Truace
Subscribe

Artificial Intelligence in Eye Disease: A Systematic Review of Screening, Diagnosis, Prognosis, Treatment-Response Prediction, and Clinical Decision Support

Background/Objectives: Artificial intelligence (AI) is expanding eye care beyond image-based detection toward prognosis, treatment-response prediction, monitoring, triage, and clinician-facing decision support. We evaluated AI across eye-disease screening, diagnosis, prognosis, treatment response, and decision support, emphasizing validation, explainability, and clinical readiness. Methods: This PROSPERO-registered review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 20…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-1490· cite & verify
Artificial Intelligence in Eye Disease: A Systematic Review of Screening, Diagnosis, Prognosis, Treatment-Response Prediction, and Clinical Decision Support
"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

A PROSPERO-registered systematic review following PRISMA 2020 mapped the literature on AI in eye care from 2015 through August 2026. It covered 270 studies across screening, diagnosis, prognosis, treatment-response prediction and clinical decision support, with diabetic retinopathy and macular oedema the largest area.

The review matters because it shows ophthalmic AI moving beyond image detection toward clinical management, yet the evidence for real-world use is uneven. Because study heterogeneity ruled out meta-analysis, the review cannot provide pooled performance estimates, and the effects on safety, workflows and patient outcomes remain largely untested.

Main points

  1. The evidence base comprised 270 studies, with diabetic retinopathy/diabetic macular oedema the largest clinical area at 82 studies.
  2. Glaucoma accounted for 46 studies, retinal/multi-disease AI for 52, and AMD/macular disease for 29.
  3. Convolutional neural network or convolutional deep learning was the largest primary method group, with 136 studies.
  4. External validation was confirmed in 130 studies, but only 49 had a strict direct clinical comparator.
  5. Explicit optometry or optometrist involvement appeared in only three studies.

The problem

Eye-disease AI is not yet clinically ready: independent validation is incomplete, direct clinician comparators are scarce, and safety, workflow and patient-outcome evidence is sparse.

The rundown

The authors searched six databases plus supplementary routes for English-language literature published between 1 January 2015 and 10 August 2026. Included studies received design-mapped Joanna Briggs Institute or adapted item-level appraisal.

Of the 130 studies with confirmed external validation, 95 had a clinical, geographic, prospective or real-world external-validation component. Of these, 86 were clinical-only and nine mixed clinical and technical validation.

Ninety studies used explainability methods, and 30 involved pediatric or pediatric-relevant evidence. The authors emphasize validation, explainability and clinical readiness as the criteria for assessing translation.

Sources

  1. Peer-reviewedJournal of Clinical Medicine2026-10-14

The debate