Improved Radiofrequency Safety Modelling in MRI Using In Vivo Measurements of Brain Conductivity.

Abstract

Tissue electrical properties are required by electromagnetic simulation software to conduct radiofrequency (RF) safety studies. Values are commonly taken from existing databases of tissue properties, where brain conductivity was measured ex vivo. We hypothesize that using in vivo brain conductivity values, as reported in the recent literature on in vivo MRI measurements, can improve the accuracy of such simulations. Sixteen subjects were scanned at 3T to obtain experimental maps of the transmit RF field, $B_1^{+}$ , of the head. Electromagnetic simulations were performed using biomodels with varying morphologies, using both a conventional ( ${\sigma }_{\mathrm{ex}\text{vivo}}=0.46S/m$ ) and a modified brain conductivity ( ${\sigma }_{\text{in vivo}}=0.70S/m$ ). A framework was developed to process the simulated $B_1^{+}$ fields, including a systematic method to combine the two excitation ports of the transmit coil model (accounting for different load impedances between simulation and experiment), geometric alignment, and scaling of the simulated fields, allowing a quantitative comparison of complex $B_1^{+}$ maps of the brain with experimental maps. Specific absorption rate (SAR) maps were also estimated by different methods, including a novel $B_1^{+}$ -derived formula. Normalized root-mean-squared errors between simulation and experiment, in the brain, were approximately 6% for $B_1^{+}$ magnitude. While head geometry mainly impacted the accuracy of $B_1^{+}$ magnitude, with errors up to 11% (p = 0.007) between the best fitting human model and the worst one, brain electrical conductivity mainly impacted $B_1^{+}$ phase, with errors reduced by 48% when using ${\sigma }_{\text{in vivo}}$ instead of ${\sigma }_{\mathrm{ex}\text{vivo}}$ (p = 0.0004). The agreement in average brain SAR, between simulation and experiment, was also improved, with differences reduced from 62% ( ${\sigma }_{\mathrm{ex}\text{vivo}}$ ) to 22% ( ${\sigma }_{\text{in vivo}}$ ). Using brain conductivity values from recent in vivo studies improves the accuracy of RF safety modelling. TRIAL REGISTRATION: ClinicalTrials.gov identifier: NCT04645628.