Retina Society 2026: AI flags retinal inflammation without dye

At the Retina Society's 59th Annual Scientific Meeting, Parisa Emami-Naeini, MD, MPH, presented data on a multimodal, dye-free inflammation severity score that combines optical coherence tomography (OCT) and ultra-widefield fundus photography (UWF FP) using machine learning. The goal, Emami-Naeini said, was to evaluate retinal inflammation more objectively in patients with uveitis, particularly those with retinal vasculitis, without relying on fluorescein angiography (FA). FA requires same-day dye injection and is invasive, time-consuming, and impractical for the frequent monitoring uveitis often requires, she said.

Model performance across OCT and widefield imaging

The OCT-based model achieved an area under the receiver operating characteristic curve (AUROC) of 0.980, with 85% sensitivity and 97% specificity, and showed strong correlation with expert FA grading (Pearson r, 0.93). The UWF FP model, designed to capture peripheral inflammatory activity, achieved an AUROC of 0.87. On external validation using a dataset from the University of Nebraska, the OCT model maintained an AUROC of 0.845, and the UWF FP model achieved an AUROC of 0.91. Combining both modalities produced an inflammation score that aligned more closely with overall disease burden than either modality alone, Emami-Naeini said, reflecting complementary central and peripheral inflammatory information.

Implications for dye-free uveitis monitoring

Emami-Naeini said the results offer proof of concept that clinically meaningful inflammation data is already present in noninvasive imaging routinely collected in clinic. Combining OCT and UWF FP data could enable more frequent, quantitative monitoring between fluorescein angiography visits, she said. She cautioned that the model is not yet ready to replace FA, noting that prospective longitudinal validation is still needed, particularly to confirm the score reliably tracks inflammation as it improves or worsens over time and can help guide treatment escalation or de-escalation.

A path toward standardized outcome measures

Looking beyond individual patient care, Emami-Naeini said a validated, standardized imaging-based severity score could offer a reproducible, continuous measure of inflammatory burden in a disease often assessed categorically. If validated across populations and imaging platforms, such a score could provide common ground for comparing disease activity across centers and specialists, and serve as an end point for future clinical trials.