Noninvasive measurement of myocardial blood flow in absolute terms (i.e., milliliters per gram per min) has been difficult to accomplish despite the intrinsically quantitative power of positron emission tomography because of the nonphysiologic nature of tracers that have been employed conventionally as well as the limited spatial resolution of currently available instruments. It was previously demonstrated that myocardial blood flow in animals can be quantitated accurately with the diffusible tracer oxygen-15-labeled water (H2(15)O) when the arterial input function and myocardial radiotracer concentration were measured directly. To extend the approach for completely noninvasive measurement of blood flow, a parameter estimation procedure was developed whereby effects of limited tomographic spatial resolution and cardiac motion were compensated for within the operational flow model. In validation studies in 18 dogs, myocardial blood flow measured with positron emission tomography after intravenously administered H2(15)O correlated closely with flow measured with concomitantly administered radiolabeled microspheres over the range of 0.29 to 5.04 ml/g per min (r = 0.95). Although regional ischemia was clearly identifiable tomographically, absolute flow could not be determined accurately in ischemic regions in four dogs because of poor count statistics related to wall thinning. Subsequently, myocardial blood flow was measured in 11 normal human subjects. Flow was homogeneous throughout the myocardium, averaged 0.90 +/- 0.22 ml/g per min at rest and increased to 3.55 +/- 1.15 ml/g per min after intravenous administration of dipyridamole. Therefore, positron emission tomography with H2 15O and the approach developed permits noninvasive measurement of myocardial blood flow in absolute terms in humans and should facilitate objective assessment of interventions designed to enhance nutritive perfusion.