AI Outperforms Traditional Methods in Predicting Ovarian Cancer Surgery Outcomes

Artificial intelligence (AI) models, namely artificial neural networks (ANNs) and machine learning (ML) models, outperform conventional statistics in predicting post–complete cytoreduction (CC) outcomes in patients with ovarian cancer (OC), according to a systematic review published in BMC Surgery.1

Currently, cytoreductive surgery and platinum-based chemotherapy serve as the first-line therapies for OC.2 The researchers emphasized that CC, defined as the absence of visible residual tumor cells post surgery, is “critical” and influenced by the surgeons’ expertise and approach.1

Although AI can potentially improve the quality of OC care, current limitations in image processing prevent its immediate application in predicting debulking outcomes. Despite these challenges, the researchers emphasized AI’s future role. Consequently, they conducted a systematic review to evaluate its accuracy compared with traditional methods in predicting CC surgery outcomes for patients with OC.

Artificial intelligence models outperform traditional methods in predicting post-complete cytoreduction outcomes in patients with ovarian cancer. | Image Credit: © putilov_denis - stock.adobe.com

The study’s outcomes included AI’s diagnostic accuracy in predicting patients’ overall survival (OS), no macroscopically residual disease (R0), length of hospital stay (LOS), and intensive care unit hospitalization. To analyze these, the researchers reviewed all relevant studies that investigated the role of AI in predicting outcomes in patients with OC post CC surgery between 2015 and February 2024; eligible studies needed to use at least 1 AI algorithm, like ANN or ML, to predict CC.

Two researchers searched PubMed, Scopus, Google Scholar, Web of Science, and Cochrane Library for the relevant studies using related keywords. They initially screened studies based on title and abstract, followed by full-text reviews before determining which they would analyze further.

The reviewers initially identified 1013 studies, which they narrowed to 10. From these studies, they extracted data on the authors, evaluated outcomes, and publication year. The 10 studies included 2842 patients whose mean (SD) age was 61.4 (4.75) years. Most studies were conducted in developed countries, like the US. Also, different AI methods were used across the studies to predict CC outcomes in patients with OC, including ANN and ML methods.

Of the studies, 3 provided quantitative data for predicting OS. The pooled estimate of these studies showed that AI achieved 69.64% (95% CI, 66.5%-71.92%) accuracy in predicting OS. Based on Egger’s test, no evidence of publication bias was reported in the studies’ results (t = 1.2; 95% CI, –1.11 to 2.04; P = .087).

Additionally, the pooled estimate of 4 studies showed that AI was 80.5% (95% CI, 71.46%-89.6%) accurate in predicting R0. Egger’s test did not report evidence of publication bias (t = –4.59; 95% CI, -14.6 to 6.1; P = .58).

The 3 remaining studies looked at AI’s ability to predict various outcomes among patients with OC post-CC, namely critical care unit (CCU) needs, LOS, and urinary tract infection. Using the ML method, one study found that AI had a 95% (range, 93%- 97%) accuracy in predicting CCU needs in this patient population . Another used an ANN to predict LOS with 93% (range, 88%-98%) accuracy.

The last study reported that AI could predict UTI risk with 86% accuracy (range, 78%-84%). The researchers considered age, body mass index, blood loss, diabetes, catheter, catheter intubation times, and hypoproteinemia as the most important predictive factors.

The researchers acknowledged their study’s limitations, including the low comparability of studies due to the high outcome variability. Consequently, they considered it “impossible” to determine which algorithm and model are superior in predicting post–CC outcomes in patients with OC. Despite their limitations, the researchers expressed confidence in AI's potential to enhance OC care and outcomes.

“…health care providers must be able to trust the predictions made by AI and use their experience to make clinical decisions based on their outcomes,” the authors concluded. “With this prerequisite fulfilled, AI’s potential to contribute to the shift toward personalized medicine and its predictive value will continue to be realized.”

References

1. Noei Teymoordash S, Zendehdel H, Norouzi AR, Kashian M. Diagnostic accuracy of artificial intelligence algorithms to predict remove all macroscopic disease and survival rate after complete surgical cytoreduction in patients with ovarian cancer: a systematic review and meta-analysis. BMC Surg. 2025;25(1):27. doi:10.1186/s12893-025-02766-3

2. Hinchcliff E, Westin SN, Herzog TJ. State of the science: contemporary front-line treatment of advanced ovarian cancer. Gynecol Oncol. 2022;166(1):18-24. doi:10.1016/j.ygyno.2022.04.021