Introduction

Hyperreflective spots or foci are small, well-circumscribed punctiform lesions with higher reflectivity similar to that of nerve fibre layer and have a scattered distribution within all layers of the neurosensory retina, on optical coherence tomography (OCT) [1]. Hyperreflective foci have been demonstrated in various retinal diseases [2,3,4,5,6,7,8,9,10,11], including diabetic macular oedema, retinal vein occlusion, age-related macular degeneration, polypoidal choroidal vasculopathy, central serous chorioretinopathy, and retinitis pigmentosa. Vujosevic et al. [12] reported that hyperreflective spots were also present in the normal retina.

In addition to the retinal layers, hyperreflective foci have been observed in various locations, such as the vitreous, subretinal, and subretinal pigment epithelium spaces, with diverse distribution patterns according to different chorioretinal diseases [7, 13,14,15,16]. In the eyes with serous retinal detachment type of macular oedema caused by diabetic retinopathy or retinal vein occlusion, hyperreflective foci were found to be attached underneath the outer border of the detached neurosensory retina [7, 15, 16], while they were clustered around the leaking point, where confirmed using fluorescein angiography of the eyes with central serous chorioretinopathy [17]. Recently, several studies have demonstrated the presence of hyperreflective choroidal foci (HCF) [18,19,20].

The advancement of spectral-domain OCT (SD-OCT) technology has enabled acquisition of detailed retinal images for various diseases. Most studies have identified hyperreflective foci on the B-scan of SD-OCT [1,2,3,4,5,6,7,8,9,10,11,12, 15,16,17,18]. Only a few studies have investigated hyperreflective foci using en face OCT images [12, 21, 22]. In addition to the enhanced depth imaging technique, the introduction of swept-source OCT (SS-OCT) has allowed acquisition of detailed choroid images [23,24,25]. The choroidal thickness and choroidal vascularity can be measured in vivo using SS-OCT. However, the factors affecting distribution of HCF in the normal eye are not yet clear.

In this study, we hypothesized that choroidal structure may affect the presence and distribution of HCF in the normal choroid. We measured HCF quantitatively using en face SS-OCT and determined the factors contributing to the distribution of HCF in normal eyes.

Methods

This retrospective study was approved by the Institutional Review Board of Korea University Anam Hospital, and all research and data collection procedures were performed in accordance with the tenets of the Declaration of Helsinki. We searched the SS-OCT database of Korea University Anam Hospital between March 2016 and December 2020 and included patients with a normal fundus. We reviewed their medical history and included only healthy eyes. The patients underwent a complete ophthalmic examination, including best corrected visual acuity measurement, biomicroscopy, intraocular pressure assessment, fundus photography, SS-OCT, and SS-OCT angiography (SS-OCTA). We excluded the eyes with pigment epitheliopathy or drusen and the eyes with high myopia (axial length ≥ 28.0 mm), diabetic retinopathy, retinal vascular diseases, a history of uveitis, central serous chorioretinopathy, ocular surgery except cataract surgery, and cataract surgery within 6 months before measurement. The eyes with poor quality SS-OCT and SS-OCTA images (image quality score in arbitrary units < 60) were excluded.

We acquired all SS-OCT and SS-OCTA images using DRI OCT Triton (Topcon Corp., Tokyo, Japan; software version 10.10), which operates at an A-scan speed of 100,000 per second and uses a central wavelength of 1050 nm. Under ideal scan conditions, this instrument had an axial resolution of 8 µm and a transverse resolution of 20 µm for examining tissues.

Selection of choroidal slab

A three-dimensional (3D) macular cube scan was acquired over a 3000 × 3000-μm area centred on the fovea during the SS-OCTA default scanning protocol. Each B-scan was produced from 320 A-scans, and each cube scan composed 320 horizontal B-scans. A 3000 × 3000-μm OCTA protocol had four repetitions of the B-scan at each position. The automated instrument’s algorithm was used to segment all retinal layers and Bruch’s membrane. For further imaging analysis, a 3000 × 3000-μm en face structural SS-OCT image, consisting of 320 × 320 pixels, was generated from this macular cube scan using the built-in software. A single experienced retinal specialist (J.O.) assessed the segmentation error, and high-quality images without significant segmentation errors were used for further analysis. To reduce the projection artefact of overlying retinal vessels and the effect of underlying large choroidal vessels, we selected a choroidal layer slab between 5.2 and 20.8 μm below the Bruch’s membrane. This choroidal layer slab improved the HCF visibility [26].

HCF on en face SS-OCT

HCF were defined as spots with reflectivity beyond two standard deviations (SDs) of the mean reflectivity in the selected slab (Fig. 1) on en face structural SS-OCT images. HCF were determined on a previously selected choroidal layer slab (Fig. 2). After the display setting of the colour of SS-OCT was set to reverse black/white, segmentation lines for the en face choriocapillaris layer image were moved to 5.2 μm and 20.8 μm below the Bruch’s membrane to obtain the choroidal layer slab of the en face image for HCF measurement. The en face images were exported and analysed using ImageJ (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health) software. After conversion of reversed images to 8-bit grayscale, the measurement scale was set from 320 pixel to 3000 μm. Thereafter, the image’s histogram was analysed, and the reflectivity threshold level for the HCF was set at two SDs below the mean of the reflectivity from the histogram. The image was processed using a binarization tool with a watershed function to separate overlapping spots. HCF were measured using an analysed particle tool to trace the boundary of the hyperreflective foci. To prevent counting noise signal as HCF, we defined the minimum diameter and area of HCF as 20 μm and 314 μm2, respectively [18]. Any particle of area from 314 μm2 to infinity on the ImageJ tool was noted. The number, total area, mean size, and circularity of HCF were calculated.

Fig. 1
figure 1

En face swept-source optical coherence tomographic images of hyperreflective choroidal foci

Full size image
Fig. 2
figure 2

The partial choroidal layer slab (lower left) including the data from a Bruch’s membrane-based 15.6-μm-thick slab positioned 5.2 μm under the Bruch’s membrane was obtained from en face SS-OCT images retrieved during SS-OCTA scanning performed in a 3000 × 3000-μm area centred on the macula (upper). Hyperreflective choroidal foci were defined as spots with reflectivity beyond two standard deviations of the mean reflectivity in the selected slab (lower middle). To prevent counting noise signal, we selected the particles with area ≥ 314 μm2 (lower right). SS-OCT, swept-source optical coherence tomography; SS-OCTA, swept-source optical coherence tomography angiography

Full size image

Choroidal thickness and choroidal vascularity index on SS-OCT

Subfoveal choroidal thickness was measured using SS-OCT images. SS-OCT images at the macula were obtained with a three-dimensional horizontal volume scan plus 5-line raster cross protocol, which was provided by the manufacturer. Among five horizontal and five vertical line scans, each averaged from 96 B-scan images, we measured the subfoveal choroidal thickness using a 9-mm horizontal line scan passing through the foveal centre. Subfoveal choroidal thickness was measured manually from the retinal pigment epithelium (RPE)/Bruch’s membrane complex to the choroidoscleral interface at the centre of the fovea using the calliper tool of the instrument’s image viewer programme, as previously reported [27, 28]. Two independent observers (Y.H.K. and J.O.) blinded to patient information independently measured subfoveal choroidal thickness, and the mean value was used for further statistical analysis.

Choroidal vascularity index (CVI) was defined as the ratio of the luminal area to the total choroidal area [29]. We measured CVI using the same 9-mm horizontal line scan image as used for the measurement of subfoveal choroidal thickness using our previously reported method (Fig. 3) [28, 30].

Fig. 3
figure 3

Measurement of choroidal vascular index (CVI). CIV was measured on the 9-mm horizontal line scan passing through the fovea, which produced from the averaging of 96 B-scans. Using a polygon tool, the total choroidal area with a width of 1500 µm, centred on the fovea, was manually selected as the region of interest (ROI). The inner and outer boundaries of the total choroidal area were traced along the RPE/Bruch’s membrane complex and the choroidoscleral junction, respectively. (upper left). After conversion to 8-bit grayscale, image binarization for demarcation of the choroidal vascular and stromal areas was performed using and auto local thresholding tool with the Niblack method (upper right). The binarized image was converted back to a red, green, blue (RGB) image, and the colour threshold tool was applied to determine the dark pixel, which represent the luminal area (lower left), and the luminal area was determined using merged images with initially selected the region of interest (lower right). Thereafter, the stromal area was calculated by subtracting the luminal area from the total choroidal area, and CVI was calculated by dividing the luminal area by the total choroidal area

Full size image

Choriocapillaris flow on SS-OCTA

The vascular density (VD) and flow void area of the choriocapillaris layer were measured on SS-OCTA images as described previously [31]. In brief, the SS-OCTA images were obtained using a 3000 × 3000-μm OCTA scan of the macula, and automated layer segmentation was obtained, as described above. The en face SS-OCTA image of the choriocapillaris was generated from the slab with a position between 0 and 10.4 μm below the Bruch’s membrane, which is a default setting provided by the instrument’s software. The en face OCTA images were exported and analysed using ImageJ software. The detailed methods for the measurements of vascular density (VD) and flow void area of the choriocapillaris are described in Online Resource 1 and our previously published study [31].

Grading of macular tessellation in fundus photography

Colour fundus photography was obtained when SS-OCT scan was performed using the built-in fundus camera of the instrument. Macular tessellation on colour fundus photography was quantitatively evaluated using the macular tessellated fundus index (TFI) [32, 33]. To evaluate the corresponding area of the en face OCT images, we made some modifications to the previously proposed method for selecting the ROI. While carefully observing the vascular shape of the macula, we selected a 3000 × 3000-μm rectangular area centred on the fovea as the ROI (Fig. 4). Thereafter, the colour fundus photographs were split into red, green, and blue pixels using the RGB stack tool. A histogram was constructed for each colour, and the mean intensities of red (R), green (G), and blue (B) pixels within the rectangular area were determined. As previously proposed by Yoshihara et al. [33], we also calculated three variations in TFIs, namely TFI-1, TFI-2, and TFI-3. The three TFIs were determined according to the following equations: TFI-1 = (R-G)/R, TFI-2 = R/(R + G + B), and TFI-3 = (R-G)/(R + G + B).

Fig. 4
figure 4

Hyperreflective choroidal foci on the B-scan image of swept-source optical coherence tomography (SS-OCT) (upper) and en face images of SS-OCT covering the rectangular area of 3000 × 3000-μm, centred on the macula (lower left), choriocapillaris layer images obtained using SS-OCT angiography (lower centre), colour fundus photography images (lower right)

Full size image

Statistical analysis

All statistical analyses were performed using IBM SPSS Statistics for Windows, version 20.0 (IBM Corp., Armonk, NY, USA). The baseline characteristics were compared using a Student’s t-test for continuous variables and χ2 test for categorical variables. Pearson’s correlation coefficient (r) was used to analyse the linear correlations of various variables with HCF. Statistical significance was set at P < 0.05.

Results

General characteristics

We investigated 51 eyes from 51 normal subjects (Table 1). The mean patient age was 56.0 ± 14.7 years, and 32 (62.7%) were female. The number and total area of HCF did not differ between the female and male patients and the right and left eyes (Table 2).

Table 1 Characteristics of the subjects
Full size table
Table 2 Comparison of the parameters of HCF in relation to sex, side of the eyes, and systemic factors
Full size table

Correlation of HCF with various parameters

The mean number of HCF was 224.6 ± 44.8 in a 3000 × 3000-μm area. The mean size of HCF was 937.3 ± 119.8 μm2. The mean number and total area of HCF did not correlate with age, subfoveal choroidal thickness, choriocapillaris VD and flow void area, and the degree of macular tessellation (Table 3). However, the mean number of HCF correlated with the subfoveal CVI (r =  − 0.363, P = 0.009) and stromal area of the choroid (r = 0.283, P = 0.045). The total area of HCF was also correlated with the stromal area of the choroid (r = 0.283, P = 0.045). The circularity of HCF correlated with age (r =  − 0.330, P = 0.018) and choriocapillaris VD (r = 0.335, P = 0.016).

Table 3 Correlation between number and area of hyperreflective choroidal foci and other parameters
Full size table

Partial correlation coefficient while controlling for the difference in image quality Several factors of HCF, such as number and circularity, tended to vary with OCT image quality, even though only high-quality OCT images were included in this study (Table 4). For this confounding factor, we analysed the partial correlation while controlling for differences in the quality of the OCT images (Table 5). We found that the number of HCF was still correlated with the subfoveal CVI (r =  − 0.364, P = 0.009) and stromal area of the choroid (r = 0.291, P = 0.040) (Fig. 5). However, the correlation between the circularity of HCF and other parameters was not significant.

Table 4 Correlation between the quality of the optical coherence tomographic images and measurements of HCF
Full size table
Table 5 Correlation between the number and circularity of HCF and other measurements while controlling for the quality of the optical coherence tomographic images
Full size table
Fig. 5
figure 5

Scattergram for the correlation of hyperreflective choroidal foci with various parameters. HCF, hyperreflective choroidal foci; r, Pearson's correlation; r*, Partial correlation

Full size image

Discussion

Hyperreflective foci have been studied in the eyes with chorioretinal diseases. They were believed to be activated microglial cells in several previous studies [2, 11, 18, 34]. Roy et al. [19] suggested that hyperreflective foci within the choroid have a composition similar to that of hyperreflective foci within the retina of the eyes with diabetic macular oedema. Romano et al. [20] hypothesized that macrophage recruitment or progressive RPE degeneration may be associated with the origin of hyperreflective foci in the eyes with choroideremia. Piri et al. [18] suggested that the predominant distribution of hyperreflective foci in the choroid and inner retina in Stargardt disease results from the spreading of lipofuscin deposits, which are by-products of the outer segment of the photoreceptor in the outer retina. However, they acknowledged several limitations present in concluding the origin, nature, and significance of HCF. In our study, we showed that hyperreflective foci exist in the choroid of normal eyes. Until now, it was not clear why hyperreflective foci were found in the normal choroid; our results may suggest an alternative possibility of their origin, which is different from inflammation or lipofuscin deposits.

In this study, we measured the number and area of HCF in the macular area between the choriocapillaris and Sattler’s layers of the choroid. We compared the distribution of HCF with choroidal parameters assessed on SS-OCT and SS-OCTA and found that the number and area of HCF were correlated with the proportion of the stromal area of the underlying choroid. The mean number of HCF was negatively correlated with the CVI. However, these parameters did not correlate with the vascularity of the choriocapillaris. These findings suggest that the distribution of HCF is related to the stroma of the choroid.

The choroid comprises both the stroma and vessels [35]. The choroidal stroma occupies the space between the dense vascular plexus and is composed of collagen and elastic fibres and several types of cells [35,36,37]. HCF may originate from some of these stromal components. In a previous human postmortem study, macrophages in the normal choroidal stroma were detected on using whole mounts of the choroid [38]. It suggests that HCF may be related to macrophages residing in normal choroid. In addition to macrophages, other candidates for the origin of HCF could be choroidal melanocytes [39,40,41,42,43,44]. Melanocytes have been reported to have high reflectivity on OCT images [45, 46]. Choroidal melanocytes are fusiform and dendritic cells that contain numerous round-to-ovoid discrete melanosomes [35,36,37]. Previous studies [35,36,37] have reported that the size of melanocyte in the choroid is larger than that in the skin, and the average length of choroidal melanocytes is approximately 20–40 μm with a diameter of approximately 20 μm. In our study, the mean size of HCF was 937.3 ± 119.8 μm2, and the diameter was estimated to be 35.5 μm. These findings suggest that HCF could be of choroidal melanocytic origin. To compare the distribution of HCF with the colour of the fundus, we measured the TFI in the macular area. However, we could not find a significant correlation between the TFI and HCF. The measurement of the TFI at the macular centre, where the retina is rich in macular pigment, could have prevented a direct correlation between choroidal pigmentation and TFI. Further studies with images of the extra-macular area are required. Another candidate for the origin of HCF could be fibroblasts dispersed throughout the choroidal stroma. However, fibroblasts are not round cells but spindle-shaped cells that often lie close to the choriocapillaris layer [47]. Further studies comparing OCT and the histology of normal eyes are needed to confirm the origin of HCF in the normal choroid.

Hyperreflective foci cannot be detected on clinical examination and have been variously defined using SD-OCT scans [10, 20]. In the current study using en face OCT images, we presented a method for quantitatively measuring HCF. B-scan images are useful for determining the relative location of hyperreflective foci in the chorioretinal layers. However, it is challenging to quantify the two-dimensional distribution of the foci in the coronal plane of the chorioretinal layers. We measured the distribution of hyperreflective foci on en face images of the choroid. It may help understand the two-dimensional distribution of foci and their relationship with fundus images. In addition, we defined hyperreflective foci based on the mean reflectivity of each en face image. This method would provide an objective quantification that is less dependent on the observers.

This study has some limitations, including its small sample size. The SS-OCT resolution of about 20 µm and the lack of histologic evidence prevented further detailed analysis of the relationship between the distribution of hyperreflective spots and specific choroidal constituents. Furthermore, we used only one type of SS-OCT and determined HCF in a slab near the choriocapillaris layer. This may have prevented further analysis of the correlation between the distribution of HCF and other parameters, such as choroidal thickness. Although we included high-quality images and statistically controlled the effect of image quality, signal noise was inevitable.

In conclusion, we found that the distribution of hyperreflective foci was associated with the underlying stromal component of the choroid in normal eyes. The results of this study can help evaluate the origin of hyperreflective foci in the normal choroid and can be used as a reference in determining abnormal hyperreflective foci in the choroid of the eyes with various diseases.