Introduction

Ultra-high field MRI systems provide significant improvements in signal to noise ratio (SNR) and parallel imaging capability¹, which can benefit SNR starved techniques such as velocity selective arterial spin labelling² (VSASL). However, as the magnetic field strength increases, so does the resonance frequency needed to excite the spins. This translates to shorter radiofrequency wavelengths, which leads to a non-uniform transmit magnetic field (B₁⁺)³. Parallel transmission (pTx)^4,5 can be used to mitigate B₁⁺ artefacts^6,7. However, pTx implementations for advanced sequences, such as VSASL, have not yet been demonstrated. Here we evaluate the use of pTx pulses to improve VSASL at ultra-high field.

Methods

pTx-Calibration: All data was collected using a Magnetom 7T Plus scanner (Siemens Healthcare, Erlangen, Germany) equipped with an 8x2kW parallel transmit system, and an 8Tx/32Rx head coil (Nova Medical, Wilmington, MA, USA) (Fig.1A). The SA2RAGE⁸ (TR=2400ms, TE=0.95ms, TD1,2=106ms,1800ms, 4x4x4mm³, iPAT=2) was used to measure the B₁⁺ field in a circularly polarised (CP) mode, which was combined with a 2D interleaved GRE (TR=300ms, TE=3ms, 4×4×4mm³, iPAT=3) pulsing on individual channels to obtain quantitative B₁⁺ maps for each channel⁹. Total calibration time was 3min.

Sequence design: The VSASL preparation module (Fig.2) consists of a 90° tip-down pulse followed by bi-polar gradients to induce a 180° phase shift on spins with flowrate V_c. A 90° tip-up pulse is then applied inducing full 180° inversion on the spins flowing at V_c, while returning the stationary spins to equilibrium². To mitigate ΔB₀ effects, refocusing pulses are placed between the tip-up and tip-down pulses. In our protocol the VSASL preparation module was followed by with a 2D EPI readout (TR=10s, TE=20ms, 1.5x1.5x5.0mm³, iPAT=3).

Design of pTx pulses: Both the tip-down and tip-up 90° pulses in the VSASL preparation module were replaced with non-selective k_T-points pulses⁷ (duration=2.62ms, 8 sub-pulses). The slice-selective excitation at the start of the EPI module was replaced with a 2-spoke¹⁰ solution (TBWP=3, sub-pulse duration=2ms). Both pTx pulses were designed with the spatial domain method¹¹ using a magnitude least squares approach⁶. Two versions of the tip-up pulse were tested, both are phase inverted but one is also time reversed (Fig.3). The fidelity of these pulse VSASL combinations were evaluated in simulation and experiment.

Simulation: Bloch simulations were performed to evaluate the benefits of pTx in the VSASL sequence. Specifically, flow was modelled by dynamically updating the spatial coordinates of the isochromats in the simulation (time step=10us, isochromats=150 per voxel). The normalised root mean square error (NRMSE) was calculated to evaluate the pulse performance.

Phantom experiments: For experimental validation, we used an in house developed flow phantom (Fig.1B). An Arduino (UNO Rev3) was used to control the rate of flow (approx. 2cm/s) in the 4 tube sections housed in the main body of the phantom. To illustrate the behaviour of the VSASL module the flow encoding gradients were varied from 0 to 60000ms*uT/m. When using the CP-mode implementation the reference voltage was calibrated to achieve a 90° flip angle excitation in the centre of the phantom.

Results & Discussion

Simulation: Simulations of the VSASL preparation module are shown in Fig.4. The B₁⁺ non-uniformities in the tip-down, tip-up and excitation pulses have a significant impact on the contrast fidelity in CP-mode (NRMSE of 0.47, see Fig.4A,B). The majority of B₁⁺ artefacts can be mitigated using k_T-points. However, simple phase inversion of the k_T-points tip-up pulse leaves considerable residual magnetisation behind (NRMSE of 0.22, see Fig.4C). As might be expected, based on the time symmetry of the design problem, the time reversed solution performed markedly better (NRMSE of 0.09, see Fig.4D).

Phantom experiment: Our pTx solution demonstrates a significant improvement in signal and contrast uniformity (Fig.5A). Using the CP-mode fails to invert the flowing spins in the outer tubes, whereas our pTx solution achieves inversion in all tubes (Fig.5B). However, a slight variation was observed in the flow-encoding gradient moments required for full inversion amongst different tubes. This may be attributed to small differences in effective flowrate due to turbulence. Looking beyond the flow velocity selective inversion fidelity, we observe that the combination of k_T-points and spokes pulses produces a much more uniform magnetisation throughout the slice, which will translate into a higher SNR in areas where the CP-mode produces a low B₁.

Although effective, the adiabatic refocusing pulses are SAR intensive. Considering that 7 Tesla is relatively SAR constrained, it would be preferable to also replace these with large-tip-angle pTx pulses¹². Moreover, this can potentially reduce the refocusing pulse durations and therefore shorten the minimum separation time between the encoding gradients. In turn, this will reduce sensitivity to B₀ inhomogeneity. The next stage would be to evaluate our pTx VSASL solution in vivo.

Conclusion

We have shown that pTx can significantly improve the SNR and contrast fidelity of VSASL at 7 Tesla, which could help make ASL more accessible and robust at ultra-high field.

Acknowledgements

This research was conducted by the Australian Research Council Training Centre for Innovation in Biomedical Imaging Technology (project number IC170100035) and funded by the Australian Government. The authors acknowledge the facilities of the National Imaging Facility at the Centre for Advanced Imaging.