFL thickness when the reflectance intensities within the slab are averaged. Therefore, these authors concluded that RNFL thickness analysis should not be replaced with analysis of reflectance intensity in OCT en face images, and the information available in OCT projection images such as fixed-thickness 52-μm slab images should be considered complementary. 
Similar to the RNFL thickness analysis, GCA depends on computer segmentation algorithms.9 Therefore, using three different approaches, we scrutinized whether the segmentation error influences the measurement of GCL+IPL thickness, whether the GCL+IPL thicknesses are accurately measured, and whether the GCL+IPL thicknesses are relatively well preserved in sections beneath high-reflective nerve fibers on the en face slab images. 
First, we carefully examined the autosegmented lines at the border between RNFL and GCL (purple lines in Figs. 8B, 8D, 8F, 8H), and at the border between IPL and inner nuclear layer (yellow lines in same) in all the B-scans of acquired OCT images. Figure 8 shows four representative B-scans randomly sampled that depicted no significant segmentation errors in all scans. Thereafter, we manually segmented these layers based on their B-scans and measured the GCL+IPL thickness at one point in each sector of the GCL+IPL sector map (Fig. 9). On calculating the average thicknesses of horizontal and vertical scans, we observed that the manually measured GCL+IPL thicknesses were remarkably similar to the average GCL+IPL thicknesses that had been automatically calculated on the sector map (Table). Therefore, autosegmentation demonstrated little impact on the GCL+IPL measurement, and it was evident that the GCL+IPL thickness exhibited substantial reductions in all the tested eyes. 
Third, we manually measured pixel-based GCL+IPL thickness in a set of B-scans of the same eyes that were horizontally traversed through hyperreflective and hyporeflective nerve fiber bundles on en face slab images (Fig. 10). We observed that the hyperreflective area on en face slab image contained hyperreflective RNFL in the B-scans, whereas the hyporeflective area did not. However, the pixel-based thickness of GCL+IPL on these B-scans of the same eyes was remarkably similar. These findings further indicate that segmentation error had minor influence on the GCL+IPL thickness measurement and that GCL+IPL thickness decreased beneath both hyporeflective and hyperreflective RNFL. 
In our study, we focused on advanced glaucomatous eyes, in which thickness analysis is typically less effective for assessing the structure–function correlation owing to floor effect. Raza et al.,2 using frequency-domain OCT, have observed that GCL+IPL thickness correlates well with localized sensitivity loss in glaucomatous eyes. They applied a simple linear model developed by Hood and Kardon1 for assessing the relationship between GCL+IPL thickness and localized sensitivity loss; findings showed a good correlation between them. Furthermore, they demonstrated that the structure–function relationship did not exhibit correlation in the advanced stage of glaucoma, where the visual field damage was worse than approximately −10 dB. As presented in the Table, 36 advanced glaucomatous eyes included in the present study had a mean deviation of −20.7 dB. Therefore, the observation that GCL+IPL thickness analysis did not correlate with visual field loss was not unexpected. Raza et al.2 found that the residual value of GCL+IPL was approximately 45 μm. In our previous study, the base level of the average GCL+IPL thickness was determined as 57.7 μm using Cirrus OCT.3 In the current study, the average GCL+IPL thickness of all the eyes was 52.8 ± 3.4 μm. Assuming the GCL+IPL thickness was accurately measured, the value was close to the floor level. As GCA measures the composite thickness of GCL+IPL, which includes various nonaxonal elements, such as ganglion cells, glial cells, vessels, and extracellular matrix, GCL+IPL thinning does not correlate perfectly linearly with the loss of RGCs. 
On the other hand, our en face slab images of approximately 50-μm thickness were reconstructed from the average signal intensity that was substantially derived from RNFLs (Fig. 7). Considering that nerve fibers are hyperreflective10 and that other nonaxonal elements, except vessels, are relatively hyporeflective, it is reasonable to believe that the en face slab images of advanced glaucomatous eyes mainly reflect the residual retinal nerve fibers. 
Raza et al.2 reported that the relationship between GCL+IPL thickness and visual field loss was weaker outside of approximately 7.2° on the retina. The elliptical annulus used in GCA is determined according to the distribution of RGCs in the macular region and corresponds to the region where the GCL is thickest in normal eyes.9 Areas outside the annulus were not analyzed in GCA. On the other hand, as shown in Figure 5, en face slab image effectively captured the high-reflective areas, mostly on the nasal side of the 6 × 6-mm cube scan. As more nerve fibers are present in the region adjacent to the optic disc, en face slab images may be much better at visualizing the residual nerve fibers in the region outside the annulus examined by GCA. 
Changes in RNFL reflectance intensity have been shown to precede changes in RFNL thickness. Huang et al.11 reported that elevated intraocular pressure caused a decrease in RNFL reflectance prior to the change in its thickness in a rat model of glaucoma. Additionally, they reported that a decrease in RNFL reflectance following an optic nerve crush occurred prior to the thinning of RNFL in a rat model.12 Additionally, Fortune et al.13,14 reported that RNFL retardance changes occurred earlier than the changes in RNFL thickness in rhesus macaques. In a clinical study by Gardiner et al.,15 reduction in RNFL reflectance was associated with a rapid functional deterioration for a given rate of RNFL thinning. Hence, they concluded that the incorporation of OCT reflectance information may provide a better characterization of the structure–function correlation in glaucomatous eyes. However, these reports assessed OCT reflectance information during the early stages of glaucoma, whereas the present study, to the best of our knowledge, is the first to assess the usefulness of OCT reflectivity analysis as a structural biomarker in patients with advanced glaucoma. 
Although the high-reflective areas in the en face slab images appeared to approximately correspond to the visual field tests, no strict spatial correspondence was observed between them despite adjusting the test points of HVF for RGC displacement (Figs. 4, 6). En face slab images reflect the residual nerve fiber bundles but not the RGCs because RNFL contains fibers passing from other locations. In comparison, visual field test sensitivity represents the function of receptive fields of RGCs. Therefore, theoretically, a strict spatial correspondence between the en face slab high-reflective bundles and visual field test points does not exist. 
Tissue reflectance varies across different scans and individuals; therefore previous studies used normalization techniques, such as RPE-referenced attenuation coefficients, for the assessment of reflectance intensity.16 However, we did not normalize the reflectance intensity of our en face slab images because we aimed to evaluate their usefulness in clinical settings with commercially available OCT. Pseudocolor en face slab images that were automatically generated using the onboard software could be readily obtained in clinical settings and appeared to exhibit sufficient ability in depicting the structural damage in advanced glaucoma. As shown in Figure 10, grayscale en face slab images are also available using Cirrus HD-OCT in clinical settings. Depending on the purpose, grayscale en face slab images may serve as an alternative to pseudocolor images. We converted the pseudocolor images into color-adjusted black and white versions to facilitate precise judgment on the presence of bundle-shaped hyperreflectivity. 
In the en face slab images, vessels are highly reflective as are retinal nerve fibers, and it is challenging to distinguish between them based on their reflectivity. However, as observed in Figure 5, vessels can be easily recognized by their paths and appearances. It has been reported that gliotic changes can manifest as an increased reflectance5 and epiretinal membranes similarly produce hyperreflective foci. Although we did not particularly focus on it, Figure 6 shows that no apparent glial alteration reflectance was observed in the present study. Patients with retinal diseases including epiretinal membranes that may affect the retinal reflectivity were excluded in this study. 
The limitations of our present study include small sample size and the semiquantitative nature of the analysis. Although OCT images are affected by axial length,17,18 the OCT device used in the present study was unable to provide magnification correction. Although eyes with axial length >27 mm were excluded, many myopic eyes with relatively long axial lengths remained. As already mentioned, tissue reflectance varies across different scans and individuals; hence we did not quantify the reflectance intensity of each image. However, Ashimatey et al.5 observed a strong correlation between entire or hemifield en face RNFL reflectance abnormalities and cpRNFL thinning by using a slab analysis, wherein the slab thickness was adjusted to minimize the putative glial alteration reflectance on visualizing RNFL reflectance in patients with early glaucomatous change with a mean MD of −3.3 dB. As stated above, no apparent glial alteration reflectance was observed in the present study. However, as indicated by Ashimatey et al.,5 a fixed slab thickness for the entire macular region analysis might have missed reflective abnormalities in the deeper layer and underestimated the correlation between thickness and reflectivity, particularly at the temporal region. In addition, the directional reflectivity of the nerve fiber bundles may affect the variability of the reflectivity.10 Further refinement is required to quantify the reflectivity and analyze the correlation between the degree of reflectivity and visual field damage. Lastly, we could not perform direct comparison between macular RNFL (mRNFL) thickness map and en face slab images with our data because the OCT system used in this study does not provide mRNFL thickness maps, and it was technically difficult to create the map by ourselves. As shown in Figure 2, the findings of en face slab images and averaged value of mRNFL thickness in each sector were correlated. Thus, similar results can be produced by using the mRNFL thickness map, as long as the segmentation is performing properly. But mRNFL per A-scan was too thin to be properly and consistently segmented and thus was practically highly challenging to measure. This is the major reason why the current version of Cirrus OCT does not report the mRNFL thickness. In clear comparison, it is quite straightforward to obtain en face slab images, which reflect both thickness and intensity information. If a more advanced device such as Cirrus HD-OCT model 5000, adaptive optics SLO, or swept-source OCT were used, as reported by Hood et al.,4 the comparison between mRNFL thickness and the slab image analysis in terms of the structure–function correlation in the advanced stage of glaucoma could be carried out. Further analysis is required to compare the ability of the mRNFL thickness map and en face slab image to detect residual functioning RGC axon bundles. 
In conclusion, en face slab images of maculae facilitated the visualization of residual nerve fiber bundles via reflectance properties, which partly corresponded to the residual normal visual function within the severely damaged visual field in advanced glaucomatous eyes despite the GCL+IPL thickness of maculae being substantially reduced. Further research is required to quantify the nerve fiber bundle reflectance for evaluating its usefulness for assessing the structure–function relationships. 
Acknowledgments
Presented in part at the 7th World Glaucoma Congress in Helsinki, Finland, June 2017. 
Supported by Grants-in-Aid No. 18K09447 from the Japan Society for the Promotion of Science (MN). 
Disclosure: M. Sakamoto, None; S. Mori, None; K. Ueda, None; T. Kurimoto, None; S. Kusuhara, None; Y. Yamada-Nakanishi, None; M. Nakamura, None 
References
Hood DC, Kardon RH. A framework for comparing structural and functional measures of glaucomatous damage. Prog Retin Eye Res. 2007; 26: 688–710.
Raza AS, Cho J, de Moraes CG, et al. Retinal ganglion cell layer thickness and local visual field sensitivity in glaucoma. Arch Ophthalmol. 2011; 129: 1529–1536.
Ueda K, Kanamori A, Akashi A, Kawaka Y, Yamada Y, Nakamura M. Difference in correspondence between visual field defect and inner macular layer thickness measured using three types of spectral-domain OCT instruments. Jpn J Ophthalmol. 2015; 59: 55–64.
Hood DC, Fortune B, Mavrommatis MA, et al. Details of glaucomatous damage are better seen on OCT en face images than on OCT retinal nerve fiber layer thickness maps. Invest Ophthalmol Vis Sci. 2015; 56: 6208–6216.
Ashimatey BS, King BJ, Burns SA, Swanson WH. Evaluating glaucomatous abnormality in peripapillary optical coherence tomography enface visualisation of the retinal nerve fibre layer reflectance. Ophthalmic Physiol Opt. 2018; 38: 376–388.
Oh J, Smiddy WE, Flynn HWJr, Gregori G, Lujan B. Photoreceptor inner/outer segment defect imaging by spectral domain OCT and visual prognosis after macular hole surgery. Invest Ophthalmol Vis Sci. 2010; 51: 1651–1658.
Drasdo N, Millican CL, Katholi CR, Curcio CA. The length of Henle fibers in the human retina and a model of ganglion receptive field density in the visual field. Vision Res. 2007; 47: 2901–2911.
Nakamura M, Ishikawa K, Nagai T, Negi A. Receiver-operating characteristic analysis of multifocal VEPs to diagnose and quantify glaucomatous functional damage. Doc Ophthalmol. 2011; 123: 93–108.
Mwanza JC, Oakley JD, Budenz DL, Chang RT, Knight OJ, Feuer WJ. Macular ganglion cell-inner plexiform layer: automated detection and thickness reproducibility with spectral domain-optical coherence tomography in glaucoma. Invest Ophthalmol Vis Sci. 2011; 52: 8323–8329.
Huang XR, Knighton RW, Cavuoto LN. Microtubule contribution to the reflectance of the retinal nerve fiber layer. Invest Ophthalmol Vis Sci. 2006; 47: 5363–5367.
Huang XR, Zhou Y, Kong W, Knighton RW. Reflectance decreases before thickness changes in the retinal nerve fiber layer in glaucomatous retinas. Invest Ophthalmol Vis Sci. 2011; 52: 6737–6742.
Huang XR, Kong W, Qiao J. Response of the retinal nerve fiber layer reflectance and thickness to optic nerve crush. Invest Ophthalmol Vis Sci. 2018; 59: 2094–2103.
Fortune B, Burgoyne CF, Cull G, Reynaud J, Wang L. Onset and progression of peripapillary retinal nerve fiber layer (RNFL) retardance changes occur earlier than RNFL thickness changes in experimental glaucoma. Invest Ophthalmol Vis Sci. 2013; 54: 5653–5661.
Fortune B, Burgoyne CF, Cull GA, Reynaud J, Wang L. Structural and functional abnormalities of retinal ganglion cells measured in vivo at the onset of optic nerve head surface change in experimental glaucoma. Invest Ophthalmol Vis Sci. 2012; 53: 3939–3950.
Gardiner SK, Demirel S, Reynaud J, Fortune B. Changes in retinal nerve fiber layer reflectance intensity as a predictor of functional progression in glaucoma. Invest Ophthalmol Vis Sci. 2016; 57: 1221–1227.
Vermeer KA, van der Schoot J, Lemij HG, de Boer JF. RPE-normalized RNFL attenuation coefficient maps derived from volumetric OCT imaging for glaucoma assessment. Invest Ophthalmol Vis Sci. 2012; 53: 6102–6108.
Ueda K, Kanamori A, Akashi A, Tomioka M, Kawaka Y, Nakamura M. Effects of axial length and age on circumpapillary retinal nerve fiber layer and inner macular parameters measured by 3 types of SD-OCT instruments. J Glaucoma. 2016; 25: 383–389.
Akashi A, Kanamori A, Ueda K, Inoue Y, Yamada Y, Nakamura M. The ability of SD-OCT to differentiate early glaucoma with high myopia from highly myopic controls and nonhighly myopic controls. Invest Ophthalmol Vis Sci. 2015; 56: 6573–6580.
Figure 1
 
Representative en face slab image of the macula in a normal eye. (A) A pseudocolor en face slab image overlaid on a fundus photograph. Areas with high reflectivity are expressed by warm color, whereas those with low reflectivity are indicated in cold color. (B) A pseudocolor horizontal cross-section image at the blue line in (A). (C) A pseudocolor vertical cross-section image at the purple line in (A). (D, E) Grayscale images of (B, C), respectively. Yellow solid lines indicate the surface of inner limiting membrane (ILM). Dotted yellow lines are 50 μm beneath and parallel to the ILM. The pseudocolor en face slab image in (A) reflects the average reflectance intensity of the “slab” sandwiched between the ILM and border 50 μm beneath and parallel to the ILM (black arrows in D, E).
Figure 1
 
Representative en face slab image of the macula in a normal eye. (A) A pseudocolor en face slab image overlaid on a fundus photograph. Areas with high reflectivity are expressed by warm color, whereas those with low reflectivity are indicated in cold color. (B) A pseudocolor horizontal cross-section image at the blue line in (A). (C) A pseudocolor vertical cross-section image at the purple line in (A). (D, E) Grayscale images of (B, C), respectively. Yellow solid lines indicate the surface of inner limiting membrane (ILM). Dotted yellow lines are 50 μm beneath and parallel to the ILM. The pseudocolor en face slab image in (A) reflects the average reflectance intensity of the “slab” sandwiched between the ILM and border 50 μm beneath and parallel to the ILM (black arrows in D, E).
Figure 2
 
The structure–function relationships in four representative cases (cases 3, 5, 7, and 16 in Figs. 5, 6). Each raw image shows five concepts for the same eye: an optical coherence tomography ganglion cell layer (GCL) + inner plexiform layer (IPL) thickness map, an average GCL+IPL sector map, an average retinal nerve fiber layer (RNFL) sector map, an en face slab image of macula, and a Humphrey visual field (HVF) 10-2 grayscale (flipped upside down). Diffuse and substantial GCL+IPL thinning indicated by the GCL+IPL thickness maps (A, F, K, P) that are almost uniformly in blue and GCL+IPL sector maps in red, with essentially similar GCL+IPL thickness values among sectors (B, G, L, Q). In comparison, the average RNFL thickness across sectors is inconsistent. Some are two or more times thicker than others (C, H, M, R). The average RNFL thickness sector maps are not color coded according to the deviation due to a lack of the normative data. En face slab images show bundle-shaped high-reflective areas, colored in orange (D, I, N, S), that approximately correspond to the thicker RNFL sectors and test points with a normal sensitivity on the HVF (E, J, O, T).
Figure 2
 
The structure–function relationships in four representative cases (cases 3, 5, 7, and 16 in Figs. 5, 6). Each raw image shows five concepts for the same eye: an optical coherence tomography ganglion cell layer (GCL) + inner plexiform layer (IPL) thickness map, an average GCL+IPL sector map, an average retinal nerve fiber layer (RNFL) sector map, an en face slab image of macula, and a Humphrey visual field (HVF) 10-2 grayscale (flipped upside down). Diffuse and substantial GCL+IPL thinning indicated by the GCL+IPL thickness maps (A, F, K, P) that are almost uniformly in blue and GCL+IPL sector maps in red, with essentially similar GCL+IPL thickness values among sectors (B, G, L, Q). In comparison, the average RNFL thickness across sectors is inconsistent. Some are two or more times thicker than others (C, H, M, R). The average RNFL thickness sector maps are not color coded according to the deviation due to a lack of the normative data. En face slab images show bundle-shaped high-reflective areas, colored in orange (D, I, N, S), that approximately correspond to the thicker RNFL sectors and test points with a normal sensitivity on the HVF (E, J, O, T).
Figure 3
 
Representative pseudocolor (0–255) en face slab images in a series of different settings of slab thickness. (A) A normal eye. (B) An eye with advanced glaucoma. Areas with high reflectance intensity are indicated in warm colors, whereas those with low reflectance intensity are indicated in cold colors. The average reflectance intensity initially increases with an increase in slab thickness from the inner limiting membrane (ILM); on approaching approximately 50-μm thickness, this intensity reaches a plateau. On approaching approximately 100-μm thickness, the intensity starts decreasing, and again increases on approaching the most hyperreflective retinal pigment epithelium at approximately 250 μm. (C, E) Pseudocolor horizontal cross-sectional images of the macular region through the fovea in (A, B), respectively. (D, F) Grayscale images of (C, E), respectively. Yellow solid lines indicate the surface of the ILM. Dotted yellow lines are 50 μm beneath and parallel to the ILM. Black arrows indicate a highly reflective RNFL within the 50-μm-thick retinal slabs. It is noteworthy that RNFL is the major component showing high reflectivity in these 50-μm-thick slabs.
Figure 3
 
Representative pseudocolor (0–255) en face slab images in a series of different settings of slab thickness. (A) A normal eye. (B) An eye with advanced glaucoma. Areas with high reflectance intensity are indicated in warm colors, whereas those with low reflectance intensity are indicated in cold colors. The average reflectance intensity initially increases with an increase in slab thickness from the inner limiting membrane (ILM); on approaching approximately 50-μm thickness, this intensity reaches a plateau. On approaching approximately 100-μm thickness, the intensity starts decreasing, and again increases on approaching the most hyperreflective retinal pigment epithelium at approximately 250 μm. (C, E) Pseudocolor horizontal cross-sectional images of the macular region through the fovea in (A, B), respectively. (D, F) Grayscale images of (C, E), respectively. Yellow solid lines indicate the surface of the ILM. Dotted yellow lines are 50 μm beneath and parallel to the ILM. Black arrows indicate a highly reflective RNFL within the 50-μm-thick retinal slabs. It is noteworthy that RNFL is the major component showing high reflectivity in these 50-μm-thick slabs.
Figure 4
 
Representative case 3 shows the correspondence of en face slab images with Humphrey visual field test (HVF) 10-2 pattern deviation (PD) plots. (A) A HVF 10-2 PD plot with original test point locations. (B) A HVF 10-2 PD plot with the test locations being adjusted after considering the retinal ganglion cell displacement. Open circles correspond to dots with a normal sensitivity in (A), whereas closed circles correspond to dashed squares with 1% level of significance below the normal sensitivity in (A). (C) A pseudocolor en face slab image of the same eye. (D) A color-adjusted black and white en face slab image converted from (C). (E) A composite image of (A) (flipped upside down) and (D). (F) A composite image of (B) (flipped upside down) and (D). Black circles in (B, E, F) indicate an analyzed zone wherein all 68 test points in HVF 10-2 are located.
Figure 4
 
Representative case 3 shows the correspondence of en face slab images with Humphrey visual field test (HVF) 10-2 pattern deviation (PD) plots. (A) A HVF 10-2 PD plot with original test point locations. (B) A HVF 10-2 PD plot with the test locations being adjusted after considering the retinal ganglion cell displacement. Open circles correspond to dots with a normal sensitivity in (A), whereas closed circles correspond to dashed squares with 1% level of significance below the normal sensitivity in (A). (C) A pseudocolor en face slab image of the same eye. (D) A color-adjusted black and white en face slab image converted from (C). (E) A composite image of (A) (flipped upside down) and (D). (F) A composite image of (B) (flipped upside down) and (D). Black circles in (B, E, F) indicate an analyzed zone wherein all 68 test points in HVF 10-2 are located.
Figure 5
 
Raw data of pseudocolor en face slab images of maculae of all 36 eyes. Areas with high reflectance intensity are indicated in warm colors, whereas those with low reflectance intensity are indicated in cold colors.
Figure 5
 
Raw data of pseudocolor en face slab images of maculae of all 36 eyes. Areas with high reflectance intensity are indicated in warm colors, whereas those with low reflectance intensity are indicated in cold colors.
Figure 6
 
Comparison between the en face slab images and Humphrey visual field test (HVF) 10-2 pattern deviation (PD) plots in 36 eyes. (A) Color-adjusted black and white en face slab images superimposed on the HVF PD plots (flipped upside down) using the original HVF test locations. (B) Images after the HVF test locations were adjusted for retinal ganglion cell displacement. Red checks are marked on three images (1, 32, and 34) that do not exhibit high reflectivity within the analyzed circle zone in en face slab images, despite the presence of normal visual field test points, indicating structure–function disagreement in these three eyes. Notably, in eyes numbered 11 and 36, all visual field test points are abnormal and no high reflectivity area is observed in the en face slab images, indicating agreement between the HVF and slab images. In the remaining 31 eyes, the two tests agree on the presence of functional areas within a circle of 10° angle radius. It is noteworthy that the adjustment of the HVF test locations for RGC displacement showed minor impact on the structure–function correlation.
Figure 6
 
Comparison between the en face slab images and Humphrey visual field test (HVF) 10-2 pattern deviation (PD) plots in 36 eyes. (A) Color-adjusted black and white en face slab images superimposed on the HVF PD plots (flipped upside down) using the original HVF test locations. (B) Images after the HVF test locations were adjusted for retinal ganglion cell displacement. Red checks are marked on three images (1, 32, and 34) that do not exhibit high reflectivity within the analyzed circle zone in en face slab images, despite the presence of normal visual field test points, indicating structure–function disagreement in these three eyes. Notably, in eyes numbered 11 and 36, all visual field test points are abnormal and no high reflectivity area is observed in the en face slab images, indicating agreement between the HVF and slab images. In the remaining 31 eyes, the two tests agree on the presence of functional areas within a circle of 10° angle radius. It is noteworthy that the adjustment of the HVF test locations for RGC displacement showed minor impact on the structure–function correlation.
Figure 7
 
En face slab image and B-scans of representative case 16. (A) A vertical section at the most nasal side of the scan cube. (B) An en face slab image. The two blue lines that are drawn at the same distance from the horizontal line via fovea represent the positions of horizontal sections shown in (CF). The upper line traverses a hyporeflective area, whereas the lower line passes through a hyperreflective area. (C) Grayscale and (D) pseudocolor OCT images of the horizontal section correspond to the upper blue line in (B). (E, F) Images of the section corresponding to the lower blue line in (B). Yellow solid lines in (C, E) indicate an inner limiting membrane (ILM), and dotted yellow lines are 50 μm beneath these ILMs. The area between these solid and dotted lines corresponds to the slab section being analyzed. Notably, the hyperreflective retinal nerve fibers, as shown in orange, are observed within the 50-μm-thick slab (F) that traverses the hyperreflective area on the en face slab image, whereas such fibers are not observed within the slab (D) that traverses the hyporeflective area on the en face slab image.
Figure 7
 
En face slab image and B-scans of representative case 16. (A) A vertical section at the most nasal side of the scan cube. (B) An en face slab image. The two blue lines that are drawn at the same distance from the horizontal line via fovea represent the positions of horizontal sections shown in (CF). The upper line traverses a hyporeflective area, whereas the lower line passes through a hyperreflective area. (C) Grayscale and (D) pseudocolor OCT images of the horizontal section correspond to the upper blue line in (B). (E, F) Images of the section corresponding to the lower blue line in (B). Yellow solid lines in (C, E) indicate an inner limiting membrane (ILM), and dotted yellow lines are 50 μm beneath these ILMs. The area between these solid and dotted lines corresponds to the slab section being analyzed. Notably, the hyperreflective retinal nerve fibers, as shown in orange, are observed within the 50-μm-thick slab (F) that traverses the hyperreflective area on the en face slab image, whereas such fibers are not observed within the slab (D) that traverses the hyporeflective area on the en face slab image.
Figure 8
 
Confirmation on the absence of segmentation error in the automatic ganglion cell analysis in four representative cases that were randomly sampled. (A, C, E, G) These are the pseudocolor images before autosegmentation, whereas (B, D, F, H) are grayscale images of (A, C, E, G), respectively, with autosegmentation in the ganglion cell analysis. Purple lines indicate a border between the retinal nerve fiber layer and ganglion cell layer. Yellow lines indicate a border between the inner plexiform and inner nuclear layers.
Figure 8
 
Confirmation on the absence of segmentation error in the automatic ganglion cell analysis in four representative cases that were randomly sampled. (A, C, E, G) These are the pseudocolor images before autosegmentation, whereas (B, D, F, H) are grayscale images of (A, C, E, G), respectively, with autosegmentation in the ganglion cell analysis. Purple lines indicate a border between the retinal nerve fiber layer and ganglion cell layer. Yellow lines indicate a border between the inner plexiform and inner nuclear layers.
Figure 9
 
Comparison of the ganglion cell layer (GCL) and inner plexiform layer (IPL) thicknesses measured by an automatic segmentation versus manual segmentation in a representative case. (A) A fundus photograph showing an ellipse in red, within which the ganglion cell analysis automatically calculated the GCL+IPL thickness; the points wherein these thicknesses were manually measured are indicated as yellow open circles. Values displayed along with the yellow open circles indicate average values of GCL+IPL thickness that were manually measured in horizontal and vertical sections at each point. (B) The GCL+IPL average sector map automatically measured by ganglion cell analysis. These values are averaged measurements of GCL+IPL thickness in each sector. (C) Vertical and (D) horizontal B-scans, respectively, at the upper middle point in (A). Yellow dashes on the blue (C) and pink (D) lines, corresponding to A-scans at the upper middle point in (A), are the manually measured GCL+IPL thicknesses. The number displayed above the dashes is the GCL+IPL thickness. Notably, the manually measured values in each sector were similar to the automeasured average values in the corresponding sector.
Figure 9
 
Comparison of the ganglion cell layer (GCL) and inner plexiform layer (IPL) thicknesses measured by an automatic segmentation versus manual segmentation in a representative case. (A) A fundus photograph showing an ellipse in red, within which the ganglion cell analysis automatically calculated the GCL+IPL thickness; the points wherein these thicknesses were manually measured are indicated as yellow open circles. Values displayed along with the yellow open circles indicate average values of GCL+IPL thickness that were manually measured in horizontal and vertical sections at each point. (B) The GCL+IPL average sector map automatically measured by ganglion cell analysis. These values are averaged measurements of GCL+IPL thickness in each sector. (C) Vertical and (D) horizontal B-scans, respectively, at the upper middle point in (A). Yellow dashes on the blue (C) and pink (D) lines, corresponding to A-scans at the upper middle point in (A), are the manually measured GCL+IPL thicknesses. The number displayed above the dashes is the GCL+IPL thickness. Notably, the manually measured values in each sector were similar to the automeasured average values in the corresponding sector.
Figure 10
 
Four representative cases of manually measured pixel-based GCL+IPL thicknesses in a set of B-scans of the same eyes [(A, B) case 3; (C, D) case 5; (E, F) case 7; (G, H) case 16, respectively]. (A, C, E, G) Grayscale en face slab images with two blue lines traversing both hyper- and hyporeflective nerve fiber bundles. (B, D, F, H) The vertical sections at purple lines in (A, C, E, G), respectively. The GCL+IPL thickness was manually measured at the intersection points of the blue and purple lines in (A, C, E, G), and the values are indicated in yellow dashes in (B, D, F, H), respectively. As the two blue lines are parallel and at the same distance from the horizontal midline of the image, the eccentricity of the two measured points in each image is the same. Notably, the hyperreflective area on the en face slab image contains hyperreflective RNFL in the B-scans, whereas the hyporeflective area does not. However, the pixel-based thickness of the GCL+IPL on the same B-scans is remarkably similar in the same eyes.
Figure 10
 
Four representative cases of manually measured pixel-based GCL+IPL thicknesses in a set of B-scans of the same eyes [(A, B) case 3; (C, D) case 5; (E, F) case 7; (G, H) case 16, respectively]. (A, C, E, G) Grayscale en face slab images with two blue lines traversing both hyper- and hyporeflective nerve fiber bundles. (B, D, F, H) The vertical sections at purple lines in (A, C, E, G), respectively. The GCL+IPL thickness was manually measured at the intersection points of the blue and purple lines in (A, C, E, G), and the values are indicated in yellow dashes in (B, D, F, H), respectively. As the two blue lines are parallel and at the same distance from the horizontal midline of the image, the eccentricity of the two measured points in each image is the same. Notably, the hyperreflective area on the en face slab image contains hyperreflective RNFL in the B-scans, whereas the hyporeflective area does not. However, the pixel-based thickness of the GCL+IPL on the same B-scans is remarkably similar in the same eyes.
Table
 
Demographic and Ophthalmic Data of the Participants
Table
 
Demographic and Ophthalmic Data of the Participants
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×