Abstract:AIM To evaluate the application value of artificial intelligence-assisted systems in diagnosing diabetic retinopathy (DR) by Meta-analysis. Methods PubMed, Web of Science, Embase, Cochrane Library, CBM, CNKI, WanFang Data and VIP database were searched to collect relevant literature on the diagnostic value of artificial intelligence-assisted systems for DR from January 2019 to September 2024. The QUADAS-2 tool was used to evaluate the quality of the included studies, and meta-analysis was performed using Stata 17.0 and Meta Disc 1.4 software. Results：A total of 23 studies were included. The results of meta-analysis showed that the pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, diagnostic odds ratio and area under the SROC curve (AUC) were 0.92 (95%CI: 0.89, 0.94), 0.94 (95%CI: 0.91, 0.96), 15.6 (95%CI: 10.6, 22.9), 0.09 (95%CI: 0.07, 0.12), 174 (95%CI: 112, 271) and 0.97(95%CI: 0.96，0.98).Meta-regression and subgroup analyses indicated that the heterogeneity of the studies originated from study type, patient type, patient source, and AI algorithm type. Deeks’ funnel plot test suggested no significant publication bias (P = 0.15 > 0.05), indicating that the results were robust. Conclusions：The artificial intelligence-assisted system demonstrates high diagnostic value for DR and can be widely implemented in the early screening and diagnosis of DR.
