This study investigates the most appropriate protocol for measuring regional myocardial blood flow (MBF) using 15O-water in clinical applications. METHODS Regional MBF, perfusable tissue fraction (PTF) and arterial blood volume (Va) were measured using 15O-water and dynamic PET on five healthy volunteers based on previously published models. Calculated values were compared for the following three tracer administration protocols: 15O-water bolus injection, 15O-water slow (2 min) infusion and 15O-carbon dioxide slow (2 min) inhalation. For the two slow administration protocols, the three parameters MBF, PTF and Va were computed by fitting the model equations to the myocardial regional time-activity curve. For the bolus injection of 15O-water, only two parameters, MBF and PTF, were fitted by using a fixed Va value obtained by a carbon dioxide blood volume scan. RESULTS All protocols provided consistent MBF values, and the calculated values were homogeneous throughout the whole myocardial segments for all subjects. PTF values were also homogeneous and consistent in the anterior and lateral wall regions, but were significantly greater in the septum (approximately 20%) when the slow 15O-carbon dioxide inhalation protocol was used. MBF and PTF values obtained from the bolus injection protocol showed the smallest intersubject and interregional variations. The simulation study also showed that the magnitude of error was smallest when the bolus injection protocol was employed. CONCLUSIONS The data suggest that the 15O-water bolus injection protocol together with the two-parameter fitting procedure provides the most accurate results for MBF and PTF. However, it requires arterial cannulation and a separate carbon monoxide scan. For clinical studies, however, the 15O-water infusion protocol would be a good alternative, providing MBF and PTF results with an acceptable degree of accuracy and without the need for arterial cannulation.