BRIDGING TECHNOLOGY AND HEALTHCARE: EVALUATING AI MODELS FOR DIABETIC RETINOPATHY SCREENING: A COMPREHENSIVE SYSTEMATIC REVIEW
Acta Innovations · 03.11.2026 · 64 (2s)
DOI: 10.62441/actainnovations.vi.3202 ↗Kurz gefasst
Systematische Übersicht in der Zeitschrift Acta Innovations, online erschienen am 03.11.2026.
Abstract (Originalsprache)
Background
Diabetic retinopathy (DR) remains a leading cause of vision loss among diabetic patients. Therefore, prompt diagnosis is crucial. Conventional methods involving human graders have certain limitations like lack of availability of experts, waiting times, and high costs. This systematic review aimed to assess the diagnostic accuracy, reliability, and effectiveness of artificial intelligence (AI) for DR screening.
Methods
A comprehensive search was undertaken in PubMed database and Google Scholar to identify relevant studies. Keywords included in search were “artificial intelligence,” “machine learning,” “deep learning,” diabetic retinopathy,” and “retinal disease”. Studies were included if they evaluated AI-based screening in adults (≥18 years) with diabetes, reported diagnostic performance metrics (sensitivity, specificity), and compared AI performance to conventional methods such as ophthalmologist grading.
Results
A total of 43 studies involving 99411 DR patients were included in the systematic review. The sample size varied considerably ranging from 115 participants to 47,269 participants. The ages of the included participants ranged from 42 to 71 years. The average of male participants in the included studies was 47.8%. Overall prevalence of DR ranged from 3.9% to 68.4%. The referral DR prevalence ranged from 3.9% to 42.5%. The prevalence of severe DR ranged from 0.04% to 18.8%. The average sensitivity for any DR was approximately 90.43%. The average sensitivity for RDR was 95.3%, with only 10 studies reporting this outcome. The average sensitivity of vision-threatening DR was 93.06%. The average specificity for any DR was approximately 86.94%. The average Positive Predictive Value (PPV) across all studies that reported this outcome was approximately 70.0%.
Conclusion
The findings of the present systematic review show that AI diagnostic performance for DR screening is quite high. The performance is high or comparable to human graders. These findings support previous literature that has demonstrated efficacy of AI in DR screening.
Zitieren
APA
(2026). BRIDGING TECHNOLOGY AND HEALTHCARE: EVALUATING AI MODELS FOR DIABETIC RETINOPATHY SCREENING: A COMPREHENSIVE SYSTEMATIC REVIEW. Acta Innovations, 64(2s). https://doi.org/10.62441/actainnovations.vi.3202
Vancouver
BRIDGING TECHNOLOGY AND HEALTHCARE: EVALUATING AI MODELS FOR DIABETIC RETINOPATHY SCREENING: A COMPREHENSIVE SYSTEMATIC REVIEW. Acta Innovations. 2026;64(2s). doi:10.62441/actainnovations.vi.3202
BibTeX
@article{euvimed2026,
title = {BRIDGING TECHNOLOGY AND HEALTHCARE: EVALUATING AI MODELS FOR DIABETIC RETINOPATHY SCREENING: A COMPREHENSIVE SYSTEMATIC REVIEW},
journal = {Acta Innovations},
year = {2026},
volume = {64},
number = {2s},
doi = {10.62441/actainnovations.vi.3202}
}
EUVIMED