SAR/B1+ calibration workflow for safe, high duty-cycle parallel transmission imaging at ultra-high field
Filiz Yetisir1, Bastien Guerin2, Lawrence Wald2,3, and Elfar Adalsteinsson1,3

1Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, 2Dept.of Radiology, Martinos Center for Biomedical Imaging, Charlestown, MA, United States, 3Harvard-MIT Division of Health Sciences Technology, Institute of Medical Engineering and Science, Cambridge, MA, United States

Synopsis

In this work, we propose a pTX safety workflow that will enable high duty cycle imaging at high field systems. Several SAR and B1+ calibration steps are suggested for a complete analysis including modeling the TX array, testing it over time and different loads and finding a safety margin to account for RF system imperfections. Good qualitative agreement was achieved between the simulated and measured B1+ maps for the TX array. 11% and 6° standard deviation was observed in the magnitude and the relative phase maps over time. A maximum difference of 16% was observed between offline and online calculated local SAR values due to RF system imperfections.

Target Audience

MR physicists, ultra-high field & parallel transmission users.

Purpose

Use of parallel transmit (pTx) requires careful monitoring of the RF waveform as well as global and local SAR, especially at high fields and high duty-cycles. Commonly, electromagnetic simulations of the TX arrays have been used to calculate local SAR [1,2,3]. Comprehensive RF safety concepts were proposed using power calibration matrices [3] or active decoupling for load compensation of the TX array [4]. In this work we propose a pTx workflow that checks B1+ and SAR variations against the normal test-retest variability of the pTx system to guarantee safe pTx imaging even at high duty-cycle.

Methods

The proposed workflow is summarized in Fig.1. Experiments were done on a 7T Siemens scanner. TX array model: First, we model our 8 channel transceiver array in the EM field simulator HFSS (Fig.2). The model of the array is loaded with a spherical phantom and it is tuned and matched to approximately the measured S-parameters of the physical array (Sii < -20 dB). Experimental and simulated transmit fields are then compared and global scale factors (GSF) as well as global phase offsets (GPO) are determined for each channel to account for coil losses and uncalibrated phase paths that are not modeled in the HFSS model, such as the interface box between the coil and the RF power amplifiers (Steps 1-3). The remaining mismatch in the B1+ maps after the application of GSFs and GPOs can be translated into a safety margin on the SAR limit using [5]. The B1+ maps are measured at different times to ensure stability of the array over time (test-retest over time, Step 4). Load sensitivity: Next, the array is loaded with another phantom that is significantly different in dimensions and shape than the sphere (elliptical phantom in Fig.2). Then GSFs and GPOs of the sphere are used in comparing the simulated and measured B1+ maps of the new phantom (Steps 5-7). If the difference between those is smaller than the test-retest variability of the B1+ maps, the workflow continues otherwise it stops. Patient body variability: Because patient-specific body models are currently impractical to use, various body models in various positions are simulated to generate different sets of VOPs [6] which then can be conservatively used to monitor SAR (Step 8). System imperfections: The SAR calculated online using measured RF waveforms and the SAR calculated offline using demanded RF waveforms are then compared for various random complex RF pulses and a statistical safety margin to account for RF system imperfections is determined (Step 9). Before each pTX session: the GSFs and GPOs of the sphere phantom should be compared to the stored values to detect unexpected changes in the TX array or the scanner hardware. (Steps 10,11).

Results/Discussion

Fig.2 shows good qualitative correlation between the magnitude and the relative phases of the simulated and experimental B1+ maps for all channels. The standard deviations of the GSFs and the GPOs are shown in Table.1. The GSFs are highly sensitive to the position of the phantom with respect to the array (up to 11% std), hence to check for their stability over time it is important to place the phantom at the same position every time (See Fig.2 for a 3D holder to be used for future studies). The fact that the actual versus the demanded RF waveforms are different lead to a maximum of 16% difference between the offline and the online calculated 10s average local SAR values for our system (Fig.3). A similar analysis can be carried out for 6 min. average local SAR and a corresponding safety factor can be applied to the SAR limit in order to have an uninterrupted scan. Future work includes the implementation of the load sensitivity steps. Furthermore, temperature measurements [1] or MR thermometry [7] can be used as an extra validation step for the TX array model.

Acknowledgements

This work was supported by R01 EB017337, R01 EB006847, P41EB015896 and Siemens MR. Also Hans-Peter Fautz and Josef Pfeuffer are acknowledged.

References

[1] Cloos et al., ISMRM 2010;p3871 [2] Deniz et al., ISMRM 2014;p4929 [3] Deniz et al., MRM 2015;00:00-00 [4] Graesslin et al., MRM 2015;74:589-598 [5] Guillaume et al., ISMRM 2015;p1862 [6] Eichfelder et al. MRM 2011;66:1468-1476 [7] Deniz et al., ISMRM 2014;p4894

Figures

Fig.1: Summary of the proposed pTX safety and calibration workflow. Steps 1-9 are to be carried out only once for the TX array, steps 10-11 are performed before every patient imaging session.

Fig.2: (Top) 8 channel pTX coil, coil model loaded with spherical and elliptical phantoms. (Bottom) Simulated vs experimental B1+ maps inside the sphere. Global scale factors and phase offsets per channel are applied to simulated fields to account for losses and uncalibrated phase paths.

Table.1: Standard deviations of the GSFs and GPOs of the sphere phantom over time.

Fig.3: Local SAR values calculated with the designed and measured RF waveforms. A maximum difference of 16% has been observed due to system imperfections.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2211