XGBoost Model With CMR Features for Prognostic Assessment in Patients With ST-Segment Elevation Myocardial...

Area of Science:

• Cardiology
• Medical Imaging
• Machine Learning

Background:

• Accurate prognosis for ST-segment elevation myocardial infarction (STEMI) is vital for clinical decision-making.
• Existing models may not fully leverage advanced imaging and machine learning techniques.

Purpose of the Study:

• To develop predictive models for long-term major adverse cardiac and cerebrovascular events (MACCEs) in STEMI patients.
• To integrate demographic, clinical, and cardiac magnetic resonance (CMR) imaging data for enhanced prediction.
• To compare the performance of different machine learning algorithms in forecasting MACCEs.

Main Methods:

• Development of four predictive models (naive Bayes, logistic regression, k-nearest neighbors, XGBoost) using 24 variables.
• Utilized CMR imaging data acquired within 1 week and 1 month post-primary percutaneous coronary intervention.
• Assessed model interpretability using Shapley values.

Main Results:

• The XGBoost model exhibited superior predictive performance for long-term MACCEs.
• Key CMR predictors included microvascular obstruction, left ventricular ejection fraction recovery, and infarct size.
• Clinical factors like Killip class, BMI, and age also significantly influenced predictions.

Conclusions:

• An XGBoost model integrating clinical and CMR data effectively predicts long-term MACCEs in STEMI patients.
• Microvascular obstruction identified via CMR is a critical prognostic factor.