The utility of Patlak graphical analysis was investigated for quantification of regional myocardial blood flow (MBF) and for generating parametric images of MBF with 13N-ammonia and dynamic PET imaging in dogs and humans. MBF was estimated by a two-compartment model fit of the initial 2 min of the kinetic data and by Patlak graphical analysis of the initial 2, 3, or 4 min of data. In 11 dog studies, MBF by compartmental model fitting linearly correlated with MBF by microspheres (correlation coefficient (r) = 0.99, slope = 0.92) and by Patlak graphical analysis (r = 0.99, slope = 0.90). In 10 normal human studies, MBF obtained by the Patlak graphical analysis agreed well with MBF obtained by the compartmental model fitting (r = 0.96, slope = 1.04). Good agreement of the MBF estimates was also observed in 10 coronary artery disease patient studies (r = 0.96). Patlak graphical analysis permitted generation of parametric images of MBF. The parametric images of MBF, in units of ml/min/g, are of good image quality and have relatively low noise levels. We conclude that regional MBF can be noninvasively and conveniently measured with dynamic 13N-ammonia PET using either a two-compartment model or Patlak graphical analysis. MBF parametric images generated with the Patlak graphical analysis both map the distribution and quantitate the magnitude of myocardial perfusion abnormalities.