Identification of High-Risk Patients for Postoperative Myocardial Injury After CME Using Machine Learning: A 10-Year...

Area of Science:

• Cardiovascular Surgery
• Oncology
• Machine Learning in Medicine

Background:

• Myocardial injury is a serious complication following complete mesocolic excision (CME).
• Predicting and preventing myocardial injury is crucial for patient prognosis after CME.

Purpose of the Study:

• To develop a machine learning model for predicting myocardial injury post-CME.
• Identify preoperative, intraoperative, and postoperative risk factors for myocardial injury.

Main Methods:

• Utilized extreme gradient boosting (XGBoost), random forest, multilayer perceptron, and k-nearest neighbor algorithms.
• Evaluated models using k-fold cross-validation, ROC curves, calibration curves, decision curve analysis (DCA), and external validation.
• Included 1198 colon cancer patients, with 133 experiencing myocardial injury.

Main Results:

• XGBoost demonstrated superior performance with an AUC of 0.997 (training) and 0.956 (validation).
• The XGBoost model showed high predictive accuracy, stability, and good extrapolative capacity (AUC 0.74 in external validation).
• DCA indicated higher benefit rates for intervention guided by the XGBoost model.

Conclusions:

• The XGBoost-based prediction model for myocardial injury after CME possesses high accuracy and clinical utility.
• This model can aid in identifying high-risk patients and guiding clinical decisions.