Artificial Intelligence in Optical Coherence Tomography‐Based Ophthalmic Imaging: From Automated Diagnosis to Personalized Disease Monitoring
TL;DR
Clinically useful OCT‐based AI should be implemented as clinician‐led decision support, with explicit intended use, external validation, calibration, post‐deployment monitoring, and lifecycle governance.
Abstract
Artificial intelligence (AI) is increasingly applied to optical coherence tomography (OCT) in ophthalmology, but evidence for automated diagnosis is stronger than evidence for longitudinal monitoring. This narrative review evaluates OCT‐based AI with emphasis on disease monitoring and progression, particularly in retinal disease and glaucoma. Literature from 2016 to 2026 was screened across major biomedical databases, and 45 verifiable sources were included. Retinal OCT models support referral triage, fluid segmentation, and treatment‐response prediction, whereas glaucoma applications face specific challenges: retinal nerve fiber layer floor effects, inconsistent progression definitions, reference‐standard circularity, spectrum bias, high myopia, anomalous discs, and device‐related domain shift. Recent meta‐analyses show high diagnostic performance for glaucoma detection, but progression prediction remains less externally validated. Clinically useful OCT‐based AI should therefore be implemented as clinician‐led decision support, with explicit intended use, external validation, calibration, post‐deployment monitoring, and lifecycle governance.