Introduction

Radiofrequency (RF) heating is a serious safety concern for patients with deep brain stimulation (DBS) implants during MRI scans¹. A reliable and accurate heating prediction (pre-scan) is essential to assess the safety of patients with DBS implants². Sadeghi-Tarakameh et al³ proposed a workflow where the heating around the electrical contacts of DBS electrodes was predicted using an MR-based current measurement⁴ method and quasi-static simulations³ (electromagnetic and thermal). In this study, we validated the workflow by predicting and measuring the heating around DBS contacts in swine, in-vivo and post-sacrifice. To generate RF heating and conduct imaging in the swine brain, a toroidal transmit-receive coil⁵ is used.

Method

The experimental setups for calibrating electrode transimpedance, a parameter vital for temperature prediction, and heating experiments are shown in Figure 1. The proximal portion of a DBS (model 6173, Abbott Laboratories, Chicago, IL) was inserted through the cavity of a toroidal transceiver for imaging in a 3T Magnetom Prisma scanner (Siemens Healthineers, Erlangen, Germany). The electrode was exposed to a 3D-TSE sequence (TR/TE = 300/3.06 ms, echo train length = 15, FA = 150°, in-plane resolution = 0.5 mm, slice thickness = 1.25 mm) and fluoroptic temperature probes (Lumasense Technologies, CA, USA) measured the heating near the most distal contact.

The equivalent transimpedance, R_DBS, was calculated as the ratio of the voltage generated at the electrical contacts, V_c, and the induced current along the electrode shaft, I_s. V_c was estimated in quasi-static EM and thermal simulations by determining the voltage boundary condition that simulated a heating curve matching the measured heating³. A new current measurement method was designed to measure I_s. When the excitation voltage of a 2D-GRE sequence (TR/TE = 300/103 ms, in-plane resolution = 0.34 – 0.5 mm) surpasses the reference voltage, a circular dark band appears around the DBS electrode. This artifact corresponds to spins experiencing a 180° flip angle. The B1+ experienced by these spins is expressed as
$$B_{1,perVolt}^{+}=\frac{1}{{2}\gamma\cdot{V_{excit}}\int_{0}^{t_p}rf(t)dt}$$
where V_exc is the excitation voltage, rf(t) denotes the time-dependent variation of the RF pulse, and t_p represents the pulse duration of the 2D-GRE sequence. Figure 2 shows the linear relationship between V_exc and r_{dark band}. The current along the shaft can be calculated as follows:
$$I_s=\frac{B_{1,perVolt}^{+}\cdot4\pi{r_{darkband}}}{\mu}\cdot{V_{sequence}}$$
where r_{dark band} is the radius of the dark band and V_sequence is the root-mean-square (RMS) voltage applied during the heating MRI sequence.

R_DBS for average swine brain tissue was calibrated before the swine experiments, in a rectangular phantom filled with the following homogenous mixtures of different conductivity and permittivity values:
1) A hydroxyethylcellulose (HEC) gel (14 g/L), CuSO4 (0.25 g/L), and varying NaCl concentration.
2) A polyvinylpyrrolidone (PVP) gel (920 g/L), NaCl (40 g/L) and 0.5% agarose concentration.
The values similar to the dielectric properties of grey and white matter were used in the calibration⁶. Table 1 summarizes the dielectric and thermal properties used in all experiments^6-7. Two temperature probes were used for calibration in HEC. Conversely, a single temperature probe was used for calibration in PVP but the DBS insertion length was varied to investigate its effect on R_DBS.

All swine experiments were compliant with a UMN IACUC-approved protocol. A 4-5 cm distal portion of the DBS electrode and a fluoroptic temperature probe were surgically implanted into the swine brain after anesthesia. Four in-vivo heating experiments were conducted with two animals, and two heating experiments were conducted post-sacrifice with one animal. The V_c for each heating experiment was calculated by multiplying the measured I_s with the calibrated R_DBS and used as an input to the EM and thermal simulations to predict heating around the contacts.

Results

Figure 3 shows the temperature matching obtained with R_DBS values calibrated for different dielectric properties. R_DBS for swine brain tissue was calculated as 161 Ω using linear interpolation. Notably, Figure 3 also shows the minimal variation RDBS (< 3%) with respect to DBS insertion length in the toroidal transceiver. Figure 4 shows the good agreement between the heating predictions and measurements in swine with NRMSE ≤ 0.09.

Discussion

Calibrating R_DBS for average swine brain tissue demonstrated robust and accurate heating predictions that agreed with measured heating around DBS contacts for both in-vivo and post-sacrifice studies. In the future, comprehensive sensitivity analyses will be conducted to quantify a safety margin that accounts for different uncertainties in temperature prediction.

Conclusion

A previously proposed temperature prediction workflow was validated with swine experiments. The equivalent transimpedance was calibrated for dielectric properties approximating the average swine brain tissue. The interpolated transimpedance resulted in heating predictions that agree with the measured heating with ≤ 0.09 NRMSE.

Acknowledgements

This work was supported by NIBIB P41 EB027061, S10OD017974-01, and NINDS R01NS115180. We extend sincere appreciation to Lindsay Knoll, Rhianna Golden, Shannon Wilks, Dr. Whitney McGee, and Dr. Giuseppe Dell'Anna of Research Animal Resources (RAR) for their support and assistance during the swine experiments.