Predicting complications after laparoscopic surgery for ureteropelvic junction obstruction using machine learning...

Area of Science:

• Urology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Postoperative complications following ureteropelvic junction obstruction (UPJO) surgery can significantly impair outcomes and lead to repeat interventions.
• Predicting these complications is crucial for optimizing patient management and surgical success.

Purpose of the Study:

• To develop and evaluate machine learning algorithms for predicting postoperative complications, specifically urinary tract infections (UTIs) and recurrence, in patients undergoing UPJO surgery.
• To identify key risk factors associated with these adverse events.

Main Methods:

• Retrospective analysis of 526 UPJO patients treated between May 2014 and May 2023.
• Screening of risk factors using multivariate logistic and Lasso regression.
• Development of prediction models using various machine learning algorithms including Logistic Regression (LR), k-nearest neighbours (KNN), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), and Neural Network (NN).

Main Results:

• Identified risk factors for complications included preoperative UTI, calculus, renal cortical thickness, collecting system status, DJ stent removal time, drainage removal, and white blood cell count.
• Logistic Regression demonstrated strong predictive performance for overall complications (AUC=0.929), UTIs (AUC=0.941), and recurrence (AUC=0.894).
• The study is the first to present a machine learning-based predictive model for total complications, UTI, and recurrence after pyeloplasty.

Conclusions:

• Machine learning models can effectively predict postoperative complications, UTIs, and recurrence in UPJO patients prior to surgical intervention.
• The developed Logistic Regression model shows promising accuracy, offering a valuable tool for clinical decision-making.
• External validation and multi-center studies are recommended to enhance the generalizability of the findings.