AI-based clinician decision support system for diagnosis of inherited retinal diseases: a multicenter, randomized trial.
An AI system analyzing eye images helps eye specialists more accurately diagnose inherited retinal diseases before genetic testing, potentially speeding care for people with rare vision loss.
This multicenter RCT (n=295 analyzed) demonstrates that the Retina4IRD AI decision support system, using Vision Transformer analysis of fundus photographs and OCT scans pretrained with RETFound, significantly improves genetic diagnosis accuracy for 17 inherited retinal disease genotypes compared to specialist assessment alone. AI assistance also improved downstream clinical management decisions, supporting this tool as a validated pre-genetic-testing workflow aid.
What the study was
- Study design
- Multicenter randomized controlled trial (n=300 randomized; 295 analyzed)
- Population
- Adults with suspected inherited retinal diseases at specialist centers in China, South Korea, and Poland
- Sample size
- 295
- Category
- Diagnostics
- Maturity
- Potentially Practice-Changing
- Journal
- Nature medicine
Why it surfaced
Nature Medicine multicenter RCT demonstrating AI decision support system significantly improves genetic diagnosis accuracy for rare inherited retinal diseases. First such RCT; primary endpoint clearly met (88.5% vs 67.3%); tool integrates with existing clinical workflow prior to genetic testing. ClinicalTrials.gov NCT06839170.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.