Diagnostic accuracy of smartphone-based artificial intelligence systems for detecting diabetic retinopathy: A systematic review and meta-analysis
Authors' conclusion
Smartphone-based AI algorithms show high diagnostic accuracy for detecting DR. However, more high-quality comparative studies are needed to evaluate the effectiveness in real-world clinical settings.
Study context
- Population
- Patients with diabetes screened for diabetic retinopathy (DR); MEDLINE, Embase, Scopus, CINAHL Plus and Cochrane searched from inception to December 2022
- Participants
- 3,931 patients6 studies
- Comparison
- Smartphone-based artificial intelligence (AI) systems vs expert human grader for DR detection
- Outcome
- Pooled sensitivity and specificity for any DR (ADR) and referable DR (RDR), analysed separately
Methodological detailsrandom-effects model
- Random-effects model
Pooled result
As reported in the abstract
Pooled diagnostic accuracy for ADR and RDR
| Group | Any DR (ADR) sensitivity | ADR specificity | Referable DR (RDR) sensitivity | RDR specificity |
|---|---|---|---|---|
| Smartphone-based AI | 88 % | 91.5 % | 98.2 % | 81.2 % |
Smartphone-based AI
Any DR (ADR) sensitivity88 %
ADR specificity91.5 %
Referable DR (RDR) sensitivity98.2 %
RDR specificity81.2 %
Further pooled results
- The overall risk of bias across the studies was low
Original publication
Diagnostic accuracy of smartphone-based artificial intelligence systems for detecting diabetic retinopathy: A systematic review and meta-analysis. Diabetes Res Clin Pract. Published 5 October 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →