The objective was to validate the measurement of myocardial perfusion in humans by ultrafast computed tomography (CT), by comparing measurements with those from single photon emission computed tomography (SPECT). Measurement of myocardial perfusion with high spatial resolution (including the differentiation of subendocardial and subepicardial perfusion) may be possible by ultrafast CT in humans. Although there are encouraging data from experiments with dogs, the technique has not been validated in humans. In 11 patients, ultrafast CT measurement of regional perfusion in a single short-axis slice was compared with that obtained by SPECT, and in 14, reproducibility of ultrafast CT was evaluated. The ultrafast CT scanner was set to acquire 20 images, gated to end-diastole. The images were divided into 32 equal segments, and the time course and extent of opacification of the left ventricular cavity and myocardium were analyzed to calculate absolute perfusion. The thallium tomograms were also divided into 32 segments for comparison. The pattern of relative perfusion by segment was represented as a curve. In 18 of 22 paired scans, the mean difference of the position of the minimum and maximum points of the curves was less than 4, indicating close agreement between the two techniques. When scoring segmental perfusion as normal or abnormal, there was agreement between the methods in 129 of 176 segments ([symbol: see text] 0.41). Reproducibility (mean difference +/- 1 SD) of basal scans was 0.005 +/- 0.2 mL/min/mL, and during adenosine vasodilation was 0.05 +/- 0.32 mL/min/mL. Absolute perfusion (mean +/- 1 SD) at rest was 0.52 +/- 0.21 mL/min/mL. During adenosine infusion, perfusion increased to a mean of 0.84 +/- 0.42 mL/min/mL. Ultrafast CT and intravenous contrast can be used to assess relative myocardial perfusion in humans, at rest and during adenosine vasodilatation, although it may underestimate absolute perfusion, particularly at high flow.