Evaluation of choroidal hyperreflective dots in acute and chronic central serous chorioretinopathy.

Purpose: To determine the association between hyperreflective dots (HRD) in the choroid and visual acuity and swept-source optical coherence tomography (SS-OCT)-derived structural parameters in central serous chorioretinopathy (CSC).
Methods: SS-OCT images (single visit) of consecutive patients with CSC were evaluated for the presence of HRDs in the choroid using an automated algorithm and manual measurements of central macular and subfoveal choroidal thicknesses were obtained.
Results: 61 eyes of 61 subjects were included in this retrospective study (32 subjects with acute and 29 with chronic CSC). Mean (± SD) choroidal HRD counts in acute and chronic CSC were 139.4 ± 29.9 and 124.9 ± 28.1, respectively (P = 0.04). In acute CSC, HRD was correlated with both age (P = 0.004) and subfoveal choroidal thickness (SFCT) (P = 0.016), but not with visual acuity or other OCT-derived measurements. In chronic CSC, HRD was correlated with central macular thickness (P = 0.011); neurosensory detachment height (P = 0.046); SFCT (P = 0.012). Considering all patients, the presence of HRDS was significantly negatively correlated with age (r = -0.401; P= 0.002) and SFCT (r = -0.332; P= 0.010).
Conclusion: HRDs are correlated with both age and SFCT in acute CSC, and with CMT, height of neurosensory detachment and SFCT in chronic CSC. Development of HRDS is associated with the remodelling of chorioretinal structures as previously noted in CSC.

Central serous chorioretinopathy (CSC) is a chorioretinal condition characterised by serous retinal detachment of the macula.[1] Developments in optical coherence tomography (OCT) have enabled high-resolution imaging to identify alteration in retinal structure and quantitative analyses of chorioretinal thicknesses in patients with retinal diseases including CSC.[23] Hyperreflective intraretinal dots have been visualised in a range of chorioretinal diseases including macular dystrophies, age related macular degeneration (AMD), retinal vein occlusion and diabetic retinopathy.[4567] These hyperreflective dots (HRDs) in the retina have been associated with alteration in visual acuity at presentation, and in some instances, visual outcome after treatment with anti-angiogenic agents.[89] Analysis of HRDs has suggested that they may be associated with genetic polymorphisms in AMD, alteration in lipid metabolism or extravasation of intracellular lipid, drusen and intraretinal migration of retinal pigment epithelial (RPE) cells.[10111213] HRDs in AMD may represent activated microglial cells developed through inflammation and could be useful to help determine treatment decisions and prognosis in AMD.[14]

The presence of HRDs has been suggested to be an important predictor of final visual acuity, especially when occurring within deeper retinal layers.[1516] The impact of the development of HRDs in chronic disease however remains to be fully elucidated. En-face OCT and adaptive optics scanning laser ophthalmoscopy has suggested that these intraretinal clusters may be cellular in nature.[17] Fundus autofluorescence has demonstrated elongated photoreceptor outer segment in CSC.[18]

Development of HRDs in the choroid in CSC may reflect alteration in chorioretinal structure but have yet to be characterised in acute or chronic disease, or indeed define their presence in association with other OCT-derived structural parameters, demographics or indeed visual acuity. The aim of this study was to investigate the association between the presence of choroidal HRDs in CSC (both in patients with acute and chronic CSC) with visual acuity, demographics and OCT structural parameters in patients with CSC.

Methods

This single centre retrospective study included evaluation of patients with acute and chronic CSC. The study was approved by the local ethics committee and adhered to the tenet set forth in the Declaration of Helsinki. Approval date November 2017. All images was conducted as part of routine patient care.

Study population

All consecutive patients attending clinic with CSC between April 2016 and November 2017 At LV Prasad Eye Insitute (Hyderabad, India) were included in this retrospective study. Exclusion criteria included the following: refractive error of more than ± 3 dioptres and axial length of more than 26mm; history of steroid treatment in the preceding year; history of other/previous chorioretinal disease; media opacity precluding adequate fundal view; previous ocular surgery (other than cataract surgery) and any other significant ocular comorbidity.

OCT imaging

Best corrected visual acuity (BCVA) and medical comorbidity was identified from medical records. All patients were adequately dilated prior to examination with 2.5% phenylephrine and 1% tropicamide. Each patient underwent a SS-OCT (Topcon, DRI OCT Triton) high definition single horizontal scan passing through the fovea at a single visit only.

SS-OCT quantitative analyses

Central macular thickness (CMT); subfoveal choroidal thickness (SFCT), defined as the distance between the RPE-Bruchs membrane complex and the chorioscleral interface; the height of the neurosensory detachment (NSD) as the perpendicular distance between the RPE and the retina were measured using caliper present on the proprietary software.

Image analysis

All SS-OCT macula images were exported in a JPEG format and HRDs were analysed using a semi-automated algorithm, using circle Hough transform in MATLAB [Fig. 1]. In particular, each OCT B-scan was first sharpened twice, using Gaussian mark of radius 4 pixels, to increase the contrast between the HRDs and the neighbouring choroid. Subsequently, thresholding was performed with an empirically chosen threshold of 175 (grey scale intensity). Finally, noting that HRDs appear as circle like structures in the binarized OCT B-scan, circle Hough transform was employed for their detection. The range for radius of circle for CHT was empirically chosen between 1 and 100 pixels. The circle Hough transform estimate may contain some spurious detections due hyperreflective regions outside choroid. In view of this, to keep the analysis confined to only the choroid region, our tool allows the observer to manually draw the choroid boundaries by selecting few control points on the choroid inner and outer boundaries. The tool then extrapolates these control points to obtain accurate choroid boundaries.

Figure 1

Optical coherence tomography automated image analysis. Automated image binarization and analysis of represented images of acute (a-c) and chronic (d-f) chronic serous chorioretinopathy (CSC). Swept-source optical coherence tomography (SS-OCT) images of patient with acute (a) and chronic (d) were acquired. Automated image binarization using automated algorithm was used to delineate choroid (b and e). Circle Hough transformation was used to delineate the presence of hyperreflective foci (c and f )

We employed circle Hough transform (CHT) available in the MATLAB (function ‘imfindcircles’). The Hough transform (HT) helps in detection of lines, circles and other structures if their parametric equation is known. In the current, we assume the hyperreflective dots as circular structures to leverage circle Hough transform which tries find circular objects with known radius. In brief, it finds the best fit circle to the edge points of a circular object where the edge points can be obtained using edge detection methods such as Canny edge operator.

In particular, in polar coordinate system, a circle with radius ri and center (x) can be described with parametric equations as: x = x + r cos(θ) and y = y + r sin(θ), where the angle θ sweeps through a full 360° range and the points (x) trace the perimeter of a circle. For an edge point, CHT votes multiple center-radius (x) triplet combinations and the optimum circle is chosen based on local maxima (maximum number of votes voted for a particular triplet). We observed that size of hypereflective dots is uniform and hence defined the radius of search space in CHT empirically and is constant for all images. The performance of our algorithm was validated based on intra-observer grading on 10 images. On a scale of 0 to 100%, proposed method achieved approximately 95% of agreement with respect to subjective intra-observer grading.

Subfoveal choroidal area (1500 μm) was segmented into luminal and stromal components using image binarization also. Choroidal vascularity index (CVI) is defined as the proportion of luminal area to total subfoveal choroidal area.[19]

Statistical analyses

Snellen visual acuity was converted into LogMAR for statistical analysis. Spearman's rank correlation coefficient was used to evaluate the correlation between the number of HRDs and various baseline parameters was evaluated in acute as well as chronic CSCR. P value of less than 0.05 was considered significant. IBM SPSS Statistics-24 was used for all the statistical analysis including descriptive and regression analysis.

Results

Subject characteristics

This study included 61 subjects including 32 eyes with acute and 29 with chronic CSC. This included 55 (90.2%) male and 6 (9.8%) female subjects. Mean (± SD) age was 43.4 ± 10.2 and best-corrected visual acuity (BCVA) was 0.28 ± 0.34logMar. Mean spherical equivalent was 0.66 dioptres (± 0.84). Subject characteristics are summarised in Table 1.

Table 1

Subject characteristics including demographics, optical coherence tomography derived chorioretinal thicknesses and hyperreflective dots

	Acute	Chronic	P
Age (mean years±SD1)	43.41±10.2	48.00±9.11	0.12
Best corrected visual acuity (mean logMar±SD)	0.28±0.34	0.38±0.44	0.03
Gender			0.56
Male n, %	29 (90.6)	26 (89.7)
Female n, %	3 (9.4)	3 (10.3)
Spherical equivalent (mean dioptres±SD)	0.66±0.84	0.82±1.07	0.31
Duration (mean days±SD)	15.55±31.69	31.22±40.80	0.02
Number of Hyperreflective dots (mean±SD)	132.74±29.81	124.93±28.13	0.09
Central Macular thickness (µm, mean±SD)	358.28±164.11	286.79±123.51	0.03
Neurosensory detachment (µm, mean±SD)	182.26±167.37	110.17±132.63	0.02
Subfoveal choroidal thickness (µm, mean±SD)	428.33±99.26	410.79±80.54	0.03
Choroidal vascularity Index	0.55±0.02	0.56±0.02	0.91

Analysis of HRDs

Mean (± SD) choroidal HRD counts in acute and chronic CSC were 139.4 ± 29.9 and 124.9 ± 28.1, respectively (P = 0.04).

Acute CSC subjects

Presence of HRDs was significantly negatively correlated with age (r = -0.365; P = 0.004) of subjects. Additionally, HRDs were negatively correlated with SFCT (r = -0.306; P = 0.016), but not any other quantitative parameter. Mean duration between onset of disease and OCT examination was 15.55 ± 31.69 days and there was no correlation with number HRDs (r = 0.099; P = 0.446). All results for both acute and chronic CSC subjects have been summarised in Table 2.

Table 2

Correlations between hyperreflective dots and demographics and chorioretinal thicknesses in acute and chronic central serous chorioretinopathy using Spearman’s rank correlation coefficient

Correlation	Acute CSC	Chronic CSC	All CSC subjects
Age	-0.365 (0.004)	-0.325 (0.085)	-0.351 (0.002)
Gender	-0.014 (0.916)	-0.167 (0.386)	-0.038 (0.710)
Duration	-0.099 (0.446)	-0.044 (0.819)	-0.053 (0.645)
Visual acuity	0.068 (0.604)	0.134 (0.487)	0.029 (0.825)
Spherical Equivalent	-0.242 (0.061)	-0.404 (0.03)	-0.238 (0.068)
Central macular thickness	0.084 (0.521)	-0.467 (0.011)	-0.069 (0.602)
Neurosensory detachment	0.115 (0.378)	-0.373 (0.046)	0.011 (0.334)
Subfoveal choroidal thickness	-0.306 (0.016)	-0.462 (0.012)	-0.332 (0.010)
Choroidal vascularity index	0.171 (0.187)	0.118 (0.543)	0.144 (0.274)

Chronic CSC subjects

Presence of HRDs was significantly associated with spherical equivalent (r = -0.404 P = 0.030). In addition, it was associated with several quantitative parameters including CMT (r = -0.467; P = 0.011), NSD height (r = -0.373; P = 0.046) and SFCT (r = -0.462; P = 0.012). Mean duration between onset of disease and OCT examination was 31.22 ± 40.80 days and there was no correlation with number HRDs (r = -0.044; P = 0.819).

All CSC subjects

Presence of HRDS was significantly negatively correlated with age (r = -0.401; P = 0.002) when analysing all subjects. Additionally, choroidal HRD count was also negatively associated with SFCT (r = -0.332; P = 0.010). Fig. 2 shows scatter plots showing correlation between HRDs and demographic, visual and structural parameters in all subjects.

Figure 2

Scatter plots showing correlation between HRDs and demographic, visual and stuctural parameters in all subjects. Corrleation between HRDs and visual acuity (a), duration (b), age (c), spherical equivalent (d), central macular thickness (e), neurosensory detachment height (f), subfoveal chroidal thickness (g) and choroidal vascularity index (h)

Discussion

OCT has enabled detailed imaging of chorioretinal structure in patients with CSC. This study suggests that presence of HRDs is correlated with SFCT in patients with acute CSC. HRDs appeared to have been associated with OCT-derived quantitative parameters including CMT, NSD and SFCT in patients with chronic CSC, but not visual acuity. Analysis of all subjects suggested that HRDs was associated with age and SFCT.

It has been suggested that there is a long-term reorganisation of choroidal structure in chronic CSC. Choroidal thickening and alteration in vascularity have been associated with eyes with CSC and fellow eyes suggesting that there is significant reorganisation of choroidal anatomy.[2021] It is interesting that HRDs was negatively correlated with SFCT in both acute and chronic forms. Perhaps the alteration of choroidal structure involves changes in choroidal vascular hyperpermeability in CSC; development of HRDs may occur due to cellular extravasation from the choroidal circulation. It may be possible to use the presence of HRDs to help assess chronicity of disease, which may be useful in guiding possible treatment decisions.

Development of HRDs in retina has been associated with the activation of microglial cells in inflammation in AMD. Indeed, adaptive optics imaging of subjects with CSC suggests that retinal HRDs in CSC may be cellular in nature.[17] In CSC, activated intraretinal microglial cells may develop hyperreflective foci in association in with the phagocytosis of photoreceptor outer segments.[11] Inflammation of the choroid in CSC may generate delay in choroidal filling (as suggested by change in choroidal vasculature on indocyanine green angiography).[22] Choroidal HRDs, however, may indeed be associated with this inflammation, or else exudation from altered choroidal vasculature. Further analysis of the cause and significance of choroidal HRDs in CSC is required. Choroidal HRDs have been correlated with disease severity in other conditions including Stargardt's disease also using optical coherence tomography perhaps due to choroidal remodelling in the chorioretinal disease.[23] As development of a neurosensory detachment is an on-going process in acute CSC, this may be responsible for the higher number of HRDs in acute than chronic choroidal CSC.

The strengths of this study include that this study involved an analysis of quantitative OCT derived chorioretinal measurements in addition to both visual acuity and demographic data in relation to HRDs.

The limitations of this study included limited sample size. It is important to understand that although subjects with spherical equivalent more than 3 dioptres and axial length more than 26mm were excluded, there are a number of other factors reported to effect choroidal thickness measurements including diurnal variation. However, imaging for all participants was performed between 9 am to 12 pm (as patients attended morning clinics only). Considering the relationship of HRDs with disease process, the number of HRDs may change with the disease duration and after intervention. We do not present such data at different time points, which could be helpful in understanding the disease pathogenesis. The relationship between the development of HRDs and disease duration and the alteration in HRDs after resolution of CSCR need further evaluation. The image analysis has some inherent limitations. Even after normalisation of images, the reflectivity of the choroid is variable especially at areas of RPE atrophy, which may have biased the results. Indeed, in future studies it would be useful to determine the variability of this choroidal reflectivity under RPE atrophy; furthermore, the effect of HRD presence in deeper retinal layers deserves further consideration. It is important to note that this study was not prospective in nature and further evaluation of HRDs in deeper layers is required, particularly in order to prognosticate the value of these features in CSC.

Conclusion

In summary, this study suggests that the presence of HRDs in choroid is correlated with age and choroidal thickness in patients with acute CSC. Additionally, it is associated with choroidal thickness, macular thickness and height of neurosensory detachment in patients with chronic CSC. Development of HRDs is associated with the remodelling of chorioretinal structures previously noted in chronic CSC. The role of HRDs as a predictor of treatment outcome in the long-term is yet to be explored.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.