Testicular salvage: using machine learning algorithm to develop a predictive model in testicular torsion.

Area of Science:

• Urology
• Medical Informatics
• Surgical Prediction Modeling

Background:

• Testicular torsion is a surgical emergency requiring prompt diagnosis.
• Predicting the need for orchiectomy (testicular removal) is crucial for patient management.
• Classical statistical methods have limitations in complex predictive tasks.

Purpose of the Study:

• To compare the predictive accuracy of a machine learning (ML) model versus a classical statistical model (Cox Regression) for orchiectomy in testicular torsion.
• To identify key preoperative parameters influencing orchiectomy decisions.

Main Methods:

• Retrospective review of patients with testicular torsion (2000-2020).
• Data collection included demographics, clinical findings, and admission details.
• Development of prediction models using Cox Regression and Random Forest (ML).

Main Results:

• Orchiectomy was performed in 28.3% of cases.
• Monocyte count, symptom duration, and prior Doppler ultrasound were significant predictors.
• Random Forest model achieved higher accuracy (AUC 0.95) than Cox Regression (AUC 0.937).

Conclusions:

• Machine learning models demonstrate superior performance in predicting orchiectomy for testicular torsion.
• ML offers a cost-effective and increasingly powerful tool for clinical application in urological emergencies.