AIMS: Both contrast enhanced (CE) two-dimensional echocardiography (2DE) and three-dimensional echocardiography (3DE) have been proposed as techniques to improve the accuracy of left ventricular (LV) volume measurements. We sought to examine the accuracy of non-contrast (NC) and CE-2DE and 3DE for calculation of LV volumes and ejection fraction (EF), relative to cardiac magnetic resonance imaging (MRI). METHODS AND RESULTS: We studied 50 patients (46 men, age 63 +/- 10 year) with past myocardial infarction who underwent echocardiographic assessment of LV volume and function. All patients sequentially underwent NC-2DE followed by NC-3DE. CE-2DE and CE-3DE were acquired during contrast infusion. Resting echocardiographic image quality was evaluated on the basis of NC-2DE. The mean LV end-diastolic volume (LVEDV) of the group by MRI was 207 +/- 79 mL and was underestimated by 2DE (125 +/- 54 mL, P = 0.005), and less by CE-2DE (172 +/- 58 mL, P = 0.02) or 3DE (177 +/- 64 mL, P = 0.08), but EDV was comparable by CE-3DE (196 +/- 69 mL, P = 0.16). Limits of agreement with MRI were similar for NC-3DE and CE-2DE, with the best results for CE-3D. Results were similar for calculation of LVESV. Patients were categorized into groups of EF (< or =35, 35-50, >50%) by MRI. NC-2DE demonstrated a 68% agreement (kappa 0.45, P = 0.001), CE-2DE a 62% agreement (kappa 0.20, P = 136), NC-3DE a 74% agreement (kappa 0.39, P = 0.005) and CE-3DE an 80% agreement (kappa 0.56, P < 0.001). CONCLUSION: CE-2DE is analogous to NC-3DE in accurate categorization of LV function. However, CE-3DE is feasible and superior to other NC- and CE-techniques in patients with previous infarction.