Diabetic Nonperfused Areas in Macular and Extramacular Regions on Wide-Field Optical Coherence Tomography Angiography

Transcription

1 Retina Diabetic Nonperfused Areas in Macular and Extramacular Regions on Wide-Field Optical Coherence Tomography Angiography Shota Yasukura, Tomoaki Murakami, Kiyoshi Suzuma, Tatsuya Yoshitake, Hideo Nakanishi, Masahiro Fujimoto, Maho Oishi, and Akitaka Tsujikawa Department of Ophthalmology and Visual Sciences, Kyoto University Graduate School of Medicine, Kyoto, Japan Correspondence: Tomoaki Murakami, Department of Ophthalmology and Visual Sciences, Kyoto University Graduate School of Medicine, 54 Shogoin-Kawaracho, Sakyo, Kyoto , Japan; Submitted: June 21, 2018 Accepted: November 10, 2018 Citation: Yasukura S, Murakami T, Suzuma K, et al. Diabetic nonperfused areas in macular and extramacular regions on wide-field optical coherence tomography angiography. Invest Ophthalmol Vis Sci. 2018;59: iovs PURPOSE. To investigate the nonperfused areas (NPAs) in each subfield segmented by large arterioles on wide-field swept-source optical coherence tomography angiography (SS-OCTA) images in diabetic retinopathy. METHODS. We retrospectively reviewed 101 consecutive eyes of 67 patients with severe nonproliferative diabetic retinopathy (NPDR) and proliferative diabetic retinopathy (PDR), for whom mm SS-OCTA images centered on the optic disc were acquired. Both eyes were included in 34 patients. NPAs in the whole retinal layers were measured in each subfield segmented by large arterioles encompassing both the superficial and deep layers. We compared the percentage of NPAs between individual subfields, considering the overlapping of the feeding arterioles. RESULTS. Extramacular areas had higher rates of NPAs than macular areas in the inner ( mm) and outer (3 5.5 mm) rings (P < in both comparisons). The arteriolar arcades contacting the NPAs on the extramacular side were significantly longer than those contacting the NPAs on the macular side (P < 0.001). In particular, the extramacular areas between two arteriolar branches had a higher percentage of NPAs than those between two arterioles. The macular NPAs were greater in eyes with PDR than in those with severe NPDR, whereas there were no differences in the NPAs in the outer ring of extramacular areas. CONCLUSIONS. Wide-field OCTA images delineated that large arterioles residing in both the superficial and deep layers appear to be the perfusion boundaries, and the overlapping perfusion mediated via collateral vessels may affect the likelihood of diabetic NPAs in each subfield. Keywords: diabetic retinopathy, optical coherence tomography angiography, nonperfused areas Among several vascular lesions in diabetic retinopathy (DR), capillary nonperfusion promotes vascular endothelial growth factor (VEGF) expression, which exacerbates angiogenic complications and the breakdown of the blood retinal barrier (BRB). 1 3 In addition, retinal ischemia per se impairs structure and function in the neuroretina. 4 Comparative studies between angiographic modalities and structural optical coherence tomography (OCT) have revealed the disorganized lamellar structure of inner layers in the nonperfused areas (NPAs). 5 7 Despite its clinical relevance, the mechanisms underlying retinal nonperfusion remain poorly understood in DR. The neuroretina is nourished by multiple layers of capillary plexuses from retinal arterioles and choroidal vessels mediated via the retinal pigment epithelium (RPE). Since the retinal structure appears to be slablike, many clinicians approximately define the location of retinal lesions on a two-dimensional map. Wide-field fluorescein angiography (FA) images showed that diabetic NPAs are more frequently seen in the midperipheral regions than in the posterior pole or far periphery, despite the gradual decrease in perfusion pressure from the optic disc to the periphery This finding implicates other anatomic or pathophysiological factors influencing NPA development and progression. Optical coherence tomography angiography (OCTA) separately visualizes several retinal capillary layers in healthy or diabetic eyes Histologic publications documented that the main trunks of the retinal vasculature originate from the optic disc and hierarchically bifurcate toward the periphery, mainly in the ganglion cell layer (GCL). 19,20 They connect to the radial peripapillary capillaries in the nerve fiber layer (NFL) or form nonhierarchical and seamless networks of deep capillary plexuses in the inner and outer borders of the inner nuclear layer (INL) This finding may allow us to speculate that multiple arterioles, for example, two arteriolar arcades and macular arterioles, overlappingly perfuse the deep capillary network in the macula. By contrast, the three-dimensional vascular structure in the extramacular areas remains poorly understood. In this study, three-dimensional OCTA images revealed that large arterioles encompass both superficial and deep capillary layers. It allows us to hypothesize that such vessels correspond to the perfusion boundaries and that the overlap of perfusing arterioles has an impact on the likelihood of NPA development. Copyright 2018 The Authors iovs.arvojournals.org j ISSN: This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

2 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5894 We therefore segmented retinal areas using large arterioles and compared the NPAs between individual subfields on wide-field OCTA images in eyes with severe nonproliferative diabetic retinopathy (NPDR) and proliferative diabetic retinopathy (PDR). 28 METHODS Patients In this retrospective, cross-sectional study, we reviewed 101 consecutive eyes of 67 patients with severe NPDR or PDR based on the international DR severity scale who visited Kyoto University Hospital from April 2017 to December Both eyes were included in 34 patients. For the subset analysis of 67 eyes of 67 patients, we randomly selected a right or left eye from these 34 patients. The eligibility criteria were severe NPDR or PDR based on fundus examination, for which OCTA images of sufficient quality were acquired using a swept-source (SS)-OCTA instrument (Plex Elite 9000; Carl Zeiss Meditec, Inc., Dublin, CA, USA). Exclusion criteria were opaque media, for example, corneal opacity, cataract, and preretinal/vitreous hemorrhage, that interfere with the quality of OCTA images; severe segmentation error due to extensive fibrovascular membrane on OCTA images; other chorioretinal diseases; glaucoma or ocular hypertension; an axial length shorter than 22.0 mm or longer than 26.0 mm; or a signal strength index of 7 or less defined by the default settings of the instrument. All research and measurements were performed in accordance with the tenets of the Declaration of Helsinki and approved by the Kyoto University Graduate School and Faculty of Medicine Ethics Committee. All participants provided written informed consent before inclusion in the study. Swept-Source OCTA After measurement of best-corrected decimal visual acuity (VA) and conversion to the logarithm of the minimum angle of resolution (logmar) VA, a comprehensive ophthalmic examination was performed. In this study, we focused on the differences in the NPAs between the extramacular and macular areas that are segmented by large arterioles as the perfusion boundaries. In order to adjust the perfusion pressure in these areas, SS-OCTA images within the mm square centered on the optic disc were acquired using Plex Elite This imaging device, which operates at 100,000 A-scans/second and is equipped with a swept-source tunable laser with center wavelength between 1040 and 1060 nm, delineates motion contrast signals using optical microangiography (OMAG) algorithm. 30 Sequential B-scans were obtained at a fixed position, followed by the calculation of variations in both intensity and phase information. Side-by-side B-scans allowed for the construction of three-dimensional OCTA images. The mm square was acquired with A-scans and digitally converted to a pixels array for further analyses. We prepared three slab images, that is, the whole retinal layers (from the inner limiting membrane [ILM] to 70 lm above the RPE), the superficial layer (from the ILM to the inner plexiform layer [IPL]), and the deep layer (from the IPL to 110 lm above the RPE) according to the default setting of the manufacturer s software. The IPL was defined as the layer at 70% of the inner thickness between the ILM and the OPL. Quantification In the normal retinas, both the superficial and deep capillary layers appeared to be divided by main trunks in the vascular arcades and extramacular regions (Figs. 1K, 1L). In particular, histologic publications documented a periarterial capillary-free zone along most arterioles at least in the superficial layer. 31,32 Since large arterioles in the extramacular areas encompass the superficial and deep layers, retinal capillaries rarely cross such vessels in the whole retinal layers. We therefore hypothesized that large arterioles corresponded to the boundaries of retinal perfusion. Furthermore, the extramacular regions could be divided into two areas according to the perfusion status, that is, the areas between arteriolar branches perfused only by a single arteriolar trunk and the extramacular areas between two large arterioles nourished by two large arteriolar trunks dually. In contrast, smaller arterioles, for example, macular arterioles and the branches from the arteriolar arcades, are more frequently seen in the macular areas where inner retinal layers are thicker. The deep capillary layer in the macula appeared to be a seamless network and may be overlappingly fed by multiple small arterioles (Figs. 1E, 1M). We therefore hypothesized that the likelihood of NPA development depends on the overlap of perfusing arterioles in each retinal area segmented by large arterioles at least partly. In this study, we calculated and compared the percentage of NPAs in each subfield segmented by the large arterioles and the distance from the optic disc using three steps: (1) the definition of the large arterioles as perfusion boundaries, (2) the determination of the subfields, and (3) the manual measurement of NPAs in the whole retinal layers. Sectional OCT images delineated that arterioles with a large diameter residing in both the superficial and deep layers corresponded to the boundaries of the retinal perfusion (Figs. 1G, 1I). By contrast, arterioles with a small diameter are present only in the superficial layer, and the deep capillary plexuses beneath them may function as collateral vessels (Figs. 1H, 1J). We prepared five en face images, that is, structural OCT in the superficial and deep layers and OCTA in the superficial and whole retinal layers and the deep layer before the removal of projection artifacts from the superficial vessels, under the default settings of the manufacturer s software. 33,34 We removed the projection artifacts according to the modified method described previously. 26 The comparison between en face structural OCT images in the superficial and deep layers allowed us to determine arterioles or venules with a small diameter, which were accompanied by shadow artifacts in the deep layer (arrowheads; Fig. 1D). We then measured the flow signal levels in these vessels in the en face OCTA images in the superficial (Fig. 1B) and deep (Fig. 1C) layers using ImageJ (National Institutes of Health, Bethesda, MD, USA) and adjusted the signal levels in the superficial layer to those of projection artifacts in the deep layer. We applied the Subtraction function of ImageJ to remove projection artifacts in the deep layer (Fig. 1E). Arterioles that were clearly delineated from the optic disc to the 11-mm-diameter circle on both the superficial and deep layers were manually determined and were referred to as large arterioles (arrows; Fig. 1). Generally, vascular arcades are the main trunks around the macula. The retinal thickness is greater in the macula than in the extramacular regions, and arterioles with a large diameter are rarely seen in the macula. We therefore defined the large arterioles nearest to the fovea as arteriolar arcades (Fig. 1L) in this study. The macular areas were circumscribed by two arteriolar arcades in the superotemporal and inferotemporal subfields from the optic disc (Fig. 2H), and areas out of the macular areas were referred to as the extramacular areas in this study (Fig. 2I). Several large arterioles originate from the optic disc and bifurcate hierarchically to the periphery in the extramacular areas. They correspond to the perfusion boundaries and

3 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5895 FIGURE 1. Large arterioles are the perfusion boundaries in a healthy eye. Wide-field OCTA (A C) or structure OCT (D) images ( mm square) centered on the optic disc. (E) The OCTA images in the deep layer after the removal of projection artifacts. (F) The merged image from OCTA of the whole layers and the deep layer. B-scan images with (I, J) and without (G, H) flow signals along the green arrows in (E) depict that large arterioles (white arrow) reside in both the superficial and deep layers, whereas small arterioles (white arrowhead) are limited in the superficial layer. Red lines in (I, J) indicate segmentation lines of the deep layer. The magnified images of the white rectangles in (F) in the extramacular region (K), the vascular arcade (L), and the macula (M). (K, L) Large arterioles (white arrows) in the extramacular regions and the vascular arcades were delineated in both the superficial and deep layers and correspond to the perfusion boundaries. (M) Small arterioles (arrowheads) resided only in the superficial layer and were accompanied by the seamless capillary network in the deep layer just beneath themselves. a, arterioles; v, venule.

4 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5896 FIGURE 2. Quantification of the NPAs in each subfield segmented by large arterioles in a case of PDR. (A E) OCTA images ( mm square) centered on the optic disc. Arrows: arteriolar arcades. (F) The magnified image of the rectangle in (E) delineates the arteriolar arcades with (arrowheads) and without (arrow) contacting the NPAs on the extramacular side. a, arterioles; v, venule. (G) The NPAs and foveal avascular zone were manually selected (red areas). The macular and extramacular areas (blue areas) (H, I) were segmented using arteriolar arcades in the inner ( mm) and outer (3 5.5 mm) rings. The extramacular areas between arteriolar branches or large arterioles (blue areas) (J, K) were defined using these vessels. The percentages of the NPAs in each subfield were quantified.

5 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5897 FIGURE 3. The NPAs in each retinal area segmented by large arterioles in a representative case of PDR. (A E) Wide-field OCTA images ( mm square) centered on the optic disc. Arrows: arteriolar arcades. Arrowheads: large arterioles in the extramacular regions. (F) The magnified image of the rectangle in (E) delineates greater rates of NPAs in the extramacular areas between two arteriolar branches (arrowheads) than in those between two large arterioles. (G I) In en face OCTA images of the whole retinal layers, the extramacular areas had greater NPAs than the macular areas (blue area) (G). The extramacular areas between arteriolar branches (blue areas) (I) had higher percentages of NPAs than those between large arterioles (blue areas) (H). (J) The deep capillary layer appeared to be a seamless network in the macular areas, whereas large arterioles in the vascular arcades and extramacular regions appear to be the perfusion boundaries in the deep capillaries (K, L). Black areas: the macular areas in (H, I, K, L).

6 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5898 TABLE 1. Patients Characteristics Parameter Values Eyes/patients 101/67 Age, median (IQR), y 59 (53 to 69) Sex Men 49 Women 18 Hemoglobin A1c, median (IQR), % 7.3 (6.5 to 8.3) Systemic hypertension Absent 29 Present 38 logmar VA, median (IQR) ( to 0.155) Lens status Phakia 70 eyes Pseudophakia 31 eyes DR severity Severe NPDR 25 eyes PDR 76 eyes Center-involving DME 35 eyes CSF thickness, median (IQR), lm 298 (269 to 334) Previous history of treatment Panretinal photocoagulation Absent 78 eyes Present 23 eyes Vitrectomy Absent 88 eyes Present 13 eyes Intravitreal injection of anti-vegf drugs Absent 95 eyes Present 6 eyes divided extramacular areas into several subfields, which were classified according to the retinal perfusion status, that is, extramacular areas between two large arterioles and those between two arteriolar branches. Arterioles after the first intraretinal bifurcation were defined as the arteriolar branches. Areas circumscribed by the branches were nourished by a single arteriole and defined as extramacular areas between arteriolar branches (Fig. 2J). By contrast, areas outside of these areas were perfused from two main arteriolar trunks and referred to as extramacular areas between large arterioles (Fig. 2K). We manually examined Bruch s membrane opening and the absence of inner retinal layers on structure OCT images to determine the optic disc and foveal center, respectively. After the optic disc (1.5-mm diameter) and foveal avascular zone (1-mm diameter) were excluded, inner and outer rings ( and mm from the center of the optic disc, respectively) were determined. Then, the en face images were divided into six subfields according to the rings and the areas segmented by large arterioles (Fig. 2). Finally, the areas without cordlike retinal vessels were manually determined as the NPAs in the en face OCTA image in the whole retinal capillary layers and their pixels were counted using ImageJ. The NPAs in the superficial and deep capillary layers were also quantified individually. We further compared the macular and extramacular NPAs along the superotemporal and inferotemporal arteriolar arcades between 0.75 and 5.5 mm from the center of the optic disc. The length of arteriolar arcades contacting the NPAs in the macular, extramacular, or both sides were manually measured using the freehand line tool of ImageJ, followed by calculation of the percentage of each length. We further measured the lengths of large arterioles contacting NPAs in the nasal, temporal, or both sides in the extramacular areas. The central subfield (CSF) thickness was measured using spectral-domain (SD)-OCT images (Spectralis; Heidelberg Engineering, Heidelberg, Germany) as described previously. 35 Briefly, three-dimensional images were obtained using the raster scan mode, and the manufacturer s software was used to calculate the mean retinal thickness within the central 1 mm. Center-involved DME was defined according to the previous publication. 36 Statistics The results are expressed as the median (interquartile range [IQR]). Mann-Whitney U test, Wilcoxon signed-rank test, and Kruskal-Wallis test with Bonferroni correction were used for populations with nonnormal distributions or unequal variance. Two independent retinal specialists performed the image processing procedures and measured subjective parameters according to the methods described above. The averages were used in further statistical analyses. The intergrader agreements were evaluated using intraclass correlation coefficient (ICC). Spearman s rank correlation coefficient was calculated for the assessment of statistical correlation. P < 0.05 was considered statistically significant. SPSS version 24.0 was used for statistical analyses (SPSS, Inc., Chicago, IL, USA). RESULTS Large Arterioles as the Boundaries of Retinal Perfusion We excluded 41 eyes in which extensive fibrovascular membrane resulted in severe segmentation error and 10 eyes that met other exclusion criteria among 152 eyes. We investigated the capillary perfusion of the whole retinal layers in retinal subfields segmented by large arterioles in 25 eyes with severe NPDR and 76 eyes with PDR (Figs. 2, 3), whose characteristics are shown in Table 1. We defined the NPAs using en face OCTA images of the whole retinal layers and TABLE 2. The Percentage of NPAs in Each Subfield on OCTA Images in All 101 Eyes With Severe NPDR or PDR Peripapillary Ring Macular Areas, Median (IQR) Extramacular Areas, Median (IQR) P Value Inner ring 0.19% ( ) 2.12% ( ) <0.001 Outer ring 1.19% ( ) 20.84% ( ) <0.001 Both rings 0.90% ( ) 15.80% ( ) <0.001 Extramacular Areas Between Two Arteriolar Branches Extramacular Areas Between Two Large Arterioles P Value Inner ring 1.85% ( ) 1.84% ( ) Outer ring 33.00% ( ) 16.49% ( ) <0.001 Both rings 26.22% ( ) 11.37% ( ) <0.001

7 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5899 FIGURE 4. The association between the NPAs in individual subfields in all 101 eyes with DR. The relationship between the percentages of NPAs in the macular and extramacular areas of the inner (A), outer (B), or both (C) rings. The correlation between NPAs in the individual extramacular areas segmented by arterioles of the inner (D), outer (E), or both (F) rings. compared the rates of NPAs between the macular (ICC ¼ or 0.895) and extramacular (ICC ¼ or 0.961) areas in the inner or outer rings centered on the optic disc (Table 2). The extramacular areas had significantly higher percentages of NPAs than the macular areas in both the inner and outer rings (P < in both comparisons). In a subset analysis using 67 eyes of 67 patients, the statistical results were repeated (Supplementary Tables S1, S2). In 75 eyes that did not receive photocoagulation, vitrectomy, or anti-vegf treatment, the NPAs in individual subfields were similar to those in all 101 eyes (Supplementary Table S3). There was a significant correlation between the NPAs in the macular and extramacular areas of the outer ring but not between those of the inner ring (Figs. 4A, 4B; Supplementary Fig. S1). The percentage of NPAs was higher in the outer ring than in the inner ring of the macular and extramacular areas (P < in both comparisons). We further investigated the NPAs in the superficial and deep layers individually. There was a significant correlation between the NPAs in the superficial and deep layers (Fig. 5). The NPAs contacted the arcade arterioles in 25 eyes, in which the lengths of arterioles with each NPA status were measured. The arterioles with NPAs on the extramacular side (ICC¼0.979) were longer than those with NPAs on the macular side (ICC ¼ 0.997; Fig. 6A). By contrast, there were no differences between the lengths of extramacular large arterioles with NPAs in the nasal (ICC ¼ 0.964) and temporal (ICC ¼ TABLE 3. Comparisons of the Percentage of NPAs in Each Subfield Between Eyes With Severe NPDR and PDR Areas Severe NPDR, 25 Eyes, Median (IQR) PDR, 76 Eyes, Median (IQR) P Value Macular area Inner ring 0.01% ( ) 0.34% ( ) <0.001 Outer ring 0.27% ( ) 1.52% ( ) <0.001 Both rings 0.23% ( ) 1.18% ( ) <0.001 Extramacular area Inner ring 0.09% ( ) 2.33% ( ) Outer ring 14.82% ( ) 23.00% ( ) Both rings 11.11% ( ) 16.76% ( ) Extramacular areas between two arteriolar branches Inner ring 0.06% ( ) 2.54% ( ) Outer ring 25.77% ( ) 34.57% ( ) Both rings 25.91% ( ) 29.80% ( ) Extramacular areas between two large arterioles Inner ring 0.07% ( ) 2.01% ( ) Outer ring 12.81% ( ) 18.42% ( ) Both rings 9.34% ( ) 12.38% ( ) 0.303

8 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5900 FIGURE 5. The correlation between the NPAs in the superficial and deep capillary layers. (A, B) Significant associations in the inner (ICC ¼ and in the superficial and deep layers, respectively) and outer (ICC ¼ and 0.889) rings of the macular areas. (C, D) The relationship in the inner (ICC ¼ and in the superficial and deep layers, respectively) and outer (ICC ¼ and 0.940) rings of the extramacular areas ) sides (P ¼ 0.593; Fig. 6B). The subset analysis using 24 eyes of 24 patients showed similar results (Supplementary Fig. S2). We further focused on the retinal perfusion in each subfield of the extramacular areas (Figs. 2, 3). The extramacular areas between two branches (ICC ¼ 0.937) had higher rates of NPAs than those between two large arterioles (ICC ¼ 0.927) in the outer ring (P < 0.001, Table 2). There was a similar trend for the NPAs of the inner ring (ICC¼0.935 or in the extramacular areas between two branches or large arterioles, respectively; P ¼ 0.011; Table 2). The NPAs in the extramacular areas between branches were correlated to those in the watershed areas between large arterioles in the inner (q ¼ 0.536, P < 0.001) or outer (q ¼ 0.740, P < 0.001) ring (Figs. 4C, 4D). Differences in the Nonperfused Areas Between Severe NPDR and PDR The comparative study revealed that eyes with PDR had higher percentages of NPAs in the macular area than those with severe NPDR (P < and P < in the inner and outer rings; Table 3). Within the extramacular areas, the percentage of NPAs was higher in eyes with PDR in the inner ring (P ¼ 0.012), whereas there were no differences in the NPAs in the outer ring (P ¼ 0.194; Table 3). In the inner ring, eyes with PDR had a greater rate of NPAs in the extramacular areas between the large arterioles or those between the branches than eyes with severe NPDR (P ¼ or P ¼ 0.010, respectively), whereas there were no differences in the outer ring (Table 3). Further statistical analyses showed no association of the NPAs with logmar VA or CSF thickness (Table 4). DISCUSSION In this study, we documented diabetic NPAs in the whole retinal layers in each subfield segmented by large arterioles and the distances from the optic disc on wide-field SS-OCTA images for the first time. Since large arterioles reside in both the

9 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5901 TABLE 4. No Association Between logmar VA and the NPAs in Each Subfield in All 101 Eyes Areas Correlation to logmar VA Correlation to CSF Thickness Macular area Inner ring q ¼ 0.017, P ¼ q ¼ 0.016, P ¼ Outer ring q ¼ 0.001, P ¼ q ¼ 0.100, P ¼ Extramacular area Inner ring q ¼ 0.155, P ¼ q ¼ 0.072, P ¼ Outer ring q ¼ 0.115, P ¼ q ¼ 0.078, P ¼ FIGURE 6. The length of large arterioles contacting the NPAs in 25 eyes with DR. The length of arteriolar arcades (A) or extramacular large arterioles (B) contacting individual NPAs. superficial and deep retinal layers, we speculated that such vessels correspond to the perfusion boundaries in slablike retinas. The extramacular areas had a higher percentage of NPAs than the macular areas, and further analyses revealed that extramacular areas between two branches had a higher percentage of NPAs than those between large arterioles. This result suggests that the extramacular area between two arteriolar branches, which is perfused by a single trunk, is the most vulnerable to diabetes-induced capillary closure. Multiple arterioles overlappingly perfuse the macula, mediated via the deep capillary plexuses as collateral vessels, and may rescue the macular capillaries from diabetic nonperfusion. Ultrawide-field FA images depicted NPAs in the midperipheral areas more frequently than in the posterior pole or the far periphery. 9,10 This result is, to some extent, consistent with a greater percentage of NPAs in the outer ring than in the inner ring on SS-OCTA images observed in this study. In particular, the greatest percentage of NPAs was found in the extramacular areas between arteriolar branches in the outer ring, which may encourage clinicians to seek the earliest capillary nonperfusion in such areas. However, there were no differences in the NPAs within this subfield between severe NPDR and PDR. This finding, to some extent, agreed with the same levels of overall NPAs on ultrawide-field FA images between these DR grades. 9 By contrast, eyes with PDR had significantly greater NPAs within the macular areas than those with severe NPDR, which suggests the pathologic processes in DR progression. Diabetic NPAs develop more frequently in the extramacular areas of the outer ring on wide-field OCTA images. When diabetes randomly induces capillary nonperfusion within the macula, multiple and overlapping feeding arterioles might compensate for retinal perfusion through the deep capillary plexuses as collateral vessels. By contrast, the extramacular areas between arteriolar branches are nourished by a single arteriolar trunk and are vulnerable to diabetes-induced capillary dropout without rescue by overlapping arterioles. The capillary beds in the extramacular areas between large arterioles are perfused by two arteriolar trunks and are moderately damaged. The first change in diabetic NPAs often occurs in the capillaries rather than in the arterioles or venules. 37 In the extramacular areas with an oriented vascular system, that is, the arteriole capillary venule circuit, capillary closure promotes further disturbance of blood flow around itself. Within the macular areas, multiple feeding arterioles may change the orientation of blood flow around the closed capillaries and prevent its propagation into the deep capillary layer. Further longitudinal studies should be planned to determine whether the extramacular areas segmented by large arterioles are clinically feasible for predicting the development of NPAs and subsequent neovascularization. The association between NPAs in individual extramacular areas of the outer ring suggests that the development or promotion of NPAs in these areas was regulated via common mechanisms, for example, systemic factors including hyperglycemia, biochemical pathways, and cytokines, at least in part. 1,38 By contrast, there was no association between NPAs in the macular and extramacular areas of the inner ring. This finding may allow us to speculate that other factors also influenced the pathogenesis of the NPAs in these regions. Macular arterioles with a small diameter might effectively rescue the capillary perfusion in the inner ring of the macular areas at least partly. Additionally, the NPAs were more frequently seen in the outer ring than in the inner ring. This finding might be explained by two pathophysiological factors, the radial peripapillary capillaries and the higher perfusion pressure in the inner ring. Two-dimensional FA images suggested that radial peripapillary capillaries may be associated with the likelihood of the midperipheral NPAs. 8 Threedimensional analyses in histologic or OCTA images revealed that the radial peripapillary capillaries lie above large arterioles in the NFL around the optic disc and might serve as collateral vessels. 22,31,32,39 As a result, peripapillary areas might be overlappingly fed by multiple arterioles and be less vulnerable to diabetic capillary nonperfusion. We compared the NPAs in the superficial layer to those in the deep layer and found a significant association between them in the macular and extramacular areas. Capillaries in both layers were dropped out in most NPAs of the macular and extramacular areas, whereas we sometimes saw the lamellar nonperfusion in either the superficial or deep layer. 7 Since the vertical vessels connect between the capillaries in the superficial and deep layers, the NPAs in these layers might extend simultaneously. 19,20,25 27 Another explanation might be that columnar neurovascular units are degenerated and concomitantly both capillary layers are dropped out. 40 The macular and extramacular retinas share these vertical neurovascular systems and may have almost the same levels of concordance of the superficial and deep NPAs. There are a few remaining questions regarding NPAs in the far periphery and temporal raphe. Ultrawide-field FA images showed that the far periphery had NPAs less frequently than the midperipheral areas in DR. 9,10 In this study, the OCTA device could not delineate the three-dimensional vascular structure in the far periphery, and it is unknown whether the arterioles correspond to the perfusion boundaries in the far periphery. NPAs often develop in the temporal raphe in DR. 41 In healthy eyes, superficial capillaries in the temporal raphe appear to be a watershed, whereas the deep capillaries appear

10 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j 5902 to be a seamless network. Further study should elucidate the multiple steps of NPA development in this area. There are several limitations to this study. Segmentation errors resulted from fibrovascular membrane in PDR or cystoid spaces or highly reflective lesions, for example, hyperreflective foci and hemorrhages, in DME. 33,34 There was the discrepancy between the outer boundaries of the deep and whole retinal slab images according to the default setting of the manufacturer s software. We often saw poor image quality in the peripheral areas in the OCTA images, and further advances in the OCTA technology are necessary for clinical use. According to the automatic segmentation in the manufacturer s software, neovascularization was often delineated in the OCTA images of the superficial layer and the whole retinal layers, which suggests that neovascularization influenced the NPA quantification. The variable blood flow at the boundary of the NPAs might lead to unstable flow signals on OCTA images. Since the en face images were not corrected for axial length, the inner or outer ring was approximately determined. The NPAs were manually and subjectively determined. These factors might affect the accuracy or reproducibility of the NPA measurement. The methods for automatic quantification of the NPA should be established in the future. We removed the projection artifacts in the deep layer according to the specified method, although it remains to be investigated whether other methods can also be applied to define large arterioles. Further comparative study between OCTA and histologic images should reveal the artifacts beneath vessels. 31 This small and retrospective study contained confounding factors, for example, DME, previous history of interventions, and a within-subjects factor of the inclusion of both eyes Future prospective multicenter studies should confirm that the results can be generalized to other populations and other OCTA devices in treatment-naïve cases. In the current study, we documented that large arterioles appeared to be the boundaries of retinal perfusion on widefield SS-OCTA images and demonstrated the different levels of diabetic NPAs in different subfields segmented by large arterioles for the first time. The findings suggest that the overlapping perfusion by multiple arterioles is a novel anatomic factor determining the likelihood of NPA development in DR. Acknowledgments Supported by Grant-in-Aid for Scientific Research of the Japan Society for the Promotion of Science (17K11423, 18K19610). Disclosure: S. Yasukura, None; T. Murakami, None; K. Suzuma, None; T. Yoshitake, None; H. Nakanishi, None; M. Fujimoto, None; M. Oishi, None; A. Tsujikawa, None References 1. Antonetti DA, Klein R, Gardner TW. Diabetic retinopathy. N Engl J Med. 2012;366: Aiello LP, Avery RL, Arrigg PG, et al. Vascular endothelial growth factor in ocular fluid of patients with diabetic retinopathy and other retinal disorders. N Engl J Med. 1994; 331: Murakami T, Frey T, Lin C, Antonetti DA. Protein kinase cbeta phosphorylates occludin regulating tight junction trafficking in vascular endothelial growth factor-induced permeability in vivo. Diabetes. 2012;61: Sim DA, Keane PA, Zarranz-Ventura J, et al. The effects of macular ischemia on visual acuity in diabetic retinopathy. Invest Ophthalmol Vis Sci. 2013;54: Byeon SH, Chu YK, Lee H, Lee SY, Kwon OW. Foveal ganglion cell layer damage in ischemic diabetic maculopathy: correlation of optical coherence tomographic and anatomic changes. Ophthalmology. 2009;116: e8. 6. Lee DH, Kim JT, Jung DW, Joe SG, Yoon YH. The relationship between foveal ischemia and spectral-domain optical coherence tomography findings in ischemic diabetic macular edema. Invest Ophthalmol Vis Sci. 2013;54: Dodo Y, Murakami T, Suzuma K, et al. Diabetic neuroglial changes in the superficial and deep nonperfused areas on optical coherence tomography angiography. Invest Ophthalmol Vis Sci. 2017;58: Shimizu K, Kobayashi Y, Muraoka K. Midperipheral fundus involvement in diabetic retinopathy. Ophthalmology. 1981; 88: Silva PS, Dela Cruz AJ, Ledesma MG, et al. Diabetic retinopathy severity and peripheral lesions are associated with nonperfusion on ultrawide field angiography. Ophthalmology. 2015;122: Fan W, Wang K, Ghasemi Falavarjani K, et al. Distribution of Nonperfusion Area on Ultra-widefield Fluorescein Angiography in Eyes With Diabetic Macular Edema: DAVE Study. Am J Ophthalmol. 2017;180: Jia Y, Tan O, Tokayer J, et al. Split-spectrum amplitudedecorrelation angiography with optical coherence tomography. Opt Express. 2012;20: Choi W, Moult EM, Waheed NK, et al. Ultrahigh-speed, sweptsource optical coherence tomography angiography in nonexudative age-related macular degeneration with geographic atrophy. Ophthalmology. 2015;122: Takase N, Nozaki M, Kato A, Ozeki H, Yoshida M, Ogura Y. Enlargement of foveal avascular zone in diabetic eyes evaluated by en face optical coherence tomography angiography. Retina. 2015;35: Ishibazawa A, Nagaoka T, Takahashi A, et al. Optical coherence tomography angiography in diabetic retinopathy: a prospective pilot study. Am J Ophthalmol. 2015;160:35 44.e Miwa Y, Murakami T, Suzuma K, et al. Relationship between functional and structural changes in diabetic vessels in optical coherence tomography angiography. Sci Rep. 2016;6: Scarinci F, Nesper PL, Fawzi AA. Deep retinal capillary nonperfusion is associated with photoreceptor disruption in diabetic macular ischemia. Am J Ophthalmol. 2016;168: Samara WA, Shahlaee A, Adam MK, et al. Quantification of diabetic macular ischemia using optical coherence tomography angiography and its relationship with visual acuity. Ophthalmology. 2017;124: Al-Sheikh M, Akil H, Pfau M, Sadda SR. Swept-source OCT angiography imaging of the foveal avascular zone and macular capillary network density in diabetic retinopathy. Invest Ophthalmol Vis Sci. 2016;57: Gariano RF, Iruela-Arispe ML, Hendrickson AE. Vascular development in primate retina: comparison of laminar plexus formation in monkey and human. Invest Ophthalmol Vis Sci. 1994;35: Snodderly DM, Weinhaus RS, Choi JC. Neural-vascular relationships in central retina of macaque monkeys (Macaca fascicularis). J Neurosci. 1992;12: Toussaint D, Kuwabara T, Cogan DG. Retinal vascular patterns. II. Human retinal vessels studied in three dimensions. Arch Ophthalmol. 1961;65: Henkind P. Radial peripapillary capillaries of the retina. I. Anatomy: human and comparative. Br J Ophthalmol. 1967; 51: Chan G, Balaratnasingam C, Yu PK, et al. Quantitative morphometry of perifoveal capillary networks in the human retina. Invest Ophthalmol Vis Sci. 2012;53:

11 Macular and Extramacular NPAs on OCTA in DR IOVS j December 2018 j Vol. 59 j No. 15 j Savastano MC, Lumbroso B, Rispoli M. In vivo characterization of retinal vascularization morphology using optical coherence tomography angiography. Retina. 2015;35: Bonnin S, Mane V, Couturier A, et al. New insight into the macular deep vascular plexus imaged by optical coherence tomography angiography. Retina. 2015;35: Park JJ, Soetikno BT, Fawzi AA. Characterization of the middle capillary plexus using optical coherence tomography angiography in healthy and diabetic eyes. Retina. 2016;36: Nesper PL, Fawzi AA. Human parafoveal capillary vascular anatomy and connectivity revealed by optical coherence tomography angiography. Invest Ophthalmol Vis Sci. 2018; 59: Wang RK, Zhang A, Choi WJ, et al. Wide-field optical coherence tomography angiography enabled by two repeated measurements of B-scans. Opt Lett. 2016;41: Wilkinson CP, Ferris FL III, Klein RE, et al. Proposed international clinical diabetic retinopathy and diabetic macular edema disease severity scales. Ophthalmology. 2003;110: Zheng F, Zhang Q, Motulsky EH, et al. Comparison of neovascular lesion area measurements from different sweptsource OCT angiographic scan patterns in age-related macular degeneration. Invest Ophthalmol Vis Sci. 2017;58: Balaratnasingam C, An D, Sakurada Y, et al. Comparisons between histology and optical coherence tomography angiography of the periarterial capillary-free zone. Am J Ophthalmol. 2018;189: An D, Balaratnasingam C, Heisler M, et al. Quantitative comparisons between optical coherence tomography angiography and matched histology in the human eye. Exp Eye Res. 2018;170: Spaide RF, Fujimoto JG, Waheed NK. Image artifacts in optical coherence tomography angiography. Retina. 2015;35: Ghasemi Falavarjani K, Al-Sheikh M, Akil H, Sadda SR. Image artefacts in swept-source optical coherence tomography angiography. Br J Ophthalmol. 2017;101: Murakami T, Ueda-Arakawa N, Nishijima K, et al. Integrative understanding of macular morphologic patterns in diabetic retinopathy based on self-organizing map. Invest Ophthalmol Vis Sci. 2014;55: Chalam KV, Bressler SB, Edwards AR, et al. Retinal thickness in people with diabetes and minimal or no diabetic retinopathy: Heidelberg Spectralis optical coherence tomography. Invest Ophthalmol Vis Sci. 2012;53: Cogan DG, Toussaint D, Kuwabara T. Retinal vascular patterns. IV. Diabetic retinopathy. Arch Ophthalmol. 1961; 66: Brownlee M. Advanced protein glycosylation in diabetes and aging. Annu Rev Med. 1995;46: Yu PK, Balaratnasingam C, Xu J, et al. Label-free density measurements of radial peripapillary capillaries in the human retina. PLoS One. 2015;10:e Gardner TW, Antonetti DA, Barber AJ, LaNoue KF, Levison SW. Diabetic retinopathy: more than meets the eye. Surv Ophthalmol. 2002;47(suppl 2):S253 S Bresnick GH, De Venecia G, Myers FL, Harris JA, Davis MD. Retinal ischemia in diabetic retinopathy. Arch Ophthalmol. 1975;93: Wessel MM, Nair N, Aaker GD, Ehrlich JR, D Amico DJ, Kiss S. Peripheral retinal ischaemia, as evaluated by ultra-widefield fluorescein angiography, is associated with diabetic macular oedema. Br J Ophthalmol. 2012;96: Sabet-Peyman EJ, Heussen FM, Thorne JE, Casparis H, Patel SJ, Do DV. Progression of macular ischemia following intravitreal bevacizumab. Ophthalmic Surg Lasers Imaging. 2009;40: Campochiaro PA, Wykoff CC, Shapiro H, Rubio RG, Ehrlich JS. Neutralization of vascular endothelial growth factor slows progression of retinal nonperfusion in patients with diabetic macular edema. Ophthalmology. 2014;121: Takamura Y, Tomomatsu T, Matsumura T, et al. The effect of photocoagulation in ischemic areas to prevent recurrence of diabetic macular edema after intravitreal bevacizumab injection. Invest Ophthalmol Vis Sci. 2014;55: Armstrong RA. Statistical guidelines for the analysis of data obtained from one or both eyes. Ophthalmic Physiol Opt. 2013;33:7 14.