Clinical and demographic data were retrieved by independent readers for each institution (P.C., S.M., and F.C.P.). In addition to demographic records, we recorded the history of steroid use from the 6 months preceding the onset of the disease episode, then whether it was the first episode, and, in case of recurrence, the time interval since the last episode had occurred. From the clinical records, we also recorded best-corrected visual acuity (BCVA) using logMAR conversion and intraocular pressure (Goldmann tonometry). All patients had undergone structural OCT and OCT angiography (Spectralis HRA+OCT; Heidelberg Engineering) with a dense volumetric acquisition (30°×20° or 20°×20° field, 97 sections, with automatic real-time function and enhanced depth imaging) mapping the entire subretinal fluid (SRF) volume. Radial OCT scans were analyzed to evaluate foveal involvement and the relationship between the leakage sites and the central macula.

All patients had FA and ICGA available in the medical records (Spectralis HRA+OCT; Heidelberg Engineering) and OCT scans passing through the fluorescein angiographic leaking points. The OCT software enables direct multimodal cross-referencing of FA guide and corresponding OCT with closely spaced B-scans, ensuring accurate spatial localization.

Fluorescein leaks were classified according to the classic textbook of Shatz, Burton, Yannuzzi, and Rabb by independent graders (P.F. and J.C.). Leaks begin as small “pinpoint” areas of hyperfluorescence at the level of the RPE in the arteriovenous phase. Throughout the angiogram, fluorescein partially pools the space under the neurosensory retina in two patterns: (1) upward from a break in the RPE, forming a “mushroom,” “umbrella,” or “smokestack” appearance; (2) gradually increasing concentrically to form a spot resembling an “ink blot” with fuzzy margins. In cases of blot-type leaks, the leakage displays variable intensity, ranging from unusually “profuse” leaks to less pronounced amounts of fluorescence. Therefore, FA patterns were graded as “smokestack” or “blot-type” and additionally as “unifocal” or “multifocal,” with a subclassification of “profuse” for unusually intense leaks, using qualitative grading.

All multimodal images were subsequently graded by the same independent graders (P.F. and J.C.). In addition to the presence of serous PED, which represents a well-established possible location of leak site, ^, graders conducted comprehensive morphological analysis of outer retinal structures on OCT B-scans passing through FA-confirmed focal leaks.

The evaluation protocol documented structural abnormalities of the outer retina in the macular area with serous detachment. Graders particularly assessed the continuity and integrity of the outer photoreceptor layer at the site of the leaking point of the RPE, with documentation of focal interruptions along an otherwise thickened outer photoreceptor layer. Graders also evaluated the localization and extent of subretinal fibrin, ^, which was defined as hyperreflective material above the RPE. Morphological changes of the outer retina and RPE in correspondence with the fibrinous deposits were evaluated. Analysis of these morphological features across the entire cohort revealed two distinct outer retinal configurations at leakage sites, which are subsequently defined and characterized in the “Results” section.

Records of all cases related to the follow-up visits were analyzed, and the date of any appearance of foveal atrophy or PCRD, defined by the presence of optically empty spaces of IRF at the posterior pole or other structural retinal changes, was recorded. An additional blinded grader (F.C.P.) was involved in case of inter-graders’ discordance.

Three-dimensional reconstruction and rendering of SRF volume were performed using custom scripts (S.H.) developed in MATLAB (v. 2024b; MathWorks). The OCT volumetric scan and corresponding segmentation boundaries—internal limiting membrane (ILM) and Bruch’s membrane—were exported in “.raw” format. Preprocessing included A-scan–wise intensity normalization to standardize brightness levels across all B-scans. The retinal region was defined as the area between the internal limiting membrane and Bruch’s membrane in each scan. A binary mask was then applied, with a threshold set at 0.1% of the maximum intensity within the retinal area. The threshold was empirically determined to optimize sensitivity to low-intensity SRF signals while minimizing noise artifacts. Voxels located within the retinal boundaries but excluded by the mask were identified as SRF volume. For volumetric rendering, the spatial coordinates of the SRF voxels were mapped to the original OCT coordinate space. These coordinates were subsequently triangulated into tetrahedra using the MyRobustCrust algorithm. The SRF volume was visualized by rendering the isosurface of the resulting tetrahedral mesh, enabling systematic analysis of distinct SRF morphological configurations. Two independent graders (P.F. and J.C.) performed comprehensive evaluation of SRF volume maps to identify and classify morphological characteristics that would subsequently inform the primary study findings.

After anonymization, OCT volumes were analyzed using the Discovery platform (Discovery OCT Fluid and Biomarker Detector, RetinAI AG) providing automated quantification of retinal and choroidal layers thickness and volumes. This software is based on a convolutional neural network architecture and was trained in a supervised manner. Specifically, the artificial intelligence (AI)-assisted volumetric analysis provides quantification of SRF and PED volumes measured in nanoliters, in analogy to previous studies. ^, When an error in automated thickness or volume was present, manual editing was performed (C.M.E.).

Statistical analyses were performed using R software (version 4.3.3). Column graphs and survival curves were created using GraphPad Prism software (version 10.0.0). Main outcome measures were (1) AI-enhanced volumetric quantification of SRF and PED between morphological groups identified in the study; and (2) development of adverse anatomic outcomes (foveal atrophy and PCRD) during the follow-up window. Quantification of SRF and PED volumes was the primary outcome, whereas follow-up analysis was considered exploratory and hypothesis-generating.

Continuous variables are presented as median and interquartile range (IQR), and nonparametric tests were applied. The Mann–Whitney U test was used for continuous variables, and categorical variables were compared using the chi-square test. Time-to-event analysis used Kaplan–Meier estimators with the log-rank test for comparative analysis between morphological groups identified during the study, with observation censored at the date of onset of PCRD or foveal atrophy, or upon the last follow-up for unaffected eyes. For descriptive purposes, patients who developed these events continued to be observed throughout the available follow-up period, with documentation of additional treatments and final visual acuity. Event-free survival probabilities with 95% CIs were calculated at landmark time points of 6, 12, and 24 months. The ratio between the number of adverse events and the total observation time was used as an estimate of the incidence rate per 100 person-years. For stratified analyses by treatment status, patients were categorized based on whether they received HD-PDT or observation during the follow-up period, and event rates were calculated for each subgroup. P values <.05 were considered statistically significant.

B-scan OCT and 3-dimensional rendering analysis of SRF volume revealed two distinct mutually exclusive morphological configurations of the outer retinal changes at the leakage sites .

• •

  Outer Retinal Erosion (ORE): This pattern was characterized by a notch-shaped erosion, or a broader loss of tissue, along an otherwise thickened outer photoreceptor layer within the macular detachment, creating a “tooth loss” appearance. The volume of SRF in cases with ORE displays a conventional dome configuration.

• •

  Outer Retinal Pinching (ORP): this pattern was defined by the presence of focal adherence or hyperreflective connecting material, presumed to be fibrin, between the RPE and the outer photoreceptor layer within the serous macular detachment. Traction on both sides of the subretinal space creates a characteristic “pinching” effect on the outer retina and the RPE, mainly appreciable on the retinal side. Consequently, the volume of SRF displays an eccentric toroidal or doughnut-like configuration.

The cohort was divided between these two patterns, with ORE identified in 43 cases (49.4%) and ORP in 44 cases (50.6%). Representative examples of the multimodal imaging and corresponding OCT morphological features are illustrated in Figure 1 . The characteristic differences in SRF volume configuration between ORE and ORP patterns are visualized through 3-dimensional rendering in Figure 2 .

Representative examples of the multimodal imaging and corresponding optical coherence tomography (OCT) patterns of the outer retina. A and B. Outer retinal erosion (ORE) cases showing focal notch-shaped erosion (A) or broader (B) photoreceptor thinning overlying pigment epithelium detachment (PED) ( blue arrowheads ) with mild fluorescein angiography (FA) leakage and focal hyper-autofluorescence due to window effect from photoreceptor layer thinning. C. Outer retinal pinching (ORP; orange arrowhead ) case with smoke-stack FA leak and clepsydra sign (ie, posterior displacement of the outer retina without clear interposition of fibrin). D. ORP case with blot-type leak, ring-shaped subretinal exudates, fibrin bridge, and hyporeflective lucency. The leak intensity is more pronounced on indocyanine green angiogram, congruous with greater degree of vascular hyperpermeability. Fundus autofluorescence reveals a focal area of nonfluorescence corresponding to the site of the retinal pigment epithelium (RPE) barrier defect.

Three-dimensional rendering of subretinal fluid (SRF) volume configurations demonstrating two distinct morphological outer retinal patterns in central serous chorioretinopathy (CSCR). A and B. ORE cases show conventional dome or conical SRF volume configuration characterized by focal erosion of the outer photoreceptor layer while maintaining separation between retinal and RPE layers within the serous neuroretinal detachment. C and D. ORP cases exhibit characteristic toroidal or doughnut-shaped SRF morphology resulting from focal contact and adherence between the RPE and outer photoreceptor layer, creating a distinctive “pinching” effect that constrains SRF flow around a focal adherence point. ORE = outer retinal erosion.

The study cohort comprised 87 eyes of 87 patients with a predominant male representation of 74.7%. The median age of patients was 46 years with an IQR of 41 to 51 years and a total range spanning from 30 to 76 years. First episodes accounted for 29 of 87 cases (33.3%), and 58 of 87 (66.6%) were recurrences. Recurrent cases had a median time from first episode of 2 years (IQR, 1-4; min-max: 1-13). Steroid-related cases represented 26 of 87 (29.9%) of the sample. Comprehensive demographics and baseline characteristics are detailed in Table 1 .

Among ORE cases, a notch-shaped erosion of the outer segments of photoreceptors ( Figure 3 , A) was present in 32 cases representing 74.4% of this group, whereas a broader erosion ( Figure 3 , B) was observed in 11 cases representing 25.6%. These lesions were differentiated by photoreceptor atrophy due to long-lasting SRF for the clear demarcation of the tissue loss in the context of a thickened outer photoreceptor layer. Thinning of the outer nuclear layer could be seen in correspondence with the outer segment loss.

Representative examples of OCT morphological grading adopted in the study. Blue arrowheads indicate ORE patterns; orange arrowheads indicate ORP patterns. A. Notch-shaped erosion: discrete interruption along an otherwise thickened outer photoreceptor layer. B. Broader erosion: broader confluent thinning of the outer photoreceptor layer. C. Clepsydra sign: posterior displacement of the outer retina indicating subclinical fibrinous adhesion. D. Subretinal fibrin bridge: well-defined proteinaceous material bridging the RPE and outer photoreceptor layer. E. Hyporeflective subretinal lucency: hyporeflective area within fibrinous deposits above the plane of the RPE.

The focal subretinal adherence characteristic of ORP cases demonstrated several distinct morphological presentations. The area of adherence and traction corresponded with a subretinal fibrin bridge with occasional hyporeflective subretinal lucency, or, less frequently, to a “clepsydra” appearance, in which the posterior displacement of the outer retina without interposition of fibrin was juxtaposed with a pointed serous RPE detachment. A fibrin bridge between the outer retina and RPE ( Figure 3 , D) was identified in 38 cases representing 86.3% of the ORP group, indicating the peculiar role of fibrinous material in establishing retinal-RPE adherence and traction. A concurrent hyporeflective lucency ( Figure 3 , E) was evident in 13 cases representing 29.5% of the ORP group and was observed in all cases in the subretinal compartment within areas of massive fibrin deposition above the RPE plane. Six of the ORP cases (13.6%) presented with a clepsydra appearance ( Figure 3 , C).

Comparative analysis between ORE (n = 43) and ORP (n = 44) groups revealed significant morphological and volumetric differences ( Table 2 ). The groups were well matched regarding demographic and clinical parameters, with no significant differences in age (median: 46.0 vs 45.5 years, P =. 677), proportion of first episodes (32.6% vs 34.1%, P =. 877), or baseline BCVA (median: 0.1 vs 0.2 logMAR, P =. 777). The prevalence of steroid-related episodes (18.6% vs 40.9%, P =. 011) was markedly higher in the ORP group.

The proportion of cases with detectable PED was significantly higher in the ORP group (77.3% vs 46.9%, P =. 027). In contrast, volumetric measurements revealed profound differences between patterns. ORP cases exhibited significantly higher median SRF volumes (1763 nL [IQR, 957-3288] vs 720 nL [IQR, 395-1197], P =. 0005) compared with ORE cases. Likewise, median PED volumes were substantially greater in ORP cases compared with ORE cases (14.5 nL [IQR, 4-41] vs 0 nL [IQR, 0-4], P =. 0002). The volumetric differences of SRF and PED between ORE and ORP cases are graphically represented in Figure 4 , A and B.

Comparison of SRF and pigment epithelial detachment (PED) volumes between ORE and ORP groups. A and B. Scatter plots with segmented y-axis demonstrate significantly higher volumes in ORP cases for both SRF ( P =.0005) and PED ( P =.0002) volumes, reflecting the association between ORP pattern and more profound choroidal hyperpermeability. C and D. Representative examples from the subgroup analysis restricted to unifocal leaks with volumes near group medians. B-scans show the segmentation boundaries with corresponding SRF and PED volume maps. Note in ORP case ( orange arrowhead ): toroidal SRF configuration with the adherence site as epicenter of subretinal exudation. ORE = outer retinal erosion; ORP = outer retinal pinching.

Subgroup analysis restricted to unifocal leaks (n = 38 per group) confirmed significant volumetric differences between the two patterns. Median SRF volume was 715 nL (IQR, 361-1287) in ORE versus 1668 nL (IQR, 673-3093) in ORP (Mann–Whitney U test, P =. 0013). Median PED volume was 0 nL (IQR, 0-4) in ORE versus 12.5 nL (IQR, 2-32) in ORP ( P =. 00019). Multifocal cases (ORE n = 5, ORP n = 6) showed median SRF volumes of 481 nL and 1205 nL, respectively, indicating that multifocality per se was not associated with higher total fluid volumes. Representative examples from the subgroup analysis restricted to unifocal leaks with volumes near group medians are shown in Figure 4 , C and D.

FA pattern analysis showed blot-type leaks in 76 of 87 patients (87.4%) and smoke-stack leaks in 11 of 87 patients (12.6%). Profuse leak intensity was observed in 16 of 87 patients (18.3%). No significant differences were observed in fluorescein leak type between the ORE and ORP groups, with blot-type leaks predominating in both groups (88.4% vs 86.4%, P =. 774) and similar distribution of smoke-stack pattern (11.6% vs 13.6%). Smokestack leaks characterized cases with the clepsydra sign in the ORP group and broader erosion in the ORE group ( Table 2 ).

Leaks were predominantly unifocal in both groups (86.0% vs 86.4%), with comparable multifocal presentation (14.0% vs 13.6%, P =. 306). However, profuse leak intensity was significantly more frequent in ORP cases (6.9% vs 29.5%, P =. 003), aligning with quantitative volumetric quantification.

In cases with multifocal FA leaks, ORE and ORP were not present in the same eyes. Each leakage site within individual patients demonstrated the same morphological pattern on OCT either exclusively ORE or exclusively ORP; an example is shown in Figure 5 . In multifocal ORE cases, the photoreceptor’s displacement was not equally pronounced at all fluorescein leak sites; this heterogeneity is consistent with the concept that not all angiographically detected leaks contribute equally to SRF accumulation.

Multifocal leakage patterns demonstrating outer retinal morphological consistency within individual patients. Top rows : ORP case with multifocal leaks (ID2). A. Color fundus photography (CFP) reveals multifocal ring-shaped exudates and translucent areas corresponding to fibrin deposition, with turbid SRF. B. FA shows multiple active blot-type leaks (profuse specification). C. Late-phase indocyanine green angiography (ICGA) demonstrates intense hyperpermeability and dye coalescence through the leak sites. I-VI. Corresponding OCT B-scans reveal multifocal hyporeflective subretinal lucency at the RPE plane, with ORP pattern thus present simultaneously at all leakage sites ( orange arrowhead ). Bottom rows : ORE case with multifocal leaks. D. CFP shows multifocal neurosensory retinal detachments without visible fibrin deposition. E. FA demonstrates multifocal leaks with modest intensity. F. Late-phase ICGA shows mild hyperpermeability, heterogeneous among the leak points identified on FA. I-IV. Corresponding OCT B-scans demonstrate multiple notch-shaped ORE areas ( blue arrowheads ); ORE features are not equally pronounced at all points since not all angiographically detected leaks contribute equally to SRF accumulation.

On FAF, in case of ORE, the zonal loss of photoreceptors outer segments was slightly hyperfluorescent ( Figure 1 ) due to the increased transmission of the RPE autofluorescence signal of the underlying RPE (G. Staurenghi, personal communication). Instead, in case of ORP, FAF usually revealed a focal area of nonfluorescence corresponding to the site of the RPE barrier defect, more obvious in cases of profuse leakage.