Deep Learning and Radiomics for Preoperative Prediction of Metastasis in Ovarian cancer: a systematic review and meta-analysis

Wuling Zhou¹Xinyuan Li¹Hua Yang^1*Mingyu Li¹Jiwei Xin²Xinlu Wang¹Yue Qi¹

• ¹Sheng Jing Hospital Affiliated, China Medical University, Shenyang, China
• ²Liaoning Provincial Health Industry Group Fukuang General Hospital, Fushun, China

Background: Ovarian cancer (OC), recognized as the gynecological malignancy with the highest mortality rate, poses significant challenges for early detection. In recent years, advancements in artificial intelligence (AI) within the medical domain have led to the development of various models designed to facilitate earlier diagnosis and preoperative prediction of metastasis, thereby contributing to a reduction in mortality rates. We assessed their performance in predicting tumor metastasis through preoperative imaging of OC by carrying out a systematic review and meta-analysis. Methods: A systematic review was conducted following PRISMA guidelines using a literature search of WEB OF SCIENCE, PubMed, EMBASE, OVID, and Cochrane Library databases up until December 2025. Studies focusing on the accuracy of radiomics models and/or deep learning techniques for predicting metastatic status in OC were included. The quality of the included studies was assessed using QUADAS-2 scale. Characterization data and diagnostic measures were extracted from each study. Pooling of the area under the curve of the receiver operating characteristic (AUROC) was synthesized in a meta-analysis to identify the strongest contributors to model performance. PROSPERO registration: February 11, 2025. Registration number: CRD420250650908 Results: Fourteen eligible studies were identified for inclusion in the systematic review, of which eleven utilized radiomics models while one employed a deep learning model and the other two incorporated both approaches. Both the radiomics and deep learning models demonstrated robust diagnostic performance; however, a certain degree of heterogeneity was observed across the included studies. Conclusion: AI-based radiomics and/or deep learning may have promising potential in predicting preoperative metastatic status of OC, exhibiting substantial predictive performance; however, variability among studies suggests further large-scale, multicenter prospective studies employing standardized methodologies and external validation are required to confirm clinical applicability.