Development and external validation of a predictive model for epithelial ovarian tumors based on MRI features and serum tumor markers

Ovarian cancer remains a major threat to women’s health, and early-stage disease often lacks specific clinical manifestations, making timely diagnosis challenging. Early and accurate preoperative assessment is essential for improving patient outcomes.To develop and externally validate a predictive model integrating magnetic resonance imaging (MRI) features and serum tumor markers for distinguishing benign from malignant epithelial ovarian tumors. A total of 222 patients from Huzhou Maternity & Child Health Care Hospital constituted the training cohort, and 102 patients from Sir Run Run Shaw Hospital, Zhejiang University School of Medicine formed the external validation cohort. Independent risk factors for ovarian cancer were identified using multivariate binary logistic regression. A predictive model was then constructed based on the weighted regression coefficients of these factors and subsequently validated in an external cohort. In the training cohort, 168 patients (183 lesions) were benign and 54 patients (68 lesions) were malignant. Multivariate analysis identified tumor morphology, diffusion-weighted imaging (DWI) signal intensity, presence of ascites, carbohydrate antigen 125 (CA125), and human epididymis protein 4 (HE4) as independent predictors of malignancy, with AUCs of 0.755, 0.887, 0.749, 0.830, and 0.803, respectively. The combined predictive model achieved an AUC of 0.966, with a sensitivity of 89.6% and specificity of 89.7%. Calibration analysis demonstrated good agreement (Hosmer–Lemeshow test, P = 0.777).In the external validation cohort, 80 patients (87 lesions) were benign and 22 patients (28 lesions) were malignant. The model achieved an AUC of 0.949, with a sensitivity of 96.6% and specificity of 89.3%. The proposed model integrating MRI features and serum tumor markers demonstrated excellent diagnostic performance and robust external validity. This practical and easily applicable tool may assist in preoperative risk stratification and clinical decision-making.