The diagnostic gain associated with image subtractions was assessed regarding contrast-enhanced 3D magnetic resonance angiography (MRA) image sets of the pelvic and lower extremity arteries. The MRA strategy combined a dedicated vascular coil with a single injection, two-station protocol. Voxel-by-voxel signal intensity subtraction was performed on MRA image sets obtained before and during dynamic infusion of a para-magnetic contrast agent. Non-subtracted and subtracted MRA image sets were assessed for the presence of occlusive (four grades) disease, using DSA as the standard of reference. In addition, SNR and CNR were recorded for each vascular segment on both the non-subtracted and subtracted images. While CNR values of subtracted images exceeded those of non-subtracted images (P < 0.05), there was no difference in diagnostic performance. For the detection of hemodynamically significant disease, non-subtracted and subtracted MRA provided overall sensitivity and specificity of 90.2%/90.3% and 95.1%/95.6%, respectively. Concordance between non-subtracted and subtracted MRA was excellent (Kappa = 0.86).