Introduction

Impaired ocular blood flow is a key factor in the pathogenesis of various retinal diseases, including diabetic retinopathy (DR) and retinal vascular occlusive disorders. In age-related macular degeneration (AMD), a leading cause of irreversible vision loss in older adults, reduced choriocapillaris (CC) perfusion plays a critical role in disease progression1. Traditionally, retinal and choroidal circulation have been assessed using dye-based angiographic methods, such as fluorescein angiography and indocyanine green angiography. However, these techniques provide only two-dimensional flow images and cannot evaluate macular blood flow quantitatively or layer-specifically.

Optical coherence tomography angiography (OCTA) is a non-invasive imaging modality that detects blood flow based on motion contrast generated by erythrocyte movement within vessels, in contrast to the static surrounding retinal tissues. With its high scanning speed and resolution, OCTA enables detailed layer-by-layer visualization of the retinal and choroidal microvasculature, offering valuable insights into the vascular alterations associated with conditions such as DR, AMD, and vascular occlusions2.

Although OCTA has primarily been used to assess pathological eyes, recent studies have increasingly focused on healthy individuals to establish normative parameters such as vessel density (VD) in the superficial capillary plexus (SCP), deep capillary plexus (DCP), and choriocapillaris flow deficits (CCFDs)3,4,5,6,7. Accurate normative data are essential to understand age-related physiological changes and distinguish them from pathological alterations. Therefore, a reliable measurement of macular blood flow in healthy eyes is a prerequisite for the identification of disease-related abnormalities.

Despite its importance, previous studies have employed heterogeneous imaging protocols and image processing methods, making it difficult to establish standardized reference data for quantitative OCTA analysis3,4,5,6,7. Differences in devices, scan areas, segmentation algorithms, and binarization techniques have contributed to inconsistencies in the reported values. For example, owing to their lower scanning speed and reduced signal penetration, spectral-domain OCTA (SD-OCTA) systems often produce suboptimal image quality, resulting in inaccurate flow measurements. Additionally, small scan fields, such as 3 × 3 mm, may not fully capture the macular circulation, and significant projection artifacts from superficial retinal vessels can obscure deeper layers, such as the DCP and CC.

In contrast, swept-source OCTA (SS-OCTA) utilizes a longer central wavelength (approximately 1060 nm) compared to the 840 nm wavelength typically used in SD-OCTA. This longer wavelength minimizes light scattering by the retinal pigment epithelium (RPE) and allows for deeper tissue penetration. These features, combined with ultra-high scanning speeds (e.g., 100,000 A-scans per second), enable more precise visualization of the choroid and choriocapillaris, which are often obscured or attenuated in SD-OCTA images5.

Furthermore, many previous studies have relied on a single OCTA acquisition, which may not adequately reflect the dynamic nature of blood flow, potentially contributing to variability in measurements4,5,7.

To address these limitations, this study aimed to quantify age-related blood flow changes in the retina and choriocapillaris using a SS-OCTA device with a wide scan protocol. Advanced image-processing techniques were applied to enhance the accuracy of flow measurements for reliable, age-stratified normative data in a healthy Korean population.

Materials and methods

This retrospective, observational, comparative study was conducted at Pusan National University Hospital between January 2023 and June 2024. The study was approved by the Institutional Review Board of Pusan National University Hospital (approval no. 2508-008-154) and adhered to the tenets of the Declaration of Helsinki.

Healthy participants without any ocular disease were recruited. To minimize the influence of systemic hemodynamic factors, participants with a history of systemic diseases such as diabetes mellitus and hypertension were excluded. Among the eligible participants, eyes with a refractive error of 1.0 diopters or less and pseudophakic eyes with an axial length of 26.0 mm or less were included. Participants who had undergone cataract surgery at least 3 months prior to the study were eligible. In addition, we reviewed structural OCT B-scans to exclude eyes with pachychoroid features, such as pachyvessels or attenuation of the inner choroidal layers.

For this retrospective analysis, the electronic medical records and OCTA imaging data were accessed for research purposes on 13 August 2025. All data were fully de-identified prior to extraction and analysis, and the investigators did not have access to any personally identifiable information during or after data collection.

As this study involved retrospective review of anonymized medical records and imaging data without any direct intervention, the requirement for written informed consent was waived by the Institutional Review Board.

Imaging

All participants underwent SS-OCTA imaging with PLEX Elite 9000 (Carl Zeiss Meditec, Inc., Dublin, CA, USA) without dilation. The Zeiss PLEX Elite 9000 is an SS-OCTA device characterized by a central wavelength of 1060 nm and a bandwidth of 100 nm, with operational specifications including 100,000 A-scans per second, an A-scan depth of 3 mm, an axial resolution of 6.3 μm, and a transverse resolution of 20 μm. It also constructs angiography images by executing four repeated B-scans. The images were captured by a trained ophthalmic photographer.

Each eye was scanned twice using a 6 × 6 mm volume scan protocol with FastTrac motion correction, centering on the fovea. The OCTA machine’s integrated software automatically segmented the retinal layers—the SCP and DCP—and the CC using a multilayer segmentation (MLS) algorithm, producing en face OCTA images for each layer. For the MLS method, the superficial retinal layer is defined as the region extending from the segmented inner limiting membrane to the segmented outer boundary of the inner plexiform layer (IPL) minus 10 μm. Similarly, the deep retinal layer is defined as the area spanning from the segmented IPL minus 10 μm to the segmented outer boundary of the inner nuclear layer (INL) plus 30 μm. The CC was segmented to a thickness of 20 μm beneath Bruch’s membrane8. In the DCP and CC, projections from overlying vessels were excluded using the embedded functionality of the device.

Image analysis

OCTA images that exhibited significant artifacts or compromised image quality were excluded. The exclusion criteria included: (1) images with a signal strength index (SSI) less than 9, (2) imprecise segmentation of the retinal layers, (3) evident residual motion artifacts, (4) inadequate foveal alignment, and (5) signal disruption due to blinking. Ultimately, only good quality images with an SSI of 9 or 10 were included in the final analysis.

Quantitative OCTA flow parameters, including the CCFDs and VD in the SCP and DCP, were measured. The en face OCTA images of all layers were binarized using FIJI software (an expanded version of ImageJ), version 1.52u (National Institutes of Health, Bethesda, MD, USA; available at http://rsb.info.nih.gov/ij/index.html), following a previously described methodology9. In brief, SCP and DCP images were first duplicated, and one set was then binarized using two distinct thresholds (“hessian” and “Huang’s fuzzy,” respectively). The other set was processed using median local thresholding. The distinct threshold images were then combined to retain only the pixels in both images9,10. This was done to avoid the confounding effects of residual projection artifacts, in line with the methodology reported by Borrelli et al.11. Large superficial retinal vessels were visualized using the MaxEntropy threshold applied to the binarized SCP images. These thresholded images were then merged with the binarized images of the DCP and CC using the FIJI software, facilitating the identification and exclusion of large superficial retinal vessels9. The VD was quantified as the percentage of the area covered by blood vessels in the resultant image.

The VD of the SCP and DCP were measured, and the CCFDs were quantified following previous literature9,10,11,12,13. The resultant en face CC image (1024 × 1024 pixels) was binarized for quantitative analysis of the CCFDs using the Phansalkar method (radius; 15 pixels, 43.94 μm). In these binarized en face CC-OCTA images, white pixels indicate areas with blood flow, represented by values exceeding a predefined threshold. The black pixels correspond to the CCFD. These values were quantified as percentages of the total image area12. The resultant en face OCTA images are illustrated in Fig. 1.

Fig. 1
Fig. 1
Full size image

Representative flow chart of the optical coherence tomography angiography image processing of the SCP, DCP, and CCFD using FIJI software (an expanded version of ImageJ), version 1.52u (National Institutes of Health, Bethesda, MD, USA). The SCP and DCP images were first duplicated, and then one set was binarized using two distinct thresholds (“hessian” and “Huang’s fuzzy”, respectively). The other set was binarized using median local thresholding. The distinct threshold images were then combined, retaining only the pixels existing in both images. The MaxEntropy threshold was applied to binarized SCP images. The obtained thresholded images were then merged with binarized DCP and CC images for identification and removal of the superficial retinal large vessels. SCP, Superficial Capillary Plexus; DCP, Deep Capillary Plexus; CCFD, Choriocapillaris Flow Deficits.

Statistical analysis

Statistical analyses were conducted using IBM SPSS Statistics for Windows (version 23.0; IBM Corp., Armonk, NY, USA). The normality of the data distribution was assessed using the Shapiro–Wilk test. While DCP VD followed a normal distribution, SCP VD and CCFDs showed deviations from normality. However, given the large sample size (n = 351), Pearson’s correlation and linear regression analysis were performed. To identify factors associated with OCTA parameters (VD of the SCP and DCP, and CCFDs), univariable and multivariable linear regression analyses were performed. Age was treated as a continuous variable. The multivariable model was adjusted for potential confounding factors, including sex, axial length, SSI, and lens status. Pearson’s correlation coefficient was calculated to determine the linear relationship between age and flow parameters.

Additionally, for descriptive comparisons and to visualize age-related trends, participants were categorized into decade-based age groups: 20s (< 29 years), 30s (30–39 years), 40s (40–49 years), 50s (50–59 years), 60s (60–69 years), 70s (70–79 years), and 80s (80–89 years). Differences in baseline characteristics and flow parameters among these groups were assessed using one-way analysis of variance (ANOVA) followed by Tukey’s post-hoc test for continuous variables, and the chi-square test for categorical variables. Data are reported as mean ± standard deviation (SD) or standardized regression coefficients (β) with 95% confidence intervals (CI). A P-value of < 0.05 was considered statistically significant.

Results

Of 436 participants, 35 were excluded owing to poor image quality. Consequently, 351 participants of Korean ethnicity, representing 351 healthy eyes, were included. The demographic and baseline ocular characteristics of the participants stratified by age group are summarized in Table 1. The mean age ± SD was 54.4 ± 18.8 years (range, 15–87 years). The sample comprised 124 males and 227 females. A total of 168 eyes were pseudophakic. The mean SSI was 9.55 ± 0.50. The participants were categorized into the following age groups for analysis: 20s (n = 57), 30s (n = 37), 40s (n = 42), 50s (n = 44), 60s (n = 76), 70s (n = 76), and 80s (n = 19). All age-related changes in OCTA flow parameters are summarized in Table 2; Fig. 2. Representative images of the SCP, DCP, and CC are illustrated in Fig. 3.

Table 1 Demographics and ocular characteristics of study participants by age group.
Full size table
Table 2 Quantitative measurement of blood flow in the SCP, DCP, and CCFDs, categorized by age group.
Full size table
Fig. 2
Fig. 2
Full size image

Scatter plots illustrating the association between age and quantitative OCTA parameters. The solid red lines represent the linear regression fit, indicating the age-related trends for each parameter. (A) Superficial Capillary Plexus (SCP) vessel density shows a significant negative correlation with age (r = − 0.317, p < 0.001). (B) Deep Capillary Plexus (DCP) vessel density exhibits a significant positive correlation with age (r = 0.282, p < 0.001). (C) Choriocapillaris Flow Deficits (CCFDs) demonstrate a strong positive correlation with age (r = 0.726, p < 0.001). r, Pearson’s correlation coefficient.

Fig. 3
Fig. 3
Full size image

Representative binarized 6 × 6 mm en face images of SCP, DCP, and CCFD OCTA in each age group. 20s, subject age under 29 years; 30s, subject age 30–39 years; 40s, subject age 40–49 years; 50s, subject age 50–59 years; 60s, subject age 60–69 years; 70s, subject age 70–79 years; 80s, subject age 80–89 years; SCP, Superficial Capillary Plexus; DCP, Deep Capillary Plexus; CCFD, Choriocapillaris Flow Deficits; OCTA, optical coherence tomography angiography.

Changes in the quantitative flow parameters of the SCP by age groups

The reference group, comprising individuals in their 20s, had a mean SCP VD of 37.3%. For those in their 30s, the VD was 37.2%; those in their 40s, 37.0%; those in their 50s, 36.7%; those in their 60s, 36.6%; those in their 70s, 36.2%; and those in their 80s, 35.6%. There were no significant differences between individuals in their 20s and those in their 30–40s. Significant changes became apparent from individuals in their 50s onwards (p < 0.05 in all cases), showing a trend of stepwise decrease.

Changes in the quantitative flow parameters of the DCP by age groups

The mean DCP VD for the 20s reference group was 26.2%. For the 30s, it was 25.7%; for the 40s, 26.7%; for the 50s, 26.5%; for the 60s, 27.2%; for the 70s, 27.7%; and for the 80s, 28.5%. There were no significant differences between individuals in their 20s and those in their 30s, 40s, or 50s. Statistically significant changes began to appear from individuals in their 60s onwards (p < 0.05 for all cases), showing a trend of stepwise increase.

Changes in the quantitative flow parameters of the CCFDs by age groups

The mean CCFDs for individuals in their 20s were significantly lower at 17.74% compared to all other age groups: 30s (19.15%), 40s (22.04%), 50s (23.01%), 60s (25.03%), 70s (27.56%), and 80s (29.30%) (p < 0.05 for all). The CCFDs exhibited a trend of stepwise increase with age.

Factors associated with quantitative OCTA flow parameters

To identify and confirm factors affecting OCTA flow parameters, univariable and multivariable linear regression analyses were performed. First, Pearson’s correlation analysis showed that age was significantly correlated with all parameters: SCP VD (r = − 0.317), DCP VD (r = 0.282), and CCFDs (r = 0.726) (all p < 0.001).

In the univariable analysis, age was significantly associated with the VDs of the SCP and DCP, as well as CCFDs. Lens status and SSI were also identified as significant factors for certain parameters. In the multivariable linear regression analysis adjusting for sex, axial length, SSI, and lens status, age remained a significant independent predictor for all flow parameters. Age showed a significant negative association with SCP VD (β = − 0.018, 95% CI − 0.025 to − 0.011, p < 0.001) and a significant positive association with DCP VD (β = 0.029, 95% CI 0.017 to 0.041, p < 0.001) and CCFDs (β = 0.181, 95% CI 0.162 to 0.200, p < 0.001).

Lens status was also identified as a significant factor across all models. Phakic eyes showed significantly higher SCP VD (β = 0.62, p < 0.001), lower DCP VD (β = − 0.48, p = 0.012), and lower CCFDs (β = − 1.42, p < 0.001) compared to pseudophakic eyes. SSI showed a significant positive association only with SCP VD (β = 0.29, p = 0.022) in the multivariable model.

Discussion

This study highlights that blood flow in the SCP and CC decreased, while that in the DCP increased with aging in healthy Korean individuals. These flow parameters were derived from en face SS-OCTA images using a 6 × 6 mm scan, with averaging of two scans per individual and masking of overlying vessels in the DCP and CC. All participants had clear media, and all OCTA scans had an SSI of 9 or higher. These conditions improved the flow image quality and minimized OCTA-related artifacts. The improved OCTA image quality enabled accurate, reliable, and reproducible VD measurements in the SCP and DCP, as well as flow deficit assessments in the CC.

We performed a regression analysis on the entire cohort to validate the findings observed across age groups. Both age and lens status (phakia vs. pseudophakia) were significantly associated with OCTA flow parameters. While age was reaffirmed as a significant determinant of macular blood flow, the association with lens status likely reflects the significant age disparity between the phakic and pseudophakic groups.

Several histological studies have reported macular and vascular alterations that occur during aging. Retinal arteries undergo arteriosclerotic changes, such as intimal thickening, hyalinization, and reduced vascular elasticity, which reduce perfusion14,15. In OCTA imaging, the retinal artery is located within the layer corresponding to the SCP. CC loss has also been well-documented, with reduced capillary density observed in histological samples16,17. However, such histological changes cannot fully explain actual in vivo blood flow, and there is a paucity of histological studies regarding changes in the DCP.

For in vivo macular blood flow assessment, several studies have employed Doppler-based techniques, enabling the quantitative evaluation of blood flow. Color Doppler imaging and scanning laser Doppler flowmetry have shown that retinal artery blood flow decreases by approximately 6% to 11% per decade with aging18. In addition, color Doppler ultrasonography studies have demonstrated significant age-related reductions in blood flow in the central retinal and posterior ciliary arteries, along with increased resistive indices in older individuals19. These findings support the notion that aging is associated with a progressive increase in vascular resistance and a decrease in ocular perfusion, even in the absence of overt vascular disease. However, these techniques have limited ability to resolve and quantify blood flow within the specific vascular plexuses of the macula. OCTA allows for depth-resolved in vivo visualization of the retinal micro-vasculature and three-dimensional assessment of various layers, including the SCP, DCP, and CC. Using this advantage, macular blood flow has been quantitatively assessed in both healthy populations and pathological conditions.

An age-related increase in CCFDs, indicating decreased blood flow in the CC, is well documented in many OCTA studies involving healthy populations5,7,9,20, consistent with our findings. The reduction in blood flow begins in individuals in their 30s compared with those in their 20s and continues to decline with age, following an approximately linear trend. These findings are consistent with histological changes that demonstrate age-related loss of the CC. With respect to retinal blood flow, decreases in VD values of the SCP and DCP have been observed in several OCTA studies14,15,16,17. Shahlaee et al.3 reported age-related decreasing trends in both SCP and DCP perfusion. Su et al.4 noted that SCP begins to decline around the age of 40 years, and DCP begins to decline at approximately 35 years. Similarly, Park et al.7 found significantly lower VD values in individuals aged > 60 years than in those aged < 20 years. In the current study, we found an age-related decrease in blood flow in the SCP, but not in the DCP. In contrast, in the current study, the DCP flow increased. This discrepancy in the DCP findings may be attributable to methodological differences among the studies. Specifically, many previous OCTA studies used spectral-domain systems with shorter wavelengths and slower scan speeds, small scan areas such as 3 × 3 mm, and lacked masking of large superficial retinal vessels in deeper layer analyses, making them prone to projection artifacts. In addition, single-image acquisitions without averaging and variability in binarization methods or segmentation boundaries could have further reduced the measurement reliability. Such methodological limitations may have influenced their quantitative results and contributed to the differences in DCP trends compared with the present study. Quantitative blood flow measurements obtained using more accurate and methodologically robust techniques are more reliable, particularly in OCTA studies. Therefore, gaining a clear understanding of this phenomenon is crucial.

Anatomical validation studies combining OCTA and histology have confirmed that the DCP is a distinct planar network of capillaries located within the deeper inner retina. Hirano et al.21 demonstrated that the DCP forms a dense, flat plexus that is structurally separated from other capillary layers and is particularly vulnerable to hypoperfusion and ischemic injury. Campbell et al.22 provided detailed microvascular mapping showing that the DCP is supplied by vertical anastomoses from the superficial retinal vasculature and forms terminal loops with limited collateral flow, potentially explaining its sensitivity to age-related or pathological changes. These structural features suggest that age-related alterations in the DCP may reflect true microvascular remodeling. Changes in the DCP with aging can be inferred from pathological conditions. For example, increased VD in the DCP has been reported in patients with retinal vein occlusion (impaired blood circulation in the superficial large retinal vein)23. Similarly, An et al.24 observed that the capillary diameter in the DCP increased in patients with diabetic retinopathy. These findings support our observation of an age-related increase in the VD of the DCP, which may reflect compensatory changes in response to age-related arteriosclerotic alterations in the retinal arteries and decreased flow in the superficial retinal circulation. Furthermore, the earlier decrease in blood flow in the SCP (beginning in individuals in their 50s), followed by an increase in the DCP (beginning in individuals in their 60s), observed in this study, supports the hypothesis of compensatory changes in the DCP in response to reduced SCP circulation.

OCTA flow parameters can be affected by many factors, including motion artifacts, projection artifacts, banding artifacts, signal intensity, and segmentation errors25. Therefore, various technical methods have been developed to overcome the limitations of quantitative blood flow measurements using en face OCTA images. In this study, we applied several techniques and conditions to improve flow measurement. First, differences in the OCTA devices significantly affected the quantitative flow assessment. For example, SD-OCTA systems, which use an 850 nm wavelength, suffer from limited resolution and significant sensitivity roll-off beneath the RPE, which may lead to inaccurate values26. In contrast, SS-OCTA, which has a faster scan rate of 100,000 A-scans per second, allows for denser sampling and better image quality for a given acquisition time5. Second, we included all images with a high SSI (9 or 10). We previously reported that a small change in SSI had a significant impact on quantitative flow parameters, particularly SCP and CC, even with SS-OCTA devices. Third, two 6 × 6 mm en face flow images with high SSIs were obtained for all eyes, and they were averaged to generate a better resultant flow image. In this step, the averaging technique provided an additional advantage by mitigating the dynamic nature of capillary circulation, which would have affected the quantitative flow assessment. Fourth, because projection artifacts generate erroneous increases in flow values, we identified and removed superficial large retinal vessels from the DCP and CC images. Based on these image acquisition and processing methods, we obtained reliable quantitative flow values of the VD; it decreased in the SCP (beginning in individuals in their 50s), increased in the DCP (beginning in individuals in their 60s), and progressively decreased in the CC.

This study has some limitations, including its retrospective nature. First, the number of participants was not evenly distributed across the age groups. Second, smoking history, which may influence ocular blood flow, was not consistently recorded due to the retrospective design and thus could not be included in the analysis. Third, the use of different binarization methods, local versus global, may have led to variability in flow measurements27,28. Fourth, although we limited inclusion to eyes with an axial length ≤ 26.0 mm, variations in axial length remain a potential confounding factor in flow analysis. Nevertheless, we employed accurate and repeatable methods—including SS-OCTA with a 6 × 6 mm scan area, multiple image acquisitions with averaging, signal strength filtering, projection artifact masking, and combined local/global binarization techniques—resulting in reliable age-stratified flow parameters for the healthy Korean population.

In conclusion, we quantitatively evaluated OCTA-derived flow parameters across age groups in healthy Korean individuals. Blood flow decreased in the SCP (beginning in individuals in their 50s) and in the CC (beginning in individuals in their 30s), while DCP flow increased (beginning in individuals in their 60s) compared to those in their 20s. These findings provide important normative data for interpreting flow changes in retinal diseases and may enhance clinical OCTA applications.