A retrospective study using machine learning to develop predictive model to identify urinary infection stones in vivo

Area of Science:

• Urology
• Medical Informatics
• Artificial Intelligence

Background:

• Preoperative diagnosis of urinary infection stones is difficult, hindering effective perioperative management and postoperative prevention strategies.
• Accurate stone composition analysis is typically only possible ex vivo, limiting real-time clinical decision-making.

Purpose of the Study:

• To develop and validate a machine learning model for the in vivo preoperative identification of infection stones.
• To provide a tool for improved risk assessment and decision support in managing patients with urinary calculi.

Main Methods:

• Retrospective analysis of clinical data from 1168 eligible patients with urinary calculi.
• Development of five machine learning models (SVM, MLP, DT, RFC, AdaBoost) using 14 preoperative variables.
• Validation of models using the area under the receiver operating characteristic curve (AUC), with AdaBoost selected as the final model.

Main Results:

• The AdaBoost model demonstrated strong discrimination with an AUC of 0.772 (95% CI, 0.657-0.887).
• Key predictors for infection stones identified were UC positivity and urine pH value.
• The model showed good sensitivity (0.522) and high specificity (0.902).

Conclusions:

• A machine learning-based predictive model can effectively identify infection stones in vivo with good predictive performance.
• This tool can optimize disease management for urinary calculi and improve patient outcomes.
• The model aids in risk assessment and decision support for infection stones.