Clinical Publications

TRITON

Improving community-based OCT screening through real-time AI-based biomarker visualization

AUTHORS:

Pichelmann J; Chopra R; Vogl WD; Unterdechler M; Leingang O; Al-khersan H; Dow E; Wykoff CC; Schmidt-Erfurth U; Whitby A

PUBLICATION:

ARVO 2026 Abstract / Conference Presentation

Study Purpose

To assess whether real-time AI-based biomarker visualization improves the accuracy and efficiency of human graders in detecting retinal biomarkers in community-based OCT screening.

Overview

Study Design

Multi-reader, multi-case crossover study in a simulated clinical workflow. Two optometrists and two reading center readers graded 112 real-world Maestro 2 widefield OCT volumes with and without AI assistance after a washout period of more than 4 weeks. An independent clinical expert established the reference standard.

# of Eyes/
Patients

112 OCT volumes (graded by 4 readers)

Study Device

Swept-source OCT (Triton, Topcon Corporation, Tokyo, Japan)

Outcome Measures

Topcon Maestro 2 widefield OCT (12 x 9 mm) with AI-based biomarker detection and visualization integrated into a single-page report / Topcon Harmony PACS workflow

Results

AI support improved overall detection performance by 42%, increasing global kappa from 0.41 without AI to 0.58 with AI. Sensitivity improved for RPEL (0.64 to 0.96), SRO (0.65 to 0.95), and PED (0.69 to 0.81), while precision improved for SRO (0.68 to 0.83) and PED (0.54 to 0.64). Standalone AI agreement with the expert for IRO was high (kappa = 0.95), although human agreement for IRO remained modest. AI support also reduced average false-positive detections by 60%, improved inter-reader consistency, and shortened grading time by 16.3% (~12.1 seconds per scan).

IMOvifa reduced measurement time by

39%

Conclusions

Real-time AI-based biomarker visualization can improve the quality, consistency, and efficiency of community-based OCT screening by helping graders identify key retinal biomarkers more accurately and faster. These findings support the potential role of AI-assisted OCT interpretation in eye-care screening and referral workflows.