BACKGROUND: The accuracy, reproducibility, and test-retest reliability of 3-dimensional (3D) echocardiography (3DE) with 3D reconstruction (3DR) and real-time (RT) imaging (RT 3DE) exceed that of 2-dimensional echocardiography (2DE). However, image quality with RT 3DE is inferior to 2DE and we sought to determine whether this justified ongoing use of 3DR. METHODS: Unselected patients (n = 30, 22 men, age 66 +/- 7 years) presenting to the echocardiography laboratory for left ventricular (LV) evaluation were studied with 2DE and RT 3DE; 3DR images were obtained using external localization. The 3D measurements and reconstructions were obtained offline. Magnetic resonance images (MRI) were obtained using true free induction, steady state precession during breath hold and 3D volumes and ejection fraction (EF) were measured using 3D software. A separate cohort of 20 patients (13 men, age 60 +/- 12 years) was measured for test-retest variation. RESULTS: All echocardiographic measures underestimated LV volumes and EF compared with MRI, but this was least with RT 3DE. End-diastolic volume by MRI (168 +/- 54 mL) was underestimated by RT 3DE (-15 +/- 31, P = .02), 3DR (-26 +/- 33, P < .01), and 2DE (-57 +/- 40, P < .01). Similarly, end-systolic volume by MRI (86 +/- 50 mL) was underestimated by RT 3DE (-15 +/- 31, P = .02), 3DR (-26 +/- 33, P < .01), and 2DE (-57 +/- 40, P < .01). However, EF measurements were similar with each method. Test-retest variation was less and interobserver and intraobserver correlations were better with RT 3DE for volumes and EF, compared with 3DR and 2DE. CONCLUSIONS: Despite limitations of image quality, RT 3DE is the most feasible and accurate approach for LV volume and EF measurements and follow-up LV assessment in daily practice.