BACKGROUND: Perfusion imaging techniques intended to identify regional limitations in coronary flow reserve in viable myocardium need to identify 2-fold differences in regional flow during coronary vasodilation consistently. This study evaluated the suitability of current first-pass magnetic resonance approaches for evaluating such differences, which are 1 to 2 orders of magnitude less than in myocardial infarction. METHODS AND RESULTS: Graded regional differences in vasodilated flow were produced in chronically instrumented dogs with either left circumflex (LCx) infusion of adenosine or partial LCx occlusion during global coronary vasodilation. First-pass myocardial signal intensity-time curves were obtained after right atrial injection of gadoteridol (0.025 mmol/kg) with an MRI inversion recovery true-FISP sequence. The area under the initial portion of the LCx curve was compared with that of a curve from a remote area of the ventricle. Relative LCx and remote flows were assessed simultaneously with microspheres. The ratio of LCx and remote MRI curve areas and the ratio of LCx and remote microsphere concentrations were highly correlated and linearly related over a 5-fold range of flow differences (y=0.96 x+/-0.07, P<0.0001, r(2)=0.87). The 95% confidence limits for individual MRI measurements were +/-35%. Regional differences of >/=2-fold were consistently apparent in unprocessed MR images. CONCLUSIONS: Clinically relevant regional reductions in vasodilated flow in viable myocardium can be detected with 95% confidence over the range of 1 to 5 times resting flow. This suggests that MRI can identify and quantify limitations in perfusion reserve that are expected to be produced by stenoses of >/=70%.