Application of Machine Learning Algorithms in Predicting Major Adverse Cardiovascular Events after Percutaneous...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Developing predictive models for major adverse cardiovascular events (MACE) is crucial for patients undergoing percutaneous coronary intervention (PCI).
• New-onset ST-segment elevation myocardial infarction (STEMI) patients represent a high-risk population requiring precise risk stratification post-PCI.
• Machine learning (ML) algorithms offer potential for enhancing the accuracy of cardiovascular event prediction.

Purpose of the Study:

• To develop and validate a predictive model for MACE following PCI in new-onset STEMI patients.
• To compare the performance of four distinct ML algorithms in predicting MACE risk.
• To identify key predictors of MACE and visualize risk using a clinical tool.

Main Methods:

• Retrospective data analysis of 250 new-onset STEMI patients undergoing PCI.
• Application of four ML algorithms: K-nearest neighbors (KNN), support vector machine (SVM), Complement Naive Bayes (CNB), and logistic regression.
• Feature selection using Boruta algorithm, model evaluation via AUC, sensitivity, specificity, and risk visualization with a nomogram based on SHAP analysis.

Main Results:

• Logistic regression demonstrated superior performance with an AUC of 0.814 (training) and 0.776 (validation).
• Seven key predictors for MACE were identified: Killip classification, Gensini score, blood urea nitrogen (BUN), heart rate (HR), creatinine (CR), glutamine transferase (GLT), and platelet count (PCT).
• The constructed nomogram provided accurate risk predictions, showing strong agreement between predicted and observed outcomes.

Conclusions:

• The developed logistic regression model is effective for predicting MACE risk in STEMI patients post-PCI.
• The nomogram serves as a valuable and practical tool for clinicians, facilitating personalized risk assessment.
• Improved clinical decision-making and patient management can be achieved through this predictive tool.