Comparative Analysis of a Locally Resampling MR Elastography Reconstruction Algorithm in Liver Fibrosis

Area of Science:

• Medical Imaging
• Biophysics
• Quantitative MRI

Background:

• Magnetic Resonance Elastography (MRE) precision relies on the ratio of mechanical wavelength to spatial resolution.
• This ratio can be suboptimal due to variations in actuation frequency, patient-specific factors, and organ heterogeneity.

Purpose of the Study:

• To introduce an MRE reconstruction algorithm that dynamically adjusts the shear wavelength to pixel size ratio.
• To evaluate its performance against existing methods in diverse clinical and phantom scenarios.

Main Methods:

• A prospective study involving phantoms, healthy volunteers, and patients with nonalcoholic fatty liver disease.
• Utilized a 3T gradient-echo elastography sequence at 40 Hz, 60 Hz, and 80 Hz.
• Compared stiffness accuracy, repeatability, and diagnostic performance using linear regression, Bland-Altman analysis, and ROC analysis.

Main Results:

• The proposed algorithm (MARS) demonstrated strong performance across frequencies, particularly at 40 Hz (AUC=0.88), 60 Hz (AUC=0.91), and 80 Hz (AUC=0.90) for advanced fibrosis detection.
• MARS showed the best diagnostic performance and competitive phantom accuracy.
• Repeatability was best with MDEV (23%) and lowest with k-MDEV (53%).

Conclusions:

• The MARS algorithm offers superior diagnostic performance for detecting advanced liver fibrosis in MRE.
• This adaptive resampling technique optimizes the wavelength-to-resolution ratio for improved quantitative MRE accuracy.