Introduction

There has been a rapidly growing interest in pursuing high resolution MRI at ultrahigh field (UHF) owing to promised improvements in SNR, tissue contrast and parallel imaging performances. However at UHF it is necessary to address the challenges of the increased transmit B1 (B1+) inhomogeneity and RF power deposition. One effective solution to these two challenges is to use parallel transmission (pTx)^1,2.
Recently, 3D simultaneous multi-slab (SMSlab) imaging has been proposed to increase the SNR efficiency of high-resolution whole-brain diffusion MRI at 3T^3,4. The aim of this study was to explore how slab-wise multiband (MB) pTx RF pulses can be designed for 3D SMSlab at 7 Tesla (7T).

Methods

Here, we considered imaging with the commercial Nova 8-channel transmit 32-channel receive coil (Nova Medical Inc.). Calibration data including dB0 maps and 8-channel B1+ maps (Fig. 1) were obtained experimentally from healthy volunteers on a 7T MR scanner (Magnetom DotPlus, Siemens). For each volunteer, we obtained calibration data in 60 axial slices covering the whole brain.
We designed MB multi-spoke pTx pulses for SMSlab image acquisition with a MB factor of 2 (MB2). All pulses were designed using the generalized slab-wise design framework². Specifically, six contiguous slabs (Fig. 2) were prescribed that jointly covered the brain in the slice direction. For each slab, band-specific (or slab-specific) multi-spoke single-band pTx pulses were designed and the single-band pulses corresponding to the same slab group were used to form the multi-spoke MB pTx pulses. We investigated two ways of using B1+ mapping slices (Fig. 2): one was only using the middle B1+ mapping slice per slab (1 slice-referred design) and the other using three B1+ mapping slices per slab (3 slices-referred design).
We designed both 2- and 3-spoke pTx pulses in the small tip angle regime. The pulse design problem was formulated as a regularized magnitude least square minimization⁵ and was solved using the variable exchange algorithm.
The gradient waveforms were designed assuming the maximum gradient strength of 80 mT/m and maximum slew rate of 100 T/m/s. The basic RF waveform was Gaussian shaped with TBWP = 4.
For improved excitation performances, the spoke placement in the excitation k-space was optimized by exhaustive search. The total pulse length was kept constant for both 2-spoke and 3-spoke pTx RF pulse designs.
The performances of our MB multi-spoke pTx pulses were compared against those of the single channel transmit circular polarization (CP) mode and single-spoke MB pTx pulse (corresponding to band and channel specific RF shimming).

Results and Discussion

Figure 2 shows the example gradient waveforms and the corresponding MB2 pTx RF pulses designed with 2 and 3 spokes.
The use of our designed multi-spoke MB pTx pulses appeared to improve the excitation uniformity across the brain (Fig. 3) in comparison to either the CP mode or the RF shimming. The multi-spoke pTx pulses designed with a single B1+ mapping slice per slab however did not perform as well as those with three B1+ mapping slices per slab, presenting reduced excitation uniformity near the edge of the target slab.
As expected, pulses designed with 3 spokes outperformed those with 2 spokes, on average reducing the excitation error by ~25%. The 3-spoke pulses designed using three B1+ mapping slices per slab was found to provide the best excitation performance.

Conclusion

This study demonstrates that multi-spoke and multi-slice-referenced pTx RF pulses have achieved satisfactory results in the excitation mode of SMSlab. A detailed comparison between 2-spoke and 3-spoke RF pulses has been made to show that 3-spoke RF pulses have better performance, and its peak amplitude increases only slightly while the pulse duration remains equal to 2-spoke RF pulses.

Acknowledgements

This work was supported in part by the NIH grants NIBIB P41 EB027061 and NIH U01 EB025144.