Optimization of Management Plan with a machine learning model for Ovarian Torsion Cases: Operative versus Conservative

Background: Ovarian torsion (OT) is a gynecologic emergency requiring prompt and accurate management to preserve ovarian function and fertility. Determining the need for operative intervention versus conservative management remains challenging due to overlapping clinical and imaging features. This study aimed to develop and validate a machine learning (ML)–based prognostic model to assist clinical decision-making in suspected OT. Methods: A retrospective analysis was conducted on 219 females (1 month–75 years) presenting with suspected OT at a tertiary center between 2022 and 2024. Clinical, demographic, laboratory, and imaging variables were analyzed. Predictors of management type (operative vs. conservative) were identified using comparative statistics and Spearman correlation. Principal component and cluster analyses were used for dimensionality reduction and patient stratification. Supervised ML algorithms—including Decision Tree, Random Forest, Neural Network, Gradient Boosting, and Logistic Regression—were trained with oversampling techniques (ROSE, SMOTE, SMOTE-ENN) to address class imbalance. Model performance was assessed using the area under the ROC curve (AUC), sensitivity, and specificity. Results: Operative management was required in 83.6% of cases, while 16.4% were managed conservatively. Significant predictors of conservative management included prior oral contraceptive use (protective; r = −0.199, p = 0.003), absence of a pelvic mass (r = 0.134, p = 0.048), and lower symptom burden (r = 0.148, p = 0.029). PCA identified eight clinical domains, and clustering revealed two distinct phenotypic subgroups. The weighted Decision Tree achieved the best-balanced performance (AUC = 0.76), while Random Forest with ROSE oversampling achieved perfect performance (AUC = 1.00), indicating potential overfitting. Conclusions: Machine learning models can enhance clinical decision-making by stratifying OT patients suitable for conservative management and identifying those requiring urgent surgery. Integration of this model into clinical decision support systems may reduce unnecessary surgeries, optimize fertility-preserving care, and improve