Sex differences in choroidal vessels using novel wide-field choroidal en-face images from optical coherence tomography

Abstract

This study aims to develop a method to quantify choroidal vessels in normal eyes using wide-field optical coherence tomography (OCT) en-face images. The study included participants with normal eyes in whom wide-angle OCT images were acquired to generate planarized choroidal en-face and thickness map images. The images were segmented into central, midperipheral, and peripheral areas, and the midperipheral and peripheral areas were further segmented into supratemporal, infratemporal, supranasal, and infranasal sectors. The mean planarized choroidal-vessel density (p-CVD), planarized choroidal-vessel size (p-CVS), and choroidal thickness (CT) were calculated in each sector. Sex differences were analyzed using the Mann–Whitney U test. The study included 162 participants comprising 84 female (mean age, 43.5 years; axial length, 24.0 mm) and 78 male (mean age, 44.4 years; axial length, 24.2 mm) participants with no significant differences in demographics (P ≥ 0.107). Men had a higher mean p-CVD in all regions (P < 0.001). The mean p-CVS was greater in men in all regions except for the supratemporal sector (P < 0.001). No significant differences in sex in the mean CT were observed in all regions (P ≥ 0.106). The p-CVD and p-CVS in normal eyes differ between sexes. This finding may contribute to the understanding of the pathophysiology of choroidal diseases.

Introduction

The choroid is a highly vascularized tissue essential for retinal health and implicated in various diseases, including pachychoroid-spectrum disorders like central serous chorioretinopathy (CSC)^1–6. Notably, CSC incidence is six times higher in men than in women⁷, yet the impact of this sex disparity remains unclear. Exploring choroidal differences in normal eyes across sexes is crucial for understanding potential sex-related influences on disease mechanisms. Previous studies have provided mixed results on sex differences in choroidal structure, possibly due to limited assessment areas^8–19. Recent advancements in optical coherence tomography (OCT) enable comprehensive imaging of the choroid, offering new insights into its vascular characteristics^{14,16,20–23}. This study aims to utilize wide-angle choroidal en-face images to compare choroidal structure between sexes in normal eyes, potentially shedding light on the pathophysiology of choroid-related diseases.

Results

Participant characteristics

The study included 162 eyes of 162 participants (average age, 43.9 ± 18.4 years), with 84 eyes (51.9%) from female participants. All participant data are presented in Online Resource 1. In instances where neither eye met the exclusion criteria, the right eye was selected for analysis. For the subset of 25 cases with data available for both eyes, an inter-eye comparison of parameters was conducted. The analysis revealed no statistically significant inter-eye differences in the measured parameters. The data are provided in Online Resource 2.

CT values in the central area were 295.0 ± 84.8 µm; CT values in the middle area were 299.2 ± 76.7 µm in the ST sector, 244.7 ± 71.1 µm in the SN sector, 264.5 ± 71.6 µm in the IT sector, and 189.6 ± 57.6 µm in the IN sector. In the peripheral area, CT values were 230.5 ± 46.0 µm in the ST sector, 219.1 ± 59.5 µm in the SN sector, 204.4 ± 43.9 µm in the IT sector, and 161.5 ± 38.1 µm in the IN sector. En-face p-CVD values in the central area were 43.9 ± 4.3% and En-face p-CVD values in the middle area were 43.7 ± 4.2% in the ST sector, 41.2 ± 3.7% in the SN sector, 43.8 ± 3.6% in the IT sector, and 40.8 ± 3.2% in the IN sector. En-face p-CVD values in the peripheral area were 43.9 ± 3.5% in the ST sector, 42.0 ± 3.2% in the SN sector, 44.4 ± 3.3% in the IT sector, and 41.1 ± 2.9% in the IN sector. In middle and peripheral areas, the p-CVD was higher in ST and IT than in other sectors, with IN showing the smallest p-CVD (Middle: SN vs. IN, P = 0.510; ST vs. IT, P = 0.900; others, P < 0.001. Peripheral: ST vs. IT, P = 0.075; others, P < 0.001). En-face p-CVS values in the central area were 10.1 ± 1.0 pixels and En-face p-CVS values in the middle area were 10.2 ± 1.1 pixels in the ST sector, 9.5 ± 0.9 pixels in the SN sector, 10.1 ± 1.0 pixels in the IT sector, and 9.1 ± 0.8 pixels in the IN sector. En-face p-CVS values in the peripheral area were 9.8 ± 0.9 pixels in the ST sector, 9.3 ± 0.8 pixels in the SN sector, 9.7 ± 0.9 pixels in the IT sector, and 8.6 ± 0.7 pixels in the IN sector. In middle and peripheral areas, the p-CVS was larger in ST and IT than that in other sectors, with IN showing the smallest p-CVS (Middle: ST vs. IT, P = 0.110; others, P < 0.001. Peripheral: ST vs. IT, P = 0.800; others, P < 0.001). All these results are shown in Online Resource 3.

Participant characteristics stratified by sex

No significant differences in age and axial length were found between male and female participants. However, male participants exhibited a higher choroidal luminal proportion in the macula. Significant sex differences in en-face p-CVD were observed in all peripheral regions, with males showing higher values. En-face p-CVS also showed significant differences in most regions, indicating larger vessel sizes in males. These findings are detailed in Tables 1 and 2, and Fig. 1.

Table 1.

Comparisons of participant characteristics between male and female participants.

Characteristics	Male, N = 78	Female, N = 84	P-values
Age (years)	44.4 ± 19.1; 42.5	43.5 ± 17.8; 42.0	0.803
Axial length (mm)	24.2 ± 0.9; 24.2	24.0 ± 09; 24.0	0.107
Central choroidal thickness (μm)	312.5 ± 89.2; 313.0	301.2 ± 101.0; 299.0	0.247
Choroidal luminal proportion (%)	69.4 ± 4.0; 68.5	66.9 ± 3.1; 67.4	< 0.001

P-values were calculated through logistic regression analysis using the firth bias reduction method for qualitative variables and Mann–Whitney U test for continuous variables.

Table 2.

Comparisons of choroidal parameters (thickness, en-face p-CVD, and en-face p-CVS) between male and female participants.

Characteristics	Male, N = 78	Female, N = 84	P-values
Central area: thickness (μm)	303.5 ± 87.3; 294.0	287.0 ± 82.1; 286.0	0.212
Middle ST: thickness (μm)	304.7 ± 76.6; 297.5	294.1 ± 77.0; 281.5	0.319
Middle SN: thickness (μm)	243.4 ± 68.6; 238.5	245.8 ± 73.7; 237.0	0.841
Middle IT: thickness (μm)	270.0 ± 72.3; 269.0	259.4 ± 71.0; 256.5	0.250
Middle IN: thickness (μm)	191.4 ± 61.1; 181.0	187.8 ± 54.4; 181.5	0.908
Peripheral ST: thickness (μm)	235.5 ± 44.2; 232.5	225.8 ± 47.4; 214.0	0.106
Peripheral SN: thickness (μm)	214.3 ± 57.4; 205.5	223.6 ± 61.5; 220.0	0.106
Peripheral IT: thickness (μm)	209.9 ± 44.7; 204.0	199.3 ± 42.7; 192.5	0.123
Peripheral IN: thickness (μm)	159.0 ± 38.4; 155.0	163.8 ± 37.8; 157.0	0.448
Central area: en-face p-CVD (%)	45.8 ± 4.1; 46.1	42.2 ± 3.8; 42.0	< 0.001
Middle ST: En-face p-CVD (%)	45.3 ± 4.3; 46.0	42.2 ± 3.5; 42.1	< 0.001
Middle SN: en-face p-CVD (%)	42.5 ± 3.7; 42.5	40.1 ± 3.3; 40.2	< 0.001
Middle IT: en-face p-CVD (%)	45.1 ± 3.5; 45.0	42.7 ± 3.3; 42.8	< 0.001
Middle IN: en-face p-CVD (%)	41.8 ± 3.2; 41.5	40.0 ± 3.0; 39.6	< 0.001
Peripheral ST: en-face p-CVD (%)	45.4 ± 3.4; 45.9	42.6 ± 3.0; 42.9	< 0.001
Peripheral SN: en-face p-CVD (%)	43.3 ± 2.9; 43.4	40.7 ± 2.9; 40.7	< 0.001
Peripheral IT: en-face p-CVD (%)	45.8 ± 3.1; 46.1	43.1 ± 3.0; 43.1	< 0.001
Peripheral IN: en-face p-CVD (%)	42.3 ± 2.7; 42.5	40.0 ± 2.6; 40.3	< 0.001
Central area: en-face p-CVS (pixels)	10.4 ± 1.1; 10.2	9.9 ± 0.9; 9.8	0.005
Middle ST: en-face p-CVS (pixels)	10.5 ± 1.1; 10.8	10.0 ± 1.0; 10.0	0.001
Middle SN: en-face p-CVS (pixels)	9.6 ± 0.9; 9.7	9.4 ± 0.9; 9.4	0.169
Middle IT: en-face p-CVS (pixels)	10.3 ± 1.0; 10.4	9.9 ± 0.9; 9.9	0.023
Middle IN: en-face p-CVS (pixels)	9.2 ± 0.8; 9.2	9.0 ± 0.8; 8.9	0.049
Peripheral ST: en-face p-CVS (pixels)	10.0 ± 0.9; 10.1	9.5 ± 0.8; 9.5	< 0.001
Peripheral SN: en-face p-CVS (pixels)	9.4 ± 0.8; 9.3	9.1 ± 0.8; 9.1	0.068
Peripheral IT: en-face p-CVS (pixels)	9.9 ± 0.9; 9.8	9.5 ± 0.9; 9.4	0.005
Peripheral IN: en-face p-CVS (pixels)	8.8 ± 0.7; 8.7	8.5 ± 0.7; 8.4	0.006

P-values were calculated through logistic regression analysis using the Firth bias reduction method for qualitative variables and Mann–Whitney U test for continuous variables.

p-CVD planarized choroidal vessel density, p-CVS planarized choroidal vessel size, ST supratemporal sector, IT infratemporal sector, SN supranasal sector, IN infranasal sector.

Figure 1.

Box-and-whisker plot and frequency polygonal line of en-face planarized choroidal-vessel density in the peripheral area by sex. Male participants demonstrate significantly higher en-face planarized choroidal-vessel density in all regions than female participants.

Discriminative ability of choroidal parameters in distinguishing sex

The AUC analysis for distinguishing sex revealed that while CCT had moderate discriminative ability, choroidal luminal proportion in the macula showed higher discrimination potential. En-face p-CVD and p-CVS in middle and peripheral areas demonstrated significant discriminative ability between sexes, with detailed AUC values provided in Table 3 and Online Resource 4.

Table 3.

Area under the curve for sex differentiation based on choroidal parameters.

Characteristics	AUC	95% confidence intervals	P-values*	P-values†
Central choroidal thickness (μm)	0.553	0.464–0.642	–	0.063
Choroidal luminal proportion: macula (%)	0.672	0.584–0.760	0.063	–
En-face p-CVD: center area (%)	0.747	0.671–0.823	< 0.001	0.206
En-face p-CVD: middle area, ST (%)	0.726	0.646–0.806	< 0.001	0.373
En-face p-CVD: middle area, SN (%)	0.695	0.613–0.777	0.005	0.706
En-face p-CVD: middle area, IT (%)	0.695	0.614–0.776	0.002	0.704
En-face p-CVD: middle area, IN (%)	0.680	0.598–0.763	0.015	0.891
En-face p-CVD: peripheral area, ST (%)	0.742	0.665–0.820	< 0.001	0.230
En-face p-CVD: peripheral area, SN (%)	0.737	0.661–0.814	< 0.001	0.274
En-face p-CVD: peripheral area, IT (%)	0.728	0.651–0.805	< 0.001	0.348
En-face p-CVD: peripheral area, IN (%)	0.730	0.653–0.808	< 0.001	0.331

En-face p-CVD values were compared with central choroidal thickness and macular choroidal luminal proportion.

p-CVD planarized choroidal vessel density, AUC area under the curve.

*Comparisons with central choroidal thickness, DeLong test.

^†Comparisons with the choroidal luminal proportion in the macula, DeLong test.

Multivariable analysis for choroidal morphological differences by sex

Adjusting for potential confounders highlighted significant differences in en-face p-CVD and p-CVS between sexes across all examined regions, with male participants showing larger values. Additionally, a negative correlation with age was observed for p-CVD and p-CVS in the peripheral area. These multivariable analysis results are presented in Online Resources 5, 6, alongside representative cases for each sex in Fig. 2.

Figure 2.

Representative cases for male and female participants. The post-binarization choroidal en-face images, average en-face choroidal-vessel density and en-face choroidal-vessel size, choroidal thickness map, and average choroidal thickness in each region are presented. (A) The participant was a 31-year-old man with axial length of 24.17 mm. (B) The participant was a 33-year-old woman with an axial length of 24.18 mm. Male participants showed higher values in average en-face choroidal-vessel density and en-face choroidal-vessel size in all regions.

Discussion

This study’s employment of WF-OCT facilitated a thorough evaluation of choroidal morphology in normal eyes, enabling the examination of sex differences. Although CT showed no significant variation across all regions between males and females, the analysis revealed that en-face p-CVD and choroidal luminal proportion in the macula were significantly higher in males across all regions. Furthermore, except for the SN region, en-face p-CVS was larger in males, a difference that persisted even after adjusting for potential confounders, indicating a marked sex distinction.

Previous studies on CT variation between sexes in normal eyes have reported mixed outcomes, often limited by the narrow regional focus under the fovea or at the posterior pole^16,19. Consistent with these findings, our study, leveraging wide-angle OCT, found no significant CT differences between sexes, including CCT, corroborating earlier reports¹⁴.

The choroid’s stromal and luminal components, which vary proportionally in different diseases (e.g., Vogt–Koyanagi–Harada disease vs. CSC)²⁴, underscore limitations of CT-based analyses for understanding the choroid’s impact on retinal diseases and sex differences. Our findings of a higher luminal: stromal ratio in men align with previous reports, suggesting that parameters like choroidal luminal proportion in the macula may offer a more accurate reflection of the choroid’s relationship with disease states^10,18.

Advancements in OCT technology now permit the creation of en-face images for comprehensive choroidal-layer analysis. The strength of en-face imaging lies in its ability to non-invasively visualize the choroidal structure across a wide area, enhancing our understanding of choroidal findings^21,23. Our study adopts this approach, utilizing the Bruch membrane to the chorioscleral boundary as the reference plane for generating en-face images of the entire choroid, achieving high reproducibility and consistent standards.

Introducing new parameters, en-face p-CVD, and p-CVS, our study uncovered significant sex differences in choroidal structure, posteriorly and peripherally, surpassing CCT in distinguishing between sexes even after adjusting for confounders. These results align with our previous research indicating a greater number of vortex vein ampullas in men²⁵ and higher fundus tessellation, suggesting larger choroidal vessels in males²⁶. Our study further supports the notion that quantitative measurements from wide-angle choroidal en-face images effectively capture sex differences in the choroid. These findings have implications for understanding the choroidal involvement in pachychoroid-spectrum disorders and age-related macular degeneration, as highlighted by Lee et al.²⁷ and Kogo et al.²³ The significant sex disparity in CSC incidence rates, especially among middle-aged men⁷, might be partly explained by choroidal structural differences identified in our analysis, offering insights into the disease mechanism.

However, our study has limitations, such as its focus on Asian participants and exclusion of individuals younger than 15 years, affecting the generalizability and depth of age-related insights. The approximation of choroidal density and vessel diameter also poses challenges owing to the eye's variable shape among individuals, necessitating cautious interpretation of results, especially with wide-angle imaging devices varying in units. Additionally, external factors such as age, axial length, smoking²⁸, caffeine intake²⁹, and diurnal variations³⁰ affect choroidal morphology warrant further investigation. In this study, we calculated the choroidal luminal proportion within a 3 mm area centered on the macula using B-scan images. However, technical limitations precluded the calculation of the choroidal luminal proportion over a broader area. Ideally, it is essential to extend the analysis to encompass a wider region beyond the macular area. Addressing these technical challenges will be a crucial objective for future research efforts.

In conclusion, our innovative approach to analyzing wide-angle choroidal en-face images has revealed significant sex differences in choroidal-vessel structure, providing a foundational perspective for future pathological studies in macular diseases.

Methods

Participants

This cross-sectional study included 162 eyes of 162 Japanese patients with no intraocular disease who visited the Department of Ophthalmology at Kagoshima University Hospital from April 2021 to June 2023. The study was conducted with the approval of the Ethics Committee of Kagoshima University Hospital (Kagoshima Prefecture, approval no. 160121, dated August 3, 2021). All study participants provided informed consent, and this study was performed in accordance with the relevant guidelines and regulations (https://www.nature.com/srep/journal-policies/editorial-policies#experimental-subjects). The examinations performed were refraction (RM8900, Topcon Corporation, Tokyo, Japan), corrected visual acuity, ocular axial length measurement (Tomey GmbH, Nuremberg, Germany), and wide-field OCT (WF-OCT) imaging (Xephilio OCT-S1, Canon Medical Systems Corporation, Tochigi, Japan). Patients were excluded if they had an ocular axial length < 20 mm or > 27 mm, equivalent sphere < -6 D or >  + 3 D, or unclear WF-OCT images (signal strength ≤ 5). Additionally, patients with a history of buckle surgery or vitrectomy, or examination results indicating uveitis, choroidal retinal disease, glaucoma, posterior uveitis, pregnancy, or neoplastic disease were excluded. In cases where neither eye met the exclusion criteria, data from the right eye were used for analysis.

OCT imaging protocol

WF-OCT images were captured using near-infrared light (1010–1110 nm, scanning laser ophthalmoscope, 780 nm) at a scanning speed of 100 000 A-scans/s. Three-dimensional volume data encompassed a 20 mm (vertical) × 20 mm (horizontal) area (696 B-scans by 696 pixels) with a 5.3 mm scanning depth (1396 pixels). To minimize the impact of diurnal variations in choroidal structure, images were acquired between 14:00 and 16:00. En-face images' reference lines were determined based on the Bruch membrane and chorioscleral interface^14,23. Segmentation was initially performed automatically using the built-in software supported by artificial intelligence (Xephilio OCT-S1; Canon Medical Systems Corporation), followed by manual correction of segmentation lines in all slices by two evaluators (NM and RF).

Measurement of choroidal thickness and luminal proportion in the macula.

The vertical distance from the Bruch membrane to the chorioscleral boundary was defined as the choroidal thickness (CT). A CT map was created from the captured three-dimensional volume data. The choroidal map was divided into three areas: central (0–20°), middle (20–40°), and peripheral (40–70°). Middle and peripheral areas were further divided into four sectors: the supratemporal (ST), infratemporal (IT), supranasal (SN), and infranasal (IN) sectors. The optic disc was excluded from the analysis area. The analysis tool calculated the average CT for the nine created sectors. The CT beneath the fovea was defined as the central choroidal thickness (CCT). The choroidal luminal proportion in the macula was obtained by quantifying a single OCT B-scan image (EDI/2-frame averaging/horizontal) within a 3000-μm diameter range around the fovea. The detailed quantification method was reported previously³¹.

Quantification of choroidal en-face images

As reported previously^21,23, we created en-face images of the entire planarized choroid using the Bruch membrane to the chorioscleral boundary as the reference plane. The resulting en-face images of the entire planarized choroid were produced as outputs of 1392 × 1392 pixel bitmap images (Fig. 3A). To extract choroidal vessels from output images, binarization was necessary. However, owing to wide-angle imaging and inconsistent contrast between the lumen and stroma, simple binarization was challenging. Specific examples are shown in (Online Resource 7A–C). Therefore, ImageJ’s plugin filters were used for preprocessing original images before binarization. The binarization process is detailed in (Online Resource 8A–F). A binarized image of choroidal vessels is presented in (Fig. 3B). As reported previously²³, the choroidal-vessel area was first calculated and its proportion to the total area was defined as the en-face planarized choroidal-vessel density (p-CVD). The following thinning process was conducted to quantify vessel size. After vessels were first thinned (Fig. 4A,B), data obtained from the thinning were line drawings (Fig. 4C). These were classified into isolated points, endpoints, bifurcation points, and passing points; segments defined by endpoints or bifurcation points were divided. These segments were approximated to line segments using a two-division method. Specifically, a line segment connecting two points, namely A and B, on a segment was first created. The point C, farthest from the perpendicular line of the segment curve, was identified. The same process was applied to line segments composed of A and C, and B and C and repeated until the perpendicular distance from the line segment was < 3 pixels (Fig. 4D). The vector direction orthogonal to the obtained line segment was calculated; the distance from the line segment to the boundary between the lumen and the stroma was measured at 1-pixel intervals (Fig. 4E). The average of these distances in each area was calculated and defined as the en-face planarized choroidal-vessel size (p-CVS). These methods were developed on the basis of the software previously developed for posterior choroidal en-face image analysis²².

Figure 3.

Binarization process of choroidal en-face images. (A) An original choroidal en-face image was created using wide-field optical coherence tomography. (B) The choroidal en-face image is presented after binarization.

Figure 4.

Quantification of vessel diameter. (A) An overlay image of the choroidal en-face image after binarization and thinning process is presented. (B) The image shows only the thinning process. (C) The data image obtained by the thinning process, consisting of a collection of 1-pixel points forming curves rather than straight lines, is presented. (D) The image illustrates the approximation of curves to line segments using a two-division method. (E) The image depicts the process of vessel diameter calculation, where vectors orthogonal to the obtained line segments were calculated. The distance from the line segment to the boundary between the lumen and stroma was measured at 1-pixel intervals.

Statistical analysis

Descriptive statistics of participant characteristics and choroidal parameters were examined; sex differences for each factor were assessed. For cases with data available from both eyes, inter-eye comparisons and correlation analyses were conducted for each parameter. The Wilcoxon signed-rank test was employed to evaluate differences between the left and right eyes, and Pearson correlation coefficients were computed to assess the inter-eye correlations. To adjust for multiple comparisons, the Bonferroni correction was applied. Regional differences in each choroidal parameter were examined using the Nemenyi–Wilcoxon–Wilcoxon all-pairs test with multiple comparison correction. Patient characteristics and outcomes were compared before adjusting for confounders. For continuous variables, the Mann–Whitney U test was used; categorical variables were assessed through Fisher’s exact test. To examine the discriminatory ability of each choroidal parameter between sexes, the area under the curve (AUC) of the receiver operating curve was calculated. Additionally, the ability to distinguish between sexes using en-face p-CVD and p-CVS derived from wide-field OCT en-face images was compared to classic choroidal morphological parameters using the DeLong test. Sex differences in choroidal parameters were also examined after adjusting for potential confounders¹⁰. A P-value of 0.05 indicated statistical significance. All analyses were performed using R software (version 4.3.1; R Core Team).

Ethics approval

The study was conducted with the approval of the Ethics Committee of Kagoshima University Hospital (Kagoshima Prefecture, Approval No. 160121, dated August 3, 2021). All study participants provided informed consent.

Author contributions

Conceptualization and study design: Naohisa Mihara Ryoh Funatsu, and Hiroto Terasaki; Data collection: Naohisa Mihara Takato Sakono and Ryoh Funatsu; Data analysis: Naohisa Mihara and Ryoh Funatsu ; Interpretation of data: Naohisa Mihara, Ryoh Funatsu, Takato Sakono, Hideki Shiihara, Shozo Sonoda, Hiroto Terasaki and Taiji Sakamoto; Writing—Original draft preparation: Naohisa Mihara and Ryoh Funatsu; Writing––review and editing: Naohisa Mihara, Ryoh Funatsu, Hiroto Terasaki, Shozo Sonoda and Taiji Sakamoto.

Funding

This study was supported by JSPS KAKENHI (Grant Number: 21H03095).

Data availability

The datasets generated and/or analyzed during the current study are not publicly available because it was not included in our approved research plan but are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-67671-w.