Velocity-encoded, phase-difference magnetic resonance imaging (MRI) previously has been used to measure flow in the aorta, as well as in the pulmonary, carotid, and renal arteries, but these measurements have not been validated against currently accepted invasive techniques. To determine the accuracy of velocity-encoded, phase-difference MRI measurements of cardiac output, 23 subjects (11 men and 12 women, aged 15 to 72 years) underwent velocity-encoded, phase-difference MRI measurements of cardiac output in the proximal aorta, followed immediately by cardiac catheterization, with measurement of cardiac output by the Fick principle and by thermodilution. For MRI, Fick, and thermodilution measurements, stroke volume was calculated by dividing cardiac output by heart rate. The magnetic resonance images were acquired in 1 to 3 minutes. For all patients, the agreement between measurements of stroke volume was 3 +/- 9 ml for MRI and Fick, -3 +/- 11 ml for MRI and thermodilution, and 0 +/- 8 ml for MRI and the average of Fick and thermodilution. Compared with standard invasive measurements, velocity-encoded, phase-difference MRI can accurately and rapidly determine cardiac output.