Is It Too Late? Machine Learning for Predicting Orchiectomy Versus Testicular Preservation in Acute Torsion

Area of Science:

• Pediatric Urology
• Machine Learning in Medicine
• Surgical Outcomes

Background:

• Testicular torsion is a pediatric emergency where delayed treatment increases orchiectomy risk.
• Accurate preoperative prediction of surgical outcomes remains challenging.

Purpose of the Study:

• Develop and validate machine learning models to predict orchiectomy versus testicular preservation in pediatric and adolescent testicular torsion.
• Integrate ischemic timing variables and complete blood count-derived inflammatory indices.

Main Methods:

• Retrospective analysis of 165 pediatric and adolescent patients with confirmed testicular torsion.
• Evaluation of clinical timing variables and CBC-derived inflammatory indices (e.g., neutrophil-to-lymphocyte ratio).
• Assessment of five machine learning algorithms (logistic regression, random forest, XGBoost, LightGBM, SVM) using nested cross-validation.

Main Results:

• Orchiectomy rate was 22.4%; preservation rate was 77.6%.
• Significantly longer symptom duration (48h vs 6h) and higher monocyte count were associated with orchiectomy (p < 0.001).
• XGBoost model achieved the highest discrimination (ROC-AUC 0.926), with symptom duration as the dominant predictor.

Conclusions:

• Prolonged symptom duration is the primary predictor of intraoperative orchiectomy in pediatric testicular torsion.
• Complete blood count-derived inflammatory indices offer complementary risk information.
• The internally validated XGBoost model demonstrates strong predictive performance and clinical interpretability, warranting external validation.