Diabetic retinopathy (DR) is one of the most common and serious microvascular complications of diabetes and is one of the leading causes of blindness in China (1). According to statistics, the number of people with visual impairment or blindness due to DR has increased significantly, and the prevalence has continued to rise from 1990 to 2020 (2). It is widely recognized that the risk factors for DR include diabetes duration, glycated hemoglobin (HbA1c) levels, hypertension, pregnancy, hyperlipidemia, nephropathy, and alcohol abuse (3). The study by Perais et al. demonstrated that, regardless of the severity of DR, the degree of good glycemic control helps reduce the risk of progression to proliferative diabetic retinopathy (PDR) (4). In the early stages of DR, patients are often asymptomatic. However, as the disease progresses, serious complications such as vitreous hemorrhage and tractional retinal detachment may occur, ultimately leading to blindness.

Fundus fluorescein angiography (FFA) remains the gold standard for diagnosing DR, with clear advantages for identifying microvascular abnormalities and retinal neovascularization. However, as FFA is an invasive procedure with contraindications, its application is limited. Clinically, spectral-domain optical coherence tomography (SD-OCT) for detecting biological imaging biomarkers holds broad application prospects. As early as 1988, Hee et al. found that OCT aids in diagnosing early DR microlesions, including macular retinal thickening (5). An increasing number of studies suggest that OCT-based biological imaging biomarkers can serve as new indicators of visual prognosis and treatment strategies in DR patients. These indicators include intraretinal cysts (IRC), disorganization of inner retinal layers (DRIL), hyperreflective foci (HRF) in the outer retina, and the integrity of the ellipsoid zone (EZ) and external limiting membrane (ELM), as well as diabetic macular edema (DME).

In addition, autofluorescence primarily detects retinal lipofuscin, which reflects the function of the retinal pigment epithelium (RPE). FAF is most used in age-related macular degeneration (AMD). Fleckenstein et al. summarized various early manifestations of AMD, including multifocal and mottled hyperautofluorescent dots (6). In advanced AMD, FAF can reveal hypoautofluorescent geographic atrophy lesions. Currently, there is limited research on FAF in DR, but studies have shown that petal-like macular edema can be clearly visualized on FAF.

The objective of this study was to describe the features of biological imaging biomarkers in patients with non-proliferative diabetic retinopathy (NPDR) and proliferative diabetic retinopathy (PDR) using OCT and FAF.

Of our 132 patients (219 eyes), 78 (59.09%) were male, and 54 (40.91%) were female. Patient ages ranged from 23 to 76 years (mean age 52.75 ± 14.17 years). Among these 132 patients, 120 (54.79%) eyes had NPDR, and 99 (45.21%) had PDR. Seventy-three healthy individuals with normal characteristics were selected as the control group, with an average age of 37.33 ± 15.72 years. Basic information on this study is shown in Table 1, and the optical coherence tomography features of NPDR and PDR are summarized in Table 2. Compared with NPDR, eyes with PDR showed a significantly higher prevalence of hard exudates (69.7% vs. 42.5%, p < 0.001), cotton-wool spots (48.5% vs. 22.5%, p < 0.001), and hyperreflective foci (87.9% vs. 62.5%, p < 0.001). Posterior vitreous detachment was also more frequently observed in PDR eyes than in NPDR eyes (72.7% vs. 45.0%, p < 0.001). Intraretinal cystic cavities and macular edema were significantly more common in PDR eyes (48.5% vs. 32.5%, p = 0.023; and 93.9% vs. 40.0%, p < 0.001), indicating more severe retinal structural involvement in advanced disease stages. In contrast, no significant differences were found between the two groups regarding subretinal fluid, disorganization of retinal inner layers, epiretinal membrane, or disruption of the external limiting membrane and ellipsoid zone (all p > 0.05). Notably, preretinal hemorrhage and fibrous proliferative membrane were observed exclusively in the PDR group (9.1% and 27.3%, p < 0.01), reflecting hallmark features of proliferative disease.

DR is one of the microvascular complications caused by diabetes and a leading cause of blindness in adults (1). The treatment plan also varies with DR severity. FFA remains the gold standard for diagnosing and staging DR and plays a crucial role in treatment planning. However, FFA has limitations, and some patients are unable to undergo the procedure due to contraindications, such as allergies or abnormal liver and kidney function. FAF has been widely recognized for its advantages in diagnosing age-related macular degeneration (ARMD) (6), hereditary retinal diseases (7), and central serous chorioretinopathy (8). Retinal function as reflected by FAF in DR provides new insights into diagnosis and treatment planning. SD-OCT can detect subtle retinal structural changes, which are significant for evaluating DR progression and prognosis. By providing a side-by-side multimodal comparison, this work offers practical insights for clinical risk stratification, particularly identifying subclinical structural changes that traditional examinations might miss.

In our study, microaneurysms (MA) and intraretinal hemorrhages were observed in all DR patients, indicating that they are fundamental retinal lesions in DR. In OCT, MA appears as a medium-reflective, punctate lesion in the inner retinal layers, measuring greater than 30 μm in diameter, with weak fluorescence on FAF. Intraretinal hemorrhages appear as variably sized hyperreflective foci in the superficial and/or inner retinal layers on OCT, accompanied by blocking fluorescence in FAF. OCT examination revealed that MA primarily originates from the inner nuclear layer. A prospective study has demonstrated that a high MA formation rate is a biomarker for the development of clinically significant macular edema in patients with type 2 diabetes (9). Additionally, research has shown that MAs originating from the deep capillary layer are strongly correlated with CME (10). Compared with MA, intraretinal hemorrhages are more easily detected on FAF. Our comparison of FFA and OCT reveals that some small MAs may not be visible in FAF, possibly because the blocking fluorescence caused by intraretinal hemorrhages is more distinguishable against the background of retinal autofluorescence than the weak fluorescence produced by MAs. Similarly, FAF is highly effective for detecting preretinal hemorrhages. In our study, FAF detected preretinal hemorrhages in all PDR patients.

Hyperreflective foci was first described by Coscas et al. using OCT in age-related macular degeneration (11). In OCT, HRF appears as a medium-reflective focus with a diameter of less than 30 μm in the retina. The formation is primarily attributed to the aggregation of activated microglia within the retina, or the migration of monocytes from the blood into the retina, which differentiate into macrophages (12). These aggregates, due to their high metabolic activity or lipid components, manifest as HRF in OCT. Some studies suggest that HRF may also originate from extravasated lipoproteins, cellular debris, or metabolic byproducts (13). In our research, HRF was predominantly located in the inner retinal layers of both patients with NPDR and PDR. In our study, HRF was detected in 87 eyes of the PDR group and 75 eyes of the NPDR group. Additionally, the number of HRFs was significantly higher in the PDR group than in the NPDR group. HRF cannot be distinguished in fundus photography, FFA, or FAF. HRF has been observed in patients with wet AMD, CME, retinal vein occlusion (RVO), central serous chorioretinopathy (CSC), retinitis pigmentosa, and other conditions leading to choroidal neovascularization (14, 15). In recent years, HRF has attracted significant attention as a biomarker of DME treatment efficacy. The presence and increased number of HRFs indicate a poor response to anti-VEGF therapy in patients with DME.

Studies have shown a relationship between HRF and ELM disruption (16). Since the ELM prevents the infiltration of large molecules into the retina, it suggests that HRF may originate from damage to the blood-retinal barrier. The ELM provides structural support for glial cells and photoreceptors. Its instability can lead to misalignment of photoreceptors. The EZ, or ellipsoid zone, primarily consists of mitochondria. Damage to these structures can be directly visualized in OCT as ELM/EZ disruption or absence. Studies have shown that damage to ELM and EZ not only leads to the appearance or increase of HRF but also causes a significant decline in patients’ vision and a poorer prognosis (17). In this study, ELM and EZ disruption were observed in 18 eyes of the NPDR group and six eyes of the PDR group. However, the number of eyes with HRF was significantly higher than those with ELM or EZ disruption in both groups, indicating that HRF is associated not only with ELM disruption but also with other factors. In OCT, hard exudates appear as hyperreflective foci with a diameter greater than 30 μm, and some studies suggest that HRF may be a precursor to hard exudates. Mori et al. found that increased VEGF disrupts the integrity of the EZ and ELM (18). Therefore, observing changes in the EZ and ELM via OCT is of significant value for assessing retinal function and prognosis in patients with DR.

Disorganization of retinal inner layers refers to the irregular structural disruption of the inner plexiform layer (IPL), inner nuclear layer (INL), and outer nuclear layer (ONL), a common imaging feature in DR that may be associated with blood-retinal barrier breakdown and macular ischemia. Our study found that the structural disruption in the inner retinal layers at the cotton-wool spots on OCT is similar to that in DRLL. In OCT, cotton-wool spots appear as hyperreflective thickening of the nerve fiber layer with underlying artifacts. It has been reported that DRIL can reflect retinal capillary non-perfusion areas caused by ischemia (19). DRIL and cotton-wool spots are closely associated with retinal ischemia, potentially reflecting a cascade of ischemic damage from the nerve fiber layer to the inner retinal layers.

Diabetic macular edema can be classified into DRT and CME based on OCT examination results. In this study, it was found that DR patients can develop a mixed pattern of DME, which is consistent with the findings of Kim et al. (20). Soliman et al. found that large cysts in the outer nuclear layer and/or Henle layer on OCT correspond to petaloid cystic leakage on FFA (21). In our study, FFA revealed macular cystoid edema in 6 eyes (5%) of NPDR patients and 15 eyes (16.13%) of PDR patients, findings consistent with those of FAF. This suggests that FAF can serve as an alternative to FFA and OCT for observing typical macular cystoid edema, aligning with the results of McBain et al.

It has been shown that PVD is also involved in DME development (22). In diabetic patients, increased cross-linking of vitreous collagen and thickening of the posterior vitreous cortex may lead to macular traction and macular edema. PVD has both protective effects and is a risk factor for the worsening of DME. In DR patients with PVD, the incidence of macular edema is lower, and some patients experience a reduction in macular edema after PVD. On the other hand, when PVD is accompanied by retinal proliferation, it can exacerbate macular edema or lead to retinal tears or detachment. In our study, PVD was observed in 72 eyes of PDR patients, which was higher than in the 54 eyes of NPDR patients. Additionally, the incidence of DME was higher in PDR eyes than in NPDR eyes.

The formation of SRF is associated with increased vascular permeability, driven by mechanisms such as blood-retinal barrier disruption and elevated VEGF levels (23). In OCT, SRF appears as a hyporeflective area beneath the retinal neuroepithelium, with an intact RPE layer, suggesting that its formation results from cellular barrier dysfunction rather than structural damage. This finding is also supported by Urner-Bloch et al. (24). SRF can serve as a biological marker for evaluating DME and the therapeutic efficacy of anti-VEGF treatment. In our study, SRF was observed in 15 eyes of the NPDR group, compared to 6 eyes in the PDR group. The presence of SRF is associated with visual decline in patients. DME patients with SRF may exhibit better therapeutic efficacy or anatomical recovery after anti-VEGF treatment. Therefore, investigating SRF on OCT in DR patients can also serve as a metric for assessing the effectiveness of DME treatment.

In our study, we found that the proportion of ERM was very high in DR patients, both in eyes with NPDR and in those with PDR. This phenomenon has not been clinically recognized or emphasized. SD-OCT examination revealed ERM in 96 out of 120 eyes (80%) of NPDR patients and 84 out of 99 eyes (84.8%) of PDR patients, with varying degrees of severity. The high incidence of ERM suggests a close association with DR. This high prevalence is attributed to the inclusion of early-stage, subclinical ERMs, characterized by localized hyperreflective bands on SD-OCT, which are often underdetected by fundus photography but are clearly visible on high-definition line scans. The ERM in PDR eyes is more severe than that in NPDR eyes and may be accompanied by proliferation. The formation of ERM is closely associated with Müller cells, and fibroblast growth factor 2 (FGF2) has been shown to induce Müller cell activation in patients with PDR (25). In DR patients, the retina has already been damaged by diabetic microvascular disease. The traction of the ERM may exacerbate retinal ischemia and hypoxia, leading to changes in the inner retinal reflectivity and increased macular thickness, thereby accelerating the progression of DR. Additionally, studies have found inflammatory cells in the fibrovascular membranes of patients with PDR (26), which may be one of the factors contributing to the development of ERM. In our study, the fibrovascular membrane was well visualized in FAF, appearing as weak fluorescence, while the surrounding retinal and vascular structures exhibited abnormal tractional morphology. In contrast, FFA may show strong fluorescence from neovascularization, which can obscure the membrane’s morphology.

In our study, there was no statistically significant difference in SFCT between the NPDR and PDR groups. However, SFCT was significantly thicker in both NPDR and PDR patients compared with the healthy control group. The findings of Kim et al. (27) are consistent with ours, and Savage et al. (28) further found that choroidal blood flow was significantly reduced in patients with PDR after treatment. This suggests that the physiological state of the retina and the biochemical reactions occurring therein influence the normal function of the choroidal vasculature, leading to choroidal vasodilation, increased permeability, and, consequently, increased choroidal blood flow and thickness. A deeper exploration of the potential link between choroidal thickness and the pathogenesis of DR could provide a theoretical basis for DR treatment.

This study provides a comprehensive and multimodal characterization of imaging biomarkers in DR by integrating SD-OCT, fundus autofluorescence, and FAF. Unlike previous studies that focused on single imaging modalities or isolated biomarkers, our work systematically compared the detection capability of different imaging techniques across a broad spectrum of retinal lesions in both NPDR and PDR. Notably, we demonstrated a remarkably high prevalence of epiretinal membrane in DR, which is underrecognized in routine clinical evaluation and may be an important structural indicator of disease severity. In addition, this study highlights the complementary role of FAF in visualizing retinal hemorrhage, DME, and fibrovascular proliferative membranes, suggesting its potential value as a non-invasive adjunct when FFA is contraindicated. By providing a side-by-side comparison of multimodal imaging features, this work offers practical insights for clinical assessment and may contribute to improved phenotyping and risk stratification in DR.

This study has several limitations that should be acknowledged. First, this was a cross-sectional imaging study, which limits the ability to establish causal relationships or evaluate longitudinal changes of imaging biomarkers during disease progression. Second, although multiple OCT, FAF, and FFA features were systematically analyzed, quantitative measurements such as lesion area, volume, or intensity were not performed, and the analysis mainly relied on the presence or absence of imaging features. Third, OCT examinations were primarily focused on the macula, which may have led to underestimation of peripheral retinal lesions, particularly cotton-wool spots, hard exudates, and microaneurysms located outside the posterior pole. Fourth, patients with poor image quality due to media opacity, severe vitreous hemorrhage, or tractional retinal detachment were excluded, which may introduce selection bias and limit the generalizability of the findings to more advanced stages of diabetic retinopathy. Finally, potential systemic confounders, such as diabetes duration, glycemic control, and systemic treatment, were not incorporated into the imaging analysis and should be considered in future studies. Notably, the significant age gap between the control and DR groups is a major limitation. Although multivariate analysis was performed to adjust for age, a perfectly age-matched cohort would provide more robust results. Furthermore, the lack of AL measurements is a critical omission, as AL is a known determinant of choroidal thickness. Future studies should include AL as a covariate to fully validate the findings regarding choroidal thickening in DR. Regarding the results detected by ELM/EZ, The limited macular scan area may have led to an underestimation of total retinal structural disruption in PDR eyes, particularly in the peripheral retina.

As a non-invasive retinal examination technique, FAF offers unique advantages for detecting retinal hemorrhage, cystoid macular edema, and retinal fibrovascular proliferative membranes. We have an important finding: SD-OCT revealed a significant increase in the incidence of macular epiretinal membranes in DR patients, with a higher prevalence in PDR than in DR. SD-OCT can detect multiple imaging biomarkers in DR patients. By integrating features from FAF, fundus photography, and FFA, this comprehensive analysis aids in diagnosing DR and provides effective prognostic predictions for DR patients.