Left Ventricular Thrombus in Ischemic Heart Failure: Machine-learning-based Prediction of Six-month Persistence and One-year Outcomes.

Left ventricular thrombus (LVT) in ischemic heart failure carries embolic risk; tools to anticipate persistence are limited. We studied 190 consecutive patients with imaging-confirmed LVT managed with guideline-concordant anticoagulation and serial echocardiography. The primary outcome was 6-month non-regression; 1-year MACE was secondary. We combined classical statistics with explainable machine learning. CatBoost yielded the best discrimination for non-regression (CV-AUC 0.76; test accuracy 0.79). SHAP highlighted left atrial diameter, pulmonary artery pressure, platelet count, and LV end-diastolic diameter as leading predictors. For 1-year outcomes, thrombus size and CHA2DS2-VA were independently associated with MACE (logistic AUC 0.71), whereas "regression vs persistence" alone was not. Baseline remodeling and coagulability markers, captured by an interpretable ML model, stratify early risk of LVT persistence and complement clinical decision-making for imaging follow-up and anticoagulation intensity.