Diffusion-weighted MRI-based Virtual Elastography for the Assessment of Liver Fibrosis

Background Diffusion-weighted (DW) MRI-based elastography has recently been proposed for noninvasive liver fibrosis staging but requires evaluation in a larger number of patients. Purpose To compare DW MRI and MR elastography for the assessment of liver fibrosis. Materials and Methods In this retrospective study, patients underwent MR elastography and DW MRI between November 2017 and April 2018. Shear modulus measured by MR elastography (μ ) was obtained in each patient from regions of interest placed on liver stiffness maps by two independent readers. Shifted apparent diffusion coefficient (ADC) was calculated from DW MRI ( = 200 and 1500 sec/mm ) and converted to DW MRI-based virtual shear modulus (μ ). MRI-based liver fibrosis stages were estimated from μ and μ values (F0-F4) and serum fibrosis markers were assessed. Statistical analyses included Bland-Altman plots, Bayesian prediction, and receiver operating characteristic analyses. Results Seventy-four patients (mean age, 68 years ± 9 [standard deviation]; 45 men) were evaluated. Interreader coefficient of reproducibility was 0.86 kPa for DW MRI and 1.2 kPa for MR elastography. Strong correlation between shifted ADC and μ was observed ( = 0.81; < .001), showing high agreement between μ and μ values (mean difference, -0.02 kPa ± 0.88; < .001). DW MRI-based fibrosis staging agreed with MR elastography-based staging in 55% of patients (41 of 74) and within one stage difference in 35% of patients (26 of 74). Binarization into insignificant (F0-F1) and significant fibrosis (F2-F4) showed agreement in 85% of patients (63 of 74; κ = 0.85). Compared with serum markers (area under the receiver operating characteristic curve [AUC], 0.50-0.69), μ showed better performance in discriminating fibrosis stages F0-F2 from F3-F4 (AUC, 0.79; 95% confidence interval: 0.69, 0.90), whereas serum markers showed slightly better results for F0-F1 versus F2-F4 differentiation (fibrosis stages were estimated by using MR elastography). Combining DW MRI with serum markers provided a trend toward highest discriminative performance (AUC, μ + aspartate aminotransferase-to-platelet radio index: F0-F1 vs F2-F4, 0.81 [95% confidence interval: 0.69, 0.93], = .17; F0-F2 vs F3-F4, 0.83 [95% confidence interval: 0.74, 0.92], = .07; and AUC μ + Fibrosis 4 score: F0-F1 vs F2-F4, 0.78 [95% confidence interval: 0.64, 0.92], < .30; F0-F2 vs F3-F4, 0.81 [95% confidence interval: 0.71, 0.91], = .08). Conclusion MR elastography and diffusion-weigh