Visualization of deep choroidal vasculatures and measurement of choroidal vascular density: a swept-source optical coherence tomography angiography approach

Choroid, a layer of vascular tissue, plays an important role in physiological and pathological process of human eyes. Visualization of choroidal vasculature is helpful in improving our understanding in a number of retinal and choroidal diseases. The swept-source optical coherence tomography (SS-OCT) technique permits deep choroid penetration. En-face OCT images of the choroid are useful for intuitive visualization of choroidal vasculatures and quantitative assessment of choroidal vascular density (CVD) [1, 2]. Quantification of CVD in deep choroid by en-face SS-OCT have been reported in healthy eyes [3], diabetic retinopathy [4], age related macular degeneration, polypoidal choroidal vasculopathy, and central serous chorioretinopathy (CSC) [5].

Optical coherence tomography angiography (OCTA), widely used for non-invasive imaging of retinal and choriocapillary microvasculatures [6, 7], has long been doubted for its value in deep choroid imaging. According to a previous report by Diaz et al. [8], intact retinal pigment epithelium (RPE) was regarded as a barrier for reliable detection of choroidal blood flow with OCTA. One previous study employed RTVue OCTA device for the quantification of deep choroidal vasculatures [9]. However, they used a built-in algorithm which was normally used for measuring choriocapillary flow density and was highly unlikely to quantify deep vasculatures correctly. Choi et al. [10] described a prototype SS-OCTA device at 1060 nm wavelengths with a 400 kHz A-scan rate in imaging Sattler’s layer of choroid. However, their equipment was commercially unavailable for wide application. Wang et al. [11] explored the ability of a commercially available SS-OCTA device to visualize and quantify the inner, middle, and outer choroidal vasculatures in a small series of eyes. However, there are still several gaps in current literature. First, the optimal depth for CVD measurements with SS-OCTA has not been identified. Second, the CVD measured with SS-OCTA has not been analyzed for its correlation with demographic factors. Third, the CVD measured with SS-OCTA has not been applicated in chorioretinal diseases.

To fill these gaps, we evaluated OCTA images at various choroidal depths with en-face OCT as reference standard, determined an optimal depth for visualizing choroidal vasculatures with OCTA, measured CVD with binarized OCTA image at the optimal depth, studied the OCTA derived CVD for its correlation with demographic factors in healthy subjects, and tested its usefulness in a series of CSC eyes. The purpose of this study is to improve our understanding in SS-OCTA for its capability in displaying deep choroidal vasculatures and quantifying deep choroidal vascularity.

Ethics committee of Peking Union Medical College Hospital approved this observational case series study. All participants gave informed consents, and we conducted the study in accordance with the ethical tenets of the Declaration of Helsinki.

We included 116 eyes of 58 healthy participants for SS-OCTA scans. After excluding 2 eyes with blink artefact, 3 eyes with banding artefacts, 6 eyes with low image quality, and 14 eyes with at least two motion artefacts wider than primary branch vein diameter, we included 91 eyes from 51 healthy subjects (16 males and 35 females) for analysis. We also included 22 eyes with CSC (14 males and 8 females). All of the 5 acute CSC eyes and 13 of the 17 chronic CSC eyes had subretinal fluid overlying the scanned area. None of the CSC eyes had pachychoroid neovasculopathy. The characteristics of all included healthy eyes, CSC eyes and matched controls are shown in Table 1.

By systematically evaluating deep choroidal OCTA images in healthy and CSC eyes, the current study not only suggested the capability of SS-OCTA in displaying choroidal vessels, but also identified CVD at 100 μm beneath BM as a useful OCTA parameter. This parameter was stable across various age and axial length groups, and was valuable in reflecting choroidal vascular status.

There had been an impression that OCTA might not be suitable for imaging the choroid beneath choriocapillaris. According to Diaz et al., motion contrast signals produced bright vessel outlines only in areas with overlying RPE atrophy, and dark signals beneath intact RPE was deemed as absence of blood flow [8]. According to our study, within a range of choroidal depths, dark signals in OCTA images corresponded to vessel lumen in OCT B-scan, and deep choroidal vessels manifested as dark stripes in OCTA images. This phenomenon could be related with several underlying mechanisms including fringe washout effect, signal attenuation, and threshold masking. Fringe washout is the signal loss due to high blood flow velocity exceeding system imaging speed [16‐18]. Attenuation is the signal blockage by RPE which reduces a significant portion of light entering choroid [13, 19]. Threshold masking is a process of signal removal in OCTA image building which ensures that only valid motion contrast signals were included and potential noises were excluded [19]. In our study, OCTA images and en-face OCT images displayed choroidal vessels as similar dark stripes within a range of choroidal depths. Therefore, the notion that OCTA cannot be used for deep choroid imaging might be biased and incomprehensive.

We identified 100 μm beneath BM as the optimal depth for the current SS-OCTA device in depicting choroidal vessels for two reasons. First, in all healthy subjects, OCTA images at 100 μm beneath BM were able to display choroidal vessels appropriately. Second, OCTA and en-face OCT images had greatest similarity at 100 μm beneath BM in almost all eyes with various choroidal thickness. The above facts suggested that capability of SS-OCTA in depicting choroidal vessels was likely to be determined by the intrinsic penetration of the SS-OCTA device, rather than the imaging preference of specific choroidal layers. Morphological studies from autopsy eyes showed that the thickness of BM and choriocapillaris ranged from 2.0 μm to 4.7 μm and from 6.5 μm to 9.8 μm, respectively [20]. In vivo investigation showed that subfoveal thickness of Sattler’s layer was 87 ± 56 μm [21]. We thereby inferred that the choroidal OCTA image at 100 μm beneath BM was likely to describe vasculatures around the boundary between Sattler’s and Haller’s layer.

Because deep choroidal vessels were displayed as dark stripes in OCTA images, the quantification of vascular density should be conducted by manual calculation of the proportion of dark signals instead of the automatic measurement of area of bright pixels generated by the built-in software. Chan et al. reported their OCTA measurements of CVD in deep choroid of CSC patients before and after photodynamic therapy (PDT). They agreed that the dark stripes in deep choroidal OCTA images reflected blood vessels. However, they measured CVD with a built-in algorithm which calculated the proportion of bright signals, normally used for measuring choriocapillary blood flow [9]. They reported a reduction in choroidal vessel diameter but an increase in vascular density after PDT, which was self-contradictory and inconsistent with a number of previous observations on PDT treatment [22, 23].

We revealed that CVD measured by SS-OCTA at 100 μm beneath BM is a useful parameter for quantifying choroidal vascular status. First, no manual calibration of SCT is needed in calculating CVD. The fixed depth, or 100 μm beneath BM, for image generation is likely to minimize potential measurement bias among different technicians. Second, this parameter was correlated with choroidal thickness and was relatively stable across various age and refraction groups. One previous study by Fujiwara et al. [3] utilized en-face OCT for the measurement of CVD at 50% SCT beneath BM and reported that CVD was significantly correlated with SCT but not with AL, which is similar to our findings. However, they reported that CVD measured with en-face OCT was negatively correlated with age, which might be due to relatively large sample size and broad sample age. Third, the value of this parameter in reflecting choroidal vascularity was also tested and verified in a series of CSC eyes. A plenty of studies have shown increased choroidal blood flow or dilated choroidal vessels in CSC patients [12, 24]. In our study, CVD measured by SS-OCTA at 100 μm beneath BM was increased in CSC eyes, consistent with previous findings. One may argue that the increased CVD value in CSC eyes might be related with thick choroid. To eliminate the potential influence from SCT, we compared CVD between groups after adjusting for SCT and the difference remained significant. One may also argue that the presence of subretinal fluid or pigment epithelium detachment may produce shadowing artifacts in OCTA images of underlying structures and may result in an overestimation of CVD in deep choroid. However, the round dark “shadow” produced by subretinal fluid was almost invisible in deep choroid, because there was mild, universal, and uniform signal loss in deep choroid. Based on the above considerations, we believe that CVD at 100 μm beneath BM measured by OCTA is a potentially valuable and convenient parameter in future studies.

Despite the potential value of SS-OCTA in visualizing choroidal vessels and measuring CVD at 100 μm beneath BM, we have to admit that the capability of SS-OCTA in visualizing deep choroidal vasculatures is limited in extremely deep choroid where signal attenuation is significant. Signal attenuation in OCTA images results in more dark signals than actual and thereby causes overestimation of CVD [11, 18].

The current SS-OCTA device was useful in visualizing choroidal vasculatures within a range of choroidal depths. Notably, this SS-OCTA device was able to measure CVD at 100 μm beneath BM, which was a useful parameter in quantifying choroidal vascular status. The study improved our understanding in OCTA technique and expanded its application in deep choroid imaging.