Introduction

With the goal of improving sensitivity and contrast for MRI of human brain, we are currently developing transmit and receive RF hardware for the 11.7 T head-only MRI system being developed at our institution. As part of this, we are performing RF safety assessment of any novel hardware. A critical point in this assessment is the accuracy of the coil model used to simulate transmit B₁ and associated electrical field. In this work, we present electromagnetic field simulations, benchtop measurements and power tests performed with different coil-phantom setups to validate the model of a 500 MHz inductive birdcage transmit-receive head coil^1,2. Towards comprehensive safety analysis, we plan to extend our electromagnetic field simulations with realistic human brain models (virtual family³) at various position inside the coil.

Methods

A detailed electrical model of the coil was developed in FDTD-based EM simulation software (Remcom, XFdtd 7.9) as illustrated in Figure 1a and b. Optimization of the model and mesh were performed through several iterations based of bench measurements of scattering parameters using the physical hardware. The expected symmetry in the scattering parameter matrix of the coil model was tested first with a head-size sphere. After this first validation, simulations were run with a model of a head-shaped phantom filled with a polyvinylpyrrolidone-based solution⁴ having a conductivity and permittivity of average brain tissue at 500 MHz (ε=48.42 and σ=0.63 S/m, https://amri.ninds.nih.gov/phantomrecipe.html). A first experimental power validation of the model was performed with the quadrature ports of the coil connected to 2 independent Tx channels of a 11.7 T animal system (11.7T Bruker Avance Neo Console). Total average power at the coil input was 82 W. Based on simulated SAR maps, optical temperature probes were placed at three different locations inside the head phantom centered inside the coil. SAR at each location was estimated from the slope of the temperature curve and compared with the value obtained in simulation at that location. A second experimental power validation was performed using MRI thermometry at 3 T. In this setup, each port of the resonator was connected to a 50 W broadband RF power amplifier (LZY, 20-512 MHz, Mini-Circuits) through a 90° hybrid for quadrature excitation. The head resonator and phantom were located inside the scanner bore at isocenter. A sliding support allowed moving the coil out the integrated body coil to image the phantom for MR thermometry⁵. Gel capsules were placed at the edge of the resonator to determine phantom placement relative to the coil using imaging. Three oil tubes were placed around the phantom to obtain a reference for post-processing correction of frequency drifts. A baseline MRI acquisition was performed (1-min temporal, 4 mm spatial resolution) followed by cycles of 2- or 5-minute 500 MHz RF excitation of total average power of 85 W, interleaved with cool-down periods. MRI thermometry was performed in between and after heating cycles. Temperature was also monitored using 2 optical fiber probes. For this experiment, 2% agarose was added to the polyvinylpyrrolidone-based solution to reduce heat convection.

Results

Simulated and measured frequency responses (S-parameter measurement) were in good agreement. Figure 1c shows as example the coupling between couplers C1 and C3 for all resonance modes. Figure 1d shows the homogenous mode measured on the bench and simulation space with an RF probe located at the center of the resonator and only C1 and C3 exciting the linear mode. In the first power validation test, we were able to readily predict the location of hot and cold spots based on simulated SAR maps. Figure 2a shows simulated sagittal SAR maps and approximate locations of temperature probes inside phantom (Figure 2b). Figure 2c shows temperature curves for the 5-minute heating and cooling cycle. Figure 2d shows the initial slopes during the heating cycle used for SAR estimation. Measured and simulated heating patterns across multiple axial slices with 16 mm separation are shown in Figure 3a. Probe location could be assessed accurately, as they were visible in the MR images. Linear regression of temperature curves at 2 locations during the first part of an RF heating cycle is shown in Figure 3b.Measured and simulated SAR ratios at these probe locations, for both RF power tests, are shown in Figure 3c.

Discussion

Preliminary simulated and experimental data are presented for 500 MHz coil model validation. Model adjustment was done manually with values of lumped elements set within 10 % of values implemented in hardware. Simulated reflection coefficient amplitudes were within 8.5 % of measured values. Co-simulation methods could be implemented for finer model adjustment as previously presented by others^6-8. To determine absolute measured SAR, we will confirm PVP solution heat capacity estimation with measurements. MR thermometry data will be validated using an increased number of temperature probes. This validation workflow was designed based on the hardware frequency response and electrical field distribution across a head phantom. Currently, validation of the B₁ distribution is not possible since our 11.7 T system is not at field yet. Further workflow optimization will allow us to complete validation of this single channel transmitter and extend to validation of multi-channel transmitters.