Magnetic resonance elastography of the supraspinatus muscle: A preliminary study on test-retest repeatability and wave quality with different frequencies and image filtering.

The purpose of this study was to determine an optimal condition (vibration frequency and image filtering) for stiffness estimation with high accuracy and stiffness measurement with high repeatability in magnetic resonance elastography (MRE) of the supraspinatus muscle. Nine healthy volunteers underwent two MRE exams separated by at least a 30 min break, on the same day. MRE acquisitions were performed with a gradient-echo type multi-echo MR sequence at 75, 100, and 125 Hz pneumatic vibration. Wave images were processed by a bandpass filter or filter combining bandpass and directional filters (bandpass-directional filter). An observer specified the region of interest (ROI) on clear wave propagation in the supraspinatus muscle, within which the observer measured the stiffness. This study assessed wave image quality according to two indices, as a substitute for the assessment of the accuracy of the stiffness estimation. One is the size of the clear wave propagation area (ROI size used to measure the stiffness) and the other is the qualitative stiffness resolution score in that area. These measurements made by the observer were repeated twice at least one month apart after each MRE exam. This study assessed the intra-examiner and observer repeatability of the stiffness value, ROI size and resolution score in each combination of vibration frequency and image filter. Repeatability of the data was analyzed using the intraclass correlation coefficient (ICC) and 95% limits-of-agreement (LOA) in Bland-Altman analysis. The analyses on intra-examiner and observer repeatability of stiffness indicated that the ICC and 95% LOA were not varied greatly depending on vibration frequency and image filter (intra-examiner repeatability, ICC range, 0.79 to 0.88; 95% LOA range, ±23.95 to ±32.42%, intra-observer repeatability, ICC range, 0.98 to 1.00; 95% LOA range, ±5.10 to ±10.99%). In the analyses on intra-examiner repeatability of ROI size, ICCs were rather low (ranging from: 0.03 to 0.69) while 95% LOA was large in all the combinations of vibration frequency and image filter (ranging from: ±62.66 to ±83.33%). In the analyses on intra-observer repeatability of ROI size, ICCs were sufficiently high in the total combination of vibration frequency and image filter (ranging from 0.80 to 0.87) while the 95% LOAs were better (lower) in the bandpass-directional filter than the bandpass filter (bandpass directional filter vs. bandpass filter, ±28.81 vs. ±54.83% at 75 Hz; ±25.63 vs. ±37.83% at 100 Hz; ±34.51 vs. ±43.36% at 125 Hz). In the analyses on intra-examiner and observer repeatability of resolution score, the mean difference (bias) between the two exams (or observations) was significantly low and there was almost no difference across all the combinations of vibration frequency and image filter (range of bias: -0.11-0.11 and -0.17-0.00, respectively). Additionally, effects of vibration frequency and image filter on wave image quality (ROI size and resolution score) were assessed separately in each exam. Both mean ROI size and resolution score in the bandpass-directional filter were larger than those in the bandpass filter. Among the data in the bandpass-directional filter, mean ROI size was larger at 75 and 100 Hz, and mean resolution score was larger at 100 and 125 Hz. Taking into consideration with the results of repeatability and wave image quality, the present results suggest that optimal vibration frequency and image filter for MRE of the supraspinatus muscles is 100 Hz and bandpass-directional filter, respectively.