Diagnostic performance of artificial intelligence–based computed tomography methods in detecting pancreatic cancer: A systematic review and meta-analysis.

Abstract

e16372 Background: Pancreatic cancer is among the deadliest tumors. Artificial intelligence (AI) methods could enhance the identification of pancreatic lesions. The aim of this study was to investigate the diagnostic performance of AI-based computed tomography (CT) in the diagnosis of pancreatic cancer. Methods: A comprehensive search was conducted in the following databases to find relevant papers research: PubMed/Medline, the Cochrane Library, Web of Science, Embase, Scopus, CINAHL, and Google Scholar. The true positive, false positive, true negative, and false negative values were extracted or calculated from the included papers. Results: Ten studies with a total of 33,174 patients met the inclusion criteria. The pooled diagnostic sensitivity and specificity were 0,92 (95% CI: 0,92-0,93) and 0,98 (95% CI: 0,98-0,98), respectively. The area under the summary receiver operating characteristics curve was 0,959, and the diagnostic odds ratio was 179,57 (95% CI: 57,98-556,16), showing excellent diagnostic precision in the identification of pancreatic cancer. Although a significant heterogeneity was detected, it was determined that threshold effect did not contribute to differences in accuracy estimations between different studies. Furthermore, meta-regression analysis failed to find the source of heterogeneity. On the other hand, no evidence of publication bias was detected. Conclusions: These encouraging initial results suggest that AI-based CT may be a valuable tool for computer-aided pancreatic cancer diagnosis. Additional testing on a larger dataset is needed to confirm these findings.