PURPOSE: To determine whether interpretation of internal carotid arter
y (ICA) stenosis from source partitions is more accurate than interpre
tation from maximum-intensity projections (MIPs) from three-dimensiona
l (3D) time-of-flight (TOF) magnetic resonance (MR) angiography. MATER
IALS AND METHODS: The percentage of diameter ICA stenosis was measured
on source images and MIPs from sagittal (n = 150) and transverse (n =
140) 3D TOF MR angiography. Measurements were compared with those fro
m conventional angiography. RESULTS: Sensitivity and specificity for d
istinguishing 70%-99% stenosis were 96% and 78%, respectively, for sag
ittal MIPs, 88% and 90% for sagittal source images, 92% and 86% for tr
ansverse MIPs, and 92% and 95% for transverse source images. Areas und
er the receiver operating characteristic curves statistically signific
antly increased (P < .05) with interpretation from source images. Comp
lete loss of intravascular signal was not encountered on source partit
ions except within a greater than 85% stenosis. CONCLUSION: Interpreta
tion of source partitions rather than MIPs reduces the tendency for ov
erestimation of stenosis with MR angiography and improves the specific
ity for discriminating 70%-99% stenosis.