AI-ASSISTED OCT BIOMARKER QUANTIFICATION REVEALS DISTINCT RETINAL FLUID TRAJECTORIES IN CENTRAL RETINAL VEIN...

To characterize longitudinal patterns of intraretinal (IRF) and subretinal fluid (SRF) dynamics using artificial intelligence-assisted optical coherence tomography (OCT) biomarker quantification and to evaluate their association with long-term visual outcomes in eyes with central retinal vein occlusion (CRVO).

Patients with previously treatment-naïve CRVO and a documented 24-month follow-up were retrospectively included. All eyes received intravitreal aflibercept administered according to a treat-and-extend protocol. Spectral-domain OCT scans were analyzed at baseline and after 3 and 24 months. IRF and SRF volumes were quantified using RetinAI Discovery OCT Biomarker Detector (Ikerian AG). Linear regression models were used to assess the association between fluid volumes and 24-month corrected distance visual acuity (CDVA). Patients were clustered according to their IRF trajectories over time to identify distinct fluid-response phenotypes.

Out of 173 patients, 64 eyes (64 patients) were included. Baseline IRF and SRF volumes showed no significant association with 24-month CDVA. Cluster analysis of longitudinal IRF trajectories identified four distinct fluid-response patterns. Eyes with low or rapidly resolving IRF achieved the best long-term visual outcomes (median CDVA 73 letters), whereas those with persistent or recurrent IRF showed poorer outcomes (median 53 letters; P = 0.008). Early IRF resolution tended to predict superior visual recovery at 24 months.

Artificial intelligence-based volumetric OCT analysis suggests that distinct retinal fluid trajectories in CRVO may have prognostic implications for long-term visual acuity, although these findings are preliminary and require validation. Dynamic IRF resolution patterns may be more informative than static fluid volumes for predicting functional outcomes after anti-vascular endothelial growth factor therapy.