Peripapillary Bruch’s membrane Opening-Minimum Rim Width (BMO-MRW) and microvascular changes in early diabetic retinopathy

Abstract

Background

To evaluate changes in peripapillary Bruch’s Membrane Opening-Minimum Rim Width (BMO-MRW) and its association with peripapillary microvascular in patients with early diabetic retinopathy (DR).

Methods

This observational cross-sectional study included 105 eyes from 105 diabetic patients, comprising 55 eyes without diabetic retinopathy (No-DR group) and 50 eyes with mild diabetic retinopathy (Mild-DR group). An additional 50 eyes from 50 healthy individuals served as the Control group. All eyes underwent optical coherence tomography (OCT) radial scanning to assess Bruch’s membrane opening-minimum rim width (BMO-MRW) and OCT-angiography (OCTA) with a 6 mm × 6 mm scan centered on the optic disc to evaluate perfusion density (PD) and vessel density (VD). OCTA measurements were calibrated based on the FoBMO axis and analyzed using the ETDRS grid. The associations between BMO-MRW and PD/VD were subsequently investigated.

Results

The mean BMO-MRW values were significantly lower in the No-DR (305.78 ± 35.12 μm) and Mild-DR groups (299.42 ± 37.33 μm) compared to the control group (323.56 ± 40.33 μm, P = 0.005), with significant thinning in the superotemporal and inferotemporal quadrants (P = 0.039, 0.047). PD in the inner ETDRS grid decreased from (56.01 ± 10.53)% in the control group to (52.16 ± 8.75)% in the No-DR group and (47.91 ± 12.95)% in the Mild-DR group (P = 0.001), while in the outer ETDRS grid, it declined from (42.92 ± 6.70)% to (40.20 ± 7.24)% and (38.13 ± 8.78)%, respectively (P = 0.008). VD in the inner ETDRS grid showed a reduction from (15.41 ± 2.68) mm⁻¹ in the control group to (14.39 ± 2.17) mm⁻¹ in the No-DR group and (13.55 ± 2.98) mm⁻¹ in the Mild-DR group (P = 0.001), while in the outer ETDRS grid, it decreased from (13.74 ± 2.26)mm⁻¹ to (12.78 ± 2.53)mm⁻¹ and (12.09 ± 2.83)mm⁻¹, respectively (P = 0.006). Significant reductions in PD were observed in the nasal quadrant of the inner ring and the supratemporal quadrant of the outer ring, while VD showed significant decreases in the nasal quadrants of the inner ring and the supratemporal and inferotemporal quadrant of the outer ring. Moreover, BMO-MRW was positively correlated with PD (r = 0.583, P = 0.001) and VD (r = 0.499, P = 0.001) in the inner ring, with positive correlations observed between BMO-MRW and inner PD and VD in the counterpart quadrants.

Conclusions

Peripapillary BMO-MRW is significantly reduced in early DR, particularly in the superotemporal and inferotemporal quadrants, alongside declines in PD and VD. The positive correlation between BMO-MRW and microvascular parameters suggests its potential as a biomarker for early DR detection and monitoring.

Background

Diabetic retinopathy (DR), the leading cause of preventable vision loss in working-age populations [1], is traditionally classified as a microvascular disorder [2]. Emerging evidence indicates that retinal neurodegeneration precedes clinically detectable microvascular changes, manifesting as reduced contrast sensitivity, visual field defects, and electrophysiological abnormalities in diabetic patients [3, 4]. Preclinical studies in diabetic models and human donors reveal early neurodegenerative features—reactive gliosis, neuronal dysfunction, and apoptosis—prior to microaneurysm formation [5,6,7]. The neurovascular unit (NVU), a concept introduced by Rafael et al. [8], highlights the interdependence of neurons, glia, and micro-vessels in maintaining blood-retinal barrier integrity and regulating retinal perfusion [9]. Despite this, early DR lacks distinct clinical signs, and interventions remain largely reactive, targeting advanced stages with overt complications [10]. Improved understanding of early pathogenesis is critical for developing preventive strategies. Advances in optical coherence tomography (OCT) enable high-resolution retinal imaging, revealing subclinical neurodegeneration in diabetics without DR [11]. Thinning of the ganglion cell-inner plexiform layer (GC-IPL) and retinal nerve fiber layer (RNFL) precedes vascular damage, with accelerated RNFL thinning rates observed longitudinally [11,12,13]. Hyperglycemia disrupts NVU homeostasis via basement membrane thickening, pericyte loss, and neuroglial activation, exacerbating microcirculatory dysfunction [14]. Recent OCT innovations, such as Bruch’s membrane opening-minimum rim width (BMO-MRW), demonstrate superior sensitivity to neurostructural damage compared to RNFL thickness (RNFLt) [12], though its utility in DR remains underexplored. Optical coherence tomography angiography (OCTA) studies confirm preclinical microvascular alterations in diabetics without DR, including reduced superficial/deep vessel density (VD) and perfusion density (PD) in paracentral, pericentral, and peripapillary regions [14,15,16]. However, existing OCTA protocols suffer from limitations: restricted scan areas (4.5 × 4.5 mm) [17, 18], crude quadrant-based analyses [19, 20], and failure to account for individual anatomical variability in the fovea-to-BMO (FoBMO) axis [21]. Standardized vertical/horizontal partitioning misaligns with patient-specific FoBMO angles (ranging 2.5° to -17.5°), compromising spatial accuracy [22].

This study employs FoBMO-axis calibrated OCTA to evaluate BMO-MRW and microvascular changes in early DR, investigating structural and vascular alterations concurrently.

Methods

This observational cross-sectional study was conducted in the Department of Ophthalmology at Shanghai Sixth People’s Hospital from January 2021 to March 2025. This study conformed to the tenets of the Helsinki Declaration and was approved by the Ethics Committee of the Shanghai Sixth People’s Hospital (No.2021-KY-091(K)). Informed consent was obtained from all the participants. Diabetic patients were recruited from Shanghai Sixth People’s Hospital. Healthy controls were recruited from the medical examination center at the same hospital during the same period. For each diabetic patient, a detailed medical history, including diagnoses, medications, disease duration and laboratory tests, was recorded. We extracted the clinical data from recorded free text clinical notes and inpatient medical records. Inclusion criteria were (1) aged 18 years or older, (2) Whether T2DM was coupled with no DR or mild DR was evaluated by a retinal specialist through indirect fundoscopy and ultra widefield fundus photographs (Optomap Panoramic 200; Optos PLC). Mild DR was defined as microaneurysms only, according to the International Clinical Diabetic Retinopathy Disease Severity Scale.(3) had a refractive error within − 3.00 diopters or + 3 diopter-equivalent spheres. (4)BCVA ≤ 0.1logMAR; (5)22 mm< axial length<26 mm; (6) intraocular pressure<21mmHg. The exclusion criteria were as follows: (1) the presence of diabetic macular edema (DME); (2) having undergone previous intraocular surgery; (3) presenting with a disease affecting retinal function, including glaucoma, advanced cataracts or significant media opacity, or with other serious systemic disease.

All patients underwent comprehensive ophthalmic examination, including the ETDRS (Precision vision) protocol visual acuity, slitlamp biomicroscopy, intraocular pressure measurement (TX-20 Canon, JAPAN) indirect stereoscopic fundus examination, and OCTA Spectralis HRA + OCT (Heidelberg Engineering, Germany).

Image acquisition and analysis

The radial scanning mode in the glaucoma module was selected, and 24 radial B-scan patterns were selected, with an angle of 7.5° between each B-scan line, and each image could also consist of 100 images superimposed on each other. A 6-mm × 6-mm macular scan was performed on both eyes of all participants using a Heidelberg OCTA Spectralis HRA + OCT (Heidelberg Engineering, Germany). The light source had a wavelength of 870 nm and contained 512 B-scans, with five consecutive scans for each B-scan of the same area. The participant’s jaw was fixed in the jaw holder, and the refractive compensation was adjusted using the machine’s refractive buttons through the internal fixation and kinesio-tracking modes of technology, and the images were acquired when the image was clear. A single 3D volumetric scan could acquire SLO, En-Face images, OCT images, and angiograms at the same time. All examinations are performed by the same physician(JZ). The scanning procedure was as follows: adjust the intracamera to the nasal side, select the OCTA mode, select a 6 × 6 mm rectangular scanning frame, and adjust the scanning frame to be centered on the optic papilla to perform the OCTA scan, which consists of 516 B-scans with an interval of 12 μm between each B-scan.

Peripapillary microvascular assessment was performed at the level of the superficial vascular complex (SVC). According to the parameter settings of the Heidelberg OCTA scanning equipment, the SVC layer was defined as 17 μm from the inner limiting membrane (ILM) to the IPL, and therefore the optic disc vascular density in this study included the radial peripapillary capillary (RPC), the superficial vascular plexus (SVP), and the peripapillary capillary (PPC). Images with obvious artifacts and signal intensity index < 20 were excluded, and only images with good imaging quality were included.

OCT angiography parameters

Image analysis was performed using FIJI, an open-source distribution of the program ImageJ. (i) Open the OCTA images in ImageJ, convert the OCTA images to 8-bit, and then analyze the OCTA images after the automatic “default” image analysis provided by the ImageJ software. The OCTA image is converted to 8-bit, automatically thresholded by the “default” threshold provided by the ImageJ software to neutralize the background noise, then converted to a binarized black-and-white image (Fig. 1A). This binarized image is used to calculate the Perfusion Density (PD): number of vascular pixels/total pixels in the analyzed area (%). (ii) A skeletonized image was then created from the binarized image, in which all vessels (including large vessels) were normalized to one pixel width, and all pixels in the skeleton image were measured and the vessel density (VD) was calculated as the number of vascular pixels × scanning width (mm)/area (mm²) [17]. The region of interest (ROI) function of ImageJ software was used to partition the binarized and skeletonized optic disc OCTA images: first, the macula-optic disc axis (FoBMO axis) was marked on the original OCTA image and the FoBMO angle of the image was calculated, and then the rotational angle of the corresponding ROIs was set using this angle to perform ETDRS grid partitioning. ETDRS grid partitioning (three concentric circles with diameters of 1-, 3-, and 6-mm are used to divide the image into an inner ring between 1 and 3 mm and an outer ring between 3 and 6 mm. The inner and outer rings were divided into six quadrants of Superonasal, supratemporal, lateral temporal, Inferotemporal, Inferonasal, and lateral nasal, with a total of twelve regions using ROIs adjusted according to the FoBMO-axis (Fig. 1B), and the PD and VD values of the inner and outer rings as well as those of each region were acquired separately(Fig. 1C and D). All images were saved and analyzed anonymously.

Statistical analysis

Baseline data of the three groups of subjects were compared and quantitative variables were expressed as (mean ± standard deviation). Qualitative variables were expressed as number, for the included studies data were analyzed by Kolmogorov-Smirnov test for obeying normal distribution, one-way ANOVA was used for continuous normally distributed variables, and differences between groups were performed using Bonferroni post hoc test. Independent samples t-test Mann-Whitney U test was performed to compare continuous variables. Categorical variables were tested using chi-square test.

Correlation between variables was assessed using Pearson correlation test. All analyses were performed using the statistical version of the software SPSS 26.0 (IBM Corporation, Chicago, IL, USA) and differences were considered statistically significant at P < 0.05.

Fig. 1

Image processing and measurement of PD and VD. A, Original grayscale OCT angiography image of peripapillary vascular. B, a total of twelve regions using ROIs adjusted according to the specific FoBMO angle. C, calculated twelve regions of perfusion density after binarized. D, calculated twelve regions of vessel density after skeletonized

Results

The study included 105 eyes from 105 diabetic patients, comprising 55 eyes without diabetic retinopathy (No-DR group) and 50 eyes with mild diabetic retinopathy (Mild-DR group), and 50 healthy eyes from 50 participants of the control group. The baseline information is shown in Table 1. There was no statistically significant difference in age, gender, laterality, spherical equivalent, intraocular pressure, FoBMO angle and axial length between the three groups. The BCVA of control group was significantly better than No-DR group and Mild-DR group (P = 0.004), and the difference in the duration of diabetes mellitus and glycosylated hemoglobin (HbA1c) between the two groups of diabetic patients was not statistically significant (P = 0.825, 0.077, respectively).

Table 1 Demographic and clinical characteristic of three groups

BMO-MRW

The average BMO-MRW and values for the ETDRS quadrants in the control, No-DR, and Mild-DR groups are presented in Table 2. The average BMO-MRW were (323.56 ± 40.33) µm, (305.78 ± 35.12) µm, and (299.42 ± 37.33) µm for the control, No-DR, and Mild-DR groups, respectively (P = 0.005). Both the No-DR and Mild-DR groups had significantly lower BMO-MRW compared to the control group (P = 0.017, 0.002, respectively), though no significant difference was observed between the No-DR and Mild-DR groups (P = 0.388). In the ETDRS quadrants, significant decreases in BMO-MRW were observed between the three groups, particularly in the supratemporal and inferotemporal quadrants. In the supratemporal quadrant, BMO-MRW decreased from (325.16 ± 43.30) µm in the control group to (308.35 ± 39.72) µm in the No-DR group, reaching the lowest value of (304.52 ± 45.33) µm in the Mild-DR group (P = 0.046, 0.017), with no significant difference between the No-DR and Mild-DR groups (P = 0.648). In the inferotemporal quadrant, the BMO-MRW values were (357.18 ± 60.07) µm, (336.62 ± 44.94) µm, and (332.68 ± 53.39) µm for the control, No-DR, and Mild-DR groups, respectively (P = 0.047). Both the No-DR and Mild-DR groups were significantly lower than the control group (P = 0.049, 0.002), but no significant difference was found between the No-DR and Mild-DR groups (P = 0.489). No statistically significant differences in BMO-MRW were found between the three groups in the superonasal, lateral nasal, inferonasal, and temporal quadrants.

Table 2 Comparison of the BMO-MRW among three groups

Peripapillary perfusion density (PD)

The distribution of peripapillary vascular perfusion density (PD) in the optic disc for the control, No-DR, and Mild-DR groups is presented in Table 3. The average PD in the inner ring was (56.01 ± 10.53)%, (52.16 ± 8.75)%, and (47.91 ± 12.95)% for the control, No-DR, and Mild-DR groups, respectively (P = 0.001). The Mild-DR group exhibited significantly lower PD compared to the control group (P < 0.001) and the No-DR group (P = 0.046), with no significant difference observed between the No-DR and control groups (P = 0.071). In the inner ring of the ETDRS grid, significant differences were found in the nasal quadrant (P < 0.001), although no significant differences were observed in other quadrants. For the outer ring of the ETDRS grid, the average PD values were (42.92 ± 6.70)%, (40.20 ± 7.24)%, and (38.13 ± 8.78)% for the three groups, respectively (P = 0.008). The Mild-DR group showed significantly lower PD compared to the control group (P = 0.002), but no significant difference was found between the Mild-DR and No-DR groups (P = 0.069) or between the Mild-DR and control groups (P = 0.168). The only significant difference in the outer ring was found in the superotemporal quadrant (P = 0.024), where the PD values in the Mild-DR group were significantly lower than those in the control group (P = 0.007). No significant difference was observed between the Mild-DR group and No-DR or between the Mild-DR and control groups (P = 0.272, 0.091).

Peripapillary vascular density (VD)

The distribution of peripapillary vascular density (VD) in the optic disc across the control, No-DR, and Mild-DR groups was also presented in Table 3. The average VD in the inner ring was (15.51 ± 2.66) mm⁻¹, (14.52 ± 2.20) mm⁻¹, and (12.89 ± 2.97) mm⁻¹ for the control, No-DR, and Mild-DR groups, respectively (P = 0.002). The Mild-DR group showed a significantly lower VD than the control group (P < 0.001), but no significant difference was found between the Mild-DR and No-DR groups (P = 0.052) or between the Mild-DR and control groups (P = 0.102). In the outer ring, a significant difference in VD was observed only in the nasal quadrant among the three groups (P < 0.001). The average VD values in the outer ring were (13.74 ± 2.26) mm⁻¹, (12.78 ± 2.53) mm⁻¹, and (12.09 ± 2.83) mm⁻¹ for the control, No-DR, and Mild-DR groups, respectively (P = 0.006). Significant differences in VD in the outer ring of the ETDRS grid were found in the inferotemporal and superotemporal quadrants (P = 0.025, 0.044), with VD in the Mild-DR group significantly lower than in the control group (P = 0.010, 0.013). In the inferotemporal quadrant, a significant difference was observed between the control and No-DR groups (P = 0.025), although no significant difference was found between the Mild-DR and control groups. In the superotemporal quadrant, no significant difference was observed between the Mild-DR and No-DR groups or the control group (P = 0.314, 0.122).

Table 3 Comparison of peripapillary perfusion density and vessel density among three groups

Correlation analysis of BMO-MRW and perfusion density/vessel density

Table 4 shows correlations between BMO-MRW values (average and separately in each quadrant) and PD/VD. The average PD in the inner ring was positively correlated with the average BMO-MRW(r = 0.583, P = 0.001). Superonasal, nasal, Inferonasal, inferotemporal, temporal, and supratemporal quadrant inner ring PDs were all positively correlated with the BMO-MRW of the corresponding quadrants. PDs were positively correlated with the BMO-MRW of the outer ring of the supratemporal quadrant (r = 0.202, P = 0.013). The average VD in the inner ring was positively correlated with the average BMO-MRW(r = 0.499, P = 0.001). The Superonasal, nasal, Inferonasal, inferotemporal, temporal, and supratemporal quadrant inner ring VD were all positively correlated with the BMO-MRW of the corresponding quadrants, VD were positively correlated with the BMO-MRW of the outer ring of the supratemporal quadrant (r = 0.233, P = 0.004).

Discussion

This study employed Heidelberg OCTA to evaluate changes in BMO-MRW and the superficial capillary layer of the peripapillary retina in healthy individuals, patients without diabetic retinopathy (No-DR), and those with early diabetic retinopathy (Mild-DR). Our study focused on the peripapillary optic disc region and used BMO-MRW, a more sensitive indicator than RNFLt, to assess nerve fiber changes and analyze the relationship between optic nerve thickness and microvascular alterations [23, 24]. Our findings indicate that, compared to healthy individuals, diabetic patients exhibited a significant reduction in BMO-MRW, with this decrease already evident in the No-DR group. Using the ETDRS grid, we further analyzed BMO-MRW of the optic disc in six quadrants. In both the No-DR and Mild-DR groups, BMO-MRW in the superior temporal and inferior temporal quadrants was significantly thinner than in the control group. Additionally, all data were partitioned based on the FoBMO axis, as the FoBMO angle varies significantly between individuals and is highly susceptible to examination posture, potentially affecting DR progression assessment. Compared to previous methods, our approach aligns more closely with anatomical structures and may more accurately reflect the true pathophysiological changes in diabetic patients [25, 26].

Table 4 Correlation analyses between BMO-MRW and perfusion density/vessel density (average and six quadrant)

Sabry et al. [27] also investigated BMO-MRW changes in patients without diabetic retinopathy. Their results showed no significant difference in the mean BMO-MRW between healthy controls and diabetic patients. In the supratemporal quadrant, BMO-MRW decreased from (342.94 ± 52.19) µm in the No-DR group to (323.60 ± 52.42) µm in the healthy control group, while in the inferotemporal quadrant, it decreased from (362.84 ± 60.24) µm to (356.92 ± 60.24) µm. However, no significant difference was found between the two groups. Considering that this study included only 19 cases (38 eyes) in both the control and No-DR groups, the small sample size may have influenced the results.

Vascular analysis of the optic disc showed that in both the inner and outer rings, vessel density in the superior and inferior temporal quadrants was higher than in other quadrants, indicating a relatively richer blood supply in these regions. Anatomically, nerve fibers from the macula and peripheral retina converge in an arc toward the optic disc along the superior and inferior sides of the macula-optic disc axis. Consequently, the superior temporal and inferior temporal regions of the optic disc exhibit denser vasculature than other areas [27]. Additionally, the superficial capillary layer of the optic disc, particularly the radial peripapillary capillaries (RPC), runs parallel to the nerve axons, forming relatively straight vessels with limited anastomoses, making them less tolerant to hypoxia and damage [28]. This anatomical arrangement renders them more susceptible to hypoxic injury. The high neural density in the superotemporal and inferotemporal quadrants further contributes to an increased oxygen demand, potentially exacerbating their vulnerability to structural alterations under hyperglycemic conditions, as reflected by the observed BMO-MRW thinning. Notably, this change was not observed in PD in the inner ring of the superotemporal and inferotemporal quadrants, whereas vessel density (VD) alterations were detected in the outer ring. As a skeletonized image, VD standardizes the influence of both large and small vessels, reducing the effect of vessel diameter on retinal perfusion measurements. This makes VD theoretically more accurate than PD in assessing vascular changes [29, 30].

Most previous studies have focused on the macular region in DR patients, identifying structural changes across different retinal layers [7, 31, 32]. Some have also explored microvascular changes around the optic disc in diabetic patients, demonstrating a significant correlation between RPC vessel density and retinal nerve fiber layer thickness (RNFLt) [33,34,35]. In an OCTA-based study, Cao et al. compared optic disc vessel density and RNFLt in normal individuals and hyperglycemic patients without diabetic retinopathy [36]. They found that all quadrants of the No-DR group exhibited significant vessel density changes, whereas RNFLt showed significant changes only in the superior nasal, inferior nasal, and nasal quadrants—differing from the BMO-MRW changes observed in our study. These discrepancies may stem from differences in imaging equipment, observation area size, and partitioning methods. In our study, changes were detected in the superior temporal and inferior temporal quadrants, whereas in their study, these regions were partly classified under the inferior nasal and superior nasal quadrants. This underscores the importance of considering partitioning methods when comparing results across studies and highlights the need for standardized optic disc segmentation.

Additionally, our study found that in No-DR patients, PD and VD changes in the nasal and inferonasal quadrants of the inner ring occurred earlier than BMO-MRW changes. This aligns with previous findings indicating thinning in these regions, suggesting that nasal quadrant blood flow alterations warrant further investigation. Earlier research using RNFL to study diabetic patients divided the optic disc region into superior, nasal, inferior, and temporal quadrants, identifying early RNFL thinning in the inferior quadrant of diabetic patients without DR. Our study used BMO-MRW, a more sensitive parameter than RNFLt, and similarly found that BMO-MRW in the superior temporal and inferior temporal quadrants was already significantly thinner in diabetic patients without DR compared to healthy individuals. However, in these groups, PD and VD changes around the optic disc were not statistically significant compared to normal individuals. This suggests that BMO-MRW may serve as a valuable biomarker for distinguishing and monitoring diabetic retinopathy progression in diabetic patients.

We further investigated the correlation between BMO-MRW and PD/VD around the optic disc and found a positive correlation in the corresponding quadrants of the standardized ETDRS inner ring. This may be because the inner ring is anatomically closer to the BMO-MRW, leading to a stronger correlation. The observed thinning of neural tissue and decreased vascular density in early DR suggest a potential relationship between neural tissue integrity and vascular density, reinforcing the concept of the “optic nerve vascular unit.” Diabetic patients without fundus lesions exhibit lower VD and PD in the optic disc region than healthy individuals, indicating that optic disc vascular density could serve as an indicator of microvascular changes in diabetes.

There are still several limitations in this study. First, as a cross-sectional observational study, it cannot establish a causal relationship between neurodegenerative changes and microvascular alterations in the optic disc region of diabetic patients. Future prospective longitudinal studies are necessary to clarify this temporal relationship. Second, our vascular density measurement method differs from previous studies; we used a 6 mm × 6 mm blood flow density assessment of the optic disc region, simultaneously observing the superficial capillaries, radial peripapillary capillary layer, and macrovascular structures. Currently, no ideal method exists to evaluate the peripapillary capillary layer separately. Future advancements in ophthalmic imaging technology may enable the development of techniques and algorithms capable of distinguishing capillary and macrovascular layers. Furthermore, based on the observed neural and vascular structural changes in DR patients, a feasible preclinical grading system for DR could be established. Through close follow-up and precise interventions, we may delay or even prevent DR progression, ultimately protecting more patients from severe vision loss and irreversible retinal damage.

Conclusions

This study demonstrates that peripapillary BMO-MRW is significantly reduced in early diabetic retinopathy, particularly in the superotemporal and inferotemporal quadrants. Concurrent declines in peripapillary perfusion density and vessel density were observed, notably in the nasal, superotemporal, and inferotemporal quadrants. The positive correlation between BMO-MRW and both PD and VD suggests a strong association between neuroretinal rim thinning and microvascular changes in early DR. These findings underscore the potential of BMO-MRW and OCTA-derived microvascular metrics as early biomarkers for DR detection and progression monitoring.