Machine Learning Predicts LVEF Recovery After PCI
Researchers retrospectively analyzed 520 chronic coronary syndrome patients from the CLIDAS database to develop and compare 48 machine-learning models predicting left ventricular ejection fraction (LVEF) recovery within 6–12 months after percutaneous coronary intervention (PCI). Top models (RFE+XGBoost, LASSO+XGBoost) reached AUCs of 0.93, 0.79, 0.88, and 0.84 across subgroups; SHAP highlighted biomarkers and ECG predictors including uric acid, BNP, HbA1c, creatinine, baseline LVEF, LVEDD, heart rate, and V5/V6 R-waves. These interpretable models may support post-PCI risk stratification and follow-up.
Key Points
- 1Developed 48 ML models on 520 CCS patients to predict LVEF recovery after PCI
- 2Achieved strong discrimination (AUCs up to 0.93) stratified by baseline LVEF subgroups
- 3Identified interpretable biomarkers and ECG features to guide risk stratification and follow-up decisions
Scoring Rationale
Novel clinical ML application with strong internal validation; limited external validation and clinical deployment evidence.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
