AI-based image analysis detects referable diabetic retinopathy

HTA syd pooled 13 studies and found, with moderate certainty, that AI grading of retinal photographs detects referable diabetic retinopathy with about 92% sensitivity and 91% specificity.

  • Not rated an organization

    The review pooled accuracy against human graders, with only a scenario of 1,400 staff hours a year for a region that did not use AI grading.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Reported benefit.
Evidence

foundstudiedreportedclaimed

An independent assessment report supports the benefit.

Compared with

Human graders reading the same fundus photographs. Measured.

The system’s part
Not isolated. The pooled studies compared AI grading with human graders' reading of the same photographs, which measures the software's accuracy. Only two studies compared the time spent with and without AI. Several included studies used deep-learning models, such as an Inception-ResNet model in Rwanda and the locally trained THEIA in New Zealand.
Not yet checked
LimitsCosts and harmsLearns from dataLater outcome

Download data Cite Suggest a correction

Benefit type
Improved Task Accuracy
System
AI-based image analysis of fundus photographs for diabetic retinopathy screening (products in the included studies such as EyeArt, IDx-DR, SELENA, RetCAD and Aireen)
Made by
Several vendors, among them Eyenuk (EyeArt), Digital Diagnostics (IDx-DR) and EyRIS (SELENA)
Used by
Clinics and screening programs in the comparative studies that HTA syd reviewed
Where
Studies in Europe, Asia, Africa, Oceania and North America, assessed for Sweden's Southern Healthcare Region
Stage
In a pilot
Date
, when the source was published
Kind
Aggregate result
Broad benefit
Identifying Objects, Actions, and Events
Collections
Scrutiny and findings
Last revised
1 November 2026

Sources

  1. Assessment report, by authors with no tie to the system.

    HTA syd, Region Skåne. AI-baserad bildanalys vid screening för diabetesretinopati: en systematisk översikt och utvärdering av medicinska, hälsoekonomiska, etiska och organisatoriska aspekter. HTA syd Rapport 2025:4, published 17 December 2025. Abstract (Results), p. 2.

    AI-based image analysis of fundus photographs demonstrated high diagnostic accuracy for detecting referable diabetic retinopathy, with pooled sensitivity of approximately 92% and specificity of 91% (moderate certainty).

Details

AI system

Workflow

The software graded fundus photographs taken in screening, most of them without dilating the pupil. Two to three human graders in consensus read the same photographs as the reference standard.

Follow-up

Conditions for replication

The software must use referral thresholds that fit Swedish grading and screening intervals. An ophthalmologist must review within a few days the images that the software sorts as most likely severe.

Entry details

Sector

Healthcare (diabetic retinopathy screening).

Purpose

Vendors such as Eyenuk, Digital Diagnostics and EyRIS built AI-based fundus image analysis software to grade retinal photographs for diabetic retinopathy and flag patients who need referral.

31 more questions not yet checked

model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.

Cite this record

Peter H. Vartanian. (2026). AI-based image analysis detects referable diabetic retinopathy (AIBD-1033). Artificial Intelligence Benefits Database, edition of 1 November 2026. https://petervartanian.xyz/aibd/records/AIBD-1033.html (archived at https://doi.org/10.7910/DVN/AADSII, version 7.0)

BibTeX Data Field definitions