The choroidal vascular system represents the largest blood supply for the outer retinal layers. Furthermore, it is essential for the removal of metabolites from the retinal pigment epithelium (RPE) and thus directly linked to the maintenance of the retinal function and aging processes of the posterior eye.1 Therefore, choroidal tissue is involved in the development of several pathologic alterations in the retina such as age-related macular degeneration (AMD) or diabetic retinopathy (DR), which are among the leading causes of permanent vision impairment in industrialized countries.2–4
It is assumed that changes in the choriocapillaris (CC) emerge prior to the onset of clinical symptoms such as irreversible changes in the RPE and retinal layers.2,5 Studies confirmed that in AMD, the formation of macular drusen might follow a pattern of underlying choroidal nonperfusion.6,7 Therefore, an early detection and preventive treatment of retinal diseases is of great interest, and quantitative parameters to objectify the choroidal perfusion are required.
Optical coherence tomography angiography (OCTA) represents a noninvasive imaging technique to obtain in vivo information on CC blood flow. Although it is not possible to visualize dye leakage of vessels as can be seen in fluorescein or indocyanine green angiography, it does allow for a higher contrast and axial resolution imaging of the choroidal vascular plexus.8 However, lateral resolution suffers from the relatively large illumination spot size, so the choroid capillaries do not represent as clearly differentiable vessels on OCTA.4 System-integrated algorithms of common OCTA devices give information on CC perfusion, although the evaluation of the obtained parameters should always be considered critically with regard to artifacts.
The aim of this study is to investigate whether a correlation between choroidal blood flow density and age can be detected in healthy subjects, analyzing directly extracted values from a commercial spectral-domain OCTA (SD-OCTA) device.
Patients and Methods
The study population included 94 volunteer subjects with a total of 183 eyes. These included employees of the clinic and the medical faculty, as well as their relatives. Besides OCTA, each subject underwent funduscopy and ocular axial length (AL) measurement (IOL Master; Carl Zeiss Meditec, Jena, Germany). Eyes with ALs of 21.9 mm or less and 24.6 mm or greater were excluded. Inclusion criteria was a manufacturer signal strength index (SSI) of 55 or greater based on the inclusion criteria of Spaide.4 Exclusion criteria were any history of vitreoretinal pathologies or surgery, as well as general disorders with potential affection on the eye such as arterial hypertension, diabetes, or autoimmune diseases. Cataract surgery in the medical history was not considered exclusion criteria. Eyes with cataract were included if the SSI was still 55 or greater. Any severe motion, projection, or vitreous shadowing-related artifacts led to an exclusion of the respective eye. Five eyes of five subjects were excluded due to the following reasons: three eyes had an SSI less than 55 due to cataract, one eye had an epiretinal membrane (ERM), and one eye had vitreoretinal surgery with vitrectomy due to retinal detachment. The study was approved by the ethics committee of the Technical University of Munich and was performed in accordance with the Declaration of Helsinki. Each subject gave his or her written consent.
OCTA was performed without drug dilatation in a darkened room using the AngioVue software of the Avanti RTVue XR SD-OCT device (Optovue, Fremont, CA). En face projection macular frame scans of 3 mm × 3 mm in size were obtained with a rate of 70,000 A-scans per second, containing 304 × 304 A-scans with two consecutive B-scans each. The split-spectrum amplitude-decorrelation algorithm (SSADA), which is a mathematical algorithm splitting the OCTA signal in several single frames to increase the number of usable images, was implemented to minimize the signal-to-noise ratio of the chorioretinal flow signal.8–10 The retinal and choroidal layers were automatically segmented by the manufacturer software in different en face projection images.
The principle of the software segmentation algorithm relies on prominent anatomical structures such as the internal limiting membrane (ILM) and the RPE. For the superficial retinal capillary plexus (SCP), the software detected perfusion within an inner boundary of −3 μm to −15 μm below the ILM. The CC flow signal was acquired within an inner and outer boundary from −31 μm to −60 μm below the RPE.
Each 3 mm × 3 mm en face image was checked for record quality regarding artifacts, either due to motion or projection, by two independent observers. Any image with a segmentation error greater than 3 lines was excluded. Furthermore, the manufacturer SSI was defined to be 55 or greater in each evaluated image.
To determine the subfoveal choroidal thickness (SCT), a 12-mm single-line widefield OCT scan (1,024 × 1) was performed with enhanced high-definition (HD) imaging using the AngioVue software.
In the first step, we determined the center of the foveal avascular zone (FAZ) using en face 3 mm × 3 mm scans and its corresponding cross-sectional B-scans of the SCP. Therefore, we manually adjusted the intersection of the red and green lines in the middle of the FAZ (Figure 1A). The software's automatic segmentation function was used to switch into the CC layer within the same OCTA scan, and the manually set intersection remained in the center of the FAZ (Figure 1B). The manufacturer software provides an integrated tool for the CC layer, which allows for the measurement of CC blood flow area (mm2) within a selected area, defined as a circle of 3.14 mm2 (Figure 1C). The CC flow area was automatically calculated by the software and was defined as: CC flow area (mm2 = selected area (mm2) - area of absent flow signal (mm2).
(A) The superficial retinal capillary plexus, where the green and red lines intersect in the center of the foveal avascular zone (FAZ). (B) Image shows the same intersection but in the choriocapillary layer. (C) The selected area within a circle of 3.14 mm2 from the center of the FAZ.
The CC flow area describes an area with a certain vessel density, where flow signal has been recorded as movement in the moment of image acquisition.
We manually positioned the center of the selected area in the center of the FAZ, where the intersection of the red and green lines was set before. Images could be exported as so-called raw images. Areas with absent flow signal appear in black, whereas areas with flow signal appear in white (Figure 2).
Choriocapillaris raw images of the selected area. The area of flow signal (white) is larger in younger eyes.
To ensure accordance of the measurements, all determination of flow area values was performed by two independent investigators. Using a Bland-Altman blot, the mean difference between the two independent measurements of the CC flow area was 0.0001 ± 0.00268 mm2 with limits of agreement ranging from −0.00481 to 0.00494 mm2. There was not any proportional bias. Since the deviations in the measurements range within ±0.005 mm2, the calculated limits were acceptable for highly according values.
In order to get a more usable parameter, we set the CC flow area (mm2) in relation to the selected area (mm2), obtaining a ratio that describes the obtained flow signal as CC flow density per selected area in percentage: CC flow density (%) = CC flow area (mm2)/selected area (mm2).
Statistical analysis was performed with SPSS software version 25.0 (SPSS, Chicago, IL). Generalized estimating equations (GEEs) were calculated for analysis of the correlation between OCTA parameters CC flow density and SCT, each with the dependent variable of age. Inter-eye correlations were adjusted if both eyes of one subject were included in the statistical analysis.
Values are given as mean values with standard deviations (SD) if not stated otherwise. For all tests, P<.05 was considered to be significant.
A total of 183 eyes of 94 subjects were enrolled in this study. Five eyes of five subjects were excluded due to cataract-associated reduced signal strength(n = 3), presence of an ERM (n = 1), and retinal detachment in medical history (n = 1). Sixty-six subjects were female (70.21%). Patients' ages ranged from 21 years to 82 years (mean: 43.43 ± 17.63). The mean AL of the globe was 23.65 mm ± 0.78 mm. The mean CC flow density was 62.71% ± 1.01%. The mean manufacturer SSI was 72.99 ± 6.86. Table 1 shows the distribution of age in decades.
Distribution of Age in Decades With Mean Values and Standard Deviation
Table 2 shows GEE analysis of the correlation between CC flow density and age as a dependent variable. A significant negative correlation (P < .001) between both variables with a mean yearly flow decrease of 0.026% for each eye could be detected. Figure 3 shows this correlation with CC flow density on the Y axis and age on the X axis in a simple scatter plot. Each dot in the chart represents one single eye. The fit line is decreasing with increasing age.
Generalized Estimated Equation of Mean CC Flow Density Withthe Dependent Variable of Age
Mean choriocapillaris flow density in percent correlated with the dependent variable of age. There is a significant negative correlation between flow density and advancing age (P < .001). The mean yearly regression per eye was 0.026%.
SCT was obtained in 139 eyes. Forty-four eyes were excluded from analysis due to inaccurate presentation of the choroidal outer boundary, which was not clearly separable from the sclera in enhanced HD imaging. The mean SCT was 290.32 μm ± 64.70 μm. Correlation between SCT and age was not significant (P = .069) (Figure 4). Correlation between SCT and CC flow density was not significant either (P = .810).
Mean subfoveal choroidal thickness in microns correlated with the dependent variable of age. Correlation did not reach a level of significance (P = .069).
In the present study, we show that alterations in the subfoveal CC flow density are significantly negative correlated with increasing age in healthy subjects. Thus, our results are in accordance with the results of other studies using diverse techniques working on CC flow characterization. In earlier studies comparable results have been shown using both SD-OCTA and swept-source OCTA (SS-OCTA) devices.11,4 Spaide showed measurable alterations in CC blood flow following a power law distribution in association with increasing age by investigating areas of absent flow signal, called flow voids. He described that larger regions of absent flow signal more likely occur in older subjects, associated with arterial hypertension and also with late AMD in the fellow eye.4 But similar results could also be found in studies using techniques other than OCTA. Straubhaar et al. investigated choroidal microcirculation using laser Doppler flowmetry and detected a decreased flow velocity with advancing age in healthy subjects.12 However, scientific approaches on CC flow characterization are complex. From functional side, the CC flow follows a segmental, lobular pattern.13 It is known that age-related drusenoid alterations and choriocapillary ghost vessels cause a lack of blood flow in the affected segment, but not mandatory in the whole lobule.14,4 Histological findings from Mullins et al. described a positive correlation of the number of choriocapillary ghost vessels with increasing age in early AMD.15 Biesemeier et al. showed similar histological results with choriocapillary loss in both early AMD and age-appropriate controls, although the number of ghost vessels was higher in eyes with drusenoid alterations.5 This leads to the assumption that the CC blood flow could have an influence on the genesis of retinal diseases or might even be a mirror for the general microvasculature status in the human body.
In our study, we set a manufacturer software-automated value of flow area (mm2) in proportion to a manually selected subfoveal area (mm2). The obtained flow values were directly extracted from the built-in software without any further processing. Although software-automated values could be easy to obtain in clinical practice, they do not analyze the actual structure and pattern of the flow characteristics.4 The extent to which the partial information obtained via the CC vascular density measurement can actually be of benefit for early detection of retinal diseases such as AMD in clinical context remains in need of further assessment.
Another limitation of OCTA interpretation remains the presence of artifacts. Retinal pathologies have a high susceptibility to segmentation artifacts, not only in the layers in which they actually appear.16 But even in healthy eyes, automated segmentation mainly depends on the signal intensity of the most prominent anatomical structures, which are mainly the ILM and the RPE. All layers in between or even below, like the CC, tend to have a significantly higher probability to produce segmentation errors due to their limited signal intensity in SD-OCTA.17 However, studies have shown that SD-OCTA represents a useful method to obtain images of the CC layer in normal eyes with no severe RPE elevation.4,12,18 On the other hand, the SS-OCTA has demonstrated excellent repeatability of CC measurements even under drusenoid lesions. Therefore, the SS-OCTA has a clear advantage over the SD-OCTA in eyes with morphological RPE alterations regarding image artifacts.19 Although our inclusion criteria minimized the likelihood of artifacts caused by errors of automated software segmentation algorithms, the actual presence of artifacts, especially projection artifacts due to overlaying retinal vessels, might not be adequately accounted for with the software-automated value. This could potentially confound measurements of the CC flow density.
However, in addition to artifact-related limitations, there are other challenges in assessing choroidal blood flow. For now, no standardized inter-device method for flow determination has been described in the literature. In contrast to the retinal vascular plexus, which allows a contrast-rich presentation of individual vessels, the CC in commercial OCTA devices presents itself as a reticulate pattern, which does not allow the differentiation of individual capillary blood flow in their resolution.4
Measurements of the SCT in our study showed similar results compared with reproducible SCT measurements of a study with a large cohort (n = 3,233) with ages ranging from 50 years to 93 years.20 The SCT in our study decreased with advancing age, but without reaching a level of significance. These results are also in accordance with the results of studies describing correlation of SCT with age.20,21
We conclude that SD-OCTA yields a reproducible and noninvasive assessment of CC flow density in healthy subjects. Our results show a significantly negative correlation between the CC flow signal and advancing age, which is in accordance to other studies using diverse materials and methods. Hence, the CC flow density obtained with SD-OCTA might be a quantitative parameter for the choroidal perfusion although software-automated values of the vascular density must not be able to adequately analyze the CC flow characterization due to the lack of information on structure and pattern.4 On the other hand, it might be a useful parameter to detect and observe early choroidal alterations prior to their occurrence in both RPE and retina, where they ultimately show clinical signs and cause visual impairment.5 Further investigations are necessary to create a norm-scaled index with an accurate intra- and inter-device age-appropriate classification.
- Schmetterer L, Kiel JW. Ocular Blood Flow. Berlin, Germany: Springer Berlin Heidelberg; 2012. doi:10.1007/978-3-540-69469-4 [CrossRef]
- McLeod DS, Grebe R, Bhutto I, Merges C, Baba T, Lutty GA. Relationship between RPE and choriocapillaris in age-related macular degeneration. Invest Ophthalmol Vis Sci. 2009;50(10):4982–4991. doi:10.1167/iovs.09-3639 [CrossRef]
- Hidayat AA, Fine BS. Diabetic choroidopathy. Light and electron microscopic observations of seven cases. Ophthalmology. 1985;92(4):512–522. doi:10.1016/S0161-6420(85)34013-7 [CrossRef]
- Spaide RF. Choriocapillaris flow features follow a power law distribution: Implications for characterization and mechanisms of disease progression. Am J Ophthalmol. 2016;170:58–67. doi:10.1016/j.ajo.2016.07.023 [CrossRef]
- Biesemeier A, Taubitz T, Julien S, Yoeruek E, Schraermeyer U. Choriocapillaris breakdown precedes retinal degeneration in age-related macular degeneration. Neurobiol Aging. 2014;35(11):2562–2573. doi:10.1016/j.neurobiolaging.2014.05.003 [CrossRef]
- Sarks SH, Arnold JJ, Killingsworth MC, Sarks JP. Early drusen formation in the normal and aging eye and their relation to age related maculopathy: A clinicopathological study. Br J Ophthalmol. 1999;83(3):358–368. doi:10.1136/bjo.83.3.358 [CrossRef]
- Lengyel I, Tufail A, Hosaini HA, Luthert P, Bird AC, Jeffery G. Association of drusen deposition with choroidal intercapillary pillars in the aging human eye. Invest Ophth Vis Sci. 2004;45(9):2886–2892. doi:10.1167/iovs.03-1083 [CrossRef]
- Spaide RF, Fujimoto JG, Waheed NK. Image artifacts in optical coherence tomography angiography. Retina. 2015;35(11):2163–2180. doi:10.1097/IAE.0000000000000765 [CrossRef]
- Jia Y, Tan O, Tokayer J, et al. Split-spectrum amplitude-decorrelation angiography with optical coherence tomography. Opt Express. 2012;20(4):4710–4725. doi:10.1364/OE.20.004710 [CrossRef]
- Huang D, Jia Y, Gao SS, Lumbroso B, Rispoli M. Optical coherence tomography using the optovue device. Dev Ophthalmol. 2016;56:6–12. doi:10.1159/000442770 [CrossRef]
- Lauermann JL, Heiduschka P, Nelis P, et al. Comparison of choriocapillaris flow measurements between two optical coherence tomography angiography devices. Ophthalmologica. 2017;237(4):238–246. doi:10.1159/000464355 [CrossRef]
- Straubhaar M, Orgül S, Gugleta K, Schötzau A, Erb C, Flammer J. Choroidal laser Doppler flowmetry in healthy subjects. Arch Ophthalmol. 2000;118(2):211–215. doi:10.1001/archopht.118.2.211 [CrossRef]
- Hayreh SS. Segmental nature of the choroidal vasculature. Br J Ophthalmol. 1975;59(11):631–648. doi:10.1136/bjo.59.11.631 [CrossRef]
- Curcio CA, Messinger JD, Sloan KR, McGwin G, Medeiros NE, Spaide RF. Subretinal drusenoid deposits in non-neovascular age-related macular degeneration: Morphology, prevalence, topography, and biogenesis model. Retina. 2013;33(2):265–276. doi:10.1097/IAE.0b013e31827e25e0 [CrossRef]
- Mullins RF, Johnson MN, Faidley EA, Skeie JM, Huang J. Choriocapillaris vascular dropout related to density of drusen in human eyes with early age-related macular degeneration. Invest Ophthalmol Vis Sci. 2011;52(3):1606–1612. doi:10.1167/iovs.10-6476 [CrossRef]
- Ghasemi Falavarjani K, Al-Sheikh M, Akil H, Sadda SR. Image artefacts in swept-source optical coherence tomography angiography. Br J Ophthalmol. 2017;101(5):564–568. doi:10.1136/bjophthalmol-2016-309104 [CrossRef]
- Niu S, Chen Q, De Sisternes L, Rubin DL, Zhang W, Liu Q. Automated retinal layers segmentation in SD-OCT images using dual gradient and spatial correlation smoothness constraint. Comput Biol Med. 2014;54:116–128. doi:10.1016/j.compbiomed.2014.08.028 [CrossRef]
- Zhang Q, Chen CL, Chu Z, et al. Automated quantitation of choroidal neovascularization: A comparison study between spectral-domain and swept-source OCT angiograms. Invest Ophthalmol Vis Sci. 2017;58(3):1506–1513. doi:10.1167/iovs.16-20977 [CrossRef]
- Zhang Q, Zheng F, Motulsky EH, et al. A novel strategy for quantifying choriocapillaris flow voids using swept-source OCT angiography. Invest Ophthalmol Vis Sci. 2018;59(1):203–211. doi:10.1167/iovs.17-22953 [CrossRef]
- Shao L, Xu L, Chen CX, et al. Reproducibility of subfoveal choroidal thickness measurements with enhanced depth imaging by spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci. 2013;54(1):230–233. doi:10.1167/iovs.12-10351 [CrossRef]
- Jonas JB, Forster TM, Steinmetz P, Schlichtenbrede FC, Harder BC. Choroidal thickness in age-related macular degeneration. Retina. 2014;34(6):1149–1155. doi:10.1097/IAE.0000000000000035 [CrossRef]
Distribution of Age in Decades With Mean Values and Standard Deviation
|Age Range (Years)||N||Mean Age (Years)|
|21 to 82||94||43.43 ± 17.63|
|21 to 29||30||25.63 ± 2.82|
|30 to 39||18||33.94 ± 2.30|
|40 to 49||12||44.92 ± 3.73|
|50 to 59||14||53.57 ± 2.65|
|60 to 69||10||64.40 ± 2.63|
|70 to 82||10||76.90 ± 3.32|
Generalized Estimated Equation of Mean CC Flow Density Withthe Dependent Variable of Age
|95% Wald Confidence Interval||Hypothesis Test|
|Parameter||B||Standard Error||Lower||Upper||Wald Chi-Square||df||Sig|