人工智能深度学习模型对早期胃癌诊断价值的系统评价与meta分析OA

Objective To systematically evaluate the effectiveness and clinical value of artificial intelli-gence(AI)deep-learning models in the auxiliary diagnosis of early gastric cancer.Methods Literature sear-ches were conducted in the PubMed,Embase,Cochrane Library,and Web of Science databases to identify rele-vant studies on AI deep-learning models for diagnosing early gastric cancer and to extract data.The quality and risk of bias of the included studies were assessed using the QUADAS-2 scale.A bivariate mixed-effects re-gression model was used for the primary meta-analysis;subgroup analyses,sensitivity analyses,and heteroge-neity testing were also performed.Results A total of 11 studies were included.The pooled sensitivity,speci-ficity and AUC of the AI deep-learning models in diagnosing early gastric cancer were 0.89(95%CI:0.80-0.94),0.92(95%CI:0.84-0.97)and 0.96(95%CI:0.94-0.97),respectively.Subgroup analysis revealed that the imaging modality and data intensity were the key factors influencing the diagnostic efficacy.The AUC based on other endoscopic models was slightly higher than that of white light endoscopy models(0.95 vs.0.93),and the AUC of the model with≥5 000 training set images was superior to that of the model with<5 000 training set images(0.97 vs.0.95).Conclusion AI deep-learning models demonstrate promising per-formance in assisted diagnosing early-stage gastric cancer,with clear potential for clinical translation.