Cardiovascular magnetic resonance (CMR) with flow velocity mapping has emerged as a noninvasive method to measure flow in saphenous vein coronary artery bypass grafts. The aim of the current study was to retrospectively test two previously described analysis methods on a large CMR data set and to compare their diagnostic accuracy in detecting diseased vein grafts. In 125 vein grafts of 68 patients, volume flow parameters (volume flow, systolic and diastolic peak flow, diastolic-to-systolic flow ratio at rest and during adenosine stress, and flow reserve) were derived from the velocity maps. Method 1 implemented basal flow < 20 ml/min or flow reserve < 2, yielding a sensitivity and specificity of 70% and 38% in the detection of a diseased graft or recipient vessel. Method 2 used receiver operating characteristic (ROC) curve analysis and implemented all significant volume flow parameters in a logistic regression model, yielding a sensitivity of 74% with a specificity of 68% in the detection of a diseased graft or recipient vessel. Evaluating single and sequential grafts separately, this method yielded a sensitivity and specificity of 79% and 87% for single grafts, and 62% and 94% for sequential grafts in the detection of > or = 50% stenosis in grafts or recipient vessels. Cut-off values were formulated for the respective volume flow parameters, which maximally separate grafts with and without > or = 50% stenosis. Using ROC curve analysis with logistic regression the specificity of the analysis method improved considerably. For the current data set the best results were acquired when single and sequential grafts were separately analyzed.