Abstract
Early prediction of microvascular obstruction (MVO) occurrence in acute myocardial infarction (AMI) patients undergoing percutaneous coronary intervention (PCI) can facilitate personalized management and improve prognosis. This study developed a prediction model for MVO occurrence using preoperative clinical data and validated its performance in a prospective cohort. A total of 504 AMI patients were included, with 406 in the exploratory cohort and 98 in the prospective cohort. Feature selection was performed using random forest recursive feature elimination (RF-RFE), identifying five key predictors: High-Sensitivity Troponin T, Neutrophil Count, Creatine Kinase-MB, Fibrinogen, and Left Ventricular Ejection Fraction. Among the models developed, logistic regression demonstrated the highest predictive performance, achieving an AUC score of 0.800 in the exploratory cohort and 0.792 in the prospective cohort. This model has been integrated into a user-friendly online platform, providing a practical tool for guiding personalized perioperative management and improving patient prognosis.