To gauge the accuracy of ultrafast CT in measuring cardiac output and myocardial perfusion in humans, measurements of continuous and pulsatile flow were made in a large asymmetrical phantom. The variation in the relationship between Hounsfield number and contrast concentration was assessed in a human thorax phantom. Radiopaque contrast medium was injected during perfusion of the phantom at a range of flow rates between 1.5 and 8 L/min. The phantom was scanned in two modes (50 and 100 ms) during continuous and pulsatile flow and with the phantom surrounded by air and by water. Flow in the tubes was calculated using indicator dilution theory, and flow in the tissue-equivalent chamber was calculated by applying first-pass distribution principles. The standard deviation of the difference between calculated and measured flow varied from 0.2 to 0.6 L/min, giving 95% limits of agreement from 0.4 to 1.2 L/min. The constant (K) relating Hounsfield unit number to iodine concentration varied widely both in different locations within the phantom and under different scan conditions (17.2-27.6 HU/mg I). Within a human thorax phantom, K varied from 14.15 to 23.18 HU/mg I and was dependent on location within the thorax phantom, the scan mode, and the cross-sectional diameter of the phantom. These data suggest that though the ultrafast CT scanner can measure continuous and pulsatile flow accurately in tubes, precise measurements of cardiac output in humans will require K to be assessed for each subject. Measurements of flow in tissue should be possible.