Highlight
-
•
OCT using minimum intensity projection detected FA-defined inflammatory changes.
-
•
MIP images showed higher sensitivity for FA-defined changes than thickness maps.
-
•
Macular-centered MIP en face OCT detected both central and peripheral FA activity.
-
•
MIP en face OCT may serve as a noninvasive adjunct for uveitis assessment.
Purpose
To evaluate the diagnostic performance of en face optical coherence tomography (OCT) using minimum intensity projection (MIP) to detect fluorescein angiography (FA)–defined inflammatory retinal vascular changes.
Design
Prospective, single-center reliability and validity analysis.
Participants
A total of 318 eyes from 177 patients with uveitis who underwent same-day ultra-widefield FA and macula-centered 12 × 12–mm spectral-domain OCT imaging.
Methods
Inflammatory retinal vascular changes were defined as vascular leakage and/or vascular wall staining on late-phase FA. En face OCT images were generated using MIP slabs spanning 10% to 70% of total retinal thickness. MIP images and retinal thickness maps were independently graded in a masked fashion and compared with FA as the reference standard. Diagnostic performance metrics were calculated. Inter-grader agreement was assessed using Cohen’s κ coefficient.
Main outcome measures
Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), F1 score, and inter-grader agreement for detecting FA-defined inflammatory retinal vascular changes.
Results
FA demonstrated inflammatory retinal vascular changes in 208 of 318 eyes (65%), including 94 with activity within the central 12 × 12–mm field and 114 with activity confined to the peripheral retina. Overall, MIP en face OCT achieved a sensitivity of 0.82, specificity of 0.55, PPV of 0.77, NPV of 0.62, and F1 score of 0.80 for detecting FA-defined inflammatory retinal vascular changes. Compared with retinal thickness maps, MIP demonstrated substantially higher sensitivity (0.82 vs 0.50) and F1 score (0.80 vs 0.57). MIP en face OCT identified FA-positive eyes with higher detection rates than retinal thickness maps for both central activity (98% vs 60%) and peripheral-only activity (68% vs 42%). Inter-grader agreement was higher for MIP images (κ = 0.75) than for thickness maps (κ = 0.37). In a subset of eyes with longitudinal follow-up, MIP positivity preceded or persisted beyond changes in FA leakage.
Conclusions
En face OCT using MIP demonstrated high sensitivity for detecting FA-defined inflammatory retinal vascular changes. Although not a replacement for FA, this approach may provide a noninvasive adjunct for clinical assessment and longitudinal evaluation of uveitic disease.
INTRODUCTION
Retinal vasculitis is a sight-threatening intraocular inflammation affecting the retinal vessels. Because vascular biopsy is not feasible, its diagnosis relies on clinical criteria and findings on fluorescein angiography (FA), and the definition of retinal vasculitis itself remains a subject of ongoing debate. ,, Accordingly, in the present study, we use the term inflammatory retinal vascular changes to describe FA-defined vascular leakage and/or vascular wall staining associated with intraocular inflammation.
Although optical coherence tomography angiography (OCTA) enables noninvasive visualization of retinal blood flow and areas of nonperfusion, it remains limited in its ability to identify vascular leakage. , The gold standard for detection of disease activity in retinal vasculitis is fluorescein leakage on FA, thus the utility of OCTA in this context is limited, and various alternative approaches using optical coherence tomography (OCT) have been explored. ,,, In particular, retinal thickness maps have been investigated in several prior studies as potential tools for assessing disease activity. , However, none of these approaches has yet been established as a clinically useful diagnostic or monitoring modality.
Minimum intensity projection (MIP) is an en face OCT technique that highlights hyporeflective structures by projecting the lowest intensity signals within a defined retinal depth range. In healthy retinas, minimum intensity (MI) values are derived predominantly from the outer nuclear layer (ONL) or Henle fiber layer (HFL), which appear similarly hyporeflective unless fibers are oriented nearly perpendicular to the optical axis. An increase in MI therefore indicates increased ONL or HFL reflectivity or localized ONL loss. In the setting of inflammatory retinal vascular changes, alterations such as vascular dilation, vessel wall thickening, and perivascular tissue changes may lead to compression or thinning of the ONL, resulting in changes in MI values within a given slab. ,
The aim of this study was to evaluate the utility of en face OCT imaging using MIP for FA-defined inflammatory retinal vascular changes and to compare its diagnostic performance with retinal thickness maps, in order to determine its potential as a practical and noninvasive tool for clinical assessment.
METHODS
STUDY DESIGN AND PARTICIPANTS
This prospective, single-center reliability and validity analysis was approved by the Cleveland Clinic Institutional Review Board and adhered to the tenets of the Declaration of Helsinki. All patients were evaluated at the Cole Eye Institute, Cleveland Clinic, and written informed consent was obtained from each participant prior to imaging.
Eligible eyes were those diagnosed with uveitis that were seen in the uveitis clinics of the treating physicians (S.S. and S.K.S.) between March 2025 and September 2025 and underwent both ultra-widefield FA and macula-centered 12 × 12-mm spectral-domain OCT (SD-OCT) on the same day. Eyes were excluded if they had conditions that could confound OCT image interpretation, including severe media opacities, macular edema, ischemia-related retinal thinning, or tractional epiretinal membrane (ERM), as well as other major retinal diseases such as retinal detachment, severe nonproliferative or proliferative diabetic retinopathy, inherited retinal degeneration, or a history of pars plana vitrectomy. Scans with signal strength less than 7 or severe artifacts were also excluded to ensure reliable image evaluation. Eyes with focal signal attenuation that obscured vascular visualization were also excluded.
IMAGE ACQUISITION
Ultra-widefield FA was performed using the Optos California (Optos PLC, Dunfermline, Scotland, United Kingdom), and same day SD-OCT imaging was obtained with the CIRRUS 6000 (Carl Zeiss Meditec, Dublin, California, USA). OCT data from the CIRRUS 6000 were exported to and analyzed in the ZEISS FORUM Viewer (software version 4.4.2; Carl Zeiss Meditec AG, Jena, Germany), using the Retina Workplace module (version 2.8.0.445).
IMAGE ANALYSIS AND GRADING
Inflammatory retinal vascular changes were defined as the presence of vascular wall staining or leakage on late-phase (after 5 minutes) ultra-widefield FA, in the setting of intraocular inflammation. Each FA scan was classified as demonstrating: (1) inflammatory retinal vascular changes within the 12 × 12-mm field, (2) inflammatory retinal vascular changes confined to the peripheral retina beyond this field, or (3) no inflammatory retinal vascular findings.
Inflammatory retinal vascular changes have been reported to present on OCT as vascular dilation and perivascular thickening, often associated with compression of the ONL. ,,, To visualize these changes, en face OCT imaging using MIP was employed. En face MIP images spanning 10% to 70% of total retinal thickness from the internal limiting membrane (ILM) was generated within FORUM/Retina Workplace. The 10% to 70% slab boundaries were manually defined and represent adjustable parameters within the software, rather than fixed default settings. For each A-scan, the darkest pixel within the specified slab was identified and compiled into a two-dimensional en face image. Slab depth boundaries were defined as percentages of the total retinal thickness measured outward from the ILM toward the retinal pigment epithelium. The upper boundary was set at 10% below the ILM to avoid capturing dark pixels from the vitreoretinal interface, while the lower boundary was set at 70% below the ILM, corresponding to a depth within the ONL in the normal retina, slightly external to the boundary between the outer plexiform layer (OPL) and the ONL. An example of an en face MIP image is shown in Figure 1 . Retinal thickness maps were also analyzed for comparison.
Structural basis of MIP positivity in inflammatory retinal vascular changes. (A) Late-phase FA demonstrates vascular leakage consistent with inflammatory retinal vascular changes. (B) En face MIP image generated using the 10% to 70% retinal depth slab shows focal hyperreflective signals along the corresponding vascular segments. (C) Corresponding OCT B-scan at the site of MIP positivity demonstrates vessel with compression of ONL and increased reflectivity of the surrounding perivascular tissue. The dashed magenta lines indicate the boundaries of the MIP slab, defined as 10% to 70% of the total retinal thickness measured from the ILM. Within this slab, the normally hyporeflective ONL is replaced by hyperreflective vascular and perivascular structures, resulting in increased minimum-intensity values and MIP positivity.
The en face MIP images were graded as positive if any suspicious signs of vasculitis were observed. For MIP images, vessels that appeared hyperreflective were graded as positive, and crossing points were excluded. For retinal thickness maps, images showing perivascular thickening compared with standard reference maps derived from representative OCT volume scans of eyes without inflammatory retinal vascular changes were graded as positive ( Figure 2 ).
Representative examples of retinal thickness map findings. (A) Late-phase FA demonstrates vascular leakage (arrows) consistent with inflammatory retinal vascular changes. (B) The corresponding retinal thickness map shows focal color-coded thickening (red hues) at locations concordant with the FA leakage (arrows). (C) Late-phase FA shows no evidence of inflammatory retinal vascular changes. (D) The corresponding retinal thickness map demonstrates no perivascular retinal thickening.
Inflammatory activity grading using MIP en face images and retinal thickness maps was performed in a masked and randomized fashion with respect to clinical inflammatory parameters as well as FA images and findings. All images were independently graded by a retina specialist (Y.M.) and a trained reader (V.B.) to assess inter-rater reliability. Any discrepancies were resolved by consensus.
STATISTICAL ANALYSIS
En face OCT findings were compared with FA as the reference standard. For eyes with longitudinal follow-up, changes in en face OCT findings were compared with corresponding changes in vascular leakage on FA.
First, en face image detection rates were summarized descriptively according to the location of inflammatory activity on FA scans, including activity within the central 12 × 12-mm field, activity confined to regions outside this field, and overall activity anywhere on FA.
Second, formal diagnostic performance metrics, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score with 95% confidence intervals, were calculated using FA findings as the reference standard, with FA positivity defined as the presence of inflammatory retinal vascular activity anywhere on ultra-widefield FA. Paired comparisons between MIP images and retinal thickness maps were performed using McNemar tests. Sensitivity was compared among FA-positive eyes and specificity among FA-negative eyes, with FA serving as the reference standard.
Inter-grader agreements for activity assessment using MIP images and retinal thickness maps were calculated using Cohen’s κ coefficient.
RESULTS
A total of 417 eyes were screened, of which 99 eyes were excluded. Reasons for exclusion included ischemia-related thinning (n = 25), ERM (n = 11), macular edema (n = 2), localized signal attenuation due to vitreous opacities such as floaters (n = 15), and poor image quality (n = 19), including low signal strength (<7) or generalized signal attenuation due to severe media opacity or dense vitreous haze.
A total of 318 eyes of 177 patients were included in final analysis. Among these, 82 eyes from 47 patients underwent multiple examinations during the study period. The mean age was 51.6 ± 18.3 years (range, 11-80 years) and 64% (113 of 177) of patients were female. Diagnoses included Birdshot retinopathy (n = 23), sarcoidosis (n = 15), Susac syndrome (n = 8), Behçet’s disease (n = 6), multiple sclerosis (n = 3), and other less frequent or undifferentiated uveitic entities. A majority of included eyes (221/318, 69%) were receiving some form of treatment at the time of imaging, including topical or systemic therapy.
FA showed active vasculitis in 94 eyes within the 12 × 12-mm area, 114eyes confined to regions outside this field, and 110 eyes with no FA activity. The proportions of FA-positive eyes correctly identified by en face OCT using MIP images versus retinal thickness maps were 98% vs 60% for eyes with FA activity within the 12 × 12-mm area and 68% vs 42% for eyes with activity confined to regions outside this field. Among all FA-positive eyes, including those with activity both within and outside the 12 × 12-mm area, the corresponding detection rates were 82% vs 50%. En face positivity using MIP was often observed in eyes with FA activity confined to areas outside the 12 × 12-mm field along contiguous vessel segments corresponding to peripheral FA leakage patterns ( Figure 3 ).
Comparison of late-phase FA and corresponding MIP images. (A) Late-phase FA shows no findings suggestive of inflammatory retinal vascular changes. (B) Corresponding MIP image obtained on the same day demonstrates uniformly hyporeflective vessels, without features suggestive of inflammatory retinal vascular changes. (C) Late-phase FA demonstrates vascular leakage (arrows) within the macula-centered 12 × 12-mm area corresponding to the MIP acquisition field. (D) Corresponding MIP image shows focal hyperreflective changes (arrows) along the involved vessels, suggestive of inflammatory retinal vascular changes. (E) Late-phase FA shows vascular leakage located peripheral to the 12 × 12-mm MIP acquisition field (arrowheads). (F) Corresponding MIP image demonstrates hyperreflective changes along vascular segments associated with the peripheral FA leakage, suggestive of inflammatory retinal vascular changes. (G) Late-phase FA demonstrates diffuse inferotemporal peripheral vascular leakage located outside the 12 × 12-mm MIP acquisition field (arrowheads). (H) Corresponding MIP image demonstrates hyperreflective changes along the inferior arcade vessels (arrow), suggestive of inflammatory retinal vascular changes.
When FA positivity was defined as the presence of inflammatory activity anywhere on FA, including both within and outside the central 12 × 12-mm field, MIP demonstrated a pattern of high sensitivity (0.82) with moderate specificity (0.55). MIP images showed higher sensitivity (0.82 vs 0.50) and F1 score (0.80 vs 0.57) compared with retinal thickness maps. To formally compare the diagnostic performance between the 2 methods, paired comparisons using McNemar testing demonstrated that MIP images had significantly higher sensitivity among FA-positive eyes ( P <.001), while no significant difference in specificity was observed among FA-negative eyes ( P = 1.00). Detailed diagnostic performance metrics are summarized in Table 1 . The confusion matrices are provided in Supplementary Table S1. Cohen’s κ was 0.75 for MIP images and 0.37 for thickness maps. Location-stratified diagnostic performance metrics for activity within and outside the 12 × 12-mm field are provided in Supplementary Table S2.
TABLE 1
Diagnostic Performance of Minimum–Intensity Projection (MIP) Images and Retinal Thickness Maps for Fluorescein Angiography–Defined Inflammatory Retinal Vascular Changes.
| Sensitivity (95% CI) | Specificity (95% CI) | PPV (95% CI) | NPV (95% CI) | F1 Score | |
|---|---|---|---|---|---|
| MIP images | 0.82 (0.76-0.87) | 0.55 (0.46-0.64) | 0.77 (0.71-0.82) | 0.62 (0.52-0.71) | 0.80 |
| Retinal thickness maps | 0.50 (0.43-0.57) | 0.52 (0.43-0.61) | 0.66 (0.58-0.73) | 0.35 (0.28-0.43) | 0.57 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
