<![CDATA[Hypertensive retinopathy (HR) is a retinal condition caused by chronic hypertension that often progresses silently and usually detected incidentally, making early diagnosis essential to prevent further progression. While current evaluations rely on ophthalmologists or retinal specialists, artificial intelligence (AI) offers a promising, time-efficient, and resource-effective alternative for HR diagnosis. This study aims to provide a robust and reliable evaluation of the diagnostic potential of AI for hypertensive retinopathy through a comprehensive systematic review and network meta-analysis, which has not been previously reported. An extensive literature search was conducted across six databases, including PubMed, ScienceDirect, SpringerLink, Taylor & Francis Online, ProQuest, and Sage Journals. The inclusion criteria were diagnostic studies published within the past decade that involved both healthy populations and patients with hypertensive retinopathy, in which the diagnosis of hypertensive retinopathy was performed by artificial intelligence and compared with that of ophthalmologists. The primary outcomes of interest included the sensitivity, specificity, and accuracy of AI in diagnosing HR from fundus image photography. Quality assessment was performed using the quality assessment of diagnostic accuracy studies (QUADAS) tool, and meta-analysis was conducted using STATA. From 1,530 articles screened, 10 studies were included in this analysis, comprising a total of 38,761 fundus images. A total of 38,761 fundus images from 10 studies were included in this analysis. The results indicated that AI demonstrated high sensitivity [90% (95% CI: 86% – 93%)] and specificity [95% (95% CI: 92% – 97%)] with p-value of 0.001. Based on these findings, AI shows promise as a future diagnostic option for HR due to its high sensitivity and specificity.]]>
Copyrights © 2025
