AI Excels in Diabetic Retinopathy Screening New Study

The Dawn of AI in Diabetic Eye Care

Artificial intelligence (AI) is increasingly showing its potential to revolutionize healthcare, and a recent study highlights its promising role in screening for diabetic retinopathy (DR). Diabetic retinopathy is a common complication of diabetes and a leading cause of vision loss in working-age adults, making early detection crucial. This new research suggests AI systems could be a game-changer in how we approach DR screenings.

A study led by Hasan Nawaz Tahir, MS, of Shaqra University in Saudi Arabia and Khyber Medical College Peshawar in Pakistan, conducted a systematic review and meta-analysis. The findings indicate that AI systems can achieve high sensitivity and specificity in detecting DR, especially when screening un-dilated eyes. The researchers stated, "These results highlight that AI systems, especially in un-dilated eye conditions, show promise for clinical use with reliable sensitivity and specificity, but variation exist depending on the system and clinical setting."

AI-assisted screening for diabetic retinopathy showed comparable sensitivity and specificity compared to manual screening. Image credit: AdobeStock/ipopba

Diving Deep into the Research Methodology

To arrive at their conclusions, the researchers undertook an extensive literature search using PubMed and Google Scholar. They looked for studies published between January 2015 and September 2024, with an additional search in February 2025 adding more studies to their analysis. This resulted in a total of 25 studies being included.

These studies were either observational or validation studies that evaluated AI algorithms or manual screenings for DR. The patient population ranged from 15 to 90 years old, all diagnosed with DR. A key factor for inclusion was the reporting of sensitivity and specificity outcomes for either dilated or un-dilated eye conditions. Altogether, an impressive 613,690 images used for screening were part of this comprehensive analysis.

The quality and potential bias of the included studies were rigorously assessed. The team used the QUADAS-2 tool for validation studies, the AXIS tool for cross-sectional studies, and the Newcastle-Ottawa Scale for cohort studies. Two independent reviewers also evaluated various types of bias, such as selection bias, outcome assessment bias, and reporting bias. Studies were excluded if they didn't report the necessary outcomes, if authors didn't respond to queries, or if the full text wasn't available.

AI vs Manual Screening The Numbers Speak

The meta-analysis yielded compelling statistics regarding the performance of AI compared to manual screening methods [1]:

• For un-dilated eyes:
  • AI screening: Pooled sensitivity of 0.90 (95% CI: 0.85–0.94) and pooled specificity of 0.94 (95% CI: 0.91–0.96).
  • Manual screening: Pooled sensitivity of 0.79 (95% CI: 0.60–0.91) and pooled specificity of 0.99 (95% CI: 0.98–0.99).
• For dilated eyes:
  • AI screening: Pooled sensitivity of 0.95 (95% CI: 0.91–0.97) and pooled specificity of 0.87 (95% CI: 0.79–0.92).
  • Manual screening: Pooled sensitivity of 0.90 (95% CI: 0.87–0.92) and pooled specificity of 0.99 (95% CI: 0.99–1.00).

These figures show that AI performs robustly, particularly demonstrating higher sensitivity in un-dilated eye screenings and comparable or higher sensitivity in dilated eye screenings when compared to manual methods.

What This Means for Patients and Doctors

The implications of these findings are significant for both eye care professionals and individuals with diabetes. The study authors stated, "AI-assisted screening for diabetic retinopathy shows comparable sensitivity and specificity compared to manual screening." They further elaborated, "These results suggest that AI can be a reliable alternative in clinical settings, with increased early detection rates and reducing the burden on ophthalmologists."

Earlier detection of DR can lead to more timely interventions, potentially preventing vision loss. AI systems could also help streamline the screening process, making it more efficient and accessible, especially in areas with limited access to ophthalmologists.

Navigating Limitations and Looking Ahead

While the results are promising, the study authors were careful to note certain limitations. There was heterogeneity among the included studies regarding settings, photographic protocols, and the reference standards used for manual grading. Additionally, some studies had a moderate risk of bias, which could influence the accuracy estimates.

The researchers emphasized the need for continued evaluation. "To ensure that AI systems are safe and effective for real-world use, evaluations need to be conducted in representative clinical settings. Systems should be tested on a wide range of image qualities, and medical settings," they noted. This careful approach is essential before widespread adoption in clinical practice.

The Bottom Line on AI in DR Screening

This comprehensive review and meta-analysis provides strong evidence for the potential of AI systems in diabetic retinopathy screening. With comparable, and in some aspects superior, performance to manual screening, AI stands as a promising tool to enhance early detection, improve patient outcomes, and support ophthalmologists in managing this sight-threatening condition. The ongoing refinement and validation of these AI technologies will be key to their successful integration into routine eye care.

Reference

1. Tahir HN, Ullah N, Tahir M, et al. Artificial intelligence versus manual screening for the detection of diabetic retinopathy: a comparative systematic review and meta-analysis. Front Med. 2025;12. https://doi.org/10.3389/fmed.2025.1519768