The present study retrospectively analyzed the medical records and images of patients diagnosed with treatment-näive ME associated with BRVO at the Department of Ophthalmology, Dongsan Medical Center, Keimyung University between October 2016 and March 2018. Comparative analysis was performed with the patients divided into two groups based on ME recurrence (ME recurrence and no ME recurrence) during 6 months after the resolution of initial ME. The present study was performed in accordance with the principles of the Declaration of Helsinki and approved by the Keimyung University Institutional Review Board (IRB no. 2018-09-039).
Patients with any of the following conditions were excluded from the study: 1) previous diagnosis and treatment for BRVO; 2) other retinal diseases that can affect the macular thickness, such as age-related macular degeneration, diabetic retinopathy, central retinal vein occlusion, and epi-retinal membrane; 3) high myopia (axial length ≥ 26.5 mm or refractive error ≥ -6 diopter); 4) glaucoma; and 5) history of pars plana vitrectomy. Patients were also excluded in case of voluntary termination of follow-up prior to ME resolution and difficulty in data anlysis due to poor image quality.
The locations of vein occlusion proposed by Hayreh et al.  were used to divide the occlusion into two types: major BRVO and macular BRVO. ME was diagnosed using swept source-OCT (SS-OCT; Swept Source DRI-OCT TritonTM, Topcon, Tokyo, Japan) and macular thickening was defined as a central macular thickness (CMT) of ≥300 um with intraretinal cysts or subretinal fluid in the macular region. OCT was performed during each visit to check for the presence of edema and changes in the macular thickness. Resolution of ME was defined as a CMT of <300 um with a concave macular contour. All patients were treated with intravitreal bevacizumab injections for ME. During the follow-up periods, intravitreal bevacizumab injection were repeated as needed until ME resolution was achieved.
Analysis of the perifoveal capillary network morphology using OCTA
Changes in the morphology of the perifoveal capillary network (FAZ area, perifoveal capillary ring) were analyzed using SS-OCTA (Swept Source DRI-OCT TritonTM, Topcon, Tokyo, Japan) and imaging was performed by a single experienced examiner on the same day.
The images of the macular region (3 × 3 mm) were automatically acquired with four slabs divided into the SCP, DCP, outer retina, and choriocapillaris using the IMAGEnet 6 software (version 1.17, Topcon, Tokyo, Japan). The SCP included the area from a point 2.6 um below the internal limiting membrane to a point 15.6 um below the inner plexiform layer, while the DCP included the area between 15.6 um to 70.2 um below the inner plexiform layer.
For the elimination of segmentation errors in the retinal layer due to macular swelling, analyzed images were obtained after resolution of ME. Images with signal strength intensity (SSI) values of ≥50, and with the centrally located fovea were selected, and the FAZ area and perifoveal capillary ring morphology were analyzed.
The FAZ area was obtained from values automatically calculated by two examiners, who used a caliper contained in the program to manually draw along the inner boundaries of the SCP and DCP (Figure 1). The perifoveal capillary ring was defined as the inner boundary of FAZ. Changes in the perifoveal capillary ring were divided into two types: (I) intact (0 clock hour) if the inner boundary of the capillary plexus was not broken after resolution of ME, and (II) ring loss or destruction if the inner boundary was broken. The degree of ring loss was marked as the clock hour and converted by multiplying each hour by 30° (Figure 2).
Analysis of macular capillary VD using ImageJ
VD in the macular capillary network was analyzed by the ImageJ software program (version 1.52a, National Institutes of Health, Bethesda, Maryland, USA; available at http://imagej.nih.gov/ij/) using SCP and DCP images acquired by OCTA for 3 × 3 mm macular region. For visualization of the capillaries, a 320 × 320-pixel image was converted to an 8-bit image and processed by binarization after marking with gray values between 0 and 255, and the average gray value of all pixels was used as the VD (Figure 3).
Moreover, for analysis of the hemi-VD disparities between the affected and unaffected areas, the SCP & DCP images were divided into upper and lower portions. The hemi-VD values were measured by binarization of the 320 × 160-pixel area (3 × 1.5 mm) and the hemi-VD disparity was derived from differences in VD between the hemi-superior area and hemi-inferior area (Figure 4).
The SAS program (version 9.4, SAS Institute Inc. Cary, North Carolina, USA) was used for statistical analysis; a p-value of <0.05 was considered statistically significant. The values measured by two examiners were considered reliable if the intraclass correlation coefficient was ≥0.8. The mean of SSI values derived from en face OCTA images showed no significant differences between the two groups.(59.05 vs 60.62, p=0.32)
The superficial and deep FAZ areas, average VD, and hemi-VD disparity in SCP and DCP were compared between the two groups using the Wilcoxon rank-sum test. Superficial and deep perifoveal capillary ring losses were analyzed using the Chi-square test, while the extent of capillary ring loss was analyzed using the Wilcoxon rank-sum test.