Classification of diabetic macular edema cases based on optical coherence tomography images and evaluation of morphological features

Abstract

Objectives

This study aimed to classify DME morphologically using optical coherence tomography (OCT) and to assess the impact of selected OCT biomarkers on DME subtypes.

Methods

OCT images of patients with DME categorized into three groups as diffuse macular edema (dME), cystoid macular edema (CME) and cystoid degeneration (CD). Presence of subretinal fluid (SRF), disorganization of inner retinal layers (DRIL), ellipsoid zone (EZ) damage, external limiting membrane (ELM) damage, exudates, hyperreflective dots (HRD), epiretinal membrane (ERM) and vitreomacular traction (VMT) were evaluated. The effect of OCT findings on DME type was examined and factors affecting visual acuity were evaluated in this retrospective, cross-sectional study.

Results

746 eyes of 538 patients were evaluated and 55.1% were in the CME group, 40.6% in dME and 3.8% in the CD group. It was observed that the integrity of ELM and EZ was impaired in 11.58% and 11.09%, respectively. When SRF, DRIL, HRD, exudates, VMT and ERM were examined, the rates of 16.7%, 19.23%, 28.83%, 54.40%, 2.76% and 15.8% were obtained, respectively. The CD group had the weakest visual acuity, while the dME group had the greatest (P < 0.001). While SRF, HRD, and exudates were most common in CME; DRIL, ELM and EZ damages were most common in the CD group (P < 0.001).

Conclusion

The group with the lowest visual acuity and the most damage to the outer retinal layers is the CD group. Presence of SRF, DRIL, HRD, exudate and ERM are also among the reasons that decrease visual acuity in DME.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12886-026-04943-1.

Introduction

DME (Diabetic macular edema) is described as the accumulation of fluid and serum macromolecules in the intracellular and intercellular spaces of the macula in diabetic individuals due to disruption of the blood-retina barrier [1]. DME is a major cause of visual loss in 7% of DM (Diabetes mellitus) patients [2, 3]. Defining optimal classification systems is critical for the successful management of DME. The Early Treatment Diabetic Retinopathy Study (ETDRS) remains the most widely used categorization system for macular edema, despite its numerous limitations [4]. OCT (Optical coherence tomography) -based macular edema classifications and biomarkers that predict prognosis or pathophysiology have become common in clinical practice [5].

The purpose of this study is to classify DME cases based on OCT images into three groups: CME (Cystoid macular edema), dME (Diffuse macular edema), and CD (Cystoid degeneration), and to determine the differences in visual acuity based on the groups, and to investigate the association between various biomarkers in OCT and the type of macular edema.

Subjects and methods

This retrospective, cross-sectional, single-center study followed the Helsinki Declaration. Ankara Yıldırım Beyazıt University Clinical Research Ethics Committee approved the study. (Approval number: 98, 11.11.2020). A waiver of consent was granted as it involved a retrospective review. Patients referred to Ankara Yıldırım Beyazıt University Ankara Atatürk Training and Research Hospital, Department of Ophthalmology between 2011 and 2019 were retrospectively analyzed. Patients diagnosed with type 2 diabetes (HbA1C > 6.5%, fasting plasma glucose level > 126 mg/dl, oral glucose tolerance test > 200 mg/dl) and DME on OCT were enrolled in the study. All eyes were treatment-naïve at the time of enrollment; only patients with a new diagnosis of diabetic macular edema who had never received any form of intravitreal therapy (including anti-VEGF or corticosteroids) nor undergone vitreoretinal surgery were included. Patients with corneal pathology, uveitis, glaucoma, poor OCT image quality, and macular edema due to other causes (retinal vein occlusions, Irvine-Gass syndrome, etc.) were excluded from the study.

Patients in the study were assessed for their visual acuity with an ETDRS chart and their anterior segments with a slit lamp. Cyclopentolate 1% was used for mydriasis, and then posterior segment evaluations were performed. Visual acuities were logarithm of the minimum angle of resolution (logMAR) -converted for statistical calculations. DRP (Diabetic retinopathy) is evaluated using Diabetic Retinopathy Severtiy Scale (DRSS). After adequate pupil dilation, Spectral Domain (SD)-OCT images (30° × 15°, 19 horizontal sections) of the patients were obtained with Heidelberg Spectralis (Heidelberg, Germany). All imaging was performed by a single experienced technician. Images with a complete centre focus with adequate image quality(Q > 25) were included in the analysis. The automatic eye tracking device (TruTrack and Tracking laser tomography) fixed the retina. Nine B-sections were above and nine below the central foveal portion. Macular thickness was measured using the X-Y scanning protocol for macular mapping. The macular map scanning pattern assessed a 6 × 6 mm (256 × 256) area centered on the fovea with 64 horizontal B-scan lines, each with 1024 A scans.

It was classified according to OCT images as dME, CME and CD. CRT (Central retinal thickness), SRF (Subretinal fluid), DRIL (Disorganization of retinal inner layers), HRD (Hyperreflective dot), VMT (Vitreomacular traction), ERM (Epiretinal membrane), EZ (Ellipsoid zone), ELM (External limiting membrane) and exudate structures were examined (Table 1; Fig. 1).

Table 1.

Definitons of DME (diabetic macular edema) types and biomarkers on OCT (optical coherence tomography)

	Definition	Reference
DME (diabetic macular edema)	Mean retinal thicknes in 1 mm area on OCT > 320 μm for men > 305 μm for women or intraretinal or hyporeflective area including SRF in 1 mm	Chalam et al. [6]
CRT (central retinal thichkness)	Distance between the inner retinal surface, the interface between the dark vitreous reflex and the light reflex of the ILM, and the outer retinal surface, the inner surface of the bright RPE/Bruch membrane interface. Measured as the average thickness of the central 1 mm area using OCT	Chalam et al. [6]
dME (diffuse macular edema)	Significant area of reduced hyporeflective area on OCT with increased central retinal thickness and the absence of cystoid spaces (Fig. 1)	Kim et al. [7]
CME (cystoid macular edema)	Oval or round retinal cyst-like cavities, hyperreflective surrounding, hyporeflective inside, and divided by septa, with increased central retinal thickness (Fig. 1)	Kim et al. [7]
CD (cystoid degeneration)	Loss of cyst septa and atrophic retina outside the cystoid gaps	Otani et al. [8]
EZ (elipsoid zone), ELM (external limiting membrane) integrity	Any discontinuity in the central, subfoveal 1 mm area on OCT was considered impaired (Fig. 1)	Helmy et al. [9]
DRIL (disorganization of retinal inner layers)	Disruption of at least one boundary between the ganglion cell, inner plexiform, inner nuclear, and outer plexiform layers in at least 50% of the foveal 1 mm area on OCT (Fig. 1)	Sun et al. [10]
HRD (hyperreflective dot)	Well-circumscribed lesions under 30 microns in a core 1 mm area did not generate a shadowing corridor. ≥10 HRD is present otherwise is absent (Fig. 1)	Nishijima et al. [11]
SRF (subretinal fluid)	Hyporeflective region under the posterior retinal surface with minimal tissue shadowing (Fig. 1)	Ozdemir et al. [12]
VMT (vitreomacular traction)	Perifoveal vitreous cortex detachment from the retinal surface Macular attachment of the vitreous cortex within a 3-mm radius of the fovea Association of attachment with distortion of the foveal surface, intraretinal structural changes, elevation of the fovea above the RPE, or a combination thereof, but no full- thickness interruption of all retinal layers (Fig. 1)	Duker et al. [13]
ERM (epiretinal membrane)	Hyperrefleftive band on foveal surface on OCT (Fig. 1)	Duker et al. [13]
Exudate	Well-circumscribed, highly reflecting, spherical, ovoid-shaped lesions that form a shadowing corridor under OCT were examined in the middle 3 mm

Fig. 1.

(a) Cystoid degeneration (CD) and external limiting membrane (ELM), ellipsoid zone (EZ) damage (The white arrowhead indicates loss of septa within the CD area, and ellipse highlights the region with ELM and EZ disruption. (b) Hyperreflective dots (HRD) with subretinal fluid (SRF) (The white arrowhead indicates SRF, black arrowhead indicates HRDs (c) Disorganization of the retinal inner layer (DRIL) (marked by the arrow tip) (d) Cystoid macular edema (CME) (with subretinal fluid) (The star marks the hyporeflective region corresponding to SRF, and the arrow indicates the intraretinal cysts forming CME (e) Vitreomacular traction (VMT) and epiretinal membrane (ERM) (The arrow shows VMT, the star marks ERM) (f) Hard exudates in diffuse macular edema (dME) (The arrow indicates exudates)

Descriptive statistics were presented as means ± standard deviation (SD), medians, and ranges (minimum–maximum) for continuous variables, while frequencies and percentages were used for categorical variables. The normality of data distribution was assessed using visual methods (histograms, Q-Q plots) and analytical tests, including the Kolmogorov-Smirnov and Shapiro-Wilk tests. For comparisons between groups, independent samples t-test or ANOVA was employed for normally distributed data, while Mann-Whitney U or Kruskal-Wallis tests were used for non-parametric data. Datas were analyzed using a Generalized Linear Mixed Model (GLMM) framework with patient identity included as a random intercept, thereby explicitly accounting for the non-independence of paired-eye observations. Given the mixed data structure — comprising both continuous variables (CRT, duration of DM) and binary OCT parameters (DRIL, HRD, ELM integrity, ERM) — Partitioning Around Medoids (PAM) with Gower distance was employed, as this approach is methodologically appropriate for datasets containing both continuous and categorical variables, unlike standard k-means clustering which assumes Euclidean distance and continuous data.

Statistical analyses were performed using R software (Version 4.3.2) and SPSS 23.0. All statistical tests were two-tailed, and a p-value of less than 0.05 was considered statistically significant.

Results

In the study, 746 eyes of 538 patients were evaluated. 55.17% of patients had CME, 40.67% had dME, and 3.89% had CD. ELM integrity was disrupted in 11.6% of patients, and EZ integrity was disrupted in 12%. Examination of the patients revealed that 16.73% had SRF, 19.23% had DRIL, 28.7% had HRD, 54.2% had exudates, 2.8% had VMT, and 15.9% had ERM.

We performed an exploratory cluster analysis on the study dataset. The optimal number of clusters was determined as k = 2 based on the average silhouette width criterion (maximum silhouette width = 0.21). Cluster 2 was characterized by higher CRT (411 vs. 388 μm), longer diabetes duration (13.2 vs. 12.2 years), and markedly higher proportions of DRIL (20.2% vs. 12.7%) and HRD (38.4% vs. 18.2%), indicating a profile of greater structural retinal involvement. Cluster 1 was distinguished by a substantially higher ERM prevalence (25.0% vs. 4.6%). CME was predominant in Cluster 2 (66.5% vs. 48.2% in Cluster 1), suggesting a partial correspondence between data-driven clusters and clinically defined edema subtypes. The moderate silhouette width (0.21) indicates incomplete cluster separation; these results should therefore be interpreted in an exploratory context. (Supplement Table 1, Supplement Fig. 1)

The statistical analysis of the variations between the continuous variables according to the DME types of the study participants was given in Table 2. Age did not reveal any statistically significant difference between the groups (p > 0.05). Visual acuity and CRT were found different between the groups. In terms of visual acuity, the best outcomes were observed in the dME group, whereas the worst outcomes were found in the CD group. A difference in diabetes duration was observed between the CME and dME groups, the mean value being higher in the CME group. (p < 0.05).

Table 2.

Age, Diabetes Mellitus (DM) duration, best corrected visual acuity (BCVA) and central retinal thickness (CRT) according to diabetic macular edema (DME) types

	dME		CME		CD		p	p 12	p 23
	n	Mean	n	Mean	n	Mean
Age (year)	219	69.03 ± 11.41 (62 to 77)	287	67.21 ± 9.93 (37 to 91)	23	71.39 ± 11.21 (51 to 90)	0.254
Duration of DM (year)	212	11.84 ± 7.36 (0 to 30)	293	13.49 ± 7.14 (0 to 40)	17	14.71 ± 6.48 (3 to 30)	0.020	0.035*	1.000*
BCVA (LogMAR)	236	0.34 ± 0.33 (0 to 1.80)	334	0.46 ± 0.34 (0 to 1.80)	20	0.91 ± 0.53 (0.1 to 1.80)	< 0.001	< 0.001*	< 0.001*
CRT (µm)	303	372.98 ± 78.53 (305 to 801)	411	433.93 ± 124.77 (271 to 933)	29	549.90 ± 128.92 (273 to 752)	< 0.001	< 0.001*	< 0.001*

p₁₂: Subgroup difference between dME and CME, p13: Subgroup difference between dME and CD p23: Subgroup difference between dME and CD, dME: Diffuse macular edema, CME: Cystoid macular edema, CD: Cystoid degeneration, LogMAR: Logarithm of minimum angle of resolution

*Bonferroni correction

There was no difference in sex, lens status, DRP grade, or VMT presence between the groups(p > 0.05). The rate of presence of SRF, DRIL, HRD, exudate, and ERM and the integrity of ELM and EZdiffered between the groups. It was revealed that the prevalence of SRF was highest in the CME subtype and lowest in the dME subtype. DRIL was significantly more frequent in the CD group (72.4%) compared with the other groups. Similarly, impairment of ELM and EZ integrity in the CD group were 62.1% and 65.5%, respectively, but this rate did not reach 13.4% in other groups. Compared to the other groups, the HRD was higher in the CME group, and the exudate was the same, while the ERM was higher in the dME group (28.9%). (Table 3).

Table 3.

Evaluation of sex, diabetic retinopathy (DRP) grade, subretinal fluid (SRF), disorganization of the retinal inner layer (DRIL), external limiting membrane (ELM) -ellipsoid zone (EZ) integrity, hyperreflective dots (HRD), exudates, vitreomacular traction (VMT) and epiretinal membrane (ERM) according to diabetic macular edema (DME) types

		dME		CME		CD		Total		P*
		n	%	n	%	n	%	n	%
Sex	Male	147	48.5%	200	48.7%	16	55.2%	363	48.9%	0.785
	Female	156	51.5%	211	51.3%	13	44.8%	380	51.1%
Lens status	Phakic	159	52.5%	203	49.3%	13	44.8%	375	50.5%	0.592
	Pseudophakic	144	47.5%	208	50.6%	16	55.2%	368	49.5%
DRP grade	NPDR	248	82.1%	319	78.2%	23	79.3%	590	79.8%	0.433
	PDR	54	17.9%	89	21.8%	6	20.7%	149	20.2%
SRF	-	273	91.9%	313	77.1%	24	82.8%	610	83.3%	< 0.001
	+	24	8.1%	93	22.9%	5	17.2%	122	16.7%
DRIL	-	280	95.6%	293	73.8%	8	27.6%	581	80.8%	< 0.001
	+	13	4.4%	104	26.2%	21	72.4%	138	19.2%
ELM integrity	Intact	274	94.5%	346	87.6%	11	37.9%	631	88.4%	< 0.001
	Disrupted	16	5.5%	49	12.4%	18	62.1%	83	11.6%
EZ integrity	Intact	276	95.2%	342	86.6%	10	34.5%	628	88.0%	< 0.001
	Disrupted	14	4.8%	53	13.4%	19	65.5%	86	12.0%
HRD	-	264	90.7%	227	57.5%	19	65.5%	510	71.3%	< 0.001
	+	27	9.3%	168	42.5%	10	34.5%	205	28.7%
Exudate	-	167	56.6%	148	36.9%	17	58.6%	332	45.8%	< 0.001
	+	128	43.4%	253	63.1%	12	41.4%	393	54.2%
VMT	-	287	98.0%	386	96.7%	28	96.6%	701	97.2%	0.616
	+	6	2.0%	13	3.3%	1	3.4%	20	2.8%
ERM	-	207	71.1%	369	93.4%	25	86.2%	601	84.1%	< 0.001
	+	84	28.9%	26	6.6%	4	13.8%	114	15.9%

Variables were compared using the Pearson’s chi-square test or Fisher’s exact test where appropriate. P < 0.05 was considered statistically significant. dME: Diffuse macular edema, CME: Cystoid macular edema, CD: Cystoid degeneration

It was statistically significant that ELM and EZ integrity are impaired and HRDs are more frequent in eyes with SRF. However, there was no relationship between the presence of SRF and DRIL or SRF and VMT (Supplement Table 2). In addition, the integrity of the ELM and EZ were found to be more damaged in eyes with HRD (Supplement Table 3).

Visual acuity variations in the SRF, DRIL, ELM integrity, EZ integrity, HRD, exudate, VMT, and ERM groups were examined (Table 4). In all groups except VMT, the variable was present, and its impaired state was associated with poorer visual acuity compared to the intact state.

Table 4.

Variation of best corrected visual acuity (BCVA, LogMAR) according to subretinal fluid (SRF), disorganization of the retinal inner layers (DRIL), external limiting membrane (ELM) -ellipsoid zone (EZ) integrity, hyperreflective dots (HRD), exudates, vitreomacular traction (VMT) and epiretinal membrane (ERM) status

	SRF		DRIL		ELM Integrity		EZ Integrity
	Absent	Exist	Absent	Exist	Intact	Disrupted	Intact	Disrupted
n	479	105	465	105	498	68	498	68
Mean (BCVA)	0.39	0.58	0.38	0.61	0.38	0.71	0.37	0.72
SD	0.34	0.39	0.33	0.43	0.32	0.45	0.32	0.45
Median	0.30	0.49	0.30	0.49	0.30	0.60	0.30	0.60
Min	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Max	1.80	1.80	1.80	1.80	1.80	1.80	1.80	1.80
p	< 0.001		< 0.001		< 0.001		< 0.001
	HRD		Exudate		VMT		ERM
	Absent	Exist	Absent	Exist	Absent	Exist	Absent	Exist
n	396	170	245	331	557	14	484	82
Mean (BCVA)	0.40	0.45	0.38	0.45	0.42	0.50	0.41	0.50
SD	0.36	0.35	0.34	0.37	0.36	0.37	0.36	0.36
Median	0.30	0.40	0.30	0.30	0.30	0.35	0.30	0.40
Min	0.00	0.00	0.00	0.00	0.00	0.15	0.00	0.00
Max	1.80	1.80	1.52	1.80	1.80	1.30	1.80	1.52
p	0.038		< 0.001		0.278		0.012

SD: Standart deviation, LogMAR: Logarithm of the minimum angle of resolution

BCVA values were compared using the independent samples t-test or Mann-Whitney U test, depending on the normality of data distribution. P < 0.05 was considered statistically significant

We analyzed the data using a Generalized Linear Mixed Model (GLMM) framework with patient identity included as a random intercept, thereby explicitly accounting for the non-independence of paired-eye observations. The random intercept variance was 0.373 (SD = 0.611), confirming a non-negligible degree of within-patient correlation that statistically justifies the mixed-model approach.

The dataset comprised three edema subtypes with analyzable sample sizes — dME (n = 303), CME (n = 411), and a third category (n = 29). The third category represented only 4.0% of the analytic sample, which was deemed insufficient to yield stable multinomial model estimates. Accordingly, the analysis was restricted to the two primary edema subtypes (CME vs. diffuse ME), for which binary GLMM with a logit link function was applied and results are presented in the Table 5.

Table 5.

Generalized Linear Mixed Model (GLMM) analysis of factors associated with cystoid macular edema (CME)

Variable	B	SE	z	p	OR	95% CI Lower	95% CI Upper
Duration of DM	0.047	0.017	2.788	0.005	1.048	1.014	1.083
CRT	0.007	0.002	3.813	< 0.001	1.007	1.004	1.011
DRIL (+)	1.876	0.442	4.245	< 0.001	6.530	2.746	15.529
ELM integrity (disrupted)	-1.217	0.478	-2.545	0.011	0.296	0.116	0.756
HRD (+)	1.879	0.348	5.404	< 0.001	6.545	3.311	12.936
Exudate (+)	0.394	0.235	1.675	0.094	1.483	0.935	2.352
Intercept	-3.918	0.834	-4.700	< 0.001	—	—	—

Random effect: Patient ID; Optimizer: bobyqa; AIC (Akaike Information Criterion) = 531.1; BIC (Bayesian Information Criterion) = 564.5; Log-likelihood = − 257.5; N observations = 484; N patients = 376; Random intercept variance = 0.373 (SD = 0.611). SE: Standard error; OR: Odds ratio; CI: Confidence interval; CRT: Central retinal thickness; DRIL: Disorganization of the retinal inner layers; ELM: External limiting membrane; HRD: Hyperreflective dots; DM: Diabetes mellitus. Model 1 (full model) could not be reported due to numerical instability of the Hessian matrix; Model 2 (parsimonious) is the primary reported model

A parsimonious GLMM with patient identity as a random intercept was fitted. Duration of DM (Odds ratio [OR] = 1.048; 95% Confidence interval [CI]: 1.014–1.083; p = 0.005), CRT (OR = 1.007; 95% CI: 1.004–1.011; p < 0.001), DRIL (OR = 6.530; 95% CI: 2.746–15.529; p < 0.001), and HRD (OR = 6.545; 95% CI: 3.311–12.936; p < 0.001) were positively and independently associated with CME. Disrupted ELM integrity was independently and negatively associated with CME (OR = 0.296; 95% CI: 0.116–0.756; p = 0.011). Exudate did not reach conventional statistical significance (OR = 1.483; 95% CI: 0.935–2.352; p = 0.094). Model fit indices (Akaike Information Criterion [AIC] = 531.1; Bayesian Information Criterion [BIC] = 564.5) confirmed adequate model parsimony.

Linear regression was used to predict visual acuity based on other variables (Supplement Table 4). Variables were examined and age, DRP grade, CRT, DRIL, HRD, exudates and ERM were found to be significant. It was observed that patients with PDR and CME had lower visual acuity. It was determined that visual acuity decreased as CRT increased and in the presence of DRIL, exudate and ERM. However, it was seen that visual acuity was better in the presence of HRD.

Discussion

DME has been classified according to various parameters such as retinal thickness, hard exudates, edema localization and relationship to the fovea, morphological patterns of fluid accumulation, vitreomacular interface changes, quantitative analysis of fluid accumulation, and microstructural changes. Various classifications have attempted to establish the baseline variance in visual acuity, determine the prognosis of the disease, rationalize the treatment modality, and predict treatment outcomes.

In 1999, Otani et al. evaluated DME on OCT images for the first time and observed that 88% of individuals exhibited sponge-like retinal edema, 47% had CME, and 15% had serous macular detachment [8]. Arf et al. showed CME prevalence of 68.5%, dME prevalence of 24.6%, and CD prevalence of 6.9% [14], whereas Yamamoto found a diffuse retinal thickening prevalence of 60% [15]. According to the OCT pattern, Kim et al. classified DME as diffuse thickening, CME, serous retinal detachment, and foveolar detachment with or without traction. In the study, 97% of participants had diffuse retinal thickening, 29% of participants had CME and diffuse retinal thickening, and 39.5% had only diffuse retinal thickening [7]. Numerous studies have been conducted that classified DME based on OCT morphologies, and varying proportions of DME groups have been reported in these investigations [7, 14, 16, 17]. CME had a prevalence of 55.17% in this study, making it the most common subgroup.

The formation of large cystoid spaces is caused by damage to Müller cells and neuroglia. In the study by Otani et al., both acute and chronic morphologies of DME were described, and it was stated that the disappearance of the septa causes it to merge with large cysts that can fill all retinal layers [8]. The progression of macular edema during the first three stages of CME is signified by fluid accumulation in the cystoid spaces. In the fourth stage, the cystic septa are destroyed, resulting in the formation of a large cyst comprising more than 90% of the macular thickness [9]. In light of these findings in the literature, the concept of CD has emerged. In the study by Arf et al., in which DME was classified based on OCT morphologies, CME with cysts larger than 600 microns in horizontal diameter were referred to as CD. This group has the lowest visual acuity and the longest duration of DM, with symptoms persisting for the longest period [14]. This group has the poorest visual acuity, as 3.89% of the eyes in our study are affected by CD. Additionally, the presence of DRIL and the rate of ELM and EZ integrity impairment were found greater in this group than in the other two groups. Kim, Yamamoto, and Arf found the dME group had the best visual acuity [7, 14, 15]. The group with the best visual acuity in our study was the dME group, similarly. In this group, the thinner CRT, less disruption of ELM and EZ integrity, and lower DRIL presence confirm the findings.

Although the pathophysiology of SRF is not well understood, it has been hypothesized that SRF represents fluid movement from the retina to the subretinal space during the resorption of macular edema [18]. Kang et al. postulated that substantial amounts of fluid and albumin could enter the subretinal space through the permeable ELM and could not be removed efficiently by RPE [19]. Moreover, SRF can occur without a significant accumulation of intraretinal fluid, and its severity is unrelated to retinal thickness. Observations which reported in SRF after branch retinal vein occlusion make it unlikely that passive fluid leakage from the retina into the subretinal space is the only mechanism responsible for SRF [20]. According to various publications, the incidence of SRF in DME ranges from 15% to 30% [8, 14, 21]. SRF was observed in 16.7% of DME patients and 22.9% in CME, 17.2% in CD, 8.1% in dME group in our study. In contrast to previous researches [14, 22], we found that DME patients with SRF had poorer visual acuity. We thought that this difference might be due to the higher incidence of ELM and EZ damage in patients with SRF in our subjects.

The DRCR.net algorithm for DME treatment uses CRT changes as an indicator [23]. CRT correlates with baseline and post-treatment visual acuity, but the correlation is lower than expected, usually with a correlation coefficient (R) of 0.2 to 0.3 [24–26]. Panazzo et al. found a linear correlation between macular thickness and visual acuity in randomly selected diabetic patients, despite the presence of multiple confounding factors affecting visual function [27]. Visual acuity and CRT were statistically correlated in this present study (R = 0.4). In addition, when the relationship between macular edema types and CRT was examined, a significant difference was found as CD > CME> dME (p < 0.05). However, CRT is just one of the many factors influencing visual acuity in DME.

Soliman et al. found that inner retinal layer irregularity affecting 50% or more of the central 1 mm wide region of the fovea is associated with poor visual acuity in patients with present or resolved central DME. This is valid even in eyes with impaired vision despite the resolution of edema, or in eyes with excellent visual acuity despite the presence of edema. The evaluation of DRIL in patients with existing DME may be useful for predicting future visual acuity [28] Zur et al. reported that baseline visual acuity was not associated with the presence of DRIL; however, eyes exhibiting DRIL on baseline OCT demonstrated poorer visual outcomes at the end of the follow-up period. In the same study, SRF was more common in acute DME patients without DRIL, suggesting that DRIL may indicate chronicity in DME [29]. In this study, DRIL rate was 72.4% in CD group, 26.3% in CME group and 4.4% in dME group. The mean visual acuitywas 0.61 logMAR in eyes with DRIL and 0.38 in eyes without DRIL (p < 0.001).

Bolz et al. found sharply demarcated, hyperreflective foci within intraretinal microaneurysm walls, distributed over all retinal layers, and confluent plaques in the outer plexiform layer in DME patients. These foci may represent extravasated lipoproteins or proteins, an early indicator of subclinical blood-retina barrier disruption in DME [30]. Uji et al. examined the rate of HRD in DME patients, and they declared that they saw HRD in the outer retinal layers in 53% of the patients without SRF and in the inner retinal layers in 99.1% of the patients without SRF. Moreover, the presence of hyperreflective foci in the outer retinal layers has been reported to be strongly associated with impaired ELM and IS/OS lines on OCT images, as well as decreased visual acuity in DME [31]. HRNs may also result from microglial cell activation-related inflammation [32]. In this study, 28.8% of DME patients had HRDs and HRD rate was 42.5% in CME group, 34.5% in CD group, and 9.3% in dME group. In addition, the significantly disrupted ELM and EZ integrity in eyes with HRD and the reduced visual acuity in this group support the association with photoreceptor damage.

Numerous studies have assessed the prevalence of vitreomacular interface disorder in DME. Kim et al. reported that 15.6% of patients with DME had posterior hyaloid traction, whereas Akbar Khan et al. found a rate of 12.0% partial vitreomacular detachment and a rate of 12.0% ERM. Wong et al. reported a rate of 43% for ERM involving the fovea and a rate of 63% for eccentric ERM, while Mikhail et al. reported a total rate of 18.5% for ERM and VMT. The most common stage in eyes with DME was perifoveolar AVD with foveolar attachment, which Gauscher et al. found to be part of the aging-related vitreous detachment process. This stage was observed significantly more frequently in eyes with DME than in eyes without DME [33]. Macular traction has been discussed in the literature as a potential cause of serous macular detachment. Catier et al. highlighted that serous macular detachment was not associated with a thick and taut membrane or an incompletely detached posterior hyaloid, whereas Kaiser et al. found a significant association between serous macular detachment and the presence of VMT [23, 34]. In the study by Arf et al. [14], ERM was detected in 21.4% and VMT in 2.1% of eyes with DME, but there was no correlation between these and visual acuity. ERM and VMT were observed in 15.9% and 2.9% of DME patients in our study, respectively. Although the presence of VMT did not differ between groups in our study, the dME group had the highest prevalence of ERM. At this point, the question arises as to whether ERM in DME is most frequently observed in dME type or ERM due to diabetes initially causes diffuse type of edema. According to the research conducted by Brar et al., Macular edema associated with ERM is mainly non-cystic [35]. Also, there was no relation between the presence of macular traction and SRF and visual acuity. It was found that patients with ERM had lower visual acuity in our study.

The study is limited by its retrospective design, the lack of a control group, and the smaller sample size in the CD group compared to the other groups. In addition, a portion of our patient population consists of asymptomatic patients who were referred to us by an endocrinologist for retinopathy screening after applying to the Endotem District Polyclinic of Ankara Atatürk Training and Research Hospital’s endocrinology unit due to diabetes. This may explain why the ophthalmological findings of the patients differ from those found in other studies. The strengths of this study include the evaluation of a large number of patients, the simultaneous assessment of multiple OCT findings, and the minimization of confounding factors by excluding patients who had previously received treatment.

Conclusively we found that the most common subtype of DME was CME, and it is important to differ CD from CME in this study. Moreover, SRF was mostly seen in the CME group, and it was related to poor visual acuity. We hope that this study will help design future studies to classify DME and contribute to the development of appropriate algorithms for the ophthalmological follow-up of diabetic patients.

Summary

What was known before:

• OCT is widely used to assess diabetic macular edema and its impact on retinal layers.

• Morphological classification of DME typically includes diffuse and cystoid types.

• Certain OCT findings like SRF, DRIL, and EZ damage are linked to poor visual outcomes.

What this study adds:

• This study highlights cystoid degeneration (CD) as a distinct subtype with the worst visual acuity and most outer retinal damage.

• It shows that SRF is most common in CME and strongly associated with reduced vision.

• By excluding previously treated patients, the study provides a clearer view of untreated DME morphology and its clinical relevance.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

BT was responsible for conducting the main analyses, writing the original draft, reviewing and revising the manuscript, and finalizing the manuscript, while FY, IEET, and ZOT contributed to writing and revising the manuscript, NU designed the study, controlled the study execution, provided critical feedback, and assisted in finalizing the manuscript, and BC and RE provided data contributions.

Funding

The authors declare that no external funding was received for this study.

Data availability

The datasets generated during and /or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study followed the tenets of the Declaration of Helsinki. The study was approved by the Ankara Yıldırım Beyazıt University Clinical Research Ethics Committee (Approval number: 98, 11.11.2020). The need for informed consent to participate was explicitly waived by the Ankara Yıldırım Beyazıt University Clinical Research Ethics Committee, as this study was retrospective in nature and based solely on pre-existing medical records without any direct contact with patients.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.