In this study we compared image quality obtained with sets of pre-calculated spatially selective RF pulses used at three 7T sites equipped with 8-channel pTX systems and between 2 different RF coils to assess reproducibility and robustness of such RF pulses.

All three systems were from the same vendor (Siemens Healthcare GmbH, Germany) and equipped with an 8-channel pTx system. However, basic imaging components like type of magnet and gradient coils differed between the systems, which might influence image quality and timing in pTx (e.g. gradient delay). For the measurement setup two different types of RF coils were used. Two sites (2, 3) were equipped with a commercially available 8ch transmit 32ch receive head coil (Nova Medical). An in-house-built 8ch transceive head coil was transported between the sites and used to compare data acquired at each site (1, 2, 3). The 8ch/32ch coils located at Sites 2 and 3 were used to check reproducibility on different RF coils and systems. All data were acquired with a vendor-provided spherical phantom (diameter: 165mm, T1/T2: 1150/750 ms) filled with polydimethylsiloxan oil. The imaging protocol consisted of sequences to check coil and system performance first. The receive (Rx) performance was evaluated with standard GRE sequences. The SNR of a central ROI was compared in the phantom and the noise correlation matrix was calculated. The transmit (Tx) sensitivity was measured with B1 maps obtained with a vendor-provided saturation-based turboFLASH sequence. Three different excitation patterns (Fig.1) implemented in a standard pTx FLASH sequence (TA = 6.4 s, nominal FA = 10°, TE = 15.8 ms, TR = 100 ms, single slice with thickness 5 mm) were employed. The patterns were used to check for positioning, adjust gradient delay, and finally to measure a special QA image pattern at every site consisting of triangles (FIG 1).

The different patterns were pre-calculated once for all sites using the B1 maps measured at Site 1 with the transceiver coil. Additionally, for the 8ch/32ch coil another set of pulses was calculated from B1 maps obtained at Site 2 and played out at both sites with this type of coil (2, 3). The 8ch/32ch pulses were also played out at Site 1 with the transceiver coil to check reproducibility with a different coil design. A vendor-provided Matlab script (The MathWorks, Inc) was used to calculate the RF pulses (2D spiral with variable geometry, pulse length: 19 - 20 ms, maximum voltage: 6-13V). Two measurements were performed: one by applying the manufacturer’s specified B0 shim (“Tune up”) and another one after 2nd order B0 shimming. The resultant images were compared regarding triangle position and rotation, as well as SNR, background suppression, and homogeneity with the help of a Matlab-based analysis tool¹.

The transceiver coil had similar Rx and Tx sensitivity at the different sites. The two different 8ch/32ch coils also showed high agreement, with 5% difference in SNR. The flip angle variation of the individual coil elements measured in a central ROI was less than 2% between the sites and coils (Fig.2). Target patterns played out at the different sites with the same coil showed strong correspondence (Fig.3). A good B0 shim was critical for the reproducibility of the patterns as shown in Fig. 4. QA patterns were successfully generated with the 8ch transceiver coil of different type after correction for the channel assignment to match those of the 8ch/32ch coil (Fig.5). All B0-shimmed triangles were analyzed and successfully detected by the QA software tool. QA pattern analysis revealed small differences between the coils and the sites. SNR of all triangles was 12% lower for a pulse calculated at Site 2 and played out at Site 3 with a different RF coil of the same type. The central region of the test pattern had the highest agreement. At the edges of the phantom small differences were observed. For individual triangles where B0 shim or B1 sensitivity was different, the SNR of the triangles dropped by 50% and homogeneity was approximately 20% lower.Sharing arbitrary pulses between sites was possible for QA purposes. We showed that even different RF coil types at different sites can be used to produce a highly defined QA pattern with high image quality. As pulses are calculated considering contributions from B0 field differences, a high field homogeneity has to be provided at all sites to avoid blurring and signal drops. In subsequent studies, reproducibility for even more strongly varying RF hardware, for more realistic tissue-simulating phantoms, and for in vivo imaging should be evaluated.