Current study revealed that DCP is likely to develop the earliest subclinical radiation induced microvascular insult following 106Ru plaque brachytherapy. The deep FAZ area was identified as a more critical determinant of BCVA than superficial FAZ in these patients. Among the tumor characteristics and radiation parameters, the foveal dose and the optic disc dose had the highest sensitivity and specificity to predict the burnout pattern of the retinal microvasculature. Choriocapillaris flow area was significantly decreased in the treated eyes.
RR may lead to visual morbidity and blindness following choroidal melanoma brachytherapy, in fact in cases of maculopathy.[15–17] The Collaborative Ocular Melanoma Study Group (COMS-report No.16) recorded 6 lines of vision loss in 49% of patients who were treated with 125I brachytherapy after 3 years. 20 In 43% of patients with an initial vision of 20/200 or better at the time of diagnosis, vision declined gradually to 20/200 or worse by 3 years. 
According to previous studies, the risk of RR following brachytherapy is directly linked to the overall dose of administered radiation. In the treatment of choroidal melanoma, the dose of radiation to the apex of the tumor is between 62–104 Gy based on various studies. Other factors, such as tumor height and diameter as well as the position of the tumor, are correlated with the risk of retinopathy. Our results showed that the time interval between plaque implantation and image acquisition is the only independent factor predicting RM based on fundoscopy, FA and/or OCT findings. (Table 3)
To date, RR diagnosis has been primarily based on biomicroscopic and angiographic data or OCT findings of macular edema. Few studies investigated the role of OCTA in early detection of RR. Previous studies in patients treated with 125I plaques have shown that OCTA is probably the most effective existing imaging for detecting early signs of RM.[11–14] OCTA offers a 3-dimensional volumetric scan that displays the segmented distribution of the blood in macular area—something that is not possible with conventional FA. All of our patients were treated with 106Ru plaque brachytherapy. The study showed that most of the examined OCTA-derived metrics, including the vascular density and FAZ in the SCP and DCP, had been altered in irradiated eyes compared to non-irradiated eyes.
According to this report, it appears that the evaluation of the crude data directly obtained from the OCTA instrument is not appropriate for the assessment of vascular density of capillary plexus in the macular area due to the high rate of noises obscuring the information. As reported in previous studies, imaging artifacts and noises may induce some signals and may influence the vascular density and FAZ measurement.[12, 14] Despite higher speed of current OCTA machines, these artifacts are usually present in irradiated eyes because of low vision and resulted fixation deficit and apparent structural changes. It is noteworthy that OCTA artifacts are more common in eyes treated by brachytherapy than untreated eyes. Although, improvements in density measurement could be attained with repeated imaging and fixation aids in some cases, image processing, noise reduction and binarization are usually needed to alleviate this problem. In this study, the images were processed with sufficient filters to generate high-quality valid data for analysis.
By measuring the skeleton mass of vessels and minimizing the weight of large vessels, the analysis was made more specified for small vessels that are possibly most commonly affected by radiation. The endpoints assessed after noise reduction and image processing indicated macular capillary plexus disorganization and obliteration after 106Ru brachytherapy. Since the manual segmentation and analysis of the images could be a possible source of variability, we designed automated methods to measure biomarkers like VAD, VSD and FAZ area as the strength of this study.
Although few studies have reported OCTA results after 125I brachytherapy, no comparable comprehensive study after 106Ru brachytherapy is available.[3, 4, 11, 12, 14, 24, 25] Veverka et al. showed gradual alterations of the macular microvasculature on OCTA following melanoma treatment with 125I plaque. Shields et al. reported an enlargement of the FAZ region and decreased capillary density in both SCP and DCP after 125I brachytherapy. In another study, the patients with choroidal melanoma treated with 125I plaque and normal macular ophthalmoscopy and OCT, showed a statistically significant decrease in density of both SCP and DCP. Most of these studies signified the role of OCTA in the early detection of changes even in eyes without RM. [12, 14, 24, 25] According to the present research, DCP vascular density decrease (VAD and VSD) in foveal and parafoveal areas is the first biomarker for RM occurring prior to clinical and OCT and FA signs of RM. Consistently, Matet et al  showed that after proton beam therapy, the DCP of irradiated eyes was altered more severely than the SCP. Using volume-rendering display, Spaide showed that in RM, macular edema is associated with DCP non-perfusion.
We assume that in the earlier stage of RM, the measured loss of capillary density could be secondary to a decrease in flow velocity below the predetermined decorrelation threshold of the SSADA algorithm and not a true structural loss of the vessel. Histopathological and ultrastructural studies of the human retina following radiation are extremely scarce and mainly report changes with varying degrees of retinal ischemia and atrophy in the larger retinal and choroidal vessels. Histologic studies have also confirmed the early and preferentially loss of vascular endothelial cells leading to occlusion of capillaries in which pericytes still survived. Endothelial cell loss is due to impaired cell division and free radical production.[3, 4, 22] More severely affected retina revealed acellular capillaries with the residual basement membrane tubes being typically fused, shrunken or collapsed. It can be concluded that slower than normal cellular (red and white blood cells) flow in the basement membrane walled tubes could be detected as lower vascular density in both SCP and DCP in the irradiated parts of retina.
Radio-sensitivity of the DCP is higher than that of the SCP. The smaller capillaries in DCP are more radiosensitive than larger capillaries in SCP.  The lower perfusion pressure in these capillaries as terminal vessels make them more vulnerable to occlusion after endothelial cell damage or loss. Moreover, studies suggest that blood flows from SCP through serial connections of vertically descending anastomoses and vortex-like channels to the DCP.[31, 32] Capillaries of DCP are likely to be terminal vessels and tend to be more sensitive than SCP to ischemic stress, similar to terminal capillaries in other organs, as kidney.[32, 33] Even slight changes in retinal circulation may also primarily affect DCP.
Some studies, evaluating other retinal vascular disorders such as retinal vein occlusion and diabetic retinopathy, have shown that reduced perfusion is more common in DCP than SCP.34,35 On the other hand, the DCP non-perfusion has recently been identified as a more critical determinant of BCVA than SCP nonperfusion in patients with retinal vascular disorders.[34, 35] We also revealed that the deep FAZ was significantly correlated with BCVA. Finally, DCP flow derives exclusively from SCP, therefore it may receive a larger amount of downstream inflammatory or free radicals from the upstream part of SCP after irradiation.
It seems that, 106Ru is less destructive to fovea and optic disc compared to 125I due to a higher dose gradient through the tissue and shorter penetration and lateral distribution.[16, 18, 36] No study has yet compared vascular changes found in OCTA characteristics of retina and choroid following brachytherapy with 125I and 106Ru plaques. Despite a higher dose of radiation to the tumor apex (84.29 Gy vs. 71.5 Gy) in our study compared to the report by Shields et al,14, the mean dose of radiation to fovea and optic disc was lower (45.76 Gy vs 50.6 Gy and 32.5 Gy vs 40.3 Gy, respectively). The quantitative comparison of our findings with the detailed features reported by Shields et al showed similar results.
From the eyes with documented RM, nine eyes (36%) had severe damages with very extensive macular ischemia in which capillary plexus detail was not detectable in OCTA (burnout). In ROC analysis, the foveal dose and optic disc dose were better parameters for predicting the burnout pattern. However, due to the insufficient number of cases, the macular tolerance threshold could not be calculated.
The optic disc dose was positively correlated with superficial and deep FAZ area and had an inverse correlation with foveal and parafoveal SCP vascular density. As a result, foveal and disc radiation doses appear to be the primary predictors of both macular microvascular damage and visual function (BCVA).
The limitations of our study are mostly related to the retrospective design and the small number of cases, and imaging acquisition. While extensive efforts have been made to remove the artifacts of OCTA images, the development of new algorithms has not been entirely successful in this period for patients with RR. The study removed low-quality images, a source of selection bias. This may lead to underestimation or oversimplification of the exact impact of radiation on vascular indices.
Although we chose the patients who had extra-macular tumors, but adjuvant treatments potential impact were overlooked. According to recent studies, retinal capillary density and FAZ area remain statistically unchanged after intravitreal injection of an anti-VEGF agent or PRP in patients with diabetic retinopathy.[37, 38] In addition, we also have not assessed the status of the retinal vasculature at baseline and the longitudinal changes after treatment.
In this study, using suitable image processing software added more value to the analysis by reducing abnormal noise and more precise segmentation as a strength.