Moderate evidence
More
Eye clinics make limited use of AI that often tops 90% accuracy, a review of 34 studies shows
Study title: Why aren't we using AI in eye clinics? A systematic review of barriers and solutions in AI-based fundus image diagnostics for ocular diseases
- Evidence type
- Systematic review or meta-analysis
- Sample
- n = 34
- Sources
- 1 sourcePeer-reviewed
- Published
- 1 min read
- Truvace record
- TRV-2026-1127cite or check it
Study title: Why aren't we using AI in eye clinics? A systematic review of barriers and solutions in AI-based fundus image diagnostics for ocular diseases
A systematic review of 34 studies from 2018-2025 found AI systems for fundus image diagnosis of diabetic retinopathy, glaucoma, ROP and AMD frequently exceeded 90% accuracy and matched or outperformed experts, yet adoption in eye clinics remained limited.
Read the study abstract
Artificial intelligence (AI) has shown remarkable accuracy in the diagnosis of common ocular diseases such as diabetic retinopathy (DR), glaucoma, retinopathy of prematurity (ROP), and age-related macular degeneration (AMD), often matching or even outperforming expert clinicians. Despite these advancements, AI adoption in clinical settings remains limited due to key barriers. This systematic review evaluates 34 studies (2018-2025) highlighting AI's diagnostic performance (often >90% accuracy) while pointing out…

The problem
A systematic review of 34 studies shows eye-scan AI often tops 90% accuracy yet sees limited clinical adoption, which the review attributes to poor workflow integration, lack of transparency in decisions and weak generalizability.
In brief
The gap matters because high laboratory accuracy has not translated into routine vision care; the review attributes this to workflow integration failures, lack of transparency in decision-making, and poor generalizability, leaving questions about how to achieve equitable real-world deployment.
Main points
- Systematic review covered 34 studies from 2018-2025 on AI fundus diagnostics.
- Reported AI accuracy often exceeded 90% for DR, glaucoma, ROP and AMD.
- Authors identified three deployment gaps: workflow integration, transparency, and generalizability.
The full story
The review synthesized 34 studies published between 2018 and 2025, focusing on AI for fundus image diagnosis of DR, glaucoma, ROP and AMD.
It framed the challenge as the 'last-mile gap' between research and clinical practice and proposed pathways toward equitable, scalable and trustworthy implementation.
Sources
1 on file- International Journal of Ophthalmologydoi:10.18240/ijo.2026.09.17
Talk it over
The review of 34 studies found eye-scan AI often topped 90% accuracy, yet clinics used it little. How do you weigh high accuracy against the gaps in workflow integration, transparency and generalizability?
Eye clinics make limited use of AI that often tops 90% accuracy, a review of 34 studies showshttps://truvace.com/item/why-arent-we-using-ai-in-eye-clinics-a-systematic-review-of-barriers-and-solutio
The review of 34 studies found eye-scan AI often topped 90% accuracy, yet clinics used it little. How do you weigh high accuracy against the gaps in workflow integration, transparency and generalizability?
Answer in the threadThe review covered studies from 2018-2025 on diabetic retinopathy, glaucoma, ROP and AMD. Whose eyes and which clinics might be missing from that research, and how could that change what accuracy means in practice?
Answer in the threadWho should take the lead in closing what the authors call the 'last-mile gap' between research and clinical practice: the developers, the clinics or the regulators, and why?
Answer in the thread
If you have more time
If you worked in an eye clinic you know, which of the three gaps, workflow integration, transparency or generalizability, would most likely stop an AI scan tool from being used there?
Answer in the threadWhat kind of study, beyond accuracy figures from the 34 studies, would show whether an AI tool actually changes care for people with diabetic retinopathy or glaucoma?
Answer in the thread
ace
What readers say