We sought to determine which contrast-enhanced magnetic resonance imaging (CE-MRI) parameter is the best predictor for left ventricular (LV) remodeling and clinical outcomes after ST-segment elevation myocardial infarction (STEMI). In 135 patients undergoing primary percutaneous coronary intervention (PCI) for STEMI, CE-MRI was performed at a median of 7 days after PCI. Echocardiography was performed soon after PCI and at a follow-up visit. LV remodeling was defined as an increase in end-diastolic volume index ≥20 % on follow-up echocardiography. Several CE-MRI parameters such as infarct size, transmurality, microvascular obstruction (MVO), and hemorrhagic infarction were tested using a 17-myocardial segment model. Optimal cut-off values were derived from receiver-operating characteristic curve (ROC) analysis. Twenty-eight patients (21 %) demonstrated LV remodeling. Although the addition of transmural necrotic segment count, infarct size, and MVO segment count to clinical models improved the prediction of LV remodeling in multivariable regression analysis, transmural necrotic segment count had better incremental predictive value than other CE-MRI parameters. The aggregate consideration of infarct size (cut-off ≥25 %), transmural necrotic segment count (≥5), and MVO segment count (≥2) yielded better diagnostic performance than each of the individual parameters in ROC analysis (P < 0.01). In Kaplan-Meier curve analysis, patients with transmural necrotic segment counts ≥5 had a higher incidence of major adverse cardiac event than did those without. The transmural necrotic segment count is the most important predictor of LV remodeling and clinical outcomes. The combination of CE-MRI parameters including infarct size, transmural necrotic segment count, and MVO segment count appeared to increase reliability for predicting LV remodeling.