Applying interpretable machine learning algorithms to predict risk factors for permanent stoma in patients after TME

Area of Science:

• Oncology
• Surgical Oncology
• Machine Learning in Medicine

Background:

• Total mesorectal excision (TME) is a standard surgical procedure for rectal cancer.
• Permanent stoma formation is a significant complication following TME, impacting patient quality of life.
• Predicting stoma risk preoperatively and intraoperatively is crucial for surgical planning and patient counseling.

Purpose of the Study:

• To develop and validate a machine learning model for predicting permanent stoma occurrence after TME.
• To identify key preoperative and intraoperative risk factors associated with permanent stoma.

Main Methods:

• A retrospective study included 1,163 rectal cancer patients undergoing TME.
• Data on 24 variables (demographics, medical history, surgical details) were collected.
• Four machine learning algorithms (XGBoost, RF, SVM, KNN) were employed and evaluated using cross-validation, ROC curves, calibration curves, and decision curve analysis (DCA).

Main Results:

• The XGBoost model demonstrated superior performance with high predictive accuracy (AUC 0.987 training, 0.963 validation, 0.89 external validation).
• Internal validation confirmed the XGBoost model's stability and high predictive power.
• DCA indicated potential benefits for patients managed with interventional treatment guided by the XGBoost model.

Conclusions:

• The XGBoost-derived prediction model for permanent stoma after TME in rectal cancer patients exhibits high accuracy and clinical utility.
• This model can aid clinicians in risk stratification and personalized treatment strategies.