Remaining Evidence Gaps for AI-Powered Eye Imaging Devices in Clinical Use

Recent evaluations reveal significant shortcomings in the evidence supporting the clinical performance of AI-based eye imaging devices that have received regulatory approval. A comprehensive review conducted by researchers from University College London and Moorfields Eye Hospital analyzed 36 AI tools approved across Europe, Australia, and the United States, uncovering notable gaps in the validation process. Alarmingly, nearly 20% of these devices lack peer-reviewed data on their accuracy and clinical outcomes.

The review examined 131 clinical evaluations, finding that only around half of the studies reported patient age, just over half included sex data, and a mere 21% documented ethnicity. Many validations relied on archival image datasets, which often lack diversity or thorough demographic reporting, raising concerns about the generalizability of these AI tools. Few studies compared different AI systems or evaluated their performance against human clinicians, and only 8% conducted interventional testing in real-world clinical environments.

Most AI devices focus on screening for diabetic retinopathy, but significant gaps remain in identifying other serious conditions such as glaucoma or age-related macular degeneration. The regulatory landscape shows disparities, with 97% of the reviewed devices approved in the European Union, but only 22% cleared for use in Australia, and just 8% authorized in the U.S. This inconsistency underscores the need for standardized, transparent evidence and data sharing practices that adhere to FAIR principles—Findability, Accessibility, Interoperability, and Reusability.

Lead researcher Dr. Ariel Ong emphasized that AI has substantial potential to address global eye care disparities by enabling early detection where specialists are scarce. However, for AI tools to be trustworthy and widely adopted, they must meet high standards of validation, with rigorous testing across diverse populations and real-world settings. Industry experts advocate for standardized reporting, including detailed 'model cards' and enhanced regulatory guidelines like the EU AI Act, to improve data transparency and ensure equitable performance.

The study calls on manufacturers and regulators to ensure that AI in ophthalmology is validated thoroughly and transparently, ultimately fostering confidence among clinicians and patients. The goal is a future where AI-driven eye care is both effective and equitable, reducing preventable vision loss worldwide.

This review was collaboratively conducted by institutions in the UK, Australia, and the US, highlighting the importance of global standards in AI medical device regulation.

Source: https://medicalxpress.com/news/2025-06-evidence-gaps-ai-eye-imaging.html