Synopsis

Keywords: Pulse Sequence Design, Fat, Water, pdff, Abdomen

A 3D radial-encoding MP2RAGE sequence was developed with binomial pulses within the GRE trains. These give the possibility to obtain water-specific, fat- specific or composite 3D T₁-maps of the abdominal cavity during free-breathing in less than 10 minutes. A 3D Proton Density Fat Fraction (PDFF) map can also be determined with the same scan. The T₁ values are similar with the standard MP2RAGE and the PDFF correspond to the ones obtained with MR spectroscopy. A test-retest on healthy volunteers shows the good reproducibility of the method.

Introduction

Abdominal T1 mapping is usually performed using variants of Look-Locker sequences (MOLLI, SASHA) or a Variable Flip Angle (VFA) sequence. VFA needs to be corrected for B₁ imperfection and requires multiple angles to measure a large T₁ range, lengthening its duration. T₁ measurements are biased in the presence of fat when using LL sequences, although the water-specific T₁ seems to be a more valuable information than the composite T₁ (including both water and fat T₁ in one voxel) for hepatic diseases. Also, these sequences are usually applied in 2D in order to fit their duration within a breath-hold, sacrificing the spatial resolution in at least one direction.
Consequently, we developed a MP2RAGE sequence with a 3D radial encoding to limit motion sensitivity, and with frequency-selective pulses in order to independently measure both the water-specific T₁ and the fat-specific T₁ at 3T. With these data, we also developed a method to accurately measure the Proton-Density Fat-Fraction (PDFF).

Materials and Methods

Two golden angles were implemented in order to obtain a pseudo-random filling of the k-spaces, continuing from one MP2RAGE_TR (5000ms) to the next, as described by Faller et al [1]. Binomial pulses including 6 sub-pulses (total duration of 3.9ms) were implemented in the GRE train (128 echoes), replacing the standard pulses (named NonSel thereafter). Odd trajectories were acquired with a water-frequency selective pulse and the even ones were acquired with a fat-selective pulse [2]. T₁-maps were reconstructed either with the odd (water-specific T₁) , the even (fat-specific T₁) projections or both (composite), from acquisitions lasting 12min for a total of 18432 projections. To measure the PDFF, the signal of each voxel obtained at equilibrium was first estimated for both water (S_w) and fat (S_f) tissues, from the signals measured at GRE₁ and GRE₂ and the measured T₁. Then, the PDFF was computed as follow : PDFF = S_f/(S_f+S_w).
This new sequence was applied on a commercially available phantom (MultiSample120, Gold Standard Phantoms, United Kingdom) containing home-made solutions: either increasing concentrations of Gd-DOTA, or increasing volume of pork fat, both diluted in 1.5% agarose gel. Then, the sequence was applied on 8 healthy volonteers on a Siemens 3T PRISMA scanner, using the 18-channel body and the spine coils.
Regions-of-interest (ROI) were drawn on the parametric maps using Matlab, on the liver (taking care of not including large blood vessels) and on the back muscles (representing 100% water-containing tissues), and on the subcutaneous adipose tissue (containing 100% fat).

Results

On the phantom, the binomial pulses enabled to selectively obtain water-specific T₁ in tubes without lard (Figure 1). Similar T₁ values were consequently obtained on the NonSel, Composite and Wat T₁ maps. Also, the valueswere similar to the gold-standard Inversion-Recovery (IR) sequence (<10% bias) and the vendor cartesian MP2RAGE sequence (<10% bias).
In tubes containing lard, these pulses enabled to selectively obtain water-specific T₁ or fat-specific T₁, without being affected by the T₁ of fat or water, respectively.
Fat-specific T₁ was measured appropriately in every tube (<10% compared to IR), independently of the volume of fat present in the tube. Also, the PDFF calculated was similar to MR spectroscopy (less thant 5% error).
Applied on the volunteers during free-breathing, the radial encoding enabled to obtain respiration-artefact free images and T₁ maps of the whole abdomen (Figure 2). The binomial pulses enabled a high selectivity of the tissues, even though the interface between the liver and the lungs was not appropriately excited by the water-selective pulses (representing <1% of the total liver volume). The T₁ values of the liver and the muscle in each healthy volunteer were similar on the NonSel, Composite and Wat T₁-maps (Figure 3) and were concordant with literature [3]. Also, the standard deviations were much smaller than the ones obtained with the VFA sequence after B₁ correction.
The reproducibility of the measurements was quite high (coefficient of variance of less than 9.8%).
Finally, the sequence was accelerated through retrospective undersampling (Figure 4), and showed high quality and similar liver T₁ using 12288 projections (so with acquisitions lasting only 8min).

Discussion

Here we presend a methodology to obtain 3D multi-parametric maps of the whole abdominal cavity, due to 1) radial encoding for free-breathing acquisitions and 2) binomial pulses with large coverage. Composite T₁, water-specific T₁, fat-specific T1 and PDFF can be obtained with one scan, and with measured values close to gold-standard methods and highly reproducible. The duration of this method can be minimized to 8 minutes.
Further investigations are needed to appropriately excite the interface between the liver and the lungs. Nevertheless, the T₁ of the liver can still be measured accurately in 3D.
Similar T1 measurements were obtained between the radial NonSel and the newly-developed sequence including binomial pulses, which validates its reliability. Application on pathological tissue is necessary to further demonstrate the potential of this method.

Conclusion

The new method developed here is of great interest for multiple applications, like obesity, grading of nonaloholic fatty liver diseases or cirrhosis.

Acknowledgements

This study was achieved within the context of the Laboratory of Excellence TRAIL ANR-10-LABX-57. This work was also supported by the French National Research Agency (ANR-19-CE19-0014).