Abstract
BACKGROUND: Complete Cytoreduction (CC) in ovarian cancer (OC) has been associated with better outcomes. Outcomes after CC have a multifactorial and interrelated cause that may not be predictable by conventional statistical methods. Artificial intelligence (AI) may be more accurate in predicting outcomes. This systematic review aimed to determine the accuracy of AI compared to traditional statistics in predicting outcomes after CC in OC. METHODS: PubMed, Scopus, Google Scholar, Embase, and Web of Science databases were searched with Mesh terms to find studies that investigated the role of AI in predicting outcomes after CC in EOC from the beginning of 2015 to February 2024. The outcomes included overall survival (OS), removal of all macroscopic disease (R0), length of hospital stay (LOS), and intensive care unit (ICU) admission. This systematic review was conducted based on the PRISMA guidelines. Heterogeneity between studies was evaluated using the I(2) test. Egger's test was used to check publication bias. RESULTS: Ten studies (3460 participants) were included. The pooled estimate of 3 studies showed that the accuracy of AI for predicting OS was (Mean: 69.64%, CI 95%:66.50, 72.78%, I(2):0%). The pooled estimate of 4 studies showed that the accuracy of AI for predicting R0 was (Mean: 80.5%, CI 95%:71.46, 89.6%, I(2):47.9%). The use of AI in predicting outcomes, including ICU admission, urinary tract infection (UTI), and LOS was investigated in one study, and the AUC of AI for predicting all three outcomes was approximately 90%. CONCLUSION: AI may accurately predict the outcomes after CC in OC patients. Most studies agree that Artificial Neural Networks (ANN) and Machine Learning (ML) models outperform conventional statistics in predicting postoperative outcomes.