Liver diseases due to various causes are a significant health concern. Liver biopsy is invasive, and there is a need for noninvasive methods, such as MRE, to stage liver fibrosis. The study aims to determine the accuracy and repeatability of MRE stiffness measurements in vitro in a multicenter study.

Three cylindrical phantoms made of polyvinyl chloride gel with known stiffness (phantoms 1-3: low, medium and high stiffness) were circulated between 1 reference center and 4 testing centers and scanned on 10 MRE-equipped commercial systems (1.5T or 3T) from the 3 major vendors, using 2D gradient recalled echo (GRE) and/or 2D spin echo echoplanar imaging (SE-EPI) with similar acquisition parameters (Figure 1) and hardware. Mean phantom stiffness was measured by a single observer for each phantom/acquisition. Test-retest repeatability was measured during the same session at all centers. The reference standard measurements were obtained in the reference center. Accuracy error (based on the reference standard) and test-retest repeatability (based on Bland-Altman coefficient of repeatability) of stiffness measurement were calculated.

The reference stiffness measurements for phantoms 1-3 were (in kPa): 2.45, 4.41 and 7.52. The mean accuracy error [95% confidence intervals] for all 3 phantoms and both sequences combined was 10.2% [8.1%-12.3%]. Mean accuracy error was 8.3% [3.8%-12.9%] for the soft phantom, 8.8% [5.2%-12.4%] for the medium stiffness phantom and 13.2% [10.3%-16.1%] for the stiff phantom with both sequences combined (Figure 1). For 2D GRE and 2D SE-EPI sequences, mean accuracy error (all phantoms combined) was 8.5% [6.5%-10.6%] and 13.0% [8.5%-17.6%], respectively. The mean Bland-Altman coefficients of test-retest repeatability were 2.9% [1.9%-4.0%] for all phantoms and sequences combined. Test-retest repeatability for each individual phantom was 3.0% [0.3%-5.8%] for phantom 1, 1.9% [0.7%-3.0%] for phantom 2, and 3.7% [2.1%-5.5%] for phantom 3 with both sequences combined (Figure 1). Per sequence repeatability was 3.6% [1.5%-5.7%] for 2D SE-EPI and 2.5% [1.3%-3.6%] for 2D GRE (including all phantoms).

Our multicenter phantom study shows that MRE has excellent measurement accuracy and test-retest repeatability in vitro, when using the same protocol and hardware. These results may help design future multicenter MRE studies.

Liver stiffness measured with MRE can non-invasively stage liver fibrosis with high diagnostic performance, but there is limited knowledge on measurement accuracy, especially in the multicenter setting through an in vitro study.