Introduction

Pancreatic MR Elastography (MRE) can detect viscoelastic changes in the course of pancreatitis or pancreatic cancer^1,2. Both breath-hold and free-breathing (FB) MRE is possible, the latter preferable in sick patients unable to suspend their breath multiple times. However, FB-MRE acquisition can lead to a mismatch in pancreatic location over multiple phase-offsets, potentially introducing errors in stiffness reconstruction.
Using a breathing task (BT) paradigm timed such that MRE acquisition for a given slice occurs during the same respiratory phase could potentially increase MRE quality without limiting the temporal and spatial resolution. We investigated the impact of BTs on pancreatic MRE by evaluating the pancreatic location over multiple phase-offsets. Furthermore, the influence of BTs on the octahedral shear strain signal-to-noise-ratio (OSS-SNR) and shear-wave speed (SWS) was determined.

Methods

Study protocol:
Nine healthy volunteers (♂=3,♀=6, mean age 27±2 years) were included. Subjects fasted four hours prior to MRI examination. BTs were created in ePrime (version 3.0,PST Inc.,Sharpsburg,PA,USA) and followed a natural breathing pattern(Figure 1). The breathing period was equal to TR and phase-offset delay, and instructions were synchronized with the sequence using a TTL pulse. This way, each slice with varying phase-offset is acquired at the same respiratory phase.
MRI measurements:
Multi-slice multi-frequency spin-echo echoplanar imaging (SE-EPI) MRE images were acquired with a 3.0T MRI scanner (R.5.7.1.1,Ingenia,Philips,Best,the Netherlands). Mechanical vibrations were introduced using four compressed-air driven MRE transducers at four vibrational frequencies (MREfreq=30,40,50,60Hz)². Motion encoding gradients (MEG) were applied along three orthogonal directions. Three different MRE acquisitions were recorded(Table 1). In FB-MRE scans, no BTs were displayed and standard interleaved slice ordering was used. For the two MRE scans using synchronized BTs, interleaved (BT-Int) and reversed central slice ordering (BT-Rev) were performed(Figure 1). BT-Rev was used to test if central slices containing the pancreas within the multi-slice acquisition were scanned during end-expiration.
Analysis and statistics:
To determine the stability of the pancreas location over phase-offsets for the three MRE scans, the pancreas was manually delineated in three slices aimed at head, body and tail on the magnitude MRE images for eight phase-offsets at 30 and 50 Hz. The mean dice similarity coefficient (DSC) of all combinations of the regions of interest (ROI) of all three slices was calculated. To test for differences between BT-Int and BT-Rev, the DSC for each slice was compared. Magnitude images of BT-Int and BT-Rev were visually compared to investigate the influence of slice order on image quality.
The whole pancreas was delineated on mean magnitude MRE images (mean over phase-offsets, MEG directions, frequencies). These ROIs were used to determine the pancreatic OSS-SNR and pancreatic SWS which was calculated with kMDEV inversion³. The OSS-SNR was also calculated over the whole abdomen. Repeated measures ANOVA with pairwise comparison and Bonferroni correction was used for statistical analysis. Image analysis and statistical analysis were performed in Matlab (R2021a,Mathworks,Natick,MA,USA) and SPSS (version 26,IBM,Armonk,NY,USA).

Results

Mean values of DSC, SWS and OSS-SNR are shown in Table 2. Compared with FB, DSC was significantly higher for BT-Rev (p=0.026), while no significant difference was found for DSC between BT-Int and BT-Rev or FB and BT-Int. No significant differences were found between the MRE scans for pancreatic SWS and OSS-SNR. However, OSS-SNR in the whole abdomen significantly increased for BT-Rev compared to FB and BT-Int (p-value=0.010 and 0.014 respectively).
Magnitude images over eight phase-offsets can be found in Figure 2. Comparing the magnitude image quality between BT-Int and BT-Rev, signal loss in various slices was present more pronounced in BT-Rev. Overlap images of phase-offset ROIs and corresponding elastograms are shown in Figure 3.

Discussion

Introduction of BT-Rev paradigm during pancreatic MRE significantly increases DSC compared to FB-MRE. This may be an important finding for pancreatic tumors where stromal tissue is of great interest. Pancreatic tumors can be small and a consistent location over multiple phase-offsets is needed to be able to perform sub-region analysis. Previous research showed that 2D motion correction in coronal direction improves sharpness of FB-MRE images⁴. Nevertheless, this is more difficult in an axial orientation and eddy current artifacts lead to distortions when acquiring pancreatic MRE in coronal orientation on a 3T MRI system.
BT-Rev MRE significantly improves OSS-SNR in the whole abdomen compared to FB and BT-Int, while pancreatic OSS-SNR did not significantly differ between scans. Higher abdominal OSS-SNR suggests that measurements may be more reproducible when BT-Rev is used. The pancreatic SWS did not significantly differ, and values were comparable to literature^1,2,4,5.
Magnitude image quality of BT-Int was visually higher than BT-Rev. This could be explained by signal leakage into adjacent slices typical of EPI acquisitions. Despite this finding, the increase in whole-abdomen OSS-SNR and DSC warrant future research on using BT-Rev to compare the reproducibility of SWS measurements with FB-MRE. The scan time increased two times when using BTs, giving the opportunity to acquire more/thinner slices or a higher in-plane resolution. Furthermore, future research could focus on performing motion correction in axial orientation or applying respiratory binning.

Conclusion

Although, pancreatic OSS-SNR did not improve and SWS did not significantly change, using BTs during pancreatic MRE acquisition causes a more stable pancreatic location over multiple phase-offsets leading to a higher OSS-SNR in the whole abdomen.

Acknowledgements

No acknowledgement found.