New research: https://www.sciencedirect.com/science/article/pii/S2588914126000079
AI summary:
Topic: The review evaluates three FDA-approved AI-based diabetic retinopathy screening systems (LumineticsCore, EyeArt, and AEYE-DS) and their suitability for practical healthcare settings.
Key findings and conclusions:
The focus has shifted to implementation: Autonomous AI screening has moved beyond the mere technical testing phase. What matters now is the suitability of these devices for integration into real-world care pathways.
Accuracy alone is not enough: The clinical utility of these systems depends on more than just diagnostic accuracy (sensitivity and specificity). It is also influenced by device usability, imageability, management of referral loads, and the treatment capacity of the local unit.
A seamless care pathway is crucial: AI screening is most effective at preventing vision loss only when it is a seamless part of a larger whole that connects imaging, referral, and the initiation of treatment. In the future, the evaluation of these devices should be based on patient outcomes measured in real-world conditions.