Parallel transmission (pTx) enables excellent control of the magnetization to mitigate the transmit field inhomogeneity at ultra-high field (UHF) by means of advanced RF pulse design techniques^1-4. However, to date the implementation of these methods necessitates the knowledge of the transmit RF field (B₁) and the static field (ΔB₀) distributions whereas the pulse design combined with the measurement and processing of those maps can easily cumulate 15 minutes, which decreases by the same amount the time available for acquiring clinically relevant data. Several studies on the other hand have suggested that the B₁ distribution exhibits a reproducible pattern across different subjects^5-7. In this work, we investigate numerically and experimentally for brain imaging at 7T on an 8 channel pTx system, the design of k_T-points pulses⁴ qualified as universal, i.e. whose design does not involve the subject-specific field distributions, but which yet considerably improves performance compared to the standard circularly-polarized (CP) and RF shim modes.

A so-called field database, composed of representative ΔB₀ and B₁ maps, was acquired experimentally on $$$N_s=6$$$ subjects. The universal pulses, characterized by the k_T-points trajectory $$$k$$$ and the RF pulse coefficients $$$b$$$, were then calculated by minimizing the objective function:

$$f(b,k) = \text{max}_{1\le i\le N_s} \| \mathcal{A}_i(b,k) - \alpha_t \|_2,$$

where $$$\alpha_t$$$ denotes the target flip angle (FA) and $$$\mathcal{A}_i(b,k)$$$ denotes the FA map generated by the pulse $$$(b, k)$$$ on subject $$$i$$$. The simultaneous optimization of the RF coefficients and the blipped k-space trajectory was performed under explicit power and SAR constraints to satisfy hardware and patient safety limits^8,9, for one excitation (9°) and one inversion (180°) pulse. SAR matrices¹⁰ were obtained from electric field simulations with HFSS (Ansys, Canonsburg, PA, USA) on a generic head model and for the home-made pTx coil under study, and were then compressed with the virtual observation point technique¹¹. An overall SAR safety margin of 2.7 was applied to account for modelling errors, anatomic variability and uncertainties in the SAR monitoring hardware^12-14. The universal pulses, designed with the field database, were then blindly applied on 6 additional subjects by acquiring MPRAGE images (TR/TI=1.1/2.6 s, TE=3 ms, nominal FA=9^◦, resolution= 1×1×1 mm³, TA=9 min), without any field map measurement or further pulse optimizations. The excitation and inversion pulses were designed with 5/7 k_T-points and 0.7/4 ms respectively. For comparison, two additional MPRAGE acquisitions were performed, in one case with an adiabatic inversion and rectangular excitation pulses driven in CP mode, and in the other case with similar optimized k_T-points pulses as for the universal pulses but using this time the subject-specific ΔB₀ and B₁ maps. The RF shim mode, with an adiabatic parametrization for the inversion pulse, was also investigated numerically. Results were compared quantitatively by computing the FA normalized root mean square errors (NRMSEs) and qualitatively by visual inspection of the MPRAGE images.

The measured B₁ distributions of each transmit channel are shown in Fig. 1 for subjects #1 to #12. The magnitude of the complex correlation coefficient between the subject-specific B₁ distributions is 0.95 ± 0.02 for each transmit channel, suggesting that B₁ varies only slightly from subject to subject despite the variability of head shape and position. The FA NRMSE values are reported in Fig. 2 for the 6 subjects constituting the field database and on whom the universal pulses were designed, and for the 6 additional subjects on whom the universal pulses were experimentally tested. Both the excitation and inversion universal pulses significantly outperform the CP and RF shim modes. Not surprisingly, the subject-based tailored pulses perform better than the universal pulses, the gain however being more marginal. The MPRAGE images obtained experimentally on two subjects are provided in Fig. 3. With the coil driven in CP mode, two problematic regions can be identified which are completely recovered with the use of the universal pulses. There is moreover no obvious degradation in image quality compared to the MPRAGE image acquired with the subject-based tailored pulses, despite the slightly poorer NRMSE. The same behavior was observed on all tested subjects #7-12. In that sense, the test of the universal pulses never failed.A method to mitigate the RF field inhomogeneity problem at UHF, without subject-specific B₁ and ΔB₀ map measurements and online pulse design, was presented. The proof of concept was validated at 7 T with 3D brain imaging. Preliminary results suggest that the idea could be successfully applied in 2D. The results suggest that the use of parallel transmission thereby could become completely transparent to the user, hence making this technology more accessible for routine use.