Intercenter differences in diffusion tensor MRI acquisition.
Pagani E., Hirsch JG., Pouwels PJW., Horsfield MA., Perego E., Gass A., Roosendaal SD., Barkhof F., Agosta F., Rovaris M., Caputo D., Giorgio A., Palace J., Marino S., De Stefano N., Ropele S., Fazekas F., Filippi M.
PURPOSE: To assess the effect on diffusion tensor (DT) magnetic resonance imaging (MRI) of acquiring data with different scanners. MATERIALS AND METHODS: Forty-four healthy controls and 36 multiple sclerosis patients with low disability were studied using eight MR scanners with acquisition protocols that were as close to a standard protocol as possible. Between 7 and 13 subjects were studied in each center. Region-of-interest (ROI) and histogram-based analyses of fractional anisotropy (FA), axial (D(ax)), radial (D(rad)), and mean diffusivity (MD) were performed. The influence of variables such as the acquisition center and the control/patient group was determined with an analysis of variance (ANOVA) test. RESULTS: The patient/control group explained approximately 25% of data variability of FA and D(rad) from midsagittal corpus callosum (CC) ROIs. Global FA, MD, and D(rad) in the white matter differentiated patients from controls, but with lower discriminatory power than for the CC. In the gray matter, MD discriminated patients from controls (30% of variability explained by group vs. 17% by center). CONCLUSION: Significant variability of DT-MRI data can be attributed to the acquisition center, even when a standardized protocol is used. The use of appropriate segmentation methods and statistical models allows DT-derived metrics to differentiate patients from healthy controls.