Cross-Platform Comparison of Brain MRE

Introduction Magnetic resonance elastography (MRE) is a technique for noninvasively measuring tissue stiffness [1]. MRE begins with the introduction of shear waves via an external vibration source. The resulting shear waves are imaged with a phase-contrast MRI pulse sequence incorporating motion-encoding gradients synchronized to the external motion. Finally, the shear wave images are mathematically inverted to calculate tissue stiffness. MRE is currently under investigation for its potential to aid in the diagnosis of brain diseases [2-8]. A well-developed MRE exam of the brain should be reproducible even when performed on different MR imaging platforms. The purpose of this work was to compare the reproducibility of MRE of the brain on two MR platforms manufactured by GE and Siemens. Methods MRE was performed on six male volunteers without known neurological disease after obtaining informed written consent. Coronal images were collected on a 1.5T GE Excite scanner with the following parameters: 60-Hz vibration, FOV=25.6 cm, TR/TE=83.3/20.3 ms, 2x SENSE acceleration, 30° flip angle, eight 4mm slices, 64x64 imaging matrix reconstructed to 256x256, throughplane motion encoding with a single 16.7-ms gradient with amplitude 2.3 G/cm, and 4 phase offsets over one period of motion. The images collected on the Siemens scanner (1.5T Espree) were the same except: TE=23.2 ms, 2x GRAPPA acceleration, and a 256x64 imaging matrix reconstructed to 256x256 and a motion encoding gradient amplitude of 1.6 G/cm. Each scanner was equipped with an active acoustic driver system located outside of the scan room as described in [9] and the passive component was a thin flexible driver [10] 15x21x2 cm placed under the head. The wave images were filtered with eight 2D directional filters with a radial Butterworth bandpass filter with cutoffs of 8 and 40 waves/FOV [11]. The first harmonic of the filtered images were inverted with a 3D direct inversion algorithm in an 11x11x3 window [12]. The median stiffness for each MRE exam was calculated over a region of interest that included the two central slices of the collected volume excluding 12 voxels from the edge of the brain. Regression analysis was performed with a Spearman correlation and the two groups were compared with a paired-sample T-test. Results Example images from both platforms for one volunteer are shown in Figure 1. A Pearson’s correlation yielded an R-value of 0.92 and a pvalue of 0.01 (Fig.2). The two-way T-test demonstrated no significant difference between the stiffness measured on the two platforms (p=0.63). The median and maximum percent differences in the group of six were 2.19% and 9.40%, respectively. Discussions The results indicate that the stiffness measurements on the two platforms were significantly correlated, and the mean stiffness of the two groups did not differ. The percent differences between the two platforms for the six volunteers were comparable to typical repeatability of MRE of the brain [3]. References [1] Muthupillai et al. Science 1995. 269 (5232): 1854. [2] Kruse et al. Neuroimage 2008. 39 (1): 231. [3] Sack et al. NMR Biomed 2008. 21: 265. [4] McCracken et al. Magn Reson Med 2005. 53 (3): 628. [5] Green et al. NMR Biomed 2008. 21: 755. [6] Xu et al. Acta Radiol 2007. 48 (1): 112. [7] Wuerfel et al. Neuroimage 2010. 49 (3): 2520. [8] Streitberger et al. NMR Biomed 2010. In press. [9] Yin et al. Clin Gastroenterol Hepatol 2007. 5(10):1207-1213.e2. [10] Chen et al., ISMRM 2010:1052. [11] Manduca et al. Med Image Anal 2003. 7: 465. [12] Oliphant et al. Magn Reson Med 2001. 45: 299. Figure 1. Example images. Images in the top row were collected on the GE system while the images in the bottom row were collected on the Siemens system. The first column is the magnitude image from the MRE data. The second column shows example wave images. The last column is the resulting elastogram.