Quantitative analysis of choroidal vasculature in polypoidal choroidal vasculopathy using ultra-widefield indocyanine green angiography

Polypoidal choroidal vasculopathy (PCV) is a common choroidal vascular disease particularly in Asians. However, the underlying pathogenesis of PCV is still yet to be fully elucidated, and the correlation between choroidal vasculature and treatment response of PCV are poorly understood. Accordingly, we sought to find clues to understand the pathogenesis and prognosis of PCV by quantitatively evaluating choroidal vasculature from the entire fundus using ultra-widefield (UWF) indocyanine green angiography (ICGA). In this study, 32 eyes from 29 patients with treatment naïve PCV and 30 eyes from 30 healthy control participants were enrolled. Choroidal vascular density (CVD) of PCV eyes was higher than normal eyes in majority regions including the periphery. CVD was positively correlated with choroidal thickness and choroidal hyperpermeability, supporting that the pathogenesis of PCV may include choroidal congestion and dilatation. Thicker choroid and higher CVD were also correlated with poor treatment response after anti-VEGF injections. The CVD, quantified from UWF ICGA can also be used as an effective image biomarker to predict the treatment response in PCV.

Choroidal features of unaffected eyes in unilateral PCV. A total of 22 eyes out of 29 PCV patients were available to be defined as unaffected fellow eyes and to be analyzed. The CVD of the fellow eyes (26.95 ± 1.98% for total; 26.08 ± 2.60 for MR; 25.10 ± 2.10 for NPR; 28.67 ± 2.09 for MPR; 29.14 ± 2.34 for FPR) were significantly higher than control eyes (p = 0.021 for total; p = 0.019 for MP; p = 0.033 for NPR; p = 0.027 for MPR; p = 0.044 for FPR), but did not differ significantly from the PCV eyes (p = 0.638 for total; p = 0.884 for MPR; p = 0.223 for

Correlation between CVD and baseline characteristics in PCV.
Linear correlation analysis did not find any association between the CVD with age, sex, refractory error, and baseline BCVA. However, baseline CMT, choroidal thickness, and choroidal vascular hyperpermeability were positively associated with CVD. Baseline CMT showed statistically significant correlation with the total CVD (R = 0.442, p < 0.001 for total), and even when analyzed by prespecified regions, it also showed a significant correlation with the CVD in each region (R = 0.287, p = 0.028 for MR; R = 0.437, p = 0.001 for NPR, R = 0.437, p = 0.001 for MPR; R = 0.316, p = 0.015 for FPR) ( Table 2).
Comparison of clinical features according to treatment response. Based on the response to three consecutive anti-VEGF injections, 21 eyes were considered as good responders, and 11 eyes were considered as poor responders. There was no statistically significant difference between the two groups in age, sex, baseline BCVA, baseline CMT, and types of anti-VEGF applied to the patients ( Table 3).
The choroid, especially in Haller's layer, was thicker in the poor responders than in good responders (371.60 ± 101.17 µm vs. 291.77 ± 81.90 µm, p < 0.016 for SFCT; 292.90 ± 97.48 µm vs. 219.59 ± 81.22 µm, p = 0.025 for the thickness of the Haller's layer), while the thickness of the choriocapillaris-Sattler's layer showed no statistical significant difference between the two groups (78.70 ± 17.58 µm vs. 72.18 ± 24.76 µm, p = 0.509) ( Table 3). The choroidal vascular hyperpermeability was more frequent in the poor responders than in good responders (60.0% vs. 22.7%, p = 0.040) ( Table 3). CVD was also significantly greater in the poor responders than good responders (28.39 ± 1.45% vs. 26.58 ± 1.77%, p = 0.008). CVD in all prespecified regions, except the MR, showed significant differences between the groups (26.58 ± 2.22% vs. 25 (Table 3). Depending on the results of univariate analysis, choroidal hyperpermeability, SFCT, and CVD were subjected to binary logistic regression analysis and found that the SFCT (p = 0.022) and the CVD (p = 0.007) were independent risk factors for poor response to anti-VEGF therapy (Table 3). Based on the Youden index, the optimal cut-off point of CVD for predicting a poor treatment response was 27.80%, with a sensitivity of 86.36% and specificity of 80.00% (Fig. 1).

Discussion
In the present study, we performed a precise measurement of CVD using stereographically projected UWF ICGA images in eyes with and without PCV. As expected, in the PCV group, the CVD increased significantly in a majority of regions, including the periphery. The CVD was positively correlated with the presence of choroidal vascular hyperpermeability and SFCT, especially in Haller's layer. The eyes with poor treatment response after three Table 2. Choroidal vascular density and associated factors. a Pearson correlation. BCVA, best-corrected visual acuity; CMT, central macular thickness; FPR, far-peripheral region MPR, mid-peripheral region; MR, macular region; NPR, near-peripheral region; SFCT, subfoveal choroidal thickness. www.nature.com/scientificreports/  www.nature.com/scientificreports/ monthly consecutive anti-VEGF injections presented relatively higher CVD than those with good response, and multivariate logistic regression analysis also showed that CVD was related to the response to anti-VEGF for PCV. Several previous studies have revealed choroidal characteristics of PCV, including choroidal thickening, choroidal hyperpermeability, and increased choroidal vessel engorgement 5,6,[10][11][12][13][14]22,[25][26][27][28] . Consistent with the previous studies, we also found the thicker choroid and higher CVD of the PCV eyes, which showed a significant positive correlation with each other. Since ICGA can depict medium-and large-sized choroidal vessels including Haller's layer and Sattler's layer 29 , it is not surprising that the thickness of choroid was correlated with CVD, especially in the Haller's layer.
In regional analysis, CVD in the peripheral area in the PCV group was significantly higher than the normal group, as in the macular area, suggesting diffuse outflow congestion of choroidal vessels in the PCV. Additionally, the increased CVD in the peripheral area was also correlated with choroidal thickness, measured at just below the fovea. These findings are in line with previous report that the increase in choroidal vasculature and hyperpermeability of PCV eyes were associated with increased ocular perfusion pressure and engorgement of the vortex vein 22,30 . Only CVD in MR, which was expected to be the most relevant, was not found to have a significant correlation with choroidal thickness. It is speculated that CVD in MR may be underestimated due to the relatively low resolution of the macula in the UWF imaging and blocked fluorescence caused by turbid subretinal fluid (SRF), subretinal hemorrhage (SRH), and pigment epithelial detachment.
However, regarding the pathogenesis of PCV, it is still uncertain whether the choroid vascular engorgement is simply an associated finding or the primary etiology of PCV. We observed that the CVD of the fellow eyes were significantly higher than the control eyes but did not differ significantly from PCV eyes. And the SFCT of the fellow eyes were significantly thicker than control eyes and significantly thinner than PCV eyes. Although the precise pathogenic mechanism regarding the choroidal changes in the PCV eye remains to be validated, the results support that diffuse choroidal venous congestion might be more suggestive of predisposing systemic background pathophysiology for the onset of PCV and the diffuse choroidal vascular dilatation may precede focal choroidal changes that cause PCV. Further longitudinal UWF ICGA studies with fellow eyes of unilateral PCV may be helpful in clarifying the pathogenesis mechanisms of PCV.
In addition, we showed the relative worse responsiveness to anti-VEGF in eyes with thicker choroid, choroidal vascular hyperpermeability, and higher CVD. CVD and SFCT were found to be the factors significantly associated with the response to anti-VEGF injections. The results are consistent with those of previous studies, which demonstrated that PCV eyes with thick choroids and choroidal hyperpermeability on ICGA show poor response to anti-VEGF treatment 9,17-21,28,31 . Baek et al. reported that baseline VEGF concentration showed a negative correlation with choroidal thickness and a positive correlation with CMT reduction after anti-VEGF treatment for PCV 32 . In the study, the authors speculated that in eyes with thick choroid, mechanical insult to the RPE may be the main factor. At the same time, an ischemic process mediated by the overexpression of VEGF may be the main factor in the development of PCV with the thin choroid. For eyes affected by PCV with higher CVD, photodynamic therapy, which induces remodeling of choroidal vessels, can be considered as an adjuvant treatment option, along with anti-VEGF injections. Further studies regarding correlations between CVD and VEGF levels in PCV are needed.
There are several limitations to this study. First, it could be underpowered to detect even small differences because of the limited sample size. Nevertheless, the strengths of this study is to provide the precise quantitative data on CVD in PCV eyes using UWF ICGA corrected for peripheral distortion, to compare them with normal controls and unaffected fellow eyes, and to evaluate treatment prognosis according to CVD. Second, while we did use a stereographic correction method, this was based on an emmetropic eye with an axial length of 24 mm 33 , and unfortunately, we did not evaluate axial length on all subjects in order to further correction. However, the mean refractory error for the subjects was 0.38 ± 1.44 and the difference in axial length could be neglected. Third, only high-quality images with good contrast were used, and this may have led to selection bias, but, at the same time, decreased information bias. Fourth, due to limited penetration of measuring light in SD-OCT, outermost layer of the choroid can be depicted dark especially in pachychoroid even using EDI and SFCT may have been measured thicker than the actual thickness. Further studies using swept-source OCT are needed in this regard.
In conclusion, CVD appeared to be higher in eyes with PCV than the normal controls in an extensive region, including the periphery. CVD of PCV eyes was positively correlated with choroidal hyperpermeability, subfoveal choroidal thickness, and CMT. These findings suggest that choroidal hypertension, leading to diffuse choroidal venous congestion, may precede PCV occurrence. Additionally, response to anti-VEGF treatment was worse in eyes with higher CVD. CVD > 27.80% could serve as an effective image biomarker that could predict poor treatment response after anti-VEGF treatment for PCV. Further, a larger sample size and a more extended period of longitudinal studies are needed to confirm these results. www.nature.com/scientificreports/ A clinical diagnosis of PCV was based on the presence of polypoidal choroidal vessels in the ICGA images. Exclusion criteria were as follows: (1) retinal disease that affects ocular circulation, such as diabetic retinopathy and retinal vein occlusion; (2) concomitant retinal disease, such as retinal detachment, macular hole, epiretinal membrane, and glaucoma; (3) refractory error more than three diopter; (4) massive SRH or fibrosis obscuring the choroidal vasculature on ICGA; (5) severe media opacity that could degrade image quality; and (6) history of previous treatments that can cause significant changes to choroidal status such as intraocular surgery or intravitreal injection. Cataract surgery performed more than three months previously was not considered an exclusion criterion.

Methods
Age-matched subjects who visited our clinic for a health-screening checkup were enrolled as normal controls without ocular disease, as confirmed by history and ophthalmic examinations. One of the two eyes in the normal control group was randomly selected.
Choroidal vascular analysis. UWF FA and ICGA were performed simultaneously after intravenous injection of 5 mL of 10% sodium fluorescein and 2 mL of 25 mg of indocyanine green. The images in the early phase (two minutes after dye injection) were used for the extraction of choroidal vasculature. Images were transformed into stereographic projection images using the manufacturer's software. After processing with a top-hat filter, the FA and ICGA images were binarized 29 , with noise removed. Finally, the binarized images were trimmed and evaluated for maximum area, excluding all of the following conditions, such as areas covered by eyelashes, low contrast areas obscuring retinal vasculature in UWF FA, and choroidal vasculature in UWF ICGA. Total vascular area (TVA) from the ICGA image and retinal vascular area (RVA) from the FA image were calculated in mm 2 by each pixel from the binarized vascular area 34 . CVA was calculated by subtracting RVA from TVA ( Figs. 2 and 3). CVD was also calculated by dividing CVA by total visible area, in which the peripheral extent of the blood vessel arborization was manually outlined. Two masked retinal specialists (C.M. and G.R.) analyzed the images and calculated CVA and CVD using ImageJ version 1.51j8 (National Institute of Health, Bethesda, MD, USA). The average values from the two independent gradings were used. Regional distributions of CVA and CVD were assessed using a grid with several concentric rings centered on the fovea which was applied to define four zones: MR (0.5-3 mm radius), NPR (3-10 mm), MPR (10-15 mm), and FPR (15 mm-outer gradable perfusion boundary).
Images in the mid-and late-phase (> 5 min after dye injection) were used to evaluate the incidence of choroidal vascular hyperpermeability. Choroidal vascular hyperpermeability was determined in eyes with PCV if there were multifocal hyperfluorescent areas with indistinct margins from choroidal vasculature in the mid-to late-phase ICGA.
Subfoveal choroidal thickness measurement. The choroidal thickness was measured using EDI-OCT.
The thickness values of the total choroid, choriocapillaris-Sattler's layer, and Haller's layer were measured in the subfoveal region (Fig. 4). The SFCT was defined as the vertical distance from the hyper-reflective line of Bruch's membrane to the line connecting the outer margin of the large choroidal vessel layer. The thickness of Haller's layer was defined as the vertical distance from the line connecting the inner margin of the large choroidal vessel layer to the line connecting the outer margin. The thickness of the choriocapillaris-Sattler's layer was calculated by subtracting the thickness of Haller's layer from the total choroidal thickness. Measurements were performed using the built-in caliper tool (Heidelberg Eye Explorer, Version 1.9.10, Heidelberg Engineering) at a single point below the fovea by two independent masked graders (C.M. and G.R.). When there was a disagreement between the graders, the supervising grader (M.S.) confirmed the final decision.
Treatment response. For patients with PCV, intravitreal anti-VEGF injections were performed using bevacizumab (Avastin: Roche, Kaiseraugst, Switzerland), ranibizumab (Lucentis: Genentech, San Francisco, CA), or aflibercept (Eylea: Regeneron, Tarrytown, NY). The selection of anti-VEGF agents was remained at the discretion of the treating physician. All patients were treated following a protocol that included a loading dose with three consecutive intravitreal injections of the same anti-VEGF agent at 1-month intervals. The good responders were defined as those who showed complete resolution of SRF and/or intraretinal fluid (IRF) on OCT at a month after three consecutive anti-VEGF injections. The poor responders were defined as those who showed remained SRF and/or IRF on OCT after the three consecutive injections.
Statistical analysis. The sample size was estimated by the free-software G power (version 3.1.9.4, Franz Faul, University of Kiel, Kiel, Germany). With a power of 80%, a significance level of 0.05, and effect size of 0.8, the sample size for each group was calculated to be 26. Considering dropouts, we planned to include more than additional 10% of calculated sample size.
Statistical analyses were performed using IBM SPSS V.20.0 for Windows (IBM Co., Armonk, New York, USA). The Mann-Whitney U test and the Chi-square test were used to compare numerical variables between PCV patients and age-matched healthy controls and numerical variables between good responders and poor responders, respectively. The Pearson correlation test was performed between numerical variables and CVD from each region. The Youden index (defined as "sensitivity + specificity -100") was used to determine the optimal CVD cut-off point for the best discrimination between the good and poor responders. Variables with a p-value of less than 0.05 were considered statistically significant. All results were presented as the mean ± standard deviation. www.nature.com/scientificreports/ www.nature.com/scientificreports/ www.nature.com/scientificreports/

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.