A novel technique for automating stiffness measurement and emphasizing the main wave: Coherent-wave auto-selection (CHASE)

This study aims to develop and assess a new automated processing technique in MR elastography (MRE), namely coherent-wave auto-selection (CHASE). CHASE enables automatic selection of the region of interest (ROI) for stiffness measurement by extraction of the coherent wave region (CHASE ROI), and it improves the reconstruction of stiffness by a directional filter oriented along the main wave in each pixel (CHASE filtering). In this study, MRE of a phantom and of the liver of four healthy volunteers was performed. To investigate the potential of CHASE, this study assessed the CHASE according to three indices through the phantom study: 1) agreement on the ROI settings between CHASE and expert observers, 2) noise dependency, and 3) effect of the CHASE on stiffness variability within the CHASE ROI. The agreements on the ROI settings were analyzed by Cohen's kappa coefficient (κ). The noise dependency was analyzed by the mean absolute percentage errors (MAPEs) within the ROI between low (20%–80% amplitudes) and high vibration amplitudes (100% amplitude). The stiffness variability was assessed by standard deviation (SD) within the ROI. In the volunteer study, agreements on the ROI settings (or stiffness value) and stiffness variability within the CHASE ROI were assessed using κ-value (or intraclass correlation coefficient: ICC) and coefficient of variation, respectively. The results showed close agreement on the ROI settings and stiffness (κ-value: greater than 0.61 in both the phantom and volunteer studies, ICC: 0.97 in the volunteer study). The MAPEs within the CHASE ROI were much smaller than those in the whole region of the phantom (CHASE ROI vs. the whole region at 20% amplitude: 10.3% vs. 50.8%). Moreover, in both the phantom and volunteer studies, the stiffness variation within the CHASE ROI was smaller in the elastogram processed with CHASE filtering than in the unprocessed one. Our results demonstrated that the CHASE has high robustness against noise and the potential to provide ROI settings for stiffness measurement comparable to expert observers, as well as improve the reconstruction of stiffness.