This is a prospective study comparing selective angiography and magnetic resonance (MR) imaging in the evaluation of the patency of coronary artery bypass grafts (CABGs). Twenty-eight patients with 52 grafts were studied (10 internal mammary artery grafts, 19 saphenous vein grafts including 2 sequential grafts, and 15 right coronary artery saphenous vein grafts). The mean interval between MR imaging and the surgical procedure was 13.2 months. Results obtained with angiography were as follows: 39 patent grafts; 4 patent but stenotic (greater than 50%); 9 occluded CABGs. Magnetic resonance imaging was performed with axial imaging, spin echo images, and electrocardiographic gating. Grafts that were normal on angiographic examination appeared without a signal on both the first and the second echoes in 38 of 39 cases. Grafts that were patent but presented significant stenosis as evidenced by angiography presented the same appearance. In one case, there was an aspect corresponding to a decreased blood flow (presence of a signal on the first echo that became stronger on the second echo). The CABG occlusion was determined in seven of nine cases. Occluded grafts twice presented a signal variation corresponding to an old thrombus formation. In two cases, only the origin of the graft was visible without any visualization of its distal portion. In three other cases, neither the proximal nor the distal portions of the graft were identified. Among six CABGs that could not be identified, three were patent. This study demonstrated that MR imaging makes it possible to correctly identify patent CABGs, but this modality has significant limitations because patent but stenotic CABGs may present in the same way as patent CABGs without stenosis. On the other hand, the correct identification of occluded grafts may be accurately performed using MR imaging.