Ultra-Widefield OCTA Mapping of Epiretinal Macrophage-Like Cells in Patients With Retinal Vein Occlusion

Abstract

Purpose

To characterize the ultra-widefield distribution and morphology of epiretinal macrophage-like cells (eMLCs) in retinal vein occlusion (RVO) and assess their associations with retinal ischemia and macular edema.

Methods

This cross-sectional study included 68 eyes with unilateral branch or central RVO and 68 unaffected fellow eyes as controls. All eyes underwent 24 × 20 mm ultra-widefield swept-source optical coherence tomography angiography (OCTA). The eMLC density and morphology were quantified using a machine-learning–based pipeline. Foveal avascular zone (FAZ) metrics, central subfield thickness (CST), and nonperfusion areas (NPAs) were analyzed. Exploratory duration-based subgroup analyses were also performed.

Results

Compared with fellow eyes, RVO eyes showed higher eMLC density (3.69 [2.43] vs. 2.24 [1.12] cells/mm²), greater total eMLCs area (5.31 [3.83] vs. 2.87 [1.74] mm²), greater CST (411.50 [374.75] vs. 290.50 [35.00] µm), and larger FAZ area (0.360 [0.375] vs. 0.220 [0.165] mm²; all P < 0.001). Within-eye regional analyses showed lower eMLC density in the macular region than in the remaining retina (3.09 [1.57] vs. 4.06 [2.82] cells/mm²) and in NPAs than in the remaining retinal regions of the same image (2.39 [1.67] vs. 4.16 [1.89] cells/mm²; both P < 0.001). In RVO eyes, eMLC count was independently associated with disease duration, prior anti-VEGF injection number, and ischemic status.

Conclusions

EMLCs are increased in RVO and show a spatially heterogeneous distribution pattern. In RVO eyes, eMLC count was associated with disease duration, prior anti-VEGF exposure, and ischemic status, whereas duration-based subgroup findings should be interpreted as exploratory.

Translational Relevance

Ultra-widefield OCTA–based eMLCs quantification may provide a clinically accessible imaging biomarker for inflammatory involvement in RVO.

Introduction

Retinal vein occlusion (RVO) ranks as the second most prevalent retinal vascular disorder after diabetic retinopathy and constitutes a major cause of visual loss, primarily via two mechanisms: macular edema and retinal ischemia. A substantial body of evidence implies that inflammation runs throughout the whole process of RVO, in which inflammation-related cytokines and adhesion molecules compromise the integrity of blood–retinal barrier (BRB), causing the breakdown of BRB, increasing retinal vascular leakage, and thus the consequent edema.^1,2 With the progression of RVO, retinal capillary dropout leads to retinal non-perfusion, which causes the retinal ischemia.³ Although intravitreal anti-VEGF agents and corticosteroids remain the mainstay to treat macular edema secondary to RVO, unresponsiveness in some patients and frequent injections largely decreased the compliance of the patients for further treatment, which largely impaired their visual acuity. Thus comprehensive retinal evaluation using larger fields of view (FOVs) is of particular importance in patients with RVO. Ultra-widefield optical coherence tomography (OCT)/OCT angiography (OCTA) enables assessment of inflammatory imaging biomarkers that may facilitate clinical treatment decisions and help predict disease prognosis.^4,5

Recent advances in en-face OCT/OCTA have enabled in vivo visualization and quantification of epiretinal macrophage-like cells (eMLCs)—an umbrella term including hyalocytes, perivascular macrophages and activated microglia cells at the vitreoretinal interface—thereby positioning eMLCs as a promising non-invasive surrogate marker of ocular neuroinflammation.^6,7 Multiple human studies have demonstrated reproducible detection and quantitative measurement of eMLCs—along with their morphologic characteristics—using standard clinical OCT/OCTA devices.^8–10

In the context of RVO, multiple clinical cohorts demonstrate that eMLC density is markedly elevated in affected eyes compared to the unaffected fellow or healthy eyes, and that a higher eMLC density/number correlates with increased treatment burden and poorer visual outcomes— supporting their prognostic value.¹¹ En-face OCT/OCTA investigations further reveal the phase-dependent changes of eMLCs in both acute and chronic RVO (i.e., in the acute phase [< 3 months]), eMLC density is lower in edematous regions and negatively correlated with central subfield thickness (CST), whereas in the chronic phase (>3 months), eMLC density becomes positively correlated with CST and mean retinal thickness—suggesting a phase-dependent association between inflammatory activation and retinal edema.¹² Complementary analyses report that increased eMLCs at the vitreoretinal interface reflect activation of inflammatory pathways and are associated with the presence of sub-foveal fluid, further reinforcing the link between eMLCs and clinically relevant fluid phenotypes.¹³ Spatial mapping indicates that eMLCs preferentially accumulate in perfused rather than non-perfused retinal regions, and that higher eMLC density associates with greater macular thickness, supporting a coupling between inflammation and retinal structure across different regions.¹⁴

However, most studies of eMLCs have been restricted to limited FOVs (e.g., 3 × 3 mm or 6 × 6 mm scans centered on the macula or optic disc), which limits ultra-widefield assessment of the relationships among inflammation, nonperfusion, and edema.^8,12 This limitation is especially pertinent in RVO sub-types characterized by regional heterogeneity (e.g., branch RVO [BRVO] vs. central RVO [CRVO]), in which ischemic damage and edema frequently extend beyond the macular region.

The 24 × 20 mm ultra-widefield swept source OCTA (SS-OCTA) enables a single-capture scan from the posterior pole through the mid-periphery, delivering quantitative ischemic metrics—such as FAZ, NPAs, capillary alterations and collateral vessels—that are comparable to those obtained via ultra-widefield fluorescein angiography in treatment-naïve RVO.¹⁵ Moreover, multi-cohort data validate the technical feasibility of this modality for ultra-widefield retinal ischemia assessment.¹⁶ From the standpoint of eMLCs biology, the 24 × 20 mm field offers three key advantages: (i) it encompasses the majority of the clinically relevant retinal region, enabling quantification of global eMLC burden and sectoral gradients (for example along retinal arcades or venous territories) beyond the macula; (ii) it allows co-registration of eMLC distributions with NPA, FAZ metrics, and CST in the same acquisition, thereby mitigating montage and sampling bias; and (iii) it better accommodates the differing regional lesion patterns of BRVO versus CRVO. Collectively, these features render the 24 × 20 mm SS-OCTA an optimal platform for comprehensive characterization of eMLCs in RVO and for examining ultra-widefield eMLC–ischemia–edema coupling with direct clinical relevance.^12,15,16 This study aimed to characterize the density, spatial distribution, and morphology of eMLCs in eyes with RVO using a 24 × 20 mm ultra-widefield SS-OCTA, and to investigate their associations with visual acuity, CST, and NPAs, thereby elucidating the potential role of eMLCs in the pathogenesis of RVO.

Material and Methods

Subjects

This cross-sectional study consecutively enrolled patients with RVO who presented to the Department of Ophthalmology at Shenzhen People's Hospital, China, between April 2024 and September 2025. All enrolled participants were Asian Chinese. The study protocol was approved by the Institutional Review Board of Shenzhen People's Hospital (Approval No. LL-KY-2025080) and was conducted in accordance with the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants. Both BRVO and CRVO were diagnosed and evaluated using multimodal imaging, including color fundus photography, scanning laser ophthalmoscopy, fundus fluorescein angiography, as well as ultra-widefield SS-OCTA. For exploratory subgroup analyses, RVO eyes were categorized as acute (<3 months) or chronic (>3 months) according to symptom duration; this duration-based grouping was used as an exploratory clinical stratification rather than a definitive biologic staging system. Both eyes of each participant underwent a comprehensive ophthalmic evaluation, including best-corrected visual acuity (BCVA), slit-lamp biomicroscopy, color fundus photography, and fundus fluorescein angiography/ultra-widefield fluorescein angiography. Ocular medical and treatment histories (e.g., prior intravitreal injections, laser therapy or other intraocular surgeries) were recorded. No eyes in the present cohort had received steroid treatment before OCTA imaging. Eyes were excluded for any of the following: (1) prior ocular trauma or intraocular surgery that might compromise imaging or segmentation (for example pars plana vitrectomy); (2) concurrent retinal diseases (for instance diabetic retinopathy, uveitis, glaucoma, neovascular age-related macular degeneration or pathologic myopia); (3) media opacities or poor fixation precluding reliable imaging; (4) extensive epiretinal membrane (ERM) interfering with en-face layer extraction or analysis; (5) a spherical equivalent greater than +3.0 diopter (D) or less than −6.0 D. These exclusion criteria mirror those of prior RVO cohort studies.

SS-OCTA Image Acquisition

SS-OCTA images were obtained using a commercial SS-OCT/OCTA system (BMizar 400 kHz Full-Range SS-OCT; TowardPi, Beijing, China), which features a 1060 nm swept source and ultra-widefield OCTA capability. For the primary analyses, we employed the single-capture 24 × 20 mm protocol using the highest available scan density (1536 × 1280 A-scans per volume) with eye-tracking-assisted motion correction. Projection artifact removal was performed using the device's built-in algorithm. To minimize artefacts and enhance reliability, we performed three to five repeated acquisitions at the same retinal location; only images with a quality score of ≥7 were included. The en-face slab for eMLCs evaluation was segmented at the internal limiting membrane (ILM) interface using the device's built-in algorithm and defined as a 3 µm-thick layer immediately above the ILM for optimal visualization, which minimizes interference from the retinal nerve fiber layer and optimizes visualization. Segmentation lines were manually inspected and adjusted when necessary.

Image Processing

We initially exported the en-face eMLCs slab (a 3 µm-thick layer immediately above the ILM) from the OCTA device software, and imported these images into FIJI/ImageJ for cropping and preprocessing. To mitigate low-frequency illumination gradients and vignetting artefacts, we applied the “Subtract Background” function. The processed image was duplicated, the duplicate underwent a Gaussian blur, and subtraction of the blurred image from the original produced a high-pass–filtered, eMLC-enhanced image. For cell extraction, a subset of representative images with clearly identifiable macrophage-like morphologies was manually annotated, and the Trainable Weka Segmentation plugin was used to train a pixel-level classifier to distinguish eMLCs from background. During classifier training, major retinal vessels and linear hyperreflective structures were explicitly labeled as background to prevent misclassification. The trained classifier was visually inspected and validated to ensure that vascular structures were not included in the final eMLC masks. The trained model was then batch-applied to the pre-processed image set to generate probability and label maps. All segmentation outputs were subsequently reviewed by two experienced graders as a quality-control step to confirm that the identified signals were consistent with eMLC-like punctate hyperreflective foci and to exclude major retinal vessels, linear reflective artefacts, and other obvious non-cellular hyperreflective structures before quantitative analysis. The resulting probability maps were converted to eight-bit images, threshold-binarized, and analyzed using the “Analyze Particles” tool to obtain eMLC count, total area, average size, and area fraction (Fig. 1). Particle areas were initially recorded in pixels and subsequently converted to physical units based on the manufacturer-provided nominal scan dimensions (24 × 20 mm) and corresponding image resolution (1536 × 1280 pixels), yielding a lateral scaling factor of 15.625 µm per pixel. The average size of eMLCs was expressed in square micrometers (µm²). The total eMLC area was reported in square millimeters (mm²), and area fraction was expressed as a percentage (%) of the analyzed field. EMLC density (cells/mm²) was calculated based on the calibrated physical area of the 24 × 20 mm FOVs. For exploratory regional analysis of the macular distribution, the ETDRS-defined macular region was used as a region of interest on ultra-widefield OCTA images, and the eMLC count within this region was measured on the matched binarized eMLCs map. Regional eMLC density within the macular region was calculated and compared with that in the remaining retinal region of the same image. For exploratory regional analysis of non-perfusion-related distribution, eyes with clearly identifiable retinal NPAs on ultra-widefield OCTA images were further evaluated. With reference to each patient's fluorescein angiography findings, the NPAs were manually delineated on the corresponding flow image using FIJI/ImageJ, and the eMLC count within this region was measured on the matched binarized eMLCs map. The regional eMLC density was then calculated and compared with that in the remaining retinal region of the same image.

Figure 1.

Representative ultra-widefield SS-OCTA visualization of eMLC-related signals in an eye with BRVO. (A) En face structural OCT projection of the eMLC slab (3 µm in thickness), positioned 3 µm above the ILM, processed using the automated image-processing workflow to enhance visualization of eMLCs. (B) En face OCTA projection of the superficial capillary plexus slab, extending from an ILM offset of 0 µm to an inner plexiform layer (IPL) offset of 9 µm, derived from the 24 × 20 mm ultra-widefield scan. (C) Corresponding cross-sectional structural OCT B-scan through the foveal center, illustrating retinal architecture and the vitreoretinal interface where eMLCs are localized. (D) Subtraction-based, enhancement-processed eMLC map generated using the automated image-processing workflow. (E) Magnified view of the region of interest showing individual eMLCs after denoising and morphological enhancement, with representative cells indicated by arrowheads. Arrowheads indicate representative examples for illustration only and do not denote all detectable eMLCs within the displayed region. Final quantification was based on the complete preprocessing and segmentation workflow rather than visual inspection alone.

Statistics

All statistical analyses were performed using R software (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were assessed for normality with the Shapiro–Wilk test. Data following a normal distribution are presented as mean ± standard deviation, whereas non-normally distributed data are expressed as median (interquartile range). Paired analyses were performed for comparisons between RVO eyes and fellow eyes using the Wilcoxon signed-rank test. For the exploratory regional analyses, eMLC density in the macular region and in NPAs was each compared with that in the remaining retinal region of the same image using the Wilcoxon signed-rank test. For multiple group comparisons, the Kruskal–Wallis test was applied, followed by Dunn's post-hoc test with Bonferroni correction. For comparisons between two groups, the Mann–Whitney U test was used. Categorical variables were compared using the χ² test or Fisher's exact test, as appropriate. Spearman's rank correlation was used to assess associations between eMLC counts and clinical/imaging parameters, including CST, FAZ metrics, NPAs, BCVA, disease duration, and treatment-related variables. Disease duration was defined as the interval from symptom onset to OCTA acquisition and analyzed as a continuous variable. Ischemic status was defined as fluorescein angiography–confirmed ischemic RVO. The number of prior anti-VEGF injections before OCTA imaging was analyzed as a continuous variable. Binary variables were coded as 1 (presence/yes) and 0 (absence/no), including macular edema, ischemic status, laser treatment, and binary treatment status. To further evaluate independent factors associated with eMLC counts in RVO eyes, multivariable linear regression analysis was performed including CST, disease duration, ischemic status, prior anti-VEGF injection number, and age. Given the modest sample size, a parsimonious model was used. A sensitivity analysis was additionally performed by replacing prior anti-VEGF injection number with binary treatment status, to assess the robustness of treatment-related findings under an alternative exposure definition. Robust standard errors (HC3) were used in the multivariable models. Spearman's correlations were two-tailed and computed using pairwise deletion for missing data. All tests were two-tailed, and statistical significance was defined as P < 0.05.

Results

Baseline Characteristics of the Study Population

The demographic characteristics of the study population are summarized in Table 1. A total of 68 patients diagnosed as unilateral RVO were included in this study. The mean age of the cohort was 58.91 ± 13.37 years old (range 15–84 years), and 50% of the participants were female. For each patient, the affected eye was included in the RVO group (n = 68 eyes), and the unaffected fellow eye served as the control (n = 68 eyes), yielding a total of 136 eyes for analysis. Among the 68 RVO eyes, 50 eyes (73.5%) were BRVO and 18 eyes (26.5%) were CRVO. For exploratory subgroup analyses, 38 eyes (55.9%) were categorized as acute and 30 eyes (44.1%) as chronic according to symptom duration. The interval from symptom onset to OCTA acquisition ranged from two days to five years, with a median of 15 days overall; the median disease duration was seven days in acute eyes and 210 days in chronic eyes. The acute subgroup was entirely treatment-naïve at the time of imaging, whereas 29 of 30 chronic eyes had received prior treatment before OCTA acquisition, including retinal photocoagulation and intravitreal injections of anti-VEGF agents, such as ranibizumab, aflibercept, and conbercept. Among the treated chronic RVO eyes, ten individuals (33.3%) received grid-pattern retinal photocoagulation and 27 patients (90.0%) received intravitreal injections of anti-VEGF agents (number of injections ranged from one to seven, with a median of three). Among RVO eyes, 52 out of 68 eyes still had macular edema on OCTA examination.

Table 1.

Baseline Demographic and Clinical Characteristics of the Study Population

Patient Characteristic	RVO Eyes	Fellow Eyes	P-Value
Participants, n	68
Age (y), mean ± SD	58.91 ± 13.37
Sex (female)	34 (50.0%)
DM	3 (4.4%)
HTN	27 (39.7%)
Duration (days), median (IQR)	15 (203.3)
Eyes	68	68
BRVO	50 (73.5%)
CRVO	18 (26.5%)
BCVA (logMAR), mean ± SD	0.59 ± 0.43	0.06 ± 0.07	<0.001
Anti-VEGF therapy	27 (39.7%)
Laser photocoagulation	10 (14.7%)

DM, diabetes mellitus; HTN, hypertension; IQR, interquartile range; SD, standard deviation.

Comparisons were performed using the Mann–Whitney U test or Wilcoxon test, as appropriate.

Quantitative OCTA and eMLC Differences Between RVO Eyes and Fellow Eyes

The quantitative OCTA and eMLC parameters are summarized in Table 2. In paired comparisons between RVO eyes and fellow eyes, CST was significantly greater in RVO eyes (411.50 [374.75] vs. 290.50 [35.00] µm, P < 0.001). The FAZ area was also significantly larger in RVO eyes (0.360 [0.375] vs. 0.220 [0.165] mm², P < 0.001), and the FAZ perimeter was significantly increased (2.745 [1.715] vs. 1.870 [0.858] mm, P < 0.001). In contrast, the FAZ perimeter-to-area (FAZ PA) ratio was significantly lower in RVO eyes (0.632 [0.207] vs. 0.781 [0.114], P < 0.001). Quantitative eMLCs parameters were likewise increased in RVO eyes. The total eMLCs area was significantly greater in RVO eyes (5.31 [3.83] vs. 2.87 [1.74] mm², P < 0.001), and the average eMLCs size was also higher (2987 [345] vs. 2789 [142] µm², P < 0.001). The percentage of total area occupied by eMLCs was significantly higher in RVO eyes (1.110% [0.797%] vs. 0.599% [0.360%], P < 0.001). No significant difference in vessel density within the 300-µm annulus surrounding the FAZ (FD-300) was observed between RVO eyes and fellow eyes. In addition, eMLC count and density were markedly elevated in RVO eyes compared with fellow eyes (1774.00 [1164.75] vs. 1074.00 [539.00] cells, P < 0.001; 3.69 [2.43] vs. 2.24 [1.12] cells/mm², P < 0.001).

Table 2.

Quantitative OCTA and eMLCs Parameters in Acute and Chronic RVO Eyes, as Well as the Fellow Eyes

	RVO Eyes (n = 68)
Variable	Acute RVO Eyes (n = 38)	Chronic RVO Eyes (n = 30)	All RVO Eyes (n = 68)	Fellow Eyes (n = 68)	P-Value
Disease duration (days)	7 (6)	210 (208.8)	15 (203.3)	—	<0.001a,b
CST (µm)	596.5 (372.5)	313.0 (147.3)	411.50 (374.75)	290.50 (35)	<0.001a,b
FAZ area (mm²)	0.435 (0.378)	0.220 (0.275)	0.360 (0.375)	0.220 (0.165)	<0.001a,b
FAZ perimeter (mm)	3.02 (1.87)	1.98 (1.63)	2.745 (1.715)	1.870 (0.858)	<0.001a,b
FAZ PA ratio	0.589 (0.206)	0.698 (0.196)	0.632 (0.207)	0.781 (0.114)	<0.001a,b
FD-300 (%)	0.348 (0.083)	0.361 (0.106)	0.354 (0.083)	0.354 (0.094)	Ns
NPA (mm²)	0.195 (3.709)	10.874 (85.894)	0.95 (57.66)	—	<0.001a,b
eMLC count	2168.50 (850.50)	1323.00 (1070.25)	1774.00 (1164.75)	1074.00 (539.00)	<0.001a,b
eMLC density (cells/mm²)	4.52 (1.77)	2.76 (2.23)	3.69 (2.43)	2.24 (1.12)	<0.001a,b
Total area (mm²)	6.38 (3.54)	3.80 (3.49)	5.31 (3.83)	2.87 (1.74)	<0.001
Average size (µm²)	3058 (391)	2916 (190)	2987 (345)	2789 (142)	<0.001a
					<0.01b
Area (%)	1.33 (0.74)	0.77 (0.73)	1.110 (0.797)	0.599 (0.360)	<0.001a,b

Area, percentage of the OCTA image area occupied by eMLCs; Average size, mean area of individual eMLC particles; eMLC density, number of eMLCs per mm²; FAZ PA ratio, FAZ perimeter-to-area ratio; Total area, total area occupied by eMLCs.

^aComparison between all RVO eyes and the fellow eyes.

^bComparison between acute group and chronic group.

Exploratory Within-Eye Regional Analyses of eMLC Distribution

Exploratory within-eye regional analyses showed that eMLC distribution was spatially non-uniform in RVO eyes. Compared with fellow eyes, eMLCs in RVO eyes appeared larger and more numerous across the ultra-widefield OCTA images. Within individual RVO eyes, eMLC density was significantly lower in the macular region than that in the remaining retinal region of the same image (3.09 [1.57] vs. 4.06 [2.82] cells/mm², P < 0.001). In eyes with clearly identifiable retinal NPAs, eMLC density was also significantly lower in NPAs than that in the remaining retinal regions of the same image (2.39 [1.67] vs. 4.16 [1.89] cells/mm², P < 0.001). In addition to increased cell size, eMLCs in RVO eyes showed qualitative morphologic enlargement compared with the fellow eyes (Figs. 2A, 2B), whereas fellow eyes demonstrated smaller and less numerous eMLCs with a more uniform distribution along the vitreoretinal interface.

Figure 2.

Representative ultra-widefield SS-OCTA images illustrating regional distribution patterns and qualitative morphologic variation of eMLCs in eyes with RVO. (A) En face structural OCT projection of the eMLC slab (3 µm above the inner limiting membrane) from an acute-phase CRVO eye, demonstrating pronounced eMLC accumulation. (A1) Magnified view of the yellow boxed region showing markedly enlarged eMLCs (yellow arrowheads). (B) En face structural OCT projection from an acute-phase BRVO eye without obvious enlargement. (B1) Corresponding magnified view of the boxed region demonstrating increased eMLC density with relatively preserved cell size, shown for comparison with panel A. (C) Ultra-widefield OCTA image from a CRVO eye illustrating regional variation in retinal perfusion. (C1) Corresponding B-scan structural OCT showing retinal thickening and inner retinal alterations. (C2) Binarized eMLCs map with the macular region indicated as the region of interest for exploratory regional analysis. (C3, C4) Magnified views illustrating relatively lower eMLC density in the macular region compared with the surrounding retinal region. (D) Ultra-widefield OCTA image from another CRVO eye illustrating retinal perfusion status. (D1) Corresponding structural OCT B-scan. (D2) Binarized eMLC map highlighting non-perfused areas (NPAs, blue box) and the surrounding retinal region for comparison. (D3, D4) Magnified views showing sparse or absent eMLCs within NPAs, whereas the surrounding retinal region contains more numerous eMLCs. Arrowheads indicate selected representative examples for illustration only and do not denote all detectable eMLCs within the displayed region. Final quantification was based on the complete preprocessing and segmentation workflow rather than visual inspection alone.

Associations of eMLCs With Disease Duration, Treatment, and Ischemic Status in RVO Eyes

Across all analyzed eyes, eMLC count was positively correlated with CST (r = 0.405, P < 0.001), logMAR BCVA (r = 0.348, P < 0.001), disease duration (r = 0.451, P < 0.001), and the presence of macular edema (r = 0.495, P < 0.001), and showed a negative correlation with the FAZ PA ratio (r = −0.388, P < 0.001). When analyses were restricted to RVO-affected eyes, eMLC count remained positively correlated with CST (r = 0.275, P = 0.023) and demonstrated significant associations with clinical characteristics, including negative correlations with disease duration (r = −0.357, P = 0.003), the number of anti-VEGF injections (r = −0.409, P < 0.001), ischemic status (r = −0.309, P = 0.010), and laser treatment (r = −0.358, P = 0.003). In subtype analyses, these clinical associations were predominantly observed in BRVO eyes, whereas no significant correlations with clinical variables were identified in CRVO eyes. Additional subgroup analyses suggested that similar relationships were more evident in eyes with macular edema, whereas they were less consistent in acute versus chronic subgroup comparisons and varied according to prior treatment status.

To further determine whether these associations remained after adjustment for other clinical variables, a multivariable linear regression analysis was additionally performed in RVO-affected eyes, including CST, disease duration, ischemic status, prior anti-VEGF injection number, and age. After adjustment, disease duration, ischemic status, and prior anti-VEGF injection number remained independently associated with eMLC count, whereas CST was no longer significant (Table 3). Similar findings were observed in a sensitivity model using binary treatment status instead of prior anti-VEGF injection number (Supplementary Table S1). An additional exploratory analysis using total eMLCs area as an alternative outcome definition showed broadly similar directional associations with disease duration, prior anti-VEGF exposure, and ischemic status.

Table 3.

Multivariable Linear Regression Analysis for Factors Associated With eMLC Counts in RVO Eyes

Variable	Adjusted β	95% CI	P Value
Intercept	2854.10	1720.26 to 3987.93	<0.001
CST	0.52	−0.56 to 1.59	0.346
Disease duration (days)	0.37	0.03 to 0.71	0.032
Ischemic status	−483.45	−840.06 to −126.86	0.008
Anti-VEGF injection number	−121.69	−212.18 to −31.21	0.008
Age	−14.78	−31.70 to 2.13	0.087

Multivariable linear regression with HC3 robust standard errors. Dependent variable: eMLC count. Analysis performed in RVO-affected eyes only (n = 68).

Exploratory Duration-Based Subgroup Comparison

As an exploratory duration-based subgroup analysis, we compared eMLCs and OCTA parameters between acute and chronic RVO eyes (Fig. 3). CST was higher in the acute group (596.5 [372.5] vs. 313.0 [147.3] µm, P < 0.001). The FAZ area was larger in acute RVO eyes (0.435 [0.378] vs. 0.220 [0.275] mm², P < 0.001), and the FAZ perimeter was also greater (3.02 [1.87] vs. 1.98 [1.63] mm, P < 0.001), whereas the FAZ PA ratio was higher in the chronic group (0.698 [0.196] vs. 0.589 [0.206], P < 0.001). The median NPA was numerically larger in chronic RVO eyes than in acute eyes (10.874 [85.894] mm² vs. 0.195 [3.709] mm²). For eMLC-related parameters, acute RVO eyes showed higher eMLC count and density than chronic eyes (2168.50 [850.50] vs. 1323.00 [1070.25] cells, P < 0.001; 4.52 [1.77] vs. 2.76 [2.23] cells/mm², P < 0.001). The total eMLCs area was also greater in the acute subgroup (6.38 [3.54] vs. 3.80 [3.49] mm², P < 0.001), and the eMLCs area percentage was higher (1.33% [0.74%] vs. 0.77% [0.73%], P < 0.001). The average eMLC size showed a modest but significant difference (3058 [391] vs. 2916 [190] µm², P < 0.01), whereas FD-300 did not differ significantly between the two groups.

Figure 3.

Representative en face structural OCT images illustrating eMLC-related alterations in eyes with RVO compared with the fellow eyes. (A) En face structural OCT projection of the eMLC slab from an eye with CRVO. The yellow box indicates the region selected for magnification. (B) Magnified view of the boxed region showing increased eMLC density and enlarged morphology (red arrowheads). (C) En face structural OCT projection from an eye with BRVO. (D) Corresponding magnified view demonstrating increased eMLC density with enlarged cellular morphology (green arrowheads). (E) En face structural OCT projection from a control eye showing a sparse and homogeneous distribution of eMLCs. (F) Magnified view of the control region illustrating small, evenly distributed eMLCs (blue arrowheads). Arrowheads indicate selected representative examples for illustration only and do not denote all detectable eMLCs within the displayed region. Final quantification was based on the complete preprocessing and segmentation workflow rather than visual inspection alone.

Additional exploratory analyses within the chronic subgroup did not show significant differences in eMLC density, CST, or other OCTA parameters according to treatment status, treatment modality, or the presence of macular edema (all P > 0.05). Morphologically, a small subset of acute RVO eyes (8/38) showed markedly enlarged eMLCs. These eyes had a mean disease duration of 8.63 days, none had received prior treatment, and six had macular edema with a mean CST of 692.17 µm. In this subgroup, the mean eMLCs particle size was 3651 µm², which was significantly larger than that in the remaining acute eyes (2894 µm², P < 0.001). These findings should be interpreted cautiously given the exploratory nature of the duration-based subgroup analysis and the differences in treatment exposure between subgroups.

Comparison of eMLCs and OCTA Parameters Between BRVO and CRVO

Significant differences in several OCTA parameters were found between BRVO and CRVO eyes. NPAs were larger in BRVO eyes than in CRVO eyes (6.56 [82.17] vs. 0.03 [0.07] mm², P < 0.001). In addition, vessel density within the 300-µm foveal region (FD-300) was lower in CRVO eyes than in BRVO eyes (0.323% [0.095%] vs. 0.369% [0.071%], P = 0.032). No statistically significant differences were observed in other OCTA or eMLCs parameters, including CST, FAZ metrics, eMLC count, eMLC density, total eMLC area, average eMLCs size, area percentage, or disease duration (all P > 0.05).

Discussion

To our knowledge, this study is among the first to employ ultra-widefield SS-OCTA to assess the distribution and morphology of eMLCs in patients with RVO over a broad retinal field. Using this approach, we demonstrated that eMLCs were significantly increased in RVO eyes compared with the unaffected fellow eyes and showed a spatially non-uniform distribution, with lower density in the macular region and in retinal NPAs. These observations were further supported by exploratory within-eye regional analyses, which showed significantly lower eMLC density in the macular region than that in the remaining retinal region and significantly lower eMLC density in NPAs than that in the remaining retinal regions of the same image. In analyses restricted to RVO eyes, eMLC count was associated with disease duration, prior anti-VEGF injection number, and ischemic status. In addition, a subset of eyes showed marked enlargement of eMLCs, suggesting qualitative morphologic variation in some cases. Last, CST was positively correlated with eMLC count across all eyes as well as within RVO eyes, indicating an association between eMLCs-related parameters and retinal thickening.

Distribution and Quantification of eMLCs in RVO Eyes

In this study, we used ultra-widefield OCTA and structural OCT to conduct extensive scans of the vitreoretinal interface, including both macular and peripheral regions, in RVO eyes. The results revealed that compared to the fellow eyes, the number, density, total area, and area percentage of eMLCs in RVO eyes were significantly increased (P < 0.001), and the average eMLCs cell area was slightly larger (P < 0.001). These findings demonstrate a marked accumulation of eMLCs in RVO eyes and are consistent with changes in the inflammatory microenvironment, rather than directly implying cellular activation.

This conclusion is consistent with previous studies in RVO patients. One study, comparing 72 RVO patients (43 acute and 29 chronic) with 64 fellow eyes using en face OCT slab (5 µm) for eMLCs quantification, found that both the acute and chronic groups had significantly higher eMLCs densities than that of the control group (P < 0.001).¹² Another study analyzing 36 untreated, unilateral acute RVO patients found that the eMLCs densities in the optic nerve head and macula were approximately 2.46 and 2.86 times higher than those in the fellow eye (P < 0.001).¹⁴ By expanding the field of view beyond the macula, we further confirmed the widespread increase of eMLCs in RVO eyes. This result supports the potential relevance of eMLCs as an imaging biomarker associated with inflammatory changes in RVO.

From a mechanistic perspective, RVO is characterized by venous obstruction, blood stasis, altered capillary permeability, disruption of the BRB, localized ischemia, reperfusion, and upregulation of inflammatory factors (such as cytokines and chemokines). These changes likely trigger activation, redistribution, and recruitment of macrophage-lineage cells (manifesting as eMLCs) to the ILM, where they aggregate and become activated. Previous animal model studies have shown that retinal ischemia and BRB disruption can elicit microglial activation and blood-derived monocyte infiltration.¹⁷ Combining these findings with our own data showing significant increases in eMLC count and density, our results suggest that eMLC accumulation in RVO is not confined to a single focal region, but is distributed across a broad retinal field with regional variation.

Thus our study, which combines broader scanning coverage, larger FOVs, and structural parameters with eMLCs quantification, provides further evidence of the significant increase in eMLCs in RVO eyes. This discovery further reinforces the potential of eMLCs as an observable, quantifiable, and non-invasive imaging biomarker associated with inflammatory and immune-related changes in RVO.

Exploratory Duration-Based Subgroup Findings and Regional Variation of eMLCs in RVO

In this study, duration-based subgroup comparisons showed that eyes with shorter symptom duration had higher eMLC count, density, total area, and area percentage, together with slightly larger average eMLCs size, than eyes with longer symptom duration. However, these subgroup findings should be interpreted cautiously, because the categorization was based on symptom duration and treatment exposure differed markedly between the two subgroups. In parallel, our exploratory regional analyses showed that eMLC density was significantly lower in the macular region than that in the remaining retinal region and was also significantly lower in NPAs than that in the remaining retinal regions within the same image. These spatial findings support the presence of regional heterogeneity in eMLC distribution across the ultra-widefield retinal field.

By using ultra-widefield OCTA scanning, we were able to evaluate eMLCs-related changes beyond the macular area and to compare them with regional perfusion characteristics in the same eye. The observed duration-based subgroup differences may reflect a combination of shorter symptom duration, lack of prior treatment in eyes with recent onset, and differences in ischemic or edema profiles between subgroups, rather than a definitive stage-specific biologic transition. Similarly, the lower eMLC density observed in NPAs may be related to reduced local blood flow, oxygenation, metabolic support, and redistribution of inflammatory cells in severely ischemic retinal regions, whereas eMLCs may preferentially accumulate in relatively better-perfused areas.

In addition, a small subset of eyes with shorter symptom duration showed marked enlargement of eMLCs, suggesting qualitative morphologic variation in some cases. Overall, these findings suggest that eMLC distribution and morphology may vary across different retinal regions and clinical contexts in RVO. Nevertheless, because these subgroup observations were exploratory and treatment exposure differed substantially between groups, they should not be interpreted as evidence of a definitive biologic distinction between acute and chronic RVO.

Association of eMLC Count With CST and Clinical Variables in RVO

In the present study, we observed a significant positive correlation between eMLC count and CST, indicating that higher eMLC counts are observed in RVO eyes with more pronounced macular edema. This finding suggests that eMLC accumulation tends to occur in eyes with increased retinal thickness and more severe edema.

Such an association is not unique to RVO. Previous studies in other retinal diseases, particularly diabetic macular edema, have reported positive correlations between eMLC burden and edema severity, supporting the concept that eMLCs may be related to inflammatory activity and fluid dysregulation in the retinal microenvironment.¹⁸ These observations provide indirect support for the possible relevance of eMLCs as indicators of retinal inflammatory status. However, because disease duration and prior treatment exposure were closely related in our cohort, the independent effect of any single clinical variable should be interpreted cautiously.

From a pathophysiological perspective, venous obstruction in RVO leads to blood stasis, increased capillary hydrostatic pressure, BRB breakdown, and subsequent fluid leakage into the retinal tissue, resulting in macular edema. These processes are accompanied by inflammatory activation and immune cell recruitment, during which eMLCs may accumulate within the affected retinal regions. Rather than implying a causal role, the observed correlation suggests that eMLC accumulation may reflect the inflammatory and vascular permeability changes associated with edema formation and persistence.

Taken together, the positive association between eMLC count and CST suggests that eMLCs may be associated with retinal thickening and edema-related changes in RVO. These findings suggest that eMLC counts may have potential value as a noninvasive OCTA-derived imaging biomarker for disease activity and edema severity. However, longitudinal studies are required to clarify the temporal relationship between eMLC dynamics and macular edema, as well as to determine their utility in monitoring treatment response and predicting prognosis. Notably, the direction of association between disease duration and eMLC counts differed between univariable and multivariable analyses, which likely reflects confounding by prior treatment exposure, as eyes with longer disease duration were also more likely to have received a greater number of anti-VEGF injections before imaging. Accordingly, the independent effect of disease duration should be interpreted cautiously.

Speculative Nature of Current Inferences and Future Research Directions

However, these interpretations remain speculative. Because of the current lack of histological, functional, and molecular data (such as inflammatory mediators—including cytokines and chemokines), we cannot confirm the exact cellular origin or composition of these eMLCs (e.g., microglia, blood-derived macrophages, or hyalocytes), nor their specific roles in edema formation and maintenance (whether they contribute to inflammation, vascular permeability changes, tissue repair, or debris clearance). Future research will need to combine aqueous humor and vitreous fluid inflammatory, permeability-related markers, and BRB breakdown markers (VEGF, pro-inflammatory cytokines, and complement pathway components) with longitudinal eMLCs quantification to determine whether eMLCs participate in the pathological mechanisms of edema and to define their utility as an imaging marker for treatment monitoring and prognosis. In addition, longitudinal studies with standardized imaging before and after anti-VEGF or intravitreal steroid treatment will be important to determine whether eMLCs are treatment-responsive and to further clarify their biological and clinical relevance. Because no eyes in the present cohort had received steroid treatment before imaging acquisition, steroid-related treatment effects could not be evaluated in this study.

This study has several limitations, including the small sample size, single-center design, and lack of longitudinal repeated imaging to evaluate the dynamic migration or turnover of eMLCs. In addition, the duration-based subgrouping was based on symptom duration and should be regarded as an exploratory clinical categorization rather than a definitive biologic staging system. Disease duration was also closely related to prior treatment exposure in this cohort because acute eyes were treatment-naïve whereas most chronic eyes had already received treatment before imaging. Accordingly, although disease duration, prior anti-VEGF injection number, and ischemic status were analyzed in correlation and multivariable models, residual confounding between duration and treatment exposure cannot be fully excluded. Prior treatment exposure, including anti-VEGF therapy and laser treatment, was not fully standardized relative to the imaging time point and may have confounded some subgroup comparisons, especially in chronic eyes. As a result, reliable pre-/post-treatment comparisons could not be performed in the present study. Because no eyes had received steroid treatment before imaging acquisition, steroid-related subgroup or censoring analyses were therefore not applicable in this cohort. In addition, given the modest sample size, more complex models incorporating multiple interaction terms were not performed, and these relationships should be further evaluated in larger cohorts. Therefore the influence of prior treatment on eMLC distribution and morphology cannot be fully excluded.

Conclusions

By using ultra-widefield OCTA, we showed that eMLCs were increased in RVO eyes compared with fellow-eye controls and exhibited a spatially non-uniform distribution, with lower density in the macular region and in retinal NPAs than that in the remaining retinal regions. In RVO eyes, eMLC count was associated with retinal thickening, disease duration, prior anti-VEGF exposure, and ischemic status, although the duration-based subgroup findings should be regarded as exploratory. The contribution and effects of eMLCs merit further study in RVO and other retinal vascular diseases, particularly in conjunction with objective inflammatory measurements and longitudinal treatment-response assessment.