To assess the feasibility, safety and usefulness of gradient-echo magnetic resonance imaging (MRI) combined with pharmacologic stress testing for the detection of coronary artery disease, 23 patients without previous myocardial infarction but with significant stenosis (greater than 70% diameter stenosis) of greater than or equal to 1 major coronary artery were selected for dipyridamole-MRI stress testing. Each patient underwent MRI at rest, and high-dose dipyridamole-MRI (0.75 mg/kg over 10 minutes) of corresponding basal and midventricular short-axis tomograms. Additionally, these patients performed symptom-limited exercise stress tests. All short-axis tomograms were evaluated on a standardized segmental basis by grading each segment as normal, hypokinetic, akinetic or dyskinetic. Dipyridamole-MRI was considered pathologic if segmental wall motion deteriorated by greater than or equal to 1 grade after dipyridamole. For comparison with coronary angiography, segmental wall motion gradings were related to the respective coronary artery territories in the short-axis plane. Pathologic dipyridamole-MRI was obtained in 18 of 23 (78%) patients. For 1- and 2-vessel diseases, sensitivity was 69 and 90%, respectively. Exercise stress tests were pathologic in 14 of 23 (66%) patients. For 1- and 2-vessel diseases, sensitivity of exercise stress test was 58% (7 of 12 patients) and 77% (7 of 9), respectively. Sensitivity/specificity of dipyridamole-MRI for the localization of the stenosed coronary artery was 78/100% for left anterior descending, 73/100% for left circumflex, and 88/87% for right coronary artery stenoses. It is concluded that dipyridamole-MRI is a feasible nonexercise-dependent test for detection and localization of functionally significant coronary artery disease.