Specialists using an AI tool diagnosed inherited eye diseases better in a 300-person trial

Study title: AI-based clinician decision support system for diagnosis of inherited retinal diseases: a multicenter, randomized trial

Researchers developed Retina4IRD, an AI-based clinician decision support system that predicts genotype categories from fundus photographs and OCT scans, and tested it in internal and external validation and in a 300-participant randomized controlled trial comparing AI-assisted specialists to specialists alone for suspected inherited retinal diseases.

Read the study abstract

The accurate and timely diagnosis of inherited retinal diseases (IRDs) represents an unmet clinical need in ophthalmology, as the current pathways rely on resource-intensive phenotyping, multidisciplinary expertise and genetic testing. Here we developed Retina4IRD, an artificial intelligence (AI)-based clinician decision support system (CDSS) that predicts 17 genotype categories from retina images. Retina4IRD uses a Vision Transformer model pretrained with RETFound. We then trained and validated Retina4IRD using…

By the Truvace deskJuly 27, 20261 min read1 sourceEvidence: Moderate · 70/100 — Randomized controlled trial · n = 300 · 1 independent sourcePeer-reviewedTRV-2026-0575· cite & verify

AI-based clinician decision support system for diagnosis of inherited retinal diseases: a multicenter, randomized trial — "CRE-003 OphaGo - Retinal Imaging Adapter - Huced Despro ITS" by Djoko Kuswanto is marked with CC0 1.0. To view the terms, visit https://creativecommons.org/publicdomain/zero/1.0/.

In brief

The trial matters because inherited retinal disease diagnosis currently relies on resource-intensive phenotyping and genetic testing; the results show AI assistance prior to genetic testing can increase diagnostic accuracy and improve management scores, though generalizability beyond the studied populations and long-term clinical outcomes remain to be established.

Main points

Retina4IRD is a Vision Transformer model pretrained with RETFound that predicts 17 genotype categories from retina images.
Model was trained and validated on multimodal data from 1,843 genetically confirmed patients (3,376 eyes) across China, South Korea and Poland.
Randomized trial enrolled 300 participants with suspected IRD, with 295 included in final analysis after next-generation sequencing, median age 33 years.
Secondary endpoints favored AI-assisted arm including top-1 accuracy 37.8% versus 22.4% and top-4 accuracy 81.8% versus 53.1%.

The gain

Clinicians assisted by Retina4IRD achieved higher top-5 genetic diagnosis accuracy and better downstream management scores for suspected inherited retinal disease compared to specialist-only care.

The rundown

Retina4IRD was developed to predict 17 genotype categories from color fundus photographs and optical coherence tomography scans, using a Vision Transformer pretrained with RETFound and trained on 1,843 genetically confirmed patients.

In the randomized controlled trial, 300 participants with suspected IRD were randomized 1:1 to Retina4IRD-assisted specialist versus specialist-only, with final analysis on 295 participants with available next-generation sequencing reports.

Beyond top-5 accuracy, top-1 to top-4 accuracies all favored the assisted arm, and post hoc analyses showed better management decisions with a composite downstream management score of 37.7 versus 28.5.