Improving peak local SAR prediction in parallel transmit using in situ S-matrix measurements

Purpose: Peak local specific absorption rate (pSAR10g) is an important parameter used to determine patient safety during radiofrequency transmission. pSAR10g predictions for parallel transmit MRI are affected by the level of coupling exhibited by a modeled array in the simulation environment. However, simulated array coupling is rarely equal to the physical array coupling. Accurately simulating the physical array coupling may improve the accuracy of predicted SAR levels. Methods: The scattering parameter matrix (S-matrix) of a prototype 4-channel array was measured in situ using directional couplers installed on a 7T scanner. Agreement between the simulated and measured S-matrix was achieved by using network co-simulation with a modified cost function. B1+ maps acquired in a phantom were compared to B1+ distributions determined from simulations. Results: The modified co-simulation technique forces the simulations to have S-matrices similar to the measured values. A comparison of realistically versus ideally simulated coupling conditions shows that ideally simulated coupling can result in large (> 40%) error in pSAR10g predictions, even when the array is reasonably tuned. The simulated B1+ distributions match the measured B1+ distributions better when the coupling is accurately simulated. Conclusion: Considering the measured array coupling matrix in numerical simulations eliminates a potential confound in pSAR10g prediction. Magn Reson Med 77:2040–2047, 2017.