Analysis of Retinal Nonperfusion Using Depth-Integrated Optical Coherence Tomography Images in Eyes With Branch Retinal Vein Occlusion

Abstract

Purpose.: To assess the morphology of areas of complete retinal nonperfusion in eyes with branch retinal vein occlusion (BRVO) by en face images of optical coherence tomography (OCT).

Methods.: Forty-six eyes with BRVO that underwent swept-source OCT (SS-OCT) and fluorescein angiography were enrolled. Depth-integrated images of the neural retina delineated by automated segmentation algorithm were obtained using SS-OCT. The findings in a 6 × 6-mm area on en face SS-OCT scans at the area of retinal nonperfusion were evaluated.

Results.: Retinal nonperfusion was detected in 25 eyes. Of these, 20 (80%) eyes had multiple concaves of low reflectivity within an area of reticular high reflectivity (honeycomb sign) on depth-integrated images at the area corresponding to retinal nonperfusion. The mean area of retinal nonperfusion and honeycomb sign were 6.72 ± 4.10 mm² and 4.19 ± 3.37 mm², respectively. The area of retinal nonperfusion was correlated significantly (r = 0.53, P < 0.001) with the area of the honeycomb sign. The mean retinal thickness in eyes with a honeycomb sign increased significantly (P = 0.017) compared with eyes without a honeycomb sign. Furthermore, after anti-VEGF injection, the mean area of honeycomb sign decreased significantly (P = 0.0013), from 4.23 mm² to 0.48 mm².

Conclusions.: Depth-integrated OCT images with automated layer segmentation showed a two-dimensional honeycomb-like structure in retinal nonperfusion with retinal edema. In eyes with BRVO, the thickness of retina did not decrease but increased due to retinal cyst in spite of neural retinal thinning.

Introduction

Branch retinal vein occlusion (BRVO), one of the most common retinal vascular diseases after diabetic retinopathy, is characterized by retinal ischemia in its pathogenesis. The most common cause of decreased vision in BRVO is macular edema.^1–4 Thrombosis of a proximal branch of the central retinal vein results in engorged veins accompanied by variable amounts of hemorrhage, edema, and retinal nonperfusion. Anti-VEGF therapy using ranibizumab (Lucentis, Genentech, Inc., South San Francisco, CA, USA), an antibody fragment that specifically binds all isoforms of VEGF-A, is effective for treating macular edema due to BRVO.^5,6 

Successful reduction of macular edema, retinal hemorrhages, and areas of retinal nonperfusion^7,8 by anti-VEGF therapy also has strengthened the hypothesis that VEGF contributes to the pathogenesis in eyes with BRVO. Retinal ischemia, which triggers VEGF expression, is characterized by retinal nonperfusion, and it is usually detected on fluorescein angiography (FA) images. Fluorescein angiography has obvious value for detecting both the location and the degree of retinal nonperfusion; however, detailed configurations of retinal nonperfusion have not been visualized. 

Swept-source optical coherence tomography (SS-OCT) with a longer wavelength (1050 nm) offers superior visualization of the microstructures beneath retinal hemorrhages or RPE compared with spectral-domain OCT (SD-OCT).⁹ Moreover, SS-OCT provides 100,000 A-scans per second compared with 18,000 to 40,000 A-scans per second for SD-OCT; therefore, it can create three-dimensional images in a shorter measurement time. Recently developed software, EnView (Topcon Corp., Tokyo, Japan), can provide en face images at arbitrary depths and offers depth-integrated images and acquisition for simultaneous multilayer delineation in the retina and choroid. EnView can create depth-integrated images from three-dimensional SS-ophthalmoscope (SLO) images by simply integrating each reflectivity depth profile (namely, A-scans). This map is similar to scanning laser ophthalmoscope images.^10–12 However, EnView also can create depth-integrated images of any layer between two boundaries obtained by an automated segmentation algorithm. 

Although retinal nonperfusion can be detected by retinal thinning using OCT,^13,14 detailed morphologic analysis of the retinal nonperfusion using OCT has not been reported. We hypothesized that retinal nonperfusion would result in a low-intensity signal on the depth-integrated images. Depth-integrated images also are advantageous for analyzing the retinal nonperfusion two-dimensionally in conjunction with FA. In the current study, to elucidate the pathology of retinal nonperfusion in BRVO, we examined the inner configuration of areas of retinal nonperfusion using SS-OCT depth-integrated images. 

Methods

We retrospectively studied 46 eyes of consecutive 45 patients (23 men, 22 women) with BRVO between April and December 2013 at one academic institution. The institutional review board approved the study and all patients provided informed consent. The inclusion criteria included eyes with BRVO for which SS-OCT (DRI OCT-1; Topcon Corp., Tokyo, Japan) and high-speed FA with a confocal laser scanning system (HRA-2; Heidelberg Engineering, Heidelberg, Germany) obtained images of sufficient quality and a duration of symptoms less than 24 months. Branch retinal vein occlusion was diagnosed based on a comprehensive ophthalmologic examination that included indirect ophthalmoscopy, slit-lamp biomicroscopy with a contact lens, FA, and SD-OCT (Cirrus HD-OCT; Carl Zeiss Meditec; Jena, Thuringia, Germany). The best-corrected visual acuity (BCVA) using a Landolt chart and IOP were measured at every visit. Eyes were excluded if they had undergone a previous vitrectomy or laser photocoagulation within the arcade or received an intravitreal injection of an anti-VEGF agent within 3 months before SS-OCT and FA were performed. Eyes with a coexisting ocular disease (i.e., epiretinal membrane, glaucoma, or proliferative diabetic retinopathy) were not recruited for this study. 

Fluorescein Angiography

The configurations of the retinal vessels and hypofluorescent areas were evaluated in the early phase (about 30 seconds after intravenous injection of fluorescein). The FA images were digitized using the IMAGEnet R4 system (Topcon Corp.). The foveal avascular zone (the 0.5-disc diameter area) centered on the fovea was not considered an area of retinal nonperfusion. Retinal nonperfusion assessed on FA images were those portions of the fundus devoid of retinal arterioles or capillaries, and they were detected by angiographic characteristics, such as a continuous area of hypofluorescence due to retinal capillary loss and precapillary arteriole and postcapillary venule obstruction.¹⁵ We distinguished the retinal nonperfusion from blocked fluorescein resulting from retinal hemorrhages using fundus photographs obtained at the same time. Measurement of posterior retinal nonperfusion was based on a previous report⁸ and one author (SS) measured the FAs in a masked fashion. 

Optical Coherence Tomography

The SS-OCT examination had a scanning speed of 100,000 A-scans per second. The center wavelength of the probe beam is 1050 nm, and the axial resolution of this system is 8 μm in tissue. The retinas were analyzed using 6 × 6-mm images with densities of up to 256 (horizontal) × 256 (vertical) B-scan lines around the macula. The scans were obtained using the automated averaging system. The averaging was set preliminarily and the slices were averaged automatically. We selected images with an image quality over 60. The central foveal thickness and mean retinal thickness were measured using DRI OCT-1 software version 9.01. Mean retinal thickness was calculated from the thickness throughout the 6 × 6-mm area. The presence of a serous retinal detachment and cystoid macular edema was evaluated by a B-scan image line throughout the 6 × 6-mm area. In fact, EnView can recognize distinct layers in retina, such as inner limiting membrane (ILM), ganglion cell layer, inner plexiform layer, inner segment/outer segment (IS/OS), RPE, and Bruch's membrane. We extracted depth-integrated images using the EnView software by selecting the analysis of the layer between ILM and IS/OS line in accordance with the automated layer segmentation system. The reflectivity of the A-scans within the automated layers was integrated. Furthermore, based on the integrated reflectivity, two-dimensional images were reconstructed. Because we wanted to exclude the effect of subretinal fluid on the depth-integrated image, we chose the IS/OS layer but not the RPE. The RPE also was excluded because the depth-integrated images of the layer between the ILM and RPE showed low contrast due to the excessive reflectivity of the RPE. 

Measurement on FA and OCT Images

We manually merged the shadowgram images obtained by SS-OCT (Supplementary Fig. S1) and FA images using the vessel pattern as a standard using Photoshop Elements software (Adobe Systems, Inc., San Jose, CA, USA). We then determined the 6 × 6-mm area for examination in the FA images. The area of retinal nonperfusion in the FA images and findings in the depth-integrated SS-OCT images were measured after one author (SS) manually merged the FA and OCT images using the Photoshop Elements software using ImageJ (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA) software in a masked fashion. The area calculated was computed initially in pixels and converted to square millimeters. Evaluation of the maximal or minimum thickness at any point outside of the fovea in the affected lesion was performed using the data from the heatmap mode of the EnView software, which shows the automatically calculated and exported data thickness at any point. 

Statistical Analysis

Statistical evaluations were performed using JMP software version 8.0 (SAS, Inc., Cary, NC, USA). The Student's t-test or Mann-Whitney test was used to compare the factors between two groups. The correlation between the area of retinal nonperfusion and OCT findings was assessed using Spearman's correlation coefficient; P less than 0.05 was considered significant. 

Results

Patient Characteristics

Forty-six eyes of 45 patients who underwent FA and SS-OCT were enrolled initially in the current study. Because retinal hemorrhages, which often complicate identification of retinal nonperfusion, begin to resolve several months after onset in BRVO,² we excluded six eyes because subjective symptoms were present for less than 6 months. In this series, we defined retinal nonperfusion as that exceeding 1 mm² within the corresponding 6 × 6-mm areas examined using SS-OCT. After evaluating eyes for retinal nonperfusion in this series, 15 eyes were excluded that had an area of retinal nonperfusion that was smaller than 1 mm² in the 6 × 6-mm area. Twenty-five eyes of 24 patients ultimately were included in the current study. 

The mean age of the 24 patients at the time of the examination was 67.2 ± 8.3 years (range, 49–78). The mean BCVA at the initial visit was 0.43 ± 0.25 (range, 0.04–0.9) using the Landolt C acuity chart, and 0.47 ± 0.36 (range, 0.05–1.40) in logarithm of the minimum angle of resolution (logMAR) equivalent. The total number of bevacizumab (Avastin; Genentech, Inc.) or ranibizumab injections before the examination was 1.96 ± 1.57 (range, 0–6). The mean duration of symptoms from the initial symptoms to the time FA and SS-OCT were performed was 12.8 ± 5.6 months (range, 6–22). The mean BCVA at the time FA and SS-OCT were performed was 0.57 ± 0.30 (range, 0.05–1.2) using the Landolt C acuity chart and 0.32 ± 0.31 (range, −0.08 to 1.30) in logMAR equivalent. 

Retinal Nonperfusion and SS-OCT Findings

The mean area of retinal nonperfusion was 6.72 ± 4.10 mm² (range, 1.16–18.0; median, 6.74; interquartile range, 3.00–9.39). The mean central foveal thickness was 428.2 ± 155.4 μm (range, 201–1023); the mean retinal thickness was 329.9 ± 37.2 μm (range, 256–417) when FA and SS-OCT were performed. The maximal thickness at any point outside of the fovea in the affected lesion was 560 ± 133 μm (range, 257–788). A serous retinal detachment was found in 5 (20%) of 25 eyes, and foveal cystoid spaces were found in 12 (48%) of 25 eyes. 

Using depth-integrated images on EnView, we found images that contained multiple concave areas of low reflectivity within areas of high reflectivity distributed in a reticular pattern and referred to them as honeycomb-like changes in the affected areas (Fig. 1). We defined this honeycomb sign as “the cluster of relatively low signal oval or polygonal lesion which are completely surrounded by the clear septa.” In the current study, in addition to retinal nonperfusion, we defined eyes as having a honeycomb sign when areas with a honeycomb sign exceeded 1 mm² in the 6 × 6-mm area. The mean area of the honeycomb sign was 4.19 ± 3.37 mm² (range, 0.00–10.8; median, 4.03; interquartile range, 1.30–6.31). The locations of the retinal nonperfusion and honeycomb sign overlapped and those areas were correlated significantly (r = 0.53, P < 0.001) (Fig. 2). Twenty (80%) eyes had a honeycomb sign and five (20%) did not. 

Figure 1

 

Swept-source optical coherence tomography findings 6 months after onset of BRVO in a 72-year-old woman. (A) Fluorescein angiography in the early phase. This area corresponds to a 6 × 6-mm examined area using SS-OCT. (B) The depth-integrated reflectance map from the SS-OCT images of the layer between the ILM and the IS/OS line. The image shows multiple concave areas of low reflectivity within an area of normal reflectivity distributed in a reticular pattern, the so-called honeycomb-like change. (C) A magnified image indicated by the square in (A). The arrows indicate fluorescein pooling. (D) High magnification indicated by the white square in (B). The multiple concave areas (arrows) correspond to the cluster of hyperfluorescence indicated by the arrows in (C). (E, F) The horizontal image of OCT indicated by white arrow in (B). (F) A magnified image indicated by the square in (E). Each cyst in the B-scan image corresponds to each concave area in (D), which can be traced by the white dotted lines in (D) and (F). Retinal cysts consist of a honeycomb sign in the depth-integrated image from the SS-OCT image.

Figure 2

 

Correlation between the area of retinal nonperfusion on FA images and the area of the honeycomb sign in the integrated reflectance map from SS-OCT.

We then focused on the differences in several parameters between the eyes with and without a honeycomb sign (Supplementary Fig. S2; Table 1). Among these, the mean retinal thickness differed significantly (P = 0.017) between the groups: 308 μm in the group without a honeycomb sign and 349 ± 29 μm in the group with a honeycomb sign (Table 1). The maximal thickness in the affected lesions also differed significantly (P < 0.05) between the groups: 453 ± 175 μm in the group without a honeycomb sign and 587 ± 94 μm in the group with a honeycomb sign (Table 1). The mean areas of retinal nonperfusion were 6.08 ± 3.63 mm² (range, 1.16–10.9) in the group without a honeycomb sign and 6.87 ± 4.28 mm² (range, 1.84–18.0) in the group with a honeycomb sign (P = 0.706). When we merged the 6 × 6-mm FA images and depth-integrated images and measured the overlapping area of retinal nonperfusion and the honeycomb changes using ImageJ software, the mean areas of overlap of the retinal nonperfusion and honeycomb sign were 0.0 ± 0.0 mm² in the group without a honeycomb sign and 3.84 ± 2.78 (range, 0.61–9.05) mm² in the group with a honeycomb sign. 

Table 1

 

Comparisons of Characteristics Among Eyes Classified by OCT Findings

	Honeycomb Sign (−)	Honeycomb Sign (+)	PValue
No. eyes (%)	5 (20.0)	20 (80.0)
Age, y, mean ± SD (range)	65.2 ± 7.8 (52 to 72)	67.7 ± 8.5 (49 to 78)	0.567
Sex, n (%)
Male	3 (60.0)	9 (45.0)	0.558
Female	2 (40.0)	11 (55.0)
Duration of symptoms, mo, mean ± SD (range)	13.4 ± 5.6 (7 to 22)	12.6 ± 5.8 (6 to 22)	0.772
Initial visual acuity, logMAR, mean ± SD (range)	0.40 ± 0.27 (0.22 to 0.82)	0.48 ± 0.38 (0.05 to 1.40)	0.659
Visual acuity at the examination, logMAR, mean ± SD (range)	0.27 ± 0.22 (−0.08 to 0.70)	0.36 ± 0.34 (−0.08 to 1.30)	0.327
Central foveal thickness, μm, mean ± SD (range)	358 ± 145 (201 to 520)	454 ± 126 (287 to 799)	0.154
Mean retinal thickness, μm, mean ± SD (range)	308 ± 41 (256 to 349)	349 ± 29 (297 to 409)	0.017
Maximal thickness at any point outside of the fovea, μm, mean ± SD (range)	453 ± 175 (257 to 663)	587 ± 94 (424 to 788)	0.026
No. of anti-VEGF therapy, mean ± SD (range)	2.4 ± 2.3 (0 to 6)	1.9 ± 1.4 (0 to 5)	0.683
Mean interval from last anti-VEGF injection, mo, mean ± SD (range)	5.3 ± 1.0 (4 to 6)	4.9 ± 2.6 (3 to 12)	0.652
Serous retinal detachment at the examination, n (%)	0 (0.0)	5 (25.0)	0.396
Foveal cystoid spaces at the examination, n (%)	2 (40.0)	10 (50.0)	0.695
Mean area of retinal nonperfusion, mm2, mean ± SD (range)	6.08 ± 3.63 (1.16 to 10.9)	6.87 ± 4.28 (1.84 to 18.0)	0.706
Mean area of honeycomb sign in depth integral map, mm2, mean ± SD (range)	0.09 ± 0.20 (0 to 0.45)	5.21 ± 2.97 (1.17 to 10.8)	<0.0001
Mean area of overlapping both RNP and honeycomb sign, mm2, mean ± SD (range)	0	3.84 ± 2.78 (0.61 to 9.05)	<0.0001

 RNP, retinal nonperfusion.

Effect of an Anti-VEGF Injection on the Honeycomb Sign

Retinal nonperfusion was closely related to the honeycomb sign in eyes with BRVO. However, because the retinal thickness changed in the two groups, we examined the effect of injection of an anti-VEGF drug, which is well known for causing reduction of retinal edema in BRVO.^5,6 In the group with a honeycomb sign, 10 (50%) eyes received an intravitreal injection of bevacizumab (IVB) or ranibizumab (IVR) just after the examination and were examined after 1 month using SS-OCT. In these eyes, the area of the honeycomb sign was 4.23 ± 2.94 mm² (range, 1.17–10.8) before the IVB or IVR injection and decreased significantly to 0.48 ± 0.51 mm² (range, 0–1.67) (P = 0.0013) (Figs. 3, 4). After the area of the honeycomb sign decreased, low reflectivity was seen on the integral reflectance map in the corresponding area of retinal nonperfusion in eight (80%) eyes (Fig. 3). The thinnest retina (minimum thickness) in the affected area at any point outside of the fovea after IVB or IVR was 192 ± 22 μm thick (range, 161–229). A retina thinner than 200 μm in the affected area at any point outside of the fovea was detected in seven (70%) eyes. The relationship between decrease in the area of the honeycomb sign and improvement of visual acuity was not significant (P = 0.158). 

Figure 3

 

Images of changes in the honeycomb sign after intravitreal injection of an anti-VEGF drug in a 71-year-old man with retinal nonperfusion and a honeycomb sign. (A, B) Depth-integrated reflectance map images of SS-OCT. (A) Before intravitreal injection of an anti-VEGF drug; (B) after injection. A large portion of the honeycomb sign ([A], black arrow) resolved and became an area of low reflectivity ([B], white arrow). The dotted arrows indicate registration of the horizontal scan in (A) for (C) (upper) and (D) (lower) and (B) for (E) (upper) and (F) (lower). The white arrows in (F) indicate retinal thinning. (G, H) Heat maps of retinal thickness obtained by SS-OCT. (G) Before intravitreal injection of an anti-VEGF drug; (H) after injection.

Figure 4

 

The effect of anti-VEGF therapy for the honeycomb sign. The area of honeycomb sign significantly decreased after anti-VEGF drug injection. n = 10. *P = 0.0013.

Discussion

We identified a novel approach for assessing the inner configuration of areas of retinal ischemia in eyes with BRVO; namely, by integrating the reflectivity of the neural retina from the en face SS-OCT images, honeycomb-like changes highly corresponded to retinal nonperfusion. On the B-scan images of the area with a honeycomb pattern, relatively large cysts were seen extending through all retinal layers (Figs. 1, 3). However, after the cysts were treated with anti-VEGF drugs, the honeycomb signs resolved, the retinal edema decreased, and the neural retina thinned (Figs. 3, 4). 

The honeycomb sign consisted of retinal cysts and retinal thinning. A low signal in the honeycomb sign represented retinal cysts that covered most of the thickness between the ILM and IS/OS line. Otherwise, a normal-reflectivity signal from the honeycomb sign, which divided each “nest,” represented the septum of retinal cysts in the B-scan images (Fig. 1). The integrated reflectance map indicated that each reflectivity depth profile of the A-scans, that is high reflectivity in the A-scans (e.g., the RPE, hemorrhages, or retinal nerve fiber layer between the designated layers), resulted in a strong signal. However, low reflectivity in the A-scans, such as extracellular fluid or a thin parenchyma, resulted in a low signal on the integrated reflectance map. Signals in depth-integrated images are decreased if the total amounts of reflectivity in the subject interlayer are low. Compared with the point on the retina without any pathological lesion, retina with thinned parenchyma leads to decreased reflectivity in depth-integrated map. In the current study, we hypothesized that the area of retinal thinning should be uniformly low reflectance on the en face depth-integrated maps because this map indicates each reflectivity depth profile of A-scans, and a low signal strength would be affected by retinal parenchymal thinning. However, retinal nonperfusion demonstrated a reticular, honeycomb-like structure on the en face depth-integrated maps. It originated from low reflectivity due to both extracellular fluid and retinal thinning in the retinal cysts and relatively normal reflectivity due to intercystic septa. Interestingly, of all 46 eyes with BRVO in this series, only three (6.5%) eyes with a honeycomb sign had no retinal nonperfusion and all of them had cystic edema without retinal thinning due to leaky microaneurysms (data not shown) in contrast to 20 (43.5%) eyes with a honeycomb sign with retinal nonperfusion. Honeycomb sign essentially may represent retinal cystic lesion compared with avascular change. Yamamoto et al.¹¹ and Katome et al.¹² demonstrated the utility of SLO in the modified dark-field imaging for quantifying cystoid macular edema. In eyes with CME, the multiple beams of scattered light give rise to shadows that silhouette the cystoid spaces, which then allows the SLO to obtain a clear image of each cystoid space. Drawing methods for OCT and SLO are different; however, honeycomb sign from retinal cysts with normal-thickness retina have unclear boundaries in each honeycomb compared with the one from retinal cysts with thinned retina (data not shown). This unclear boundary originated from normal signal strength without retinal parenchymal thinning. These data suggested that integrated reflectance images should be advantageous for analyzing the inner structures of retinal nonperfusion in eyes with BRVO using en face images. 

Ota et al.¹³ analyzed the thickness of the inner retina in areas of retinal nonperfusion in eyes with BRVO using SD-OCT. They reported that the inner retina in the area of capillary nonperfusion was thinner than that in normal parts of the retina and that the structure of the inner retinal layer was disorganized. However, they reported a thinning and disorganized retina only in a case report in their study. In the current study, the areas of retinal nonperfusion often included retinal cystic edema and the mean retinal thickness increased (Table 1). We quantitatively showed that areas of retinal nonperfusion do not always decrease in thickness but rather increase due to edema. Meanwhile, after intravitreal injection of an anti-VEGF drug, eyes with both retinal nonperfusion and retinal cysts (i.e., with a honeycomb sign on the depth-integrated reflectance map) had decreased retinal thickness in the area that corresponded to the retinal nonperfusion. We speculated that the retinal thickness in areas of retinal nonperfusion might decrease along with resolution of retinal edema due to thinning of the neural parenchyma. Thinning of the neural retina often accompanies a marked decrease in retinal sensitivity.^14–17 Ota et al.¹³ also reported that damage to the outer photoreceptor layer results in a moderate decrease in retinal function and damage to the inner retina due to capillary nonperfusion causes a marked decrease in retinal function, often leading to an absolute scotoma. 

Yu et al.¹⁸ found, in their experimental study of macaques, that the inner retinal layers may be particularly at risk of hypoxic insult because the retinal vasculature supplies them with oxygen, which is relatively sparse compared with the choroidal circulation that supplies most of the outer retina. Thinning of the inner retinal parenchyma should result from neural cell apoptosis due to ischemic changes. However, leakage from residual vessels also formed the retinal cysts in the areas of retinal nonperfusion.¹⁵ The honeycomb sign, which is attributable to fluid accumulation in a thinned retina, might originate from reduced interstitial pressure compared with capillary intraluminal pressure due to neural cell loss. 

Since the initial studies more than 50 years ago, FA remains the gold standard to detect retinal nonperfusion.^19,20 Due to the complexity of this examination and the possibility of drug allergy, there has been increased interest in obtaining angiographic images noninvasively. To detect the lesions in areas of retinal nonperfusion noninvasively using OCT, previous studies have visualized the retinal vasculature by using phase-variance OCT.^21–23 Although Schwartz et al.²³ reported the efficacy of phase-variance OCT angiographic imaging for visualizing the fine structures of the capillary networks, phase-variance OCT is not widely used. Recently, Imai et al.²⁴ reported the efficacy of SS-OCT for detecting retinal nonperfusion. Measuring the area of thinned inner retinal layer, size of retinal nonperfusion in FA and SS-OCT showed strong correlation and colocalization. As Imai et al.²⁴ did not show the timing of OCT examination in the course of BRVO treatment, involvement of retinal edema in retinal nonperfusion has not been clarified. Nevertheless, they show beneficial information about retinal nonperfusion analyzing by en face OCT. 

A limitation of the current study was inclusion of patients with relatively chronic BRVO, because analysis of retinal nonperfusion is difficult before retinal hemorrhages resolve. Another limitation was the retrospective nature of the study. Anti-VEGF therapy during the follow-up period could affect development of the honeycomb sign. However, we excluded eyes that received an injection within 3 months before the examination. Moreover, because anti-VEGF therapy is indispensable for managing BRVO, the current findings contribute to an understanding of the pathophysiology of BRVO in this anti-VEGF era. 

Acknowledgments

Disclosure: S. Sakimoto, None; F. Gomi, None; H. Sakaguchi, None; M. Akiba, Topcon Corporation (E); M. Kamei, None; K. Nishida, None