1. Introduction
The choroid is the most vascularized structure in the eye, and choroidal vascular structures are correlated with the pathogenesis of various diseases, including age-related macular degeneration, polypoidal choroidal vasculopathy, central serous chorioretinopathy, and myopic macular degeneration [
1,
2,
3,
4,
5]. Owing to the anatomical features of the choroid, structural analysis of the choroidal vasculature is challenging and there is a lack of qualitative and quantitative indicators. With the advent of enhanced depth imaging optical coherence tomography (EDI-OCT) technology, which is a modification of standard spectral-domain OCT (SD-OCT) and provides improved signal penetration, the study of the choroidal structure has advanced considerably [
6,
7,
8,
9]. Choroidal thickness (CT) is one of the most widely examined choroidal parameters, and has been found to be affected by different variables, such as axial length, refractive error, intraocular pressure, and systolic blood pressure, but does not provide us with information about structural changes in the choroid [
10,
11].
The application of EDI-OCT image binarization has become a research hotspot in recent years and has made it possible to explore the choroidal structure and its changes in different disease models. Choroidal vascularity index (CVI) is a novel parameter that is calculated from EDI-OCT scans via image binarization. It is defined as the ratio of vascular luminal area (LA) to total choroidal area (TCA), which is presented as a percentage. Several small-sample studies in healthy populations showed that CVI is less affected by physiological variables and has been demonstrated to be a reliable tool, which can quantitatively reflect structural changes in the choroidal vasculature [
12,
13,
14]. CVI combined with subfoveal choroidal thickness (SFCT), LA, and stromal area (SA) can analyze choroidal morphological changes and choroidal blood perfusion.
In this study, we assessed the macular choroidal structure in a large sample of healthy Chinese individuals, and analyzed the changes in CVI and the thickness of each choroidal vascular layer with age and sex. We further examined the stability and reproducibility of CVI and SFCT measurements in this healthy population, and performed univariate and multivariate regression analyses.
2. Methods
2.1. Population
This cross-sectional study was conducted in the Department of Ophthalmology, Beijing Friendship Hospital, from December 2018 to December 2019. Individuals were enrolled according to the inclusion and exclusion criteria. Inclusion criteria: (1) best corrected visual acuity of ≥0.6 and (2) spherical power between +3D and −3D. Exclusion criteria: (1) patients who cannot undergo routine mydriasis; (2) active intraocular inflammation and/or infection, history of any type of intraocular surgery (except cataract surgery); (3) those with retinal and choroidal diseases (except retinal atherosclerosis); (4) those with significant refractive interstitial opacity or abnormalities that affect OCT measurements, such as significant corneal opacity, cataract, vitreous opacity, vitreous hemorrhage, or silicone oil inside the vitreous cavity; (5) those with hypertension, hyperlipidemia, diabetes, heart disease, cerebrovascular disease, pulmonary hypertension, renal disease, and peripheral and central vasculopathy; (6) current smoking and alcohol consumption; (7) those with acute or chronic infectious diseases, infectious inflammatory diseases, and malignancies; and (8) patients who are unable to cooperate in the examination or those in whom clear results cannot be obtained or whose test results cannot be analyzed. This study was approved by the Bioethics Committee of Beijing Friendship Hospital, Capital Medical University (2018-P2-205-01), and conducted in accordance with the tenets of the Declaration of Helsinki.
2.2. Data Collection
The data included demographic information (age, sex, occupation, marital status, education, and workload), health status (hypertension, diabetes mellitus, coronary atherosclerotic heart disease, hyperlipidemia, cerebrovascular disease, malignancy, peripheral vascular disease, etc.), and daily habits (alcohol consumption and smoking). Ocular data were collected from all enrolled patients, including best-corrected visual acuity, noncontact IOP measurements, slit lamp examination, and color fundus photography. Patients underwent high-definition EDI-OCT scans in both eyes using Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany), and raster scans were performed to cover a 20 × 20° (6 × 6 mm) area. OCT images were acquired from 9:00 a.m. to 12:00 p.m. daily by the same technician (Wei Wang).
2.3. Quality Control of Images
Patients were enrolled by selecting EDI-OCT images of the left eye first and all data were collected by the same technician (Wei Wang). Three physicians in this study group (Shuang Li, Luping Wang, and Zhuohua Zhou) performed independent assessment of choroidal image clarity in both eyes. If two or more evaluators determined that the image clarity in the left eye was worse than that in the right eye, the patient was enrolled using the right eye image for analysis.
After the EDI-OCT images were acquired, three physicians from this study group performed independent evaluations. The images were considered acceptable and used for analysis when two or more investigators determined that the subfoveal choroid was clearly discernible and the choroidal and scleral boundaries were well demarcated. For images with an undefinable sclerochoroidal boundary, two physicians discussed the demarcation. If there was continuing discrepancy between them, the image was given to a third physician for arbitration and if a decision could not be reached, the image was excluded. Subsequent to this process, 27 images were eliminated, leaving 1566 images for consideration.
2.4. Range of the Subfoveal Choroidal Area
SFCT: defined as the vertical distance from the outer surface of the subfoveal retinal pigment epithelium to the sclerochoroidal junction. SFCT was measured using the built-in software caliper tool (
Figure 1).
2.5. Thickness Measurement of Each Choroidal Vascular Layer
Vascular thickness analysis of each layer of the choroid was performed manually according to Branchini’s method [
15]. The large choroidal vessels closest to the fovea were first selected. Large choroidal vessels were defined as those with a diameter of ≥100 μm. A horizontal line was drawn along the inner edge of the large choroidal vessels, and this line intersected the line used to measure the SFCT. Measurements were made from the sclerochoroidal junction to the point where this horizontal line intersected the SFCT measurement line. This length was taken as the thickness of the choroidal large vessel layer. The length of this intersection point to the Bruch’s membrane is equivalent to the thickness of the choriocapillaris–medium choroidal vessel layer (CC+MCVL), which was obtained by subtracting the large choroidal vascular layer (LCVL) thickness from the SFCT. We also calculated the ratio of LCVL to SFCT (
Figure 2).
2.6. Image Binarization and CVI Measurement
The public open-source software Image J (version 1.51) was used to process images (
Figure 3). We masked patient information during image processing. The choroidal area centered on the fovea, and nasal and temporal distances of 750 μm were marked. The total choroidal area (TCA) was calculated. We performed 8-bit conversion of the image, and adopted Niblack Auto Local Threshold tool to binarize the selected area. In the binarized image, dark pixels indicated the vascular lumen and white pixels indicated the stromal region. After converting the image to RGB (red, green, and blue) colors, the dark pixels were selected using the color threshold tool, and the LA was calculated. SA was obtained by subtracting LA from TCA. The ratio of LA to TCA constituted CVI. Image segmentation was performed by one of the authors (Shuang Li) (
Figure 3).
2.7. Interrater and Intrarater Agreement
CVI and SFCT were calculated by two examiners (Shuang Li and Zhuohua Zhou) to determine interrater agreement. The CVI and SFCT of the study cohort were calculated by one examiner (Shuang Li) after an interval of 1 week to compute intra-rater reliability. The interrater and intrarater reliability for the measurement of images was measured by the absolute agreement model of the intraclass correlation coefficient (ICC). ICCs of 0.81–1.00 indicate good agreement. ICCs < 0.3 indicate weak or poor agreement.
5. Discussion
Recent advances in OCT imaging techniques, especially the EDI-OCT, have helped us better visualize the choroidal layers. Various OCT-related parameters have been developed to evaluate the choroid in healthy and disease states, and mostly evaluating the choroidal thickness and volume [
15,
16]. CVI is a noninvasive choroidal quantitative parameter, which is easily accessible [
17]. The application of CVI has aroused great interest in studying the choroidal vascular structure [
17,
18]. The use of this quantitative parameter to assess the structural state of the choroid in healthy individuals at different ages can provide additional information for morphological and physiological structures of the choroid [
19]. Furthermore, it can aid in observing age-related choroidal changes and help us to better characterize retinal/choroidal diseases, such as age-related macular degeneration, inflammatory chorioretinal disorders, pachychoroid disease spectrum, myopia, and inherited retinal disorders [
20,
21,
22,
23,
24].
Sonoda et al. [
15,
24] first used image binarization to study LA and SA of the choroid in EDI-OCT images of healthy eyes. They reported a significant decrease in LA, SA, and LA/SA ratios with age, indicating a greater decrease in vascular LA than in SA. Ruiz-Medrano et al. [
25] studied healthy individuals in a larger age range (3–85 years) and reported a significant decrease in LA and CVI with age while SA remained stable. However, the aforementioned study included only a limited sample size (136 individuals), the number of patients in each group was small, and data were missing for the younger children and the elderly population after stratifying by age [
25].
Our study demonstrated a strong correlation between CVI and age in the macular region in this cohort of a healthy Chinese population, and the LA and CVI decreased with age. The SA was found to be smaller in the 0–10 and 11–20 age groups and gradually increased with age, reaching a maximum in the 61–70 age group, and then relatively stabilized. Previous histological studies have shown that the volume of choroidal cells and interstitial components decreases with age [
26,
27]. Several studies on choroidal immunohistochemistry reported a decrease in CT with age, accompanied by a significant decrease in collagen fibronectin, and cellular components in the choroid [
26,
27,
28]. In addition, it is known that vascular tone and the amount of endothelial nitric oxide synthase (eNOS) decrease with age, thereby leading to a decrease in circulating blood volume [
29,
30]. All of these observations support the results of the present study.
In our study population, both CVI and subfoveal LCVL/SFCT had less variability than that observed in SFCT. Moreover, previous studies showed that CVI and subfoveal LCVL/SFCT are less affected by physiological factors other than age [
31]. These observations suggest that CVI and subfoveal LCVL/SFCT are relatively stable markers for studying choroidal changes. The CV of CVI was smaller compared with LCVL/SFCT in the fovea, indicating that CVI is a better and relatively more stable marker for monitoring the choroid and provides more information than a simple SFCT measurement.
The univariate analysis of ocular and systemic factors correlated with CVI showed significant correlations with age, BMI, ocular axis, LA, SA, LCVL thickness, CC+MCVL thickness, and LCVL/SFCT ratio. However, age, LA, LCVL thickness, and LCVL/SFCT ratio were factors correlated with CVI in the multiple regression model of this study. Of these, there was a strong correlation with age and a weak correlation with all other factors. Of note in the group of 0–10 years, the ocular axis is shorter, and the CVI is significantly higher. This may be due to the children at this stage having hyperopia and thicker choroid. In groups older than 10 years, natural growth and developmental processes of the eye, such as axial elongation and choroidal thinning, might play a role in reductions in CVI. We also observed that LCVL/SFCT showed a statistically significant correlation with age and CVI. Hence, understanding the age-related changes in choroidal structure in healthy eyes is important for the clinical application of CVI. This can better explain the differences in choroidal structure in patients of different ages and help explore the pathogenesis of various diseases, especially in ocular diseases with a strong correlation to age, such as age-related macular degeneration and adolescent myopia.
Furthermore, we observed a significant correlation between BMI and CVI in the univariate analysis. Since the current study included only healthy people with a mean BMI of 23.75 ± 5.72, we intend to study the high BMI population in future research, and analyze the changes in the choroidal vascular structure in the obese population.
Agrawal et al. reported that CVI was less variable than SFCT, and SFCT was affected by more factors than CVI [
12]. In that population-based study, they included 345 healthy eyes. Except for CVI and SFCT, in the present study, we also measured LCVL, CC+MCVL, and LCVL/SFCT, expecting to find the correlation between the thickness of these two parts of the vascular layer and age. LCVL thickness was the smallest in the age group of 0–10 years, then gradually increased and peaked in the age group of 41–50 years, after which it slightly changed with age. CC+MCVL thickness was the maximum in the age group of 11–20 years, then gradually decreased with age and reached the minimum after 80 years. LCVL/SFCT was the smallest in the group of 0–10 years, then increased gradually with age and reached the maximum in the group of >80 years. LCVL thickness changed very little with age >20 years, but CC+MCVL decreased significantly with age. We inferred that some of the choriocapillaris may occlude and detach with age, which may also be correlated with the occurrence of age-related macular degeneration. The CV of LCVL/SFCT was 7.56%. It is also a relatively stable biological parameter for monitoring choroidal vascular structures, an understanding of its age-related changes in normal eyes could enable further exploration of the pathogenesis of various age-related fundus diseases. We further observed the measurement consistency of CVI and SFCT, and the repeatability of both parameters was within the confidence range. Compared to SFCT, the ICC values of CVI were significantly higher and exhibited better measurement consistency.
The strengths of our study are (1) It is a large-sample observation with a single common ethnicity, and is less likely to be confounded by ethnic heterogeneity. (2) Our study applied standardized clinical examination protocols and image processing procedures. (3) We confirmed that CVI and LCVL/SFCT had less variability than that observed in SFCT. However, this study has several limitations: (1) Binarization of the images was only conducted in the single eye of each study subject. (2) Although OCT images were binarized at standardized protocols, there was a possibility of over or underestimation of LA and SA. (3) Our study mainly focused on the subfoveal region, the extramacular choroidal vascular features in individuals have remained largely unexplored.
In conclusion, in this study based on a large sample of subjects, we found that CVI of healthy populations showed better consistency and reproducibility when compared with SFCT. In addition, we demonstrated that the choroidal vascular area and CVI significantly decrease with age. Age-related decrease in vascular components maybe dominated by a decrease in CC+MCVL. Our findings provide new insights that may be helpful in future studies on the pathophysiology of the human choroid. Larger datasets in different disease states are needed to further validate the value of these markers for application in clinical practice.