Calibration of multichannel transmit coils to reduce radio frequency (RF) field non-uniformities requires accurate knowledge of the transmit B₁ maps of each transmit channel. Existing B₁ mapping methods suffer from a variety of issues ranging from long scan times, limited B₁ dynamic range, or high specific absorption rate (SAR), and will scale poorly to high channel counts. Moreover, experimental B₁ mapping methods do not take into account the known coil geometries and orientation which can be used to aid B₁ estimation. The idea here is to generate “good enough” B₁ predictions, without mapping. In this study we apply knowledge of coil RF current and orientation relative to the sample to estimate B₁ maps of each transmit channel for rapid pre-scan calibration of parallel transmit systems.

The experiment was performed using a custom 1.5T parallel transmit and receive system, shown in Fig.1, built around a Medusa MRI console, and a cylindrical four-channel transmit/receive coil array. The system includes optically coupled RF current sensors on each coil and a 4-ch 4P4T switch matrix to route current sensor, forward voltage, or reverse voltage to the digitizers during transmit, and the preamp signal during receive. To calculate RF shim complex coefficients specific to each individual channel, the transmit array was localized by placing three fiducial markers on the conductor edge of one of the coils¹(Fig.2), and then an estimated B₁ map normalized to unit current for each coil was calculated from SEMCAD FDTD data using the fiducial projection co-registration method². Next, amplifier non-linearity was corrected by calculating look up tables using each directional coupler forward power signal. To relate the desired RF currents to the excitation vectors, coil coupling must be numerically nulled. The decoupling matrix³ was calculated using the current sensor outputs on each coil⁴. The co-registration method also requires the RF current sensor inputs to properly scale the FDTD result to the physical system. To validate estimated RF maps, Bloch-Siegert maps of each transmit coil with a mineral oil phantom were acquired in an axial slice and compared with FDTD-based simulated transverse maps² scaled by the experimental to simulated current ratio for an arbitrary orientation of coil relative to the target plane in imaging space. The Bloch-Siegert mapping sequence was performed with following image parameters: TR= 100 ms, TE = 5ms, matrix size = 128 × 128, field of view = 25 cm, and slice thickness = 6mm; the 4ms slice-select pulse was followed by a 6-ms Fermi pulse with 4 kHz off resonance.

In separate tests, RF shim complex weights were calculated for exciting an axial slice of a phantom filled with doped water with 1mM MnCl2. Imaging parameters here were: TR = 100 ms, TE = 5.4 ms, matrix size = 256 × 256, field of view = 25 cm, and slice thickness = 6mm.

Figure 3 shows the magnitude of experimental Bloch-Siegert maps compared with the simulated B₁⁺ maps for all 4 channels for arbitrary orientation of the mineral oil phantom with respect to the transmit coils. The simulated predictive maps show reasonable agreement with experimental maps in the regions of higher SNR.

Figure 4 shows the feasibility of using the prediction B₁⁺ maps in RF shimming. Figure 4a is the intensity image of axial slice of doped water phantom. The receive sensitivities were removed in the reconstruction step. The complex coefficients for the excitation pulses were computed by the proposed method. As is shown the signal intensity is uniform. Figure 4b shows the normalized estimated B₁⁺ maps that were used in this experiment to design RF pulse excitation vectors.

These feasibility tests demonstrate that estimated B₁ maps obtained by the coil locator technique combined with coil current sensor data can be used in computing RF pulse vector weights for RF shimming applications. The physical inputs and decoupling control provide the information needed to co-register and scale FDTD predictions to the physical system. The results shown here are more correctly incident field results as the phantom loading caused minor field scattering. This would suffice at 1.5T for use in interventional transmit arrays applications. In future, for more general use, one could apply a pre-computed library to estimate patient perturbations for “good enough” predictions. Second, a better matching of SAR models with patient-coil proximity would be enabled. Finally, the proposed method can be used in rapid auto-calibration of parallel transmit systems and computation of first-guess RF shim weights.