Machine learning algorithm predicts urethral stricture following transurethral prostate resection

Area of Science:

• Urology
• Medical Informatics
• Machine Learning

Background:

• Post-transurethral prostate resection (TURP) urethral stricture is a complication.
• Predictive models for this complication are crucial for patient management.

Purpose of the Study:

• To develop and evaluate machine learning algorithms for predicting post-TURP urethral stricture.
• To utilize preoperative blood parameters as predictive features.

Main Methods:

• Retrospective analysis of 109 patients undergoing bipolar-TURP.
• Development of machine learning models using preoperative blood tests and patient characteristics.
• Performance assessment using metrics like accuracy, F1 score, and ROC AUC.

Main Results:

• Statistically significant preoperative differences in Platelet Distribution Width, Mean Platelet Volume, Plateletcrit, Activated Partial Thromboplastin Time, and Prothrombin Time were observed.
• Random forests model achieved the highest prediction accuracy (0.91).
• Other models showed varying accuracies: Support Vector Machines (0.86), Decision Trees (0.82), Logistic Regression (0.82), K-Nearest Neighbors (0.82), and Naïve Bayes (0.77).

Conclusions:

• Machine learning models demonstrate high accuracy in predicting post-TURP urethral stricture.
• Preoperative blood parameters are valuable predictors.
• Future research incorporating additional variables may further improve predictive precision.