Introduction

Liver MRI could benefit from the increased SNR at 7T. However, the B1+ wavelength decreases to abdominal dimensions at this field strength, causing flip angle (FA) inhomogeneity when using simple RF pulses in circularly polarized (CP) transmission. In this work, we use multi-channel parallel transmission for whole-liver FA homogenization comparing kt-points¹ with phase shimming. This necessitates artifact-free B1+ mapping, which was done using Fourier phase-encoded (FPE) DREAM^2,3.

Methods

Eight healthy volunteers were scanned with a 7T Philips Achieva scanner (Philips, Best, The Netherlands) using an eight-channel Tx/Rx fractionated dipole antenna body array⁴ (MRCoils, The Netherlands, Figure 1c). For FPE-DREAM B1+ mapping, multiple B1+ measurements (‘modes’) are acquired with all channels emitting with continuously varying phase between modes according to $$$\exp\left(\frac{2\pi i\left(m-1\right)c}{M}\right)$$$, where m and c are the mode and channel number respectively and M is the total number of modes. In case M is higher than the number of channels, a weighting function is used to suppress unreliable measurements at low SNR from the overdetermined inverse problem³. After acquiring all modes, single-channel B1+ maps were calculated on-scanner which, together with a 2^nd order B0-shimmed 3D-GRE, served as input for phase shimming and kt-points pulse calculation. To find the minimally necessary number of modes, a 35-year old male with a relatively high BMI of 28.4 was scanned with 8, 10, 13 and 16 modes. Figure 2a shows the eight-channel B1+ magnitudes and phases calculated using 13 modes and p=0.25, q=20 in the weighting function (eq. 9)³. Figure 2b shows single-channel B1+ magnitude maps calculated from these acquisitions. Based on these results, 13 modes were determined sufficient. Using those values, FA shimming was performed on seven volunteers (BMI 18.4-23.3, weight 50-78kg, age 26-34 years, 3 males). In volunteer 1-5, 3D-GRE scans were acquired using CP (45˚ phase increments between channels), phase shimming (except for volunteer 1) and kt-points. Additionally, intra- and inter-session kt-points repetitions were performed in four volunteers. The shimming volume-of-interest (VOI) encompassed the whole liver (see exemplary VOI in Figure 3, volunteer 5, kt-points). For phase shimming, the eight transmit-channel phase offsets were individually adapted to minimize the RF standard-deviation/(mean)² over the VOI, using the MRCodeTool toolbox (MRCode, The Netherlands). Kt-points pulses were calculated with five subpulses, max. blip strength = 33 mT/m, max. blip slew rate = 166 mT/m/ms and a maximum excitation k-space extent of 1/6 cm^-1. Channel RF amplitudes and phases and the 3D excitation k-space trajectory were calculated using an interleaved greedy-local optimization algorithm⁵ in a home-built MRCodeTool extension⁶ (Figure 1b). The kt-points pulses were non-selective, requiring large FOVs to cover the entire transmission volume of the antennae. All scans were acquired in a single breath-hold, as were all separate FPE-DREAM modes. Other scan settings are listed in Figure 1a. The scans were reconstructed offline using BART⁷ and MRecon (Gyrotools, Zurich, Switzerland) in MATLAB. Receive B1 inhomogeneity was reduced using ANTs’ N4BiasFieldCorrection⁸. Magnitude images and simulated FA distributions over the VOI were plotted to assess image quality.

Results

Figure 3 shows magnitude images in three orientations for volunteers 1-5 using CP, phase shimming and kt-points. CP shows clear signal dropouts, which are removed using phase shimming and kt-points. The kt-points scans show an increased overall contrast over phase shimming and CP. These results match with the simulations. They show that with both phase shimming and kt-points, the lowest FAs in CP can be increased, eliminating signal dropouts. Kt-points is able to deliver the nominal FA of 8˚ with a smaller FA range, while phase shimming only reaches a mean of ~4˚. The same behavior is seen in all five volunteers. Figure 4 shows good intra- and inter-session repeatability for volunteer 5. In volunteers 2 and 6, one of the scans showed signal dropouts, despite promising simulation results. Volunteer 7 reported breath-hold difficulties and showed varying B1 patterns.

Discussion

With FPE-DREAM, a homogeneous small-flip-angle whole-liver GRE can be acquired at 7T using kt-points pulses. With phase shimming, signal dropouts could be removed but with roughly half the nominal flip angle over the VOI. Occasionally, kt-points acquisitions give residual signal dropouts as shown in Figure 4. From volunteers 2 and 6 it follows that repeated acquisitions led to successful shimming, pointing at motion or inconsistent breath-holds as a possible cause. This would also explain signal dropouts in volunteer 7. The BMI values of volunteers in this study were quite low. However, Figure 2a shows that artifact-free B1+ maps and good simulation results could also be achieved on a volunteer with a higher BMI, which is promising for successful application of kt-points. This will be subject of future research, as will be the application of universal pulses⁹ to avoid the time-consuming B1+ calibration phase. Finally, we intend to make the kt-points pulses slab-selective to enable smaller FOVs which, in combination with undersampling techniques like SENSE or CS¹⁰, could increase spatial resolution.

Conclusion

With FPE-DREAM, we could acquire a homogeneous small-flip-angle whole-liver GRE at 7T using kt-points, with highly increased homogeneity and mean FA compared to phase shimming and CP.

Acknowledgements

No acknowledgement found.