Introduction

Previously, we introduced a new parallel transmit (pTx) spatial-spectral (spsp) pulse design with explicit SAR management for robust uniform water excitation and demonstrated its utility for whole-brain water excitation at 7 Tesla (7T) using Bloch simulation¹. Here, we validated our design in humans, and demonstrated its usability for designing universal pulses², comparing to existing approaches.

Methods

Experiments were conducted on a Siemens Terra (Siemens, Erlangen, Germany) equipped with a body gradient (200 T/m/s slew-rate, 80 mT/m G_max).

Data were acquired using the commercial Nova 8-channel transmit 32-channel receive coil (Nova Medical Inc.). Calibration including $$$\Delta$$$B₀ and B₁+ mapping was obtained using vendor-provided pulse sequences with 3-mm resolution, 100×108 matrix size, and 80 axial slices. A pTx-enabled sequence was developed to acquire 3D GRE images.

Our pulses (Fig. 1) were designed using a 2-step approach as described previously¹. Local SAR control was informed by SAR calculation based on 1669 virtual observation points^3,4.

We started by validating our method in a phantom constructed by filling a cylinder (diameter: 10 cm; length: 13 cm) with 500 ml of oil and 500 ml of physiological saline. Our pulses’ frequency responses over the frequencies of interest were measured by incrementing RF frequency and acquiring corresponding 3D GRE images with 3-mm resolution, 2-fold acceleration, 8-ms TE and 50-ms TR.

We then validated our method in humans. Brain masking was achieved by brain extraction based on a 3D whole-brain GRE. To demonstrate the utility of our pulses, high-resolution whole-brain 3D GRE images were acquired with 1-mm resolution, 220×220×220 matrix size, 2-fold acceleration, 8-ms TE, and 50-ms TR.

We also investigated how our method would perform when used to design universal pulses (UP). A calibration dataset comprising brain masking, $$$\Delta$$$B₀ and B₁+ mapping obtained in 10 healthy volunteers (five men) were used. Simulations were conducted to evaluate the performances of our UP based on a 10-fold cross-validation. Root mean square error (RMSE) and coefficient of variation (CoV), averaged across folds, were calculated to evaluate performances for fat suppression and water excitation, respectively. Results were compared to tailored, subject-specific design.

To demonstrate the utility of our method, we designed UP using entire calibration and use it to scan a new volunteer using the same 3D GRE protocol as in the above validation.

For comparison, imaging in same volunteers with matched parameters was also performed using 1) the Nova coil operated in its circularly polarized (CP) mode mimicking single-transmit, and 2) pTx pulses designed with the interleaved binomial approach⁵.

Results

Our design was validated in the phantom (Fig. 2), achieving water-selective excitation at the water resonance frequency while producing the desired spectral selectivity across the entire frequency range of interest.

In humans, our method outperformed existing approaches (Fig. 3), improving not only water excitation (relative to excitation in the CP mode) but also fat suppression (compared to the interleaved binominal approach) across the entire brain even in challenging areas.

When used to design UP, our method was found robust against inter-subject variabilities (Fig. 4), inheriting most of water-excitation performances from the tailored design while outperforming both UP and tailored designs with the interleaved binomial approach in achieving robust simultaneous water excitation and fat suppression over a 400-Hz bandwidth.

As shown in Fig. 5, our UP when used to scan an unseen volunteer, improved water-selective excitation across the entire brain, effectively restoring signal loss in the cerebellum (compared to CP mode approaches) and robustly suppressing fat even in challenging regions (relative to UP designed with the interleaved binomial approach).

Discussion

We have validated our previously proposed pTx spsp pulse design capable of robust uniform water excitation across an extensive image volume even in the presence of strong RF inhomogeneity and large off-resonances. Our 7T human results show that our design can produce high-quality uniform water excitation across the entire brain, outperforming existing approaches. Our results also show that our method can be used to design universal pulses without sacrificing its performances, suited to be used for plug-and-play pTx.

Further SAR calculation suggests that our method can reduce local SAR by as much as ~31% relative to the interleaved binomial approach for comparable pulse lengths, thanks to explicit local SAR control incorporated in pulse design.

Part of our future work is to integrate our pulses into GRE-EPI to promote whole-brain functional MRI at ultrahigh field.

Conclusion

In humans at 7T, we have validated that our new pTx spatial-spectral pulse design can achieve quality uniform water excitation across the entire brain. We believe our new design will have many applications including high-resolution functional MRI at ultrahigh field.

Acknowledgements

The authors would like to acknowledge Patrick Liebig from Siemens Healthineers for his assistance with calculation of multichannel B₁+ mapping using the vendor sequence. KU and XW and all work carried out at the University of Minnesota were supported in part by USA NIH grants (NIBIB P41 EB027061 and U01 EB025144).