Abstract
BACKGROUND AND PURPOSE: Various magnetic resonance (MR) measurements have been proposed to aid in differentiating between progressive supranuclear palsy (PSP) and idiopathic Parkinson's disease (IPD); however, these methods have not been compared directly. The aim of this study was to determine which measurement method exhibits the highest power to differentiate between PSP and IPD. METHODS: Brain MR images from 82 IPD and 29 PSP patients were analyzed retrospectively. T1-weighted 3D volumetric axial images, or sagittal images reconstructed from those axial images were examined. MR measurements included the length from the interpeduncular fossa to the center of the cerebral aqueduct at the mid-mammillary-body level, adjusted according to the anterior commissure-posterior commissure length (MB(Tegm)), the ratio of the midbrain area to the pons area (M/P ratio) as measured by both Oba's method (Oba M/P) and Cosottini's method (Cosottini M/P), and a modified MR parkinsonism index (mMRPI). RESULTS: Receiver operating characteristic (ROC) analysis indicated that the areas under the ROC curves (AUCs) exceeded 0.70, with a high intrarater reliability for all MR measurement methods. ROC analyses of four MR measurements yielded AUCs of 0.69-0.76. At the cutoff value with the highest Youden index, mMRPI had the highest sensitivity, while Oba M/P offered the highest specificity. A comparison of the ROC analyses revealed that MB(Tegm) was superior to mMRPI in differentiating PSP from IPD (p=0.049). There was no difference in discriminating power among Oba M/P, Cosottini M/P, and MB(Tegm). CONCLUSIONS: Simple measurements of MB(Tegm) on axial MR images at the mid-mammillary-body level are comparable to measurements of the M/P ratio with regard to their ability to discriminate PSP from IPD.