Radiofrequency shimming with multiple channel excitation is a well established method to increase the transverse magnetic field homogeneity at high magnetic field strength(≥7T)(1). To harness the benefits of a parallel transmit system, the magnitude and relative phase of each transmit channel must be determined during a calibration scan, this is achieved by acquiring images with each transmit channel individually.

In cardiac and abdominal applications, sufficiently high resolution is required in order to avoid partial volume effects, particularly when the proton density images are used as a surrogate for absolute B₁⁺(2). For this purpose, we propose a new strategy to accelerate the acquisition of such calibration images. It is hypothesised that simultaneously exciting several transmit channels and reconstructing the calibration images using the technique similar to simultaneous multi-slice acquisitions(3) with controlled aliasing in order to increase the efficiency of the reconstruction and reduce the geometry factors of the SENSE reconstruction(4).

Measurements were performed on a whole-body Magnetom 7T MR-system (Siemens Healthcare, Erlangen, Germany). Data acquisition was performed with a custom built eight-channel transmit/receive TEM coil (5). Images were acquired using a dual-echo GRE sequence (TR=7.7ms,TE_1,2=1.02,2.04ms, α=14°).

Acquisition with multiple excitation channels and the controlled aliasing was simulated offline using the raw kspace data with acceleration factors of two and four (simultaneous excitation with two and four channels respectively). Controlled aliasing with a shift of FOV/2 was used with an acceleration factor of two by multiplying each even line with a factor of pi and shifts of FOV/4,FOV/2 and 3*FOV/4 with an acceleration factor of four.

Images corresponding to each transmit channel were calculated using the SENSE reconstruction. Coil sensitivities(B₁^-) were estimated using 16 phase encoding lines acquired in the centre of kspace with single channel excitation.

The images acquired with two different echo times were combined using the following formulation

$$I_combined =sqrt(M_TE1·M_TE2)·exp(j·(𝛟_TE2- 𝛟_TE1)) (Eq.1)$$

The magnitude of coil sensitivities were calculated by dividing each of the magnitude images by the sum of squares image. Furthermore, as B₁^- coil sensitivities are independent of the transmit channel, they can be calculated using a weighted mean of B₁^- for each transmit channel.

In absence of noise, the phase difference 𝛟_TE2- 𝛟_TE1(Eq.1) is the same for all receive/excitation channels and is proportional to the main magnetic field B₀ inhomogeneities. The phase difference mentioned previously was calculated using the complex sum over all transmit and receive channels. Regions with higher signal magnitude will contribute more substantially to the final phase difference than the regions with low signal magnitude. A complex image corresponding to TE=0 was calculated using the following relationship.

$$I(TE=0)_m+,n-=[M(TE_1,2)_m+,n-·exp(j𝛟(TE_1,2)_m+,n-)]·exp[-j·(𝛟_TE2- 𝛟_TE1)·TE_1,2/(TE2-TE1)] (Eq.2)$$

The phase image obtained using Eq.2 only has contributions due to the transmit channel m (C_m⁺) and reception channel n (C_n^-). For the receive sensitivity maps the phase information of the transmit sensitivities is removed using one receive channel as a reference by subtracting its phase from the other receive channels (as such the phase of each receive channel is C_n^- - C_ref^-)

For accurate RF shimming, accurate magnitude images and relative phase between channels is required, consequently, the relative phase of the transmit channels will not be affected if the phase information of the receive reference channel is present in all reconstructed images using SENSE.

Fig.1 summarizes the reconstructed transmit images obtained for a full acquisition, acceleration factors of two, and of four respectively. It can be observed that the phase difference between the full acquisition and the accelerated acquisition is very accurate especially in regions with high B₁⁺ values. In regions with lower B₁⁺ phase difference between the full acquisition and the accelerated is mainly dominated by noise and no systematic error was observed. Four representative images of transmit channels acquired with two fold acceleration are shown in fig.2.

In order to estimate the accuracy of the proposed method the simulated accelerated acquisition was compared to the fully encoded acquisition. The ratio between the magnitude images for the accelerated case and the full sampling was calculated for all acquired slices. The histogram of the ratio of magnitude images suggests that the method maintained the accuracy of the maps for acceleration factors up to four while reducing the scan time.