In the foveal region, interconnecting rings of capillaries form a specified foveal avascular zone (FAZ) in the superficial capillary network (sCN) and deep capillary network (dCN).1 This two-layered retinal microvessel architecture was first derived from histologic studies as dCN and sCN with multiple vertical connections.2,3 It has been shown that fluorescein angiography (FA) does not provide enough information about the sCN and even less for the dCN.3,4
Quantitative measurements of retinal vasculature in the healthy eye other than FA have been reported using several in vitro and in vivo techniques, including confocal microscopy,5 speckle-variance optical coherence tomography (SV-OCT),6–7 adaptive optics OCT,8,9 and swept-source OCT angiography (OCTA).10
With recent advancement in OCT techniques and application of sophisticated algorithms, OCTA has emerged as a noninvasive procedure to produce detailed images of retinal and choroidal vasculature.11,12 By capturing the dynamicity of red blood cell movements and flow in vessels, using an OCT system can yield a cross-sectional illustration of retinal capillaries in customized depth.13,14 Employing OCTA, the two vascular networks in the inner retina were described.4,15,16
With the advent of new imaging techniques, the necessity for the determination of standards and specification lines between normal and abnormal findings developed. In addition, in vivo assessment of normal anatomical structures may substantially help in defining the corresponding pathologic forms in disease states. Ethnic factors, sex, and age can also result in normal variations that should be kept in mind when considering disease processes.
In the present study, we used an automated image thresholding method in OCTA for quantification of foveal and parafoveal vascular density (VD) and blood flow area and assessed their correlations with other measured ocular and demographic factors.
Patients and Methods
This study was conducted in Farabi Eye Hospital (Tehran, Iran) during the period of January 2016 to May 2016. After approval of Tehran University of medical sciences review board, healthy subjects with best-corrected visual acuity of 20/20 or better, spherical equivalence between +1 diopter (D) and −3 D, and intraocular pressure lower than 21 mm Hg were recruited. Informed consent was obtained and complied with the requirements of the Declaration of Helsinki. Media opacity precluding imaging, significant motion, or blinking artifact; previous diagnosis of migraine; and any present or previous history of ocular and systemic disease, lens opacities, or surgery were our exclusion criteria. Any subject who had received any kind of laser treatment (corneal or retinal for any reason) or systemic medical treatments was excluded from the study.
Two trained readers from our team (FG, FB) reviewed and graded all OCT images in each session. The results were the agreed upon values in the same session of reading, and our epidemiologist (SS) performed the analysis. Angiography was done in a 3-mm diameter circle centered on fovea using OCTA (RTVue XR Avanti; Optovue, Fremont, CA) for both eyes of each subject without pharmacologic pupillary dilation. We used the split-spectrum amplitude decorrelation angiography algorithm in this study. This instrument operates at 840-nm wavelength and performs 70,000 A-scans per second to acquire OCTA volumes consisting of two repeated B-scans of 3 mm × 3 mm (304 pixels × 304 pixels) in the transverse dimension (Figure). Two volumetric raster scans, including one horizontal priority (x-fast) and one vertical priority (y-fast), were obtained consecutively, and the merging of two scans removed any motion artifacts. Low-quality scans (ie, due to blinking or motion) were excluded and repeated until good-quality scans were achieved.
Measuring the vascular density and flow surface area by optical coherence tomography angiography with the preset settings of automated segmentation. Comparison of superficial (A) and deep (B) capillary network vascular density in the foveal and parafoveal areas is presented. The superficial capillary network, deep capillary network, and choriocapillaris are obtained through the use of the automated segmentation of the structural B-scan (C). The blood flow area in 1-mm radius of foveal area (3 mm × 3 mm) at superficial (D) and deep capillary network (E) and choriocapillaris (F) was measured with a preset setting.
VD was computed by the preset automated algorithm and is presented as a percent of the area of interest occupied by retinal vasculature in selected depths (Figure). The OCTA commercial software calculated the flow area per mm2 in a 1-mm (3 × 3) and 3-mm (6 × 6) diameter circle in the superficial and deep layers of the macular capillary networks (Figure). These indices were measured in the foveal and parafoveal areas as a whole and selectively in the superior, inferior, nasal, and temporal parafoveal sectors in the macular region. In each eye, the measurements were performed in the sCN (extending 3 μm from the internal limiting membrane [ILM] with an offset of 15 μm) and the dCN (extending from 15 μm from the ILM to 71 μm) layers of retina. In case of erroneous determination by the built-in software, manual correction was performed. The calculations were performed over the foveal and parafoveal regions. The foveal region was outlined as a central circle with a 120-pixel (1.2-mm) diameter, and the parafoveal region was delineated as a ring by 91 pixels wide surrounding the foveal region.17
The FAZ area was measured in mm2. Briefly, a non-flow area measurement was preset on the Optovue instrument. Upon clicking on the center of the FAZ, the software automatically calculated the area areas on the superficial and deep vascular network.16 The user manually fine-tuned the plane to maximize the visualization of the retinal capillary bed.
We categorized the ages of the candidates into four age groups: 30 years or younger, 31 years to 40 years, 41 years to 50 years, and older than 50 years.
All quantitative variables were reported as mean with standard deviation (SD) after confirming normality of distribution with the Kolmogorov-Smirnov test. All statistical analyses were performed using SPSS (Version 21; SPSS, Chicago, IL). Two-tailed, independent samples t-test was used to compare the FAZ areas with other two-tailed variables, such as posterior vitreous detachment (PVD), laterality, and sex. Analysis of variance (ANOVA) was performed among the age groups. Multivariate linear regression analysis was used to determine the correlation between VDs and flow areas after adjusting for confounding factors including age, sex, and other correlated variables in a partial correlation test. In this study, collinearity for different variables was checked. P values less than .05 were considered statistically significant.
A total of 224 eyes of 112 individuals were recruited in this study. Mean age of the subjects was 36.39 years ± 11.31 years (range: 12 years to 67 years). Fifty-two percent (59 cases) of participants were male with a mean age of 33.42 years ± 7.99 years, and 48% (53) were female with a mean age of 40.05 years ± 13.42 years. Among the four groups, 41 cases (36.3%) were in the 30 years or younger group, 37 cases (33%) were in the 31 years to 40 years group, 20 cases (17.9%) were in the 41 years to 50 years, and 14 cases (12.5%) were in the older than 50 years groups.
The mean ± SD values of superficial and deep foveal avascular zone (sFAZ, dFAZ) surface areas were 0.27 μm2 ± 0.11 μm2 and 0.35 mm2 ± 0.12 mm2, respectively (P < .001). Central foveal thickness (CFT) was 247 μm ± 21.1 μm and parafoveal thickness (PFT) was 317 μm ± 15.6 μm. Superficial FAZ and dFAZ were highly correlated (r = 0.82; P < .001). No significant differences were found between both eyes.
Mean whole macular VD of sCN and dCN was measured as 54.22% ± 2.29% and 59.30% ± 2.03%, respectively (Table 1). In the foveal region, superficial network VD was significantly higher than deep network VD (31.1% ± 5.5% vs. 28.3% ± 7.2%; P < .001), whereas in the parafoveal area, the result was the opposite (62.24% ± 2.8% vs. 56.5% ± 2.5%; P < .001). Flow area in a 1-mm radius circle in the sCN was less than in the dCN (1.49 mm2 ± 0.08 mm2 vs. 1.56 mm2 ± 0.09 mm2; P < .001) (Table 2). This was not the case for the 3-mm radius circle. Macular whole VDs of the sCN and dCN were negatively correlated with age (r = −0.34, P < .001 and r = −0.46, P < .001, respectively, by the Pearson correlation) that was proved after adjusting for sex as a main confounder (beta = −0.32, P = .001 and beta = −0.48, P < .001). Macular sCN and dCN blood flow area was also negatively correlated with the age (r = −0.34, P < .001 and r = −0.41, P < .001, respectively) by the Pearson correlation, which was confirmed in linear regression by considering sex as a confounder (beta = −0.28, P = .004 and beta = −0.38, P < .001).
Vascular Density of Superficial and Deep Foveal and Parafoveal Capillary Plexus of Normal Eyes
Blood Flow Area of Superficial and Deep Capillary Plexus and Choriocapillaris in Macula of Normal Eyes
Gender-Based VD and Blood Flow Area Changes in Foveal and Parafoveal Area
Vascular densities of sCN and dCN of the foveal area were higher in males than females (32.8 vs. 29.2%, P < .001; and 30.4 vs. 26%, P = .001, respectively) (Table 1). These differences disappeared in linear regression after adjusting for age (P > .05). In the evaluated flow areas, the mean flow area in the central 1-mm radius of both the sCN and dCN were significantly higher in males (1.49 vs. 1.47%, P = .04 and 1.57 vs. 1.52%, P = .02, respectively) (Table 2). The observed discrepancies did not persist in regression analysis after considering the age of the cases.
Age Grouping-Based VD and Blood Flow Area in Foveal and Parafoveal Area
Results of macular VD assessments in the foveal and parafoveal areas based on age category are summarized in Table 3. Considering the first age group as the reference group, in a post hoc analysis of variance (ANOVA) test, the VD of nearly all parafoveal sectors of dCN decreased after the age of 50 years. In the sCN, a similar decrease is limited to the whole parafovea, the superior hemifield of parafoveal retina, and superior and temporal parafoveal sectors. The foveal VD was not changed over the time.
Vascular Density of Superficial and Deep Foveal and Parafoveal Capillary Plexus of Normal Eyes in the Age Groups Ranging From 12 to 67 Years
Again, considering the first age group as a reference, in the ANOVA, the flow area of sCN and dCN in the central 1-mm radius circle in the macular area was significantly decreased in the second and third age groups (Table 4). This decrease was not seen in the 3-mm radius circle.
Blood Flow Area of 1-mm and 3-mm Radius Circle of Foveal and Parafoveal of Normal Eyes in the Age Groups Ranging From 12 to 67 Years
Correlation of FAZ and CFT With VD and Blood Flow Area in Foveal and Parafoveal Area
Superficial FAZ (sFAZ): In the age- and sex-matched partial correlation, sFAZ area was positively correlated with dFAZ area (r = 0.82, P < .001), blood flow area in a 1-mm radius of the sCN (r = 0.49, P < .001), flow area in a 1-mm radius of the dCN (r = 0.56, P < .001), and CFT (r= −0.72, P < .001). On the other hand, it was negatively correlated with VD in the sCN in the foveal area (r = −0.918, P < .001), VD in the dCN in the foveal area (r = −0. 85, P < .001), VD in the temporal part of dCN in foveal area (r = −0.21, P = .05). In a multivariate linear regression analysis, after checking collinearity and adjusting for age and sex, only the correlations of sFAZ to dFAZ area (beta = 0.27; 95% CI, 0.12– 0.32; P < .001), VD in sCN of fovea (beta: −0.65; 95% CI, −0.01 to −0.10; P < .001), and gender (beta = 0.11; 95% CI, −0.01 to −0.04; P = .001) persisted.
Deep FAZ (dFAZ): In the age- and sex-considered partial correlation analysis, dFAZ area was directly correlated with sFAZ area (r = 0.82, P < .001) and flow area in the 1-mm sCN (r = 0.48, P < .001) but negatively correlated with flow area in dCN (r = −0. 70, P < .001), VD in sCN of fovea (r = −0.76, P < .001), and CFT (r = −0. 50, P < .001). In the age- and sex-adjusted multivariate regression analysis, only the correlation with sFAZ area (beta = 0.93; 95% CI, 0.73 to 1.57; P < .001) and CFT (beta = 0.26; 95% CI, 0.0 to 0.003; P = .008) remained.
Central Foveal Thickness: In the age- and sex-matched partial correlation analysis, CFT was positively correlated with sFAZ area (r = 0.72, P < .001), dFAZ (r = 0.50, P < .001), 1-mm flow area in the sCN (r = 0.46, P < .001), and 1-mm flow area in the dCN (r = 0.41, P < .001). In multivariate regression analysis, CFT had only a negative correlation with the sFAZ area (beta: −0.65; 95% CI, −159.41 to −97.43; P < .001), and sex (beta = −0.19; 95% CI, −13.95 to −1.99; P = .01), and a positive relation to age (beta = 0.19; 95% CI, 0.11 to 0.61; P = .006) was spotted.
In this study, the VD and blood flow surface area of the sCN and dCN in fovea and parafoveal area were similar in both eyes and sexes, on OCTA images. VD of foveal sCN was significantly higher than dCN, though in the parafoveal area the result was opposite. Flow area in the 1-mm radius circle in the sCN was less than in the dCN. Superficial FAZ size was negatively correlated with VD of the foveal sCN, but the dFAZ and CFT were not correlated with whole VD or blood flow area of the fovea. In this study, dFAZ was influenced by sFAZ and CFT.
Different Reports of VD
Mammo et al.7 used SV-OCT in the specifying the foveal microvessels in a 2 mm × 2 mm region in six healthy subjects with a mean age of 34 years and compared it with FA and histologic pattern, finding that the mean capillary density in FA images was 19.3% and in SV-OCT images was 31.2% of total tissue area. Their study showed the ability of SV-OCT to delineate and segment the different capillary layers within healthy subjects with good accordance with histologic cuts.
Using swept-source ultra-high-speed frequency-domain OCT optical microangiography of the regions of interest (ROI) at 250 μm and 500 μm, Kuehlewein et al. found no difference between values of superficial and deep vascular layers (superficial: 67.3% and deep: 34.5% for 250 μm ROI and superficial: 74.2% and deep: 72.3% for ROI 500 μm).10 Matsunaga et al., using the same OCT system, also found similar VD values between the inner and middle retina of the central macula.18
In a qualitative study of 52 healthy eyes using the OptoVue OCTA system, Savastano et al. described a denser and more complex vascular pattern in the deep compared to the sCN.16 In their OCTA study of sickle cell disease and control eyes, Minvielle et al. reported a parafoveal VD of 58% in sCN and 54% in dCN in nine normal control eyes in black individuals with a mean age of 38 years.19 Our result showed higher VD in the dCN compared with the sCN (62.3% vs. 56.5%).
Using OCTA on 163 healthy eyes of 122 subjects, Shahlaee et al. showed that in the foveal area, VD was 31.9% in the sCN and 27.5% in the dCN, similar to our results of 31.1% and 28.3%, respectively.17 However, they reported 46.0% in the sCN and 51.5% in dCN in the parafoveal area, which was much lower than our results (Table 1). They did not address the ethnicity of their study group. Comparable with their results, we found that VD is greater in the parafoveal area in the dCN, whereas it is greater in the foveal area in the sCN. This could be due to different layering of sCN and dCN in foveal and parafoveal areas, presumably due to local metabolic needs in the retina; however, this remains to be elucidated.
Wang et al.20 assessed VD of the retina and choriocapillaris in 105 healthy subjects with a mean age of 36 years using a similar method to ours. They measured mean density of sCN and dCN as 45.8% and 44.4%, respectively, compared to our results of 54.2% and 59.3%, respectively, in the whole network. In contrast to our study, Wang's study showed that higher sCN density was correlated with male sex, and higher dCN density was correlated with female sex and younger age; none of these findings were observed in our multivariate analysis. However, our results were similar to theirs in correlation of sCN density with dCN. Higher vascular density in the dCN compared with the sCN is reported with more recent studies.21,22
The dCN comprises the homogenous capillary vortex,23 horizontal and monolayer, and lobular pattern, converging toward an epicenter with more layer-based concentration, whereas the sCN is formed by transverse capillaries21 and more obliquely traversing vessels, which is shown less in transverse cuts in OCTA images. Savastano et al. have also noted that the pattern of the two capillary beds were different and that the deep network presented as an interconnected mesh of dense vessels.16,24
According to our findings, both the sCN and dCN present some decreases in density with distance from the disc along the maculopapillary axis.33 This concurs with previous histologic evidence reported by Snodderly et al.1
According to the effect of aging on the decreasing VD, Shahlaee et al.17 observed a decrease of the signal strength index of obtained scans with increased age (r = −0.353; P < .001). Yu et al.,25 in their study of 76 eyes from 45 normal Chinese subjects using the Optovue OCTA, reported the same negative correlation between the age and both the flow index and VD of the macular area, as well as increased FAZ. They showed average annual reductions of 0.6% and 0.4% in the flow index and VD, respectively. Burgansky-Eliash et al.26 reported the same negative correlation between age and blood flow velocity in venules of subjects aged older than 40 years. Iafe et al.22 found a negative correlation between age and vessel density, but with a smaller rate of change in VD mainly in the dCN, occurring in subjects between the ages of 40 and 49 years, as well as those between the ages of 50 and 59 years (P = .033).
Kimura et al.27 found no correlation between the blood flow and age in subjects aged younger than 42 years by scanning laser doppler flowmeter. Hence, it was concluded that the reduction in macular perfusion might be more significant in older subjects, particularly those aged older than 35 to 40 years.26,27 In our study, although the VD of most segments of parafoveal area and foveal areas decreased with increased age, this was statistically significant by post hoc test (ANOVA-Dunnett test) only in the group of patients older than 50 years. Regarding our study, it can be concluded that for each year of increase in age, VD of the sCN in the macula decreases by 0.32%, and by 0.48% (referred to the multivariate regression analysis) in the dCN. By the same analysis, foveal VD was stable during the time by central 1-mm and 3 mm × 3 mm scans.
The flow area of the sCN and dCN in the central 1-mm radius decreased by 0.28 mm2 and 0.38 mm2, respectively, for every year of increase in age. In the present study, considering the first age group as a reference, in the post hoc ANOVA test, the flow area of the sCN and dCN in the central 1-mm radius circle in the macular area was significantly decreased in the second and third age groups (Table 4). This decrease was not seen in the 3-mm radius circle. The only flow index and VD studies already reported used skeletonized images of OCTA (ImageJ protocol) in open-angle glaucoma or diabetic patients, comparing patient-related data with a small sample of age-matched healthy controls.28,29 We also noticed that the mild negative correlation of foveal and parafoveal VD values with age was in accordance with the increased FAZ and decreased parafoveal flow. Yu et al. and Shahlaee et al. documented the same observation.17,25
Our results are in line with other studies,17,25,30 showing no relation between VD or flow area and sex.
The current investigation is the first study that has evaluated the effect of VD and flow area on FAZ size. Based on our findings by multivariate analysis, sFAZ size is negatively correlated with VD of the sCN and directly affected by dFAZ size in whole foveal area.
It means that in foveal area, increasing the VD of sCN is associated with a decrease in sFAZ size and increased VD in the dCN causes greater dFAZ. This could be related to homogenous capillary vortex nature of vessels in the dCN versus transverse capillaries in sCN. These correlations were not observed in parafoveal segments. The mechanical and physiological causes remain to be elucidated. According to our study, as a new finding, the flow surface area does not affect sFAZ and dFAZ size.
No correlation between VD and CFT was found in this study. A partial correlation between flow area of the central 1 mm and 3 mm was observed; however, this disappeared in regression analysis. Our findings were very similar to those of Chui et al, who showed that the CFT decreased significantly as FAZ increased in size.31 CFT as the thinnest part of the macula can be representative of mean retinal thickness of the whole macula. Yu et al. reported a significant positive correlation between the parafoveal VD with the inner retinal thickness (P < .05) (Figure 2A), but not with full retinal thickness, in a group of healthy Chinese subjects.32 Their analyses showed that for each single SD decrease in the inner retinal thickness, the VD decreased by 1.3% to 1.6%.32
Previously, Burgansky-Eliash et al.26 failed to find a correlation between blood flow velocity in retinal vessels and retinal thickness using the Retinal Function Imager (Optical Imaging, Rehovot, Israel). By contrast, Landa et al.30 reported that retinal volume was strongly correlated with the blood flow velocity in retinal vessels. To the best of our knowledge, this study is the first study evaluating central foveal thickness in relation to the vascular density and macular flow area. VD of the sCN impacts the sFAZ size only. Overall, neither VD nor flow area affect the CFT in normal volunteers.
Our study benefits from having a large sample size, wide age range, mono-ethnic subjects, and meticulous exclusion criteria. Thanks to the current version of OptoVue software, with proper imaging technique and capillary resolution, adjustment for signal strength, and incorporation of an automated analysis method and reproducibility, we had clearer images of both the sCN and dCN, as well as quantifications.
We also acknowledge several limitations of this work. This study assessed only relatively young subjects with clear ocular media far out of normal condition. We have not validated our quantified data with another method of VD assessment. The lack of efficient algorithm for elimination of projection artifacts is another downside of present investigation.