Intoduction

DWI of the liver is a useful clinical tool for lesion detection. Further, mapping of quantitative parameters such as the apparent diffusion coefficient (ADC) may enable lesion classification or prediction of therapy response^1,2. A major challenge of liver DWI is the sensitivity of diffusion sequences to physiological motion³. Cardiac and respiratory induced liver motion cause signal dropout in DWI and artificially elevate ADC measurements⁴. This limits the reproducibility of liver DWI and ADC estimation. M1-optimized diffusion imaging (MODI) uses partially motion-compensated diffusion gradients to desensitize liver DWI to motion while preserving suppression of vessel signal which may otherwise mimic liver lesions⁵. MODI has been previously implemented on a 3.0T MR system with high performance gradients. However, its implementation on conventional gradient systems was delayed due to the longer TEs associated with the weaker gradients. Recent updates to MODI allow for reduced TEs making motion-robust liver DWI on these systems feasible⁶. Therefore, the purpose of this work is to evaluate the reproducibility of DWI and ADC measured across 3 different MR systems, including 1.5T and 3.0T systems and systems with high-performance and conventional performance gradients.

Methods

Acquisition

Eight (n=8) healthy human subjects were imaged on three GE Healthcare (Waukesha, Wisconsin) MR systems (Table 1): 3T Signa Premier (scanner 1), 3T Signa PET/MR (scanner 2), and 1.5T Signa Artist (scanner 3). On each scanner, two different DWI acquisitions were performed: 1) a conventional Stejskal–Tanner, “monopolar”, sequence (MONO) and 2) MODI with tetrahedral diffusion directions and M1 = 0.1 s/mm. The tetrahedral encoding was implemented to reduce the TE of MODI. The scan volume covered the whole liver and respiratory triggering was used. Two b-values (50 and 500 s/mm²) were acquired for each sequence (see Table 1 for additional acquisition parameters).

Analysis

A board-certified radiologist delineated one ROI in each of the 9 segments of the liver on the b=50 s/mm² images⁷. Average ADC estimates within each ROI were determined. Reproducibility of each sequence was evaluated using the within-subject coefficient of variation (CV). MONO and MODI CV distributions for each ROI were tested for significant difference using Bonferroni-corrected t-tests with a threshold of p < 0.05. Bland-Altman analyses and reproducibility coefficients (RDCs) were used to compare ADC measurements from all ROIs collectively between 1) 1.5T and 3.0T scanners and 2) high (scanner 1) and conventional (scanners 2 and 3) gradient performances.

Results

ADCs across all scanners and subjects for each liver segment are shown in Figure 1. With the MONO acquisition, ADC values in the left lobe of the liver (segments I-IVb) were elevated (mean: $$$1690 \times 10^{-6}$$$ mm²/s) compared to the right lobe (segments V-VIII, mean: $$$1170 \times 10^{-6}$$$ mm²/s). MODI ADC measurements were relatively consistent between the two lobes (left mean: $$$1410 \times 10^{-6}$$$ mm²/s, right mean: $$$1230 \times 10^{-6}$$$ mm²/s).

The within-subject CVs distributions for each liver segment can be found in Figure 2. No significant differences between MONO and MODI CVs were found in any segment in the left or right lobes. However, a general trend was observed. CVs in the left lobe tended to be similar between MONO and MODI (mean difference: -1.5%) whereas CVs in the right lobe were consistently lower (mean difference: -5.5%).

Bland-Altman analyses between all ROIs collectively showed that both MONO and MODI had a small mean difference in ADC between field strengths ($$$-18~\times~10^{-6}$$$ vs $$$-16~\times~10^{-6}$$$ mm²/s) (Figure 3), but MODI had smaller RDC ($$$825~\times~10^{-6}$$$ vs $$$579~\times~10^{-6}$$$ mm²/s). Between high and conventional gradient systems, the mean difference was also similar for MONO and MODI ($$$-32~\times~10^{-6}$$$ vs $$$-97~\times~10^{-6}$$$ mm²/s), but MODI again had smaller RDC ($$$819~\times~10^{-6}$$$ vs $$$611~\times~10^{-6}$$$ mm²/s).

Finally, representative ADC maps are shown in Figure 4, which demonstrates the similarity of ADC maps across the three different scanners. In particular, MODI shows resilience to motion-induced elevated ADC in the left liver lobe.

Discussion

This work assessed the feasibility and reproducibility of liver ADC measurements using motion-robust MODI gradient waveforms across systems with different field strength and gradient performance. Within individual ROIs, MODI achieved similar reproducibility in the left lobe and trended toward lower reproducibility in the right lobe. When all ROIs were analyzed collectively, MODI ADC saw superior agreement between field strengths and gradient performance. Moreover, MODI helped to mitigated motion-induced bias in ADC measurements typically seen in the left lobe of the liver. Therefore, MODI could be beneficial for providing reproducible ADC values while reducing motion induced ADC bias.
This study has several limitations, including the evaluation of a limited number of healthy volunteers using a single MR vendor, and the lack of comparison to other motion-robust DW waveform designs.

Conclusion

Our results suggest that MODI is feasible on both 1.5T and 3.0T MR systems with conventional gradient performance. Additionally, MODI may provide better reproducibility compared to conventional MONO diffusion. Liver DWI using MODI tended to score better on reproducibility metrics across systems than conventional DWI while also providing better interlobar agreement in ADC measurements of healthy volunteers. MODI may enable motion-robust and reproducible DWI of the liver across field strengths and gradient performance.