Introduction

Multiparametric MRI promises to provide a more comprehensive assessment of liver disease^1,2. Multiparametric studies of T1, T2 and apparent diffusion coefficient (ADC) are traditionally performed in sequential scans³, which are prone to mis-registration artifacts. Recently, Magnetic Resonance Fingerprinting (MRF) has been proposed for simultaneous T1, T2 and ADC mapping of the brain^4-6. However, single-breath-hold simultaneous T1, T2 and ADC MRF for liver tissue characterization has not been investigated. In this study we propose a triggered 16-heartbeat acquisition that enables fast, simultaneous T1, T2 and ADC quantification in a single scan of ~16s. The proposed sequence has been evaluated in phantoms and in eleven healthy subjects.

Methods

A GRE MRF sequence with golden radial trajectory and varying magnetization preparation pulses is proposed (Figure 1a). The acquisition consists of 16 blocks preceded by the following preparation pulses: IR (TI = 20ms), three T2 preparation (T2ps(repetitions) = {33(2),50(1)}ms), ten diffusion preparation (b-values(repetitions) = {10(2), 200(2), 400(2) and 600(4)} s/mm2) and two blocks without preparation. The diffusion preparation had a duration of 50ms, including two adiabatic refocusing pulses of 10.3 ms inserted in-between the diffusion gradients to increase robustness against high field inhomogeneities⁷. The pulse was optimized to reach the highest possible b-value for a given diffusion preparation, given hardware max. gradient and slew-rate specifications, and gradients were applied simultaneously along the three Cartesian axes. The presence of short eddy currents was alleviated by leaving a 1-ms gap between each gradient lobe and its consecutive RF. The gradient waveforms optimization problem had restrictions to compensate for concomitant gradients, and up to first-order moment nulling for increased robustness against body motion and blood flow. Each acquisition block contains a train of 30 shots with sinusoidally varying flip angles in the range 10°-20°. Acquisition parameters include: TE=3.3ms and TR=7.0ms, 1x1mm² in-plane resolution, 6mm slice thickness and acquisition time ~16s. The acquisition was triggered at a fixed delay of 200ms using PPU. Reconstruction was performed using low-rank inversion approach⁸ and regularized with a multi-contrast patch-based higher-order reconstruction⁹. All scans were performed on a 3T scanner (Achieva TX, Philips; Gmax=62 mT/m, Smax=100 T/m/s). The proposed approach was evaluated in phantoms: a) a standardized T1MES phantom for T1, T2 validation and b) a set of tubes filled with a fixed concentration of NiCl2 and Agarose to obtain physiological T1, T2 values, as well as a varying concentration of polyvinylpyrrolidone to generate a range of isotropic diffusion values¹⁰. Measurements were compared against SE references. The sequence was additionally tested in eleven healthy volunteers (5 female, mean±SD age=30.1±3.7y) and compared against conventional individual single breath hold maps (T1-MOLLI (2x2x6mm), T2-GRaSE (2x2x6mm), and DWI-SE (3x3x4mm)) through Bland-Altman plots. Quantification for each subject was done by measuring T1, T2 and ADC in eight 10mm-diameter ROIs manually placed in the liver avoiding contamination from visible blood vessels.

Results

Phantom evaluation of T1-MRF, T2-MRF and ADC-MRF maps against their respective references is shown in Fig.2. Pearson r² coefficients are excellent (r²>0.90 for all cases). A comparison between MRF-derived maps and their respective references for in vivo liver quantification is shown in Fig.3. For each subject, the co-registered MRF maps were obtained from a single breath-held acquisition of ~16s. The corresponding reference maps were obtained in 14s, 18s and 45s respectively. Fig. 4A shows the Quantitative evaluation of all healthy subjects on a per-subject basis, and Bland-Altman analysis of agreement between MRF and their corresponding clinical sequences are shown in Fig.4.

Discussion

MRF-derived T1, T2 and ADC quantification is in good agreement with their respective reference measurements in phantom and healthy subjects. Bland-Altman plots showed a small negative bias in T2 and a significative T1 underestimation in comparison to T2-GRaSE and T1-MOLLI, which has been previously reported^11-13. Several overestimated ADC-SE measurements are observed in some subjects. This corresponds always to ROIs within the left lobe (see yellow arrows inf Figure 3), where cardiac motion leads to incorrect diffusion preparation and introduces a positive bias in ADC quantification. This bias is not observed in the proposed ADC-MRF, demonstrating more robustness of the proposed technique against this type of artefacts. Multi-echo readouts will be explored as future work to enable water-fat estimation and a more comprehensive liver tissue characterization. Future work will also evaluate this technique in a larger cohort of healthy volunteers and patients with hepatic disease.

Acknowledgements

This work was supported by the following grants: (1) EPSRC P/V044087/1,(2) BHF programme grant RG/20/1/34802, (3) Wellcome/EPSRC Centre for Medical Engineering (WT 203148/Z/16/Z), (4) Millennium Institute for Intelligent Healthcare Engineering ICN2021_004, (5) FONDECYT 121074, 1210637 and 1210638, (6) IMPACT, Center of Interventional Medicine for Precision and Advanced Cellular Therapy, ANID FB210024, (7) PhD program in Biological and Medical Engineering of the Pontificia Universidad Católica de Chile.