A Predictive Model Based on Machine Learning Algorithm for Vein Thrombosis After Ovarian Cancer Resection

Abstract

Background: Postoperative venous thromboembolism (VTE) is the most fatal complication of ovarian cancer and adversely affects prognosis. This study aimed to develop and validate predictive models for VTE risk following ovarian cancer resection using machine learning (ML) techniques and incorporating perioperative clinical and surgical variables. Methods: Retrospective data were collected from 931 patients with ovarian cancer who underwent resection between March 2018 and April 2024 at two tertiary hospitals. The Least Absolute Shrinkage and Selection Operator (LASSO) regression was employed to identify critical predictors of VTE and seven ML models, including Logistic Regression (LR), Decision Tree (DT), Extreme Gradient Boosting Machine (XGBoost), Random Forest (RF), Support Vector Machine (SVM), Naïve Bayes (NB), and Light Gradient Boosting Machine (LGBM) were trained and optimized. Optimal hyperparameters were selected based on a 10-fold cross-validation. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), precision-recall area under the curve (PR-AUC), balanced accuracy, precision, recall, F1 score, and Brier score. The SHapley Additive exPlanation (SHAP) package was used to interpret the optimal models. Results: The incidence of postoperative VTE was 10.0% (93/931). Among the models, XGBoost demonstrated superior performance, achieving an AUC of 0.935 (95% CI: 0.902–0.963) and PR-AUC of 0.620 (95% CI: 0.457–0.809), recall of 0.849, F1 score of 0.571, and Brier score of 0.116. SHAP analysis identified residual disease, surgical duration, postoperative D-dimer levels, postoperative chemotherapy, and age as the top five contributors to postoperative VTE risk. Conclusion: The ML-based model, particularly the XGBoost algorithm, effectively predicted the VTE risk in patients with post-resection ovarian cancer. This tool may assist clinicians in early identification of high-risk individuals, thereby enabling personalized thromboprophylaxis and optimizing perioperative management to mitigate VTE-related morbidities.