Introduction Gastric cancer remains one of the leading causes of cancer-related mortality worldwide, with lymph node metastasis (LNM) being an independent prognostic factor. However, there are still challenges in the pathological diagnosis of LNM in gastric cancer (GC). The aim of this meta-analysis is to systematically evaluate the accuracy of artificial intelligence (AI) in detecting LNM in GC from whole-slide pathological images.. Methods As of March 24, 2024, a comprehensive search for studies on the pathological diagnosis of GC LNM AI was performed in the databases of PubMed, Web of Science, Cochrane Library, and CNKI. Meta-analysis of the included data was performed using Meta-Disc 1.4, Review Manager 5.4, and Stata SE 17.0 software to calculate diagnostic metrics such as overall sensitivity and specificity. The overall diagnostic performance of the AI was assessed. Meta-regression analysis explored sources of heterogeneity. Results A total of 7 articles involving 1,669 GC patients were included. The analysis showed that AI had a sensitivity of 0.90 (95% CI: 0.84-0.94) and a specificity of 0.95 (95% CI: 0.91-0.98) for the diagnosis of GC LNM, with significant heterogeneity across studies. The area under the curve was 0.97, indicating an excellent diagnostic value. Meta-regression analysis showed that the sample size and the number of study centers contributed to the heterogeneity. Conclusion AI for diagnosing LNM in GC from whole-slide pathological images demonstrates high accuracy, offering significant clinical implications for improving diagnosis and treatment strategies.
Read full abstract