Read My Leads: Subject-Specific RF Hazard Assessment and Mitigation for DBS Implants in MRI

Magnetic Resonance in Medicine RESEARCH ARTICLE OPEN ACCESS Read My Leads: Subject-Specific RF Hazard Assessment and Mitigation for DBS Implants in MRI Berk Silemek1| Frank Seifert1| Mevlüt Yalaz2| Michael Höft2| Günther Deuschl3| Arzu Has Silemek4| Rüdiger Brühl1| Bernd Ittermann1| Lukas Winter1 1Physikalisch-Technische Bundesanstalt (PTB), Braunschweig and Berlin, Germany | 2Department of Electrical and Information Engineering, Christian-Albrechts-Universität zu Kiel, Kiel, Germany | 3Department of Neurology, Christian-Albrechts-Universität zu Kiel, Kiel, Germany | 4Department of Neurology, Cedars-Sinai Medical Center, Los Angeles, California, USA Correspondence: Lukas Winter ([email protected]) Received: 20 June 2025 | Revised: 8 October 2025 | Accepted: 5 November 2025 Funding: This work was supported by the European Association of National Metrology Institutes (European Partnership on Metrology, co-financed by the European Union’s Horizon Europe Research and Innovation Programme and by the Participating States), 21NRM05 STASIS; This work was also partially supported by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) through the project T1 of the Collaborative Research Centre CRC 1261. Keywords: deep brain stimulation | impedance | implants | MR safety | RF heating | sensors ABSTRACT Purpose: To develop and validate a framework for personalized, implant-specific MRI safety assessments using feedback from commercial deep brain stimulation (DBS) systems. To further use this framework to suppress RF-induced heating with minimum compromise in imaging performance. Methods: Two off-the-shelf DBS implantable pulse generators and a commercial 8-electrode DBS lead were utilized for quantitative safety assessments. In controlled phantom experiments, (i) RF-induced voltages on the DBS lead, and (ii) temperature-dependent admittance/impedance changes in the tissue surrounding the lead’s electrodes were quantified. This information was used to suppress implant-related RF heating by calculating implant-friendly imaging modes. Experimental conditions included excitations with different RF transmit coils (8-channel 3 T and 7 T head coils, 2-channel 3 T body coil), over 1000 different exposure scenarios, different implant configurations, and the use of external reference probes (𝐸-field and temperature) for validation. Imaging performance of the applied implant-friendly mode was demonstrated in vivo on a 3 T scanner. Results: 𝐸-fields and temperature rises around the tip electrodes could be robustly detected directly from the DBS lead. Both signals quantify the momentary patient hazard. Utilizing these measurements–recorded and wirelessly transmitted by the DBS system–tissue heating was reduced up to 99% for the same transmission power with comparable imaging performance to a conventional imaging mode. Conclusion: All the information needed for full in situ control of implant heating in MRI can be read directly from the DBS device. This approach would improve both patient safety and image quality while simultaneously reducing workload and responsibilities of the clinical personnel. Please note that parts of the work have been presented at the 2024 and 2025 of Annual Meeting of the International Society for Magnetic Resonance in Medicine. ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2025 The Author(s). Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine. Magnetic Resonance in Medicine, 2025; 0:1–15 1 https://doi.org/10.1002/mrm.70186 1|Introduction Deep brain stimulation (DBS) [1, 2] is an essential medical treatment, increasingly employed to restore physiological function and manage debilitating neurological disorders. 56% of DBS indicated patients are estimated to need an MRI exam within 5 years of implantation [3]. From an MRI perspective, this is worrying [4] since the long metallic leads of DBS implants and many other active implantable medical devices (AIMD) pick up the RF fields from the MRI scanner and emit a “scattered” electric (𝐸)-field at the lead’s tip electrodes in brain tissue which can exceed the “background” 𝐸-field from the RF coil by orders of magnitude [5, 6]. Patients have been severely impaired by brain tissue burns near the electrodes [7–11], which triggered intensive research in the physics of implant heating and possible mitigation strategies [12–34]. MR compatibility, technically termed conditionality [35], of implants was pioneered for cardiac pacemakers [36], when the US Food and Drug Administration judged the case pressing enough to warrant a fast clinical approval for these devices [37]. Current MR conditionality assessments by implant manufacturers rely on hazard predictions from ex ante numerical simulations for a large variety of possible exposure scenarios, including varying anatomies, implant-lead trajectories, and patient positions [38]. One manufacturer reports the simulation of 38 000 scan conditions and 10 million simulated patient scans [39]. Nevertheless, in clinical practice a significant curtailing of the permitted RF power is often imposed, either as a SAR or 𝐵+ 1,rms restriction [11], thus trading image quality for patient safety. Not only patients suffer from this situation, but also hospitals and their personnel since they have the sole responsibility and liability for implant safety. As a consequence, long and complex pre-scan assessments need to be performed [40, 41] leading to substantial organizational challenges and costs [4] and significant delays and denials of MRI scans for implant carriers [3, 4, 40, 42, 43]. Here, we propose an alternative approach toward safe MRI scanning of patients carrying AIMD with long metallic leads: personalized measurements rather than generalized simulations. In previous work, it was investigated how small and inexpensive physical sensors attached to the implant tip can be utilized to detect and mitigate RF-induced heating [44–46]. However, this required modifications to the implant lead and its electrodes, functionally the most critical area, which manufacturers are reluctant to implement. Now, we demonstrate that unmodified, commercial DBS implant components have the capability already built in to detect and quantify the momentary RF hazard: the implant itself is the sensor we need and even the electronics to transmit this information to an external receiver are already integrated [47, 48]. Phantom experiments were performed on a commercial 8-electrode directional DBS lead and two different commercial implantable pulse generators (IPG) that were exposed to over 1000 different RF field configurations from a 3 T and a 7 T 8-channel parallel transmit (pTx) head coils. External 𝐸-field and temperature probes near the electrodes provided independent reference measures to confirm our approach. First, this work explores how the implant itself can be used to detect and quantify the momentary RF hazard for the patient. Next, it is investigated how such information can be exploited to control and mitigate the hazard. Finally, the practical feasibility of the mitigation approach is demonstrated in MRI experiments on a commercial 3 T MR scanner. 2|Methods Typical for modern DBS leads, the investigated commercial 8-electrode implantable lead (Cartesia, Boston Scientific Corporation), as illustrated in Figure 1a, consists of eight conductive wires, carrying stimulation pulses from the terminal contacts T1–T8 at the IPG (Figure 1b) to the electrodes E1–E8 over a lead extension connector (Figure 1c) to the tip end in tissue. During an MRI experiment, the scanner transmits an RF field which is collected and concentrated by the wires in the lead (“antenna effect”). The scattered 𝐸-field (𝐸𝑆)emanating from the electrodes can then cause excessive heating in the surrounding tissue. This constitutes the patient hazard, and the task is to quantify it in order to control it. Previous simulation work has shown that both 𝐸𝑆(cause) and the temperature rise (consequence) in the vicinity of the lead electrodes are suitable and largely equivalent metrics for this purpose if they can be detected in situ [46]. The implant lead and interface were immersed in an American Society of Testing Materials (ASTM)-phantom container [50] filled with polyvinylpyrrolidone (PVP) solution, with dielectric parameters (𝜀𝑟=50 and =0.33 S/m at 128 MHz) closely matching the respective values for white brain matter [51, 52]. The phantom’s head section was surrounded by either a commercial 3 T (RAPID Biomedical) or an open-source 7 T [53] MRI head coil, each with eight transmit channels (see Figure 2a). The coil in use was hooked to an implant safety testbed [44] which allows to generate eight phase-locked, continuous wave RF signals with freely selectable amplitudes and phases and up to 20W per channel. 2.1 |Detection of the Scattered E-Field Around the Electrodes: The Auto-Sense Signal The scattered 𝐸-field around the lead’s tip generates a net RF voltage between any two lead electrodes. This voltage can be detected and, crucially, this can be done “remotely” at the terminal contacts in the pulse generator since all fields induced along the lead are nearly identical in all eight wires and cancel out very effectively. This RF voltage between two electrodes is henceforth referred to as the “auto-sense” signal (𝑈AS).To detect 𝑈AS, which is not yet available from the commercial IPG, the proximal end of the lead was connected to a home-built interface to read the RF voltage between electrodes. This signal was fed to the receiver of the testbed which provided the common time base for all experiments as well as synchronized inputchannelstorecordand digitize externalprobesfor reference measurements. 𝑈AS is recorded for 1000 random RF excitation vectors (pulse lengths 100 μs), each generating a unique electromagnetic field distribution inside the phantom. Simultaneously, 𝐸𝑆-field components normal to the electrodes were measured in the phantom liquid in ∼2 mm distance from the lead by an external time-domain 𝐸-field probe (E1TDSz, SPEAG) as illustrated in Figure 2b. 2Magnetic Resonance in Medicine, 2025 15222594, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mrm.70186 by Berk Silemek , Wiley Online Library on [25/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License FIGURE 1 |(a) Schematic and photograph of the tip end of the deep brain stimulator (DBS) lead carrying eight electrodes (E1–E8). (b) Photograph of the DBS system in a saline-filled anthropomorphic phantom. The implant was realistically placed inside the phantom by an experienced DBS neurosurgeon [49]. (c) Photograph of the DBS implants’ components used in the experiments including the lead extension connector and external temperature sensors. FIGURE 2 |Experimental setup for the test bench experiments. (a) Polyvinylpyrrolidone (PVP) filled ASTM phantom with its head section inside the 8-channel RF coil. (b) Side view of the implant lead with an E-field probe at its tip (blue, coming from the top) and two fiber Bragg grating temperature probes all located within the RF coil and ASTM phantom. (c) DBS implant and electrode configuration with a detailed placement of the fiber Bragg grating (FBG) temperature sensors, the arrows point their sensitive location which is not at the tip. For the more time-consuming heating experiments, 21 of the previously used 1000 RF excitation vectors were selected, covering the full range of 𝐸Svalues. The temperature rise in PVP during 60 s of RF exposure was measured using two high-resolution (∼2 mK) fiber Bragg grating (FBG) temperature sensors (FBG-T8, imc test & measurement GmbH) with the sensitive spot in about 1 mm distance from electrode E1 (see Figure 2c). Magnetic Resonance in Medicine, 2025 3 15222594, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mrm.70186 by Berk Silemek , Wiley Online Library on [25/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 2.2 |Detection of the Temperature Around the Electrodes via Impedance Measurements For a patient specific calibration of their stimulation pulses, modern DBS systems can measure the tissue impedance between any pair of electrodes on the fully implanted device [54–56]. This occurs via an external “programming unit” (see Figure 2a), provided by the manufacturer, and uses a wireless communication link to the IPG. This built-in functionality of the off-the-shelf device was exploited to detect temperature changes in the tissue surrounding the electrodes using only existing and unmodified, manufacturer-provided hardware. The implant lead was connected to a matching commercial IPG (Vercise Gevia, Boston Scientific Corporation) and the complete DBS including a 30-cm lead extension was realistically placed within an anthropomorphic phantom by an experienced DBS neurosurgeon (see Figure 1b)[49]. Inter-electrode impedances were read from the remote DBS programmer while the power supplied to the surrounding RF coil was varied. Each RF pulse was set to 5 s to accommodate for timing uncertainties due to manual triggering. Impedance changes Δ𝑍𝑖,𝑗 =𝑍RF on 𝑖,𝑗 −𝑍RF off 𝑖,𝑗 relative to measurements without RF were calculated for each of the 28 electrode combinations 𝑖, 𝑗 of the directional lead. To investigate the mechanism behind the RF induced impedance changes, at first experiments without RF exposure were performed. The implant lead was immersed 5cm deep into a tube-shaped glass container filled with 40 mL of PVP solution and this tube placed into a temperature-controlled water bath (RE630G, Lauda). A calibrated PT-100 temperature probe inserted 2 cm deep into the phantom liquid provided precise reference temperatures. To achieve the high temporal resolution and precise measurement control required for this characterization, the proximal terminals (contacts to the pulse generator) T1 and T8 belonging to electrodes E1 and E8 were connected to a precision LCR meter (ST2829C, Sourcetronic) to measure time-dependent admittance values 𝑌=1∕𝑍at 1 kHz with a source current set to 2 mA. Admittance 𝑌was chosen over impedance 𝑍, here, since the conductivity of aqueous PVP solutions is known to increase approximately linearly with temperature [51] in the given range of interest compared to gelatin phantoms [57, 58]. Both bath and PVP temperatures (1 Hz), and admittances (2Hz) were simultaneously recorded while the bath temperature was increased from 20˚C to 45˚C at a rate of 1K per 15 min. To establish a calibration curve, the relationship between PVP temperature and admittance was modeled using ordinary least squares linear regression. To ensure steady-state conditions and mitigate thermal lag, data points were generated by averaging measurements over a 2-min window, beginning 12 min after the start of each thermal step. In the next experiment, the implant lead was placed in the ASTM phantom, again, and exposed to 60s of RF heating at different power levels while the temperature near the electrodes was measured with two FBG probes at a 5 Hz rate, and the admittance between two electrodes by the LCR meter at a 24 Hz rate. Both 3 and 7 T frequencies were applied using the aforementioned RF coils. Admittances for three different electrode pairs were investigated to check for possible electrode dependence. 2.3 |Implant Coupling Matrix and Mitigation of RF Heating If more than one RF transmitter is available—many modern 3 T MRIs have two, 5 T and 7 T systems typically have eight such “parallel transmit” (pTx) channels—the spatial distribution of the RF field can be manipulated [59, 60]. This allows reducing the RF coupling to the implant and calculating “implant-friendly” scan modes with little sacrifice in image quality [6, 29, 45, 61]. For quantitative assessments of the implant-related hazard, the aforementioned time domain 𝐸-field sensor and two temperature probes (FBG-1 and FBG-2) were placed in the ASTM liquid near the electrodes (Figure 2c). The lead was connected to another IPG (Vercise PC, Boston Scientific Corporation) with a matching extension cable and the inter-electrode impedance changes Δ𝑍𝑖,𝑗=𝑍RF on 𝑖,𝑗 −𝑍RF off 𝑖,𝑗 in response to RF exposure were measured via the DBS programmer as described before. Since DBS system and RF transmitter are not synchronized in this type of experiment, 5s long RF pulses were applied to allow for manual triggering of the IPG readout. The coupling of the individual RF channels to a given pair of lead electrodes can be described by the so-called sensor Q-matrix (𝑸𝑺) [45, 46, 61], such that any RF safety parameter of interest 𝑋, here the impedance change, can be expressed as: 𝑋=𝒖𝑯𝑸𝑺𝒖,(1) where 𝒖is the complex voltage vector applied to the coil ports. To construct the complex, Hermitian 𝑸𝑺from real-valued sensor data, 𝑁2measurements are required for an 𝑁-channel pTx system: 𝑄𝑆,kl = ⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩ (𝑋kl −𝑋𝑘−𝑋𝑙)+𝑗(𝑋† kl −𝑋𝑘−𝑋𝑙)for 𝑘≠𝑙and 𝑘<𝑙 (𝑋kl −𝑋𝑘−𝑋𝑙)−𝑗(𝑋† kl −𝑋𝑘−𝑋𝑙)for 𝑘≠𝑙and 𝑘>𝑙 2𝑋𝑘for 𝑘=𝑙 , (2) where𝑋𝑘is the sensorreading for a single coilchannel 𝑘, whereas 𝑋kl and 𝑋† kl are the readings for two channels 𝑘, 𝑙 transmitting in-phase and with a 𝜋∕2 phase difference, respectively. Recording Δ𝑍𝑖,𝑗 for all 64 exposure conditions provides the desired 𝑸𝑺for an 8-channel RF coil. 2.4 |MRI Experiments For the MRI experiments, the DBS lead was immersed in an ASTM phantom and placed inside a 3T MRI scanner (Cima.X, Siemens Healthineers). The T1 and T8 contacts were connected to a ∼3 m coaxial cable via the aforementioned home-built adapter and a low-pass filter (10 MHz cut-off). The signal is then routed to the outside of the RF cabinet and fed to the LCR meter. An RF-only sequence was applied with RMS voltages set at the scanner console ranging from 5.0 to 40.7 V. After 4s of RF exposure, 𝑍1,8was recorded. Next, the scanner’s 2-channel body coil was used to mitigate implant heating. The RF-only sequence was run with four different 4Magnetic Resonance in Medicine, 2025 15222594, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mrm.70186 by Berk Silemek , Wiley Online Library on [25/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License excitation vectors to acquire 𝑸𝑺which is now a 2 ×2 matrix. Worst case and implant-friendly “orthogonal projection” [44] modes were calculated from 𝑸𝑺and applied via the scanner’s manual adjustments. RF-induced impedance changes Δ𝑍1,8 were recorded during 3D GRE sequences (0.5 ×0.5 ×2.09 mm3, FOV =256 mm2, TE/TR =3.69/8.0 ms, 128 slices, TA =9 min, FA =20˚) and the expected temperature rise calculated. For in vivo experiments, a healthy volunteer was imaged using both the conventional, circular polarization (CP) and the orthogonal projection mode from the previous phantom measurements. 3D T1-weighted MPRAGE images were acquired using the following settings: 0.73mm isotropic, FOV =352 ×352 ×175 mm3, TE/TI/TR =3.16/1160/2360 ms, 128 slices, TA =7:23 min. All in vivo experiments were approved by our local ethics board and informed consent was obtained from the participant. For comparison of the image quality with CP excitation versus orthogonal projection mode, the volumes of various brain regions were determined using an automated procedure in FreeSurfer [62]. The procedure included the removal of non-brain tissues such as the skull, eyeballs, and skin to enable precise whole-brain segmentation. Cortical surface reconstruction methods were applied to derive regional cortical volume measurements. Furthermore, each subject was registered to the Destrieux atlas [63] using spherical registration, following the removal of white matter residuals as part of the “autorecon” processing stages in FreeSurfer. 3|Results 3.1 |Detection of the Scattered E-Field Around the Electrodes: The Auto-Sense Signal The distribution of the normalized total power of 1000 random 8-channel RF excitation vectors is shown in Figure 3a. A scatter plot of 𝑈AS versus corresponding reference 𝐸-field measured by the external field probe (Figure 3b) shows a linear correlation (𝑅2=0.949) of both quantities. This result is corroborated by the linear correlation (𝑅2=0.95) of 𝑈2 AS with reference temperature measurements (Figure 3c). In summary, Figure 3demonstrates that 𝑈AS, read directly from a commercial DBS lead, is a valid measure of the scattered 𝐸-fields in the tissue surrounding the electrodes. 𝑈AS thus quantifies the implant-related RF hazard. 3.2 |Temperature Effect on Impedance Measurements The results of the heat-bath experiments performed to investigate the temperature dependence of the inter-electrode admittance 𝑌in PVP independently of any RF exposure are shown in Figure 4, confirming a linear correspondence between admittance and temperature from 20˚C to 45˚C. During the temperature cycling both heating and cooling demonstrated a linear correlation (𝑅2=0.99) with comparable (slope deviation of 0.0016%) fitting values (Figure 4d). These results demonstrate that the electrolyte’s temperature has a substantial and well measurable effect on the measured admittance. From Figure 4c, a relative admittance change (Δ𝑌)of(Δ𝑌∕𝑌)∕Δ𝑇=2.7%∕˚C is obtained which is comparable with the literature value for our PVP formulation [51]. The absolute sensitivity for the E1, E8 electrode pair is Δ𝑌∕Δ𝑇=27.3 𝜇𝑆∕˚C. 3.3 |Detection of the Temperature Around the Electrodes via Admittance Measurements Using the complete commercial DBS system, that is, the lead connected to the IPG, a linear correlation (R2=0.98) is observed between the RF power transmitted to the coil and the admittance change Δ𝑌measured by the implant at one electrode pair as shown in Figure 5a. The increasing admittance changes reflect an increasing conductivity of the electrolyte between the electrodes in response to the RF-induced temperature rise. The DBS system has a reported precision of ±1Ω. For a typical 1-kΩtissue impedance and a temperature sensitivity of ∼2%/˚C FIGURE 3 |Detection of RF-induced heating in DBS implants using the auto-sense signal. (a) The distribution of the normalized total power at RF channels for the transmitted 1000 random RF excitation vectors. (b) Scatter plot of the reference 𝐸-field in PVP phantom measured by an external probe near the electrodes versus auto-sense signal measured between the lead’s terminal contacts T1 and T8 for 1000 random RF excitation vectors. (c) Scatter plot of the temperature increase in the phantom after 60 s of RF exposure recorded by the external fiber Bragg grating temperature probes near the electrodes versus the square of the (T1, T8) auto-sense signal. Magnetic Resonance in Medicine, 2025 5 15222594, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mrm.70186 by Berk Silemek , Wiley Online Library on [25/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License FIGURE 4 |Simultaneous measurements of temperature and admittance (complex-valued conductance) in a sample of the PVP phantom liquid. An LCR meter was connected to T1 and T8 of the implant lead, measuring the electrolyte’s admittance between electrodes E1 and E8. The sample was kept in a temperature-controlled bath whose temperature was varied in 1˚C steps every 15min. No RF was applied for this measurement. (a) Measured admittance, water bath temperature, and PVP temperature versus time within the heating cycles. (b) Detailed view from the data presented in a. (c) Measured quantities during the cooling cycle. (d) Admittance versus steady-state temperature at the end of each temperature step. (e) Detailed view of the experimental setup. [64, 65], a 1-Ωchange corresponds to a temperature change of ∼0.05˚C. That impedance changes do indeed detect temperature changes is confirmed by a control experiment (Figure 5b) using the more precise and faster responding LCR meter instead of the DBS programmer. When the implant lead in the phantom is exposed to 60 s of RF heating at different power levels, the temperature near the electrodes, measured with external FBG temperature probes, and the admittance between two electrodes, measured by the LCR meter, vary in almost perfect synchronicity. Note that the admittance shows no sudden jumps when the RF is turned on or off. This proves that there is no direct RF effect, neither by the background field from the coil nor by the scattered field in the liquid. The admittance readings reflect the temperature and nothing but the temperature in the tissue equivalent liquid. These results were reproduced for different electrode pairs, FBG probes and RF frequencies as shown in Figure 6. RF heating can reliably be detected by Δ𝑌measurements and this determination is independent of how the temperature change was created. It is thus independent of the applied RF coil or frequency being used, making Δ𝑌a generally applicable measurand for implant safety assessments. The two bottom rows in Figure 6repeat the admittance data from the two top rows but now compared to the other temperature sensor. Although both sensor readings correlate linearly (𝑅2=0.98), the slightly larger distance between electrodes and FBG-2, see 6Magnetic Resonance in Medicine, 2025 15222594, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mrm.70186 by Berk Silemek , Wiley Online Library on [25/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License FIGURE 5 |Detection of RF-induced heating in DBS implants using admittance measurements (a) Admittance change immediately after RF exposure versus total transmit power applied to a commercial 3T RF coil. Impedance values for the electrode pair E5, E7 were measured and wirelessly transmitted by the commercial DBS implant and then converted to admittances. A linear correlation is observed. (b) Temperature near the implant tip, measured by a fiber Bragg grating temperature probe, and admittance between terminal contacts T1 and T8, measured by the LCR meter, over time. RF heating at different power levels was applied from 𝑡=0–60 s. Note that the figure shows absolute values for both admittances and temperatures. The traces are segments of one consecutive measurement over 35 min, shifted in time to have the heating periods aligned. The vertical scales were chosen such that the temperature and admittance traces coincided at the beginning of the first and at the end of the last RF exposure period (greenish and black curves, respectively). No further corrections were applied. Figure 2c, results in a noticeable slower response and also lower peak temperatures, compared to FBG-1. 3.4 |Mitigation of RF Heating Using Parallel Transmission The results so far demonstrate that either of the described measurands (𝑈AS or Δ𝑌)quantifies the momentary RF hazard for an implant-carrying patient, specific for the given subject, implant trajectory, and scan conditions. Now we turn to the question of how to exploit such information to reduce RF-induced heating and this can be done most effectively by using multiple RF channels, that is, pTx. Figure 7depicts the normalized sensor Q matrices, 𝑸𝑺, for the 8-channel pTx head coil at 3 T for all 28 electrode combinations of the 8-electrode directional DBS lead. Their information content is largely redundant, however, since the RF coupling to the implant occurs along the long wires connecting electrodes and IPG terminals and the spatial separation of these wires is much smaller than the RF wavelength. The induced RF currents are always identical on all eight wires, therefore, and the relative spatial distribution of both 𝐸𝑆and the temperature rise around the electrodes are always the same, independent of the RF mode. Different electrode combinations can be used to consistently derive the same coupling matrix, except for a scaling factor, improving the robustness of the method. With 𝑸𝑺, the implant heating properties of all possible pTx modes are known and tailored modes for optimum image quality under the constraint of safe scan conditions are determined. Three exemplary modes are depicted in Figure 8a, all transmitted with the same RF power. The reference is the CP mode, the conventional mode for head imaging in 3 T MRI, while the worst case and orthogonal projection mode were constructed by utilizing 𝑸𝑺 [44]. The worst case mode maximizes RF coupling to the implant, it allows to calculate the maximum risk to the patient. Orthogonal projection, in contrast, is an “implant-friendly” imaging mode, designed to inherit good image quality from the CP mode while avoiding strongly coupling RF channels. The reference 𝐸-fields measured with the external probe are shown in Figure 8b. In implant-friendly mode, the 𝐸-field is 10 times lower and in the worst case 3 times higher, compared to the CP mode. Temperature measurements (Figure 8c) confirm these results. After 1 min of RF exposure at identical RF power levels (16.85 ±0.45 W), the temperature near the electrodes increased by 2.0˚C for the CP mode, 17.1˚C for the worst-case mode, but only 0.02˚C for the orthogonal projection mode. Implant heating was reduced by a factor of ∼100, therefore, just by utilizing the built-in impedance measurements of the implant to switch the MR scanner’s transmit mode from conventional to implant-friendly. 3.5 |MRI Experiments The practical feasibility of the proposed procedures is investigated by transferring the experiments from the testbed to a commercial 3T MRI scanner. The admittance changes between two electrodes (E1, E8) of the DBS lead in a phantom were measured at their corresponding terminal contacts (T1, T8). The accuracy of admittance readings under MR conditions is illustrated in Figure 9a. The RF-only scanner adjustment pulses result in an admittance change of 1.7 μS corresponding to a temperature change of 0.066˚C while the noise without RF corresponds to a Magnetic Resonance in Medicine, 2025 7 15222594, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mrm.70186 by Berk Silemek , Wiley Online Library on [25/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License FIGURE 6 |RF-induced heating and admittance measurements using three different electrode combinations and two different RF coils for frequencies 128 MHz (3 T) and 298 MHz (7 T). For each combination, RF was applied for 60 s at five different power levels. Admittances were measured by an LCR meter connected to the respective terminal contacts (T1-T8, etc.) of the lead, temperatures by two fiber Bragg grating (FBG) temperature probes FBG-1 and FBG-2 with FBG-1 located slightly closer to the electrodes (see Figure 2c). The admittances from the two top rows are repeated in the bottom rows. Admittances probe the phantom liquid connecting the respective electrodes; their absolute values vary because of different electrode geometries and possibly different temperatures in the connecting bulk PVP volumes. temperature error of ±0.002˚C. Consequently, the assessment of RF induced heating in the implant can effectively be performed at very low power levels with negligible implant heating. With imaging gradients but no RF (Figure 9a), the induced noise levels are Δ𝑌=±1𝜇𝑆 corresponding to Δ𝑇=±0.04˚C, suggesting that sensitive temperature monitoring is feasible even under imaging conditions and without any attempt to filter out the gradient noise. Figure 9b shows that Δ𝑌correlates quadratically with the transmitted RF voltage and linearly (Figure 9c) with the average transmitted power, both as reported by the scanner. To reproduce the mitigation experiments, the scanner’s built-in 2-channel RF body coil was used to measure 𝑸𝑺and calculate the aforementioned transmission modes. For each mode, the admittance changes during a GRE sequence are displayed in Figure 9d, corresponding to temperature rises of 4.10˚C, 0.75˚C and 0.15˚C for the worst case, CP and implant-friendly mode, respectively. Again, the implant-friendly imaging substantially reduces RF heating. To illuminate the image quality of the implant-friendly mode, the RF parameters determined above for an “implant-carrying phantom” were used to acquire in vivo images from a healthy volunteer (Figure 9e,f). The mitigation-mode image compares well with the CP reference image (structural similarity index measure SSIM =0.999). A quantitative comparison by applying a standard neuroscience analysis to the full 3D imaging datasets, namely the parcellation of the cerebral cortex, is summarized in Figure S1. The independently calculated volumes of various brain structures agree well for both acquisition modes. 4|Discussion The presented findings suggest a solution for arguably today’s biggest and most complex MR safety problem: patients with AIMD. If implemented, improved safety, better image quality, and simplified clinical workflows can be expected for implant-carrying patients. To validate the suggested approach, measurements were performed involving off-the-shelf DBS 8Magnetic Resonance in Medicine, 2025 15222594, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mrm.70186 by Berk Silemek , Wiley Online Library on [25/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License FIGURE 7 |RF coupling derived from built-in impedance measurements of the full DBS system. Testbed measurements of the sensor-Q-matrices (𝑸𝑺) for the 8-channel 3 T RF coil are shown for all 28 combinations of the 8 lead electrodes. The color bars are individually scaled to the respective maximum element indicated at each bar (in units of ohm). components, different phantom setups and implant lead routings, as well as RF coils for different MRI field strengths. Over 1000 different pTx excitations, each representing a unique electromagnetic field configuration, were tested using both a dedicated testbed and a commercial 3T MRI scanner, demonstrating the method’s robustness in detecting and mitigating RF hazards. 4.1 |Auto-Sense and 𝚫𝒀as Hazard Sensors Two independent measures for the patient hazard were found and evaluated: the auto-sense signal 𝑈AS probing the net RF 𝐸-field emanating from the tip electrodes, and admittance change, Δ𝑌, correlating with the local temperature rise near the electrodes. Either signal has the potential for a disruptive change in implant safety management. Both allow to quantify the RF hazard for an implant-carrying patient in situ, specific for the given subject anatomy, implant trajectory, and scan conditions. Instead of millions of possible exposure scenarios only the momentarily given one is analyzed. This eliminates the need for large safety factors which are so-far required due to the accumulated uncertainties when assessing the complex interactions of implant, RF coil, and patient by modeling [37]. Both signals average over the volume carrying the exploitedinformation, that is, either the scattered 𝐸-field or the low-frequency test current for impedance measurements. This volume averaging is an advantage, since both 𝐸-field and temperature vary drastically on sub-mm length scales in the immediate vicinity of a metallic implant, which limits the reproducibility and the information content of all single-point measurements in safety assessments. Volume measurements are more robust, for example, against minor positional variations, and thus provide more meaningful information in quantifying the risk for the patient. Conceptually, both auto-sense and Δ𝑌provide point-sensor measurements, they characterize the 𝐸𝑆or Δ𝑇distribution by one single value. The choice of a particular electrode combination defines the “sensor position”, that is, the location of the probing volume, relative to the implant tip. All combinations provide redundant information; they only differ in a global scaling factor. Ultimately, the only relevant metric for implant safety is tissue damage, however, and only extensive numerical simulations can provide the link from sensor reading at a given position to the resulting tissue damage. Please note that this comment also applies to all existing implant safety assessments [38]. Magnetic Resonance in Medicine, 2025 9 15222594, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mrm.70186 by Berk Silemek , Wiley Online Library on [25/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License