Home›Ophthalmology› AI models detect retinal detachment with 95.7% sensitivity and 99.2% specificity in meta-analysis
AI models detect retinal detachment with 95.7% sensitivity and 99.2% specificity in meta-analysisMachine learning models show high accuracy in detecting retinal detachment
European journal of ophthalmologyPublished September 4, 2026Study authors: Łajczak Paweł, Łajczak AnnaPubMed ↗DOI ↗Editorial oversight: Dr. Lars van Dijk, PhD · Surgical, Procedural & Diagnostic
AI-generated summary of the cited source, checked by automated accuracy review.
How we work
Share
Key Takeaway
Consider AI models as adjunctive tools for retinal detachment detection, but validate locally due to heterogeneity.
This meta-analysis synthesized diagnostic accuracy data from 69 machine learning (ML) and deep learning (DL) models for retinal detachment detection. The analysis pooled results across studies, reporting an overall sensitivity of 95.7% (95% CI: 94.1-96.9%) and specificity of 99.2% (95% CI: 98.7-99.5%). Subgroup analyses indicated that DL models achieved higher sensitivity (96.3%) and specificity (99.4%) compared with ML models (92.7% and 96.7%, respectively). Models using fundus imaging demonstrated superior performance, with sensitivity of 97.3% and specificity of 99.5%, compared with other imaging modalities.
The authors noted significant heterogeneity across studies (I > 90%), which limits the certainty of pooled estimates. They also highlighted challenges in generalizability due to biases in patient selection and data quality. These factors suggest that while the diagnostic accuracy appears high, the true performance in diverse clinical settings may vary.
The meta-analysis did not report on adverse events, follow-up, or funding sources. The findings are based on diagnostic test accuracy studies, not on patient outcomes, so the clinical impact of these models on management and visual outcomes remains uncertain.
For clinical practice, these results suggest that AI-based tools, particularly those using deep learning and fundus imaging, hold promise for assisting in retinal detachment detection. However, adoption requires addressing heterogeneity and bias through standardization and multicenter validation. Clinicians should interpret these accuracy estimates cautiously and consider them as supportive rather than definitive for diagnosis.
How this fits prior evidence
This meta-analysis extends prior coverage on retinal detachment by quantifying the diagnostic accuracy of AI models, whereas earlier items focused on surgical and adjunctive treatments. It confirms the potential of AI in detection, complementing the surgical reattachment rates reported for familial exudative vitreoretinopathy-associated detachments. However, it does not address treatment outcomes, and the high heterogeneity contrasts with the more specific surgical evidence. The findings address a gap by providing pooled accuracy estimates, but they do not resolve questions about clinical integration or comparative effectiveness against standard ophthalmoscopic diagnosis.
Retinal detachment is a serious eye condition that requires quick detection to protect a person's vision. Because speed is so important, researchers are looking at ways to use technology to help doctors spot the problem sooner. This study looked at 69 different models to see how well computer programs could identify the condition.
Researchers found that deep learning models, which are advanced types of machine learning, performed very well. These systems showed a sensitivity of 96.3% and a specificity of 99.4%. When looking specifically at fundus imaging, the accuracy was even higher. These results suggest that computer-assisted tools can be very reliable for identifying retinal issues.
While the results are promising, there are still hurdles to clear before these tools are used in every clinic. The study noted a lot of variety in the data, which makes it hard to know exactly how these tools will perform in every hospital. There are also concerns about how well these tools work across different types of patients. More testing is needed to make sure these tools work consistently for everyone.
What this means for you:
Deep learning models show high accuracy in detecting retinal detachment, especially when using fundus imaging.
Retinal detachment (RD) is a sight-threatening condition requiring rapid diagnosis to prevent vision loss. Artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), offers potential for improving diagnostic accuracy in RD, addressing current limitations of manual interpretation. A diagnostic test accuracy (DTA) meta-analysis adhering to PRISMA 2018 guidelines was conducted. Studies utilizing ML for RD detection across imaging modalities were included. Databases (PubMed, Web of Science, Scopus, Cochrane) and manual searches identified 20 studies. Statistical analyses assessed sensitivity, specificity, and area under the curve (AUC), with subgroup analyses by ML technique, imaging modality, validation method, and testing set size. From 69 models analyzed, pooled sensitivity and specificity were 95.7% (95% CI: 94.1-96.9%) and 99.2% (95% CI: 98.7-99.5%), respectively, indicating high diagnostic accuracy. DL models outperformed ML, achieving higher sensitivity (96.3% vs. 92.7%) and specificity (99.4% vs. 96.7%). Models employing fundus imaging exhibited superior performance (sensitivity: 97.3%; specificity: 99.5%). However, significant heterogeneity (I > 90%) was noted. External validation enhanced specificity but highlighted challenges in generalizability due to biases in patient selection and data quality. ML demonstrates high potential for accurate RD detection, particularly using DL and fundus imaging. Nonetheless, addressing biases, heterogeneity, and external validation is crucial for clinical adoption. Future research should focus on standardization, cost-effectiveness, and multicenter validation to ensure practical ML integration in ophthalmology.
