Predicting left ventricular remodeling post-MI through coronary physiological measurements based on computational fluid dynamics

Early detection of left ventricular remodeling (LVR) is crucial. While cardiac magnetic resonance (CMR) provides valuable information, it has limitations. Coronary angiography-derived fractional flow reserve (caFFR) and index of microcirculatory resistance (caIMR) offer viable alternatives. 157 patients with ST-segment elevation myocardial infarction (STEMI) undergoing primary percutaneous coronary intervention were prospectively included. 23.6% of patients showed LVR. Machine learning algorithms constructed three LVR prediction models: Model 1 incorporated clinical and procedural parameters, Model 2 added CMR parameters, and Model 3 included echocardiographic and functional parameters (caFFR and caIMR) with Model 1. The random forest algorithm showed robust performance, achieving AUC of 0.77, 0.84, and 0.85 for Models 1, 2, and 3. SHAP analysis identified top features in Model 2 (infarct size, microvascular obstruction, admission hemoglobin) and Model 3 (current smoking, caFFR, admission hemoglobin). Findings indicate coronary physiology and echocardiographic parameters effectively predict LVR in patients with STEMI, suggesting their potential to replace CMR. [Display omitted] •Larger infarcts, MVO, and lower caFFR predict a higher likelihood of LVR•Machine learning enhanced LVR prediction with CMR or coronary physiology parameters•Coronary physiology offers a more convenient approach than CMR for predicting LVR•Machine learning identified several key risk factors for LVR Surgery; Cardiovascular medicine; Bioinformatics