Introduction

Magnetic Resonance Elastography (MRE) is becoming an accepted biomarker of liver injury and is being widely adopted for clinical trials using the Quantitative Imaging Biomarker Alliance (QIBA) recommendations. However, higher failure rates of the gradient‐recalled‐echo (GRE) MRE sequences are observed, particularly in patients with fatty liver disease and/or iron overload. The iron-mediated T₂* effects are more prominent at 3 T than 1.5 T, and have been suggested to lead to failure rates of up to 15 % for MRE at 3 T [1]. Here we assess the effect of liver T₂* and acquired spatial resolution on GRE MRE stiffness maps.

Methods

5 healthy volunteers (3M/2F, BMI 23.1 ± 1 kg/m², age range 24-27 years) and 2 patients with liver disease (both Non-Alcoholic SteatoHepatitis, NASH) (1M/1F, BMI 29.9 ± 4 kg/m², age range 31 - 51 years) were recruited. Participants attended an MRI scan after an overnight fast. Data was collected on a 3 T Philips Ingenia DDAS scanner (DS Anterior coil + posterior bed coil) to assess liver stiffness as assessed by MRE, liver tissue T₂*.
Data acquisition:
MRE based liver stiffness was measured using the QIBA recommendation with a 2D GRE MRE sequence (FOV 360x375 mm, voxel 1.5x4.5x10 mm³, TE = 20 ms, TR = 50 ms, 4 axial slices through the liver, slice gap 1 mm, 1 slice acquired per 18 second breath hold) [2]. For comparison, data was also collected at 4.5 mm isotropic spatial resolution with matched FOV, TE/TR and slice positioning. For both acquisitions, the passive acoustic driver (Resoundant, Rochester, MN) was placed against the lower right chest at the level of the xiphoid in the midclavicular line, and continuous vibrations of 60 Hz were applied. Liver T₂* maps were collected using multi-echo fast-field echo (mFFE) data acquired at 12 echo times (TE1 = 2.5 ms, ΔTE = 2.5 ms, 9 axial slices, 3x3x8 mm³ voxel).
Data Analysis:
T₂*: Manual ROIs were drawn on the scanner computed T₂* map images, avoiding large blood vessels, bile ducts, and edges of the liver parenchyma. Histogram analysis was then performed to assess the distribution of T₂* within the liver, the mode of the distribution was used to represent tissue T₂* and FWHM to assess heterogeneity.
MRE: From the acquired magnitude and phase images, the scanner computed elastograms depicting the spatial distribution of the shear stiffness (ie, magnitude of the complex modulus, |G*|) in kPa, and confidence masks computed from voxels with confidence values >0.95. Data were then reshaped to match the resolution of the T₂* maps. Liver stiffness measurements were calculated based on mode values from the elastograms within the intersection of manually drawn ROIs and confidence masks. The percentage of the liver in which a confidence mask was formed was computed. In addition liver stiffness values were computed within each confidence mask for each spatial resolution of MRE data.

Results

Figure 1 shows example liver stiffness maps computed for the non-isotropic and isotropic resolutions in two participants, one participant with a short T₂* (12 ms mode) and one with a longer T₂* (22 ms mode). For shorter liver T₂* values, the percentage of the liver fit as defined by the confidence map is reduced when using the non-isotropic GRE-MRE protocol compared with the isotropic protocol, Fig. 2. Mean liver stiffness for the non-isotropic GRE-MRE protocol was 2.4 ± 2 kPa and 1.9 ± 1.4 kPa for the isotropic protocol.

Discussion

Participants who have shorter liver T₂* associated with higher iron content within the liver tissue, have a smaller percentage of the liver which fits with confidence (confidence >0.95), Fig. 2. The GRE acquisition is limited to a TE of 20 ms as the MRE motion encoding gradient frequency is set to match the mechanical driving frequency of 60 Hz. As a result, signal-to-noise ratio is limited in participants with a shorter liver T₂* as rapid dephasing of the signal occurs during the echo time. We show that by choosing an isotropic resolution with larger voxel volume (91.1 ml isotropic compared to 67.5 ml for the non-isotropic QIBA recommendation), the goodness of fit in the stiffness maps is improved, and not dependent on the T₂* of the liver, allowing successful spatial mapping of stiffness over a larger area of the liver (Fig. 2).

Conclusion

Using a isotropic voxel size with larger volume as compared to the QIBA recommendation for the GRE MRE acquisition provides a higher SNR which in turn results in a higher proportion of voxels being fit for stiffness with confidence >0.95 within the liver tissue.

Acknowledgements

We would like to thank the NIHR Nottingham BRC research nurses who conducted patient enrolment