Report on the translation of in vitro safety testing results to realistic exposure conditions considering combined GC and RF induced heating

Confidentiality Status: PU - Public, fully open (remember to deposit public deliverables in a trusted repository) Deliverable Cover Sheet Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or EURAMET. Neither the European Union nor the granting authority can be held responsible for them. The project has received funding from the European Partnership on Metrology, co-financed from the European Union’s Horizon Europe Research and Innovation Programme and by the Participating States. 1 of 46 21NRM05 STASIS D5: Report on the translation of in vitro safety testing results to realistic exposure conditions considering combined GC and RF induced heating Organisation name of the lead participant for the deliverable: Forschungsstiftung für Informationstechnologie und Gesellschaft (IT’IS) Due date of the deliverable: 30 September 2025 Actual submission date of the deliverable: 30 September 2025 2 of 46 Glossary MRI Magnetic Resonance Imaging MR Magnetic Resonance RF Radiofrequency SAR Specific Absorption Rate DBS Deep Brain Stimulator IPG Implantable Pulse Generator AIMD Active Implantable Medical Device CP Circular Polarization GC Gradient Coil RMS Root Mean Square TSM Tissue Simulating Medium TABLE OF CONTENTS 21NRM05 STASIS.............................................................................................................................1 1 Summary.....................................................................................................................................3 2 Motivation....................................................................................................................................4 3 Activity A3.3.1.............................................................................................................................5 4 Activity A3.3.2...........................................................................................................................10 5 Activity A3.3.3...........................................................................................................................13 5.1 Passive Implants.................................................................................................................14 5.2 Active Implant.....................................................................................................................15 6 Activity A3.3.4...........................................................................................................................17 7 Activity A3.3.5...........................................................................................................................18 8 Activity A3.3.6...........................................................................................................................19 9 Summary of A3.3......................................................................................................................28 10 Activity A3.4.1...........................................................................................................................29 11 Activity A3.4.2...........................................................................................................................32 11.1 Validation and Confidence Interval of RF-Heating.......................................................... 32 11.2 Validation and Confidence Interval of GC Heating..........................................................38 12 Activities A3.4.3 and A3.4.4, and Conclusions.........................................................................44 3 of 46 1 Summary In this deliverable, gradient-induced heating concepts and procedures proposed in ISO/TS 10974 are extended to passive orthopedic implants and fixation devices. A Tier 2 approach is developed to reduce the overestimation of the safety limits in case of anisotropic implants, and to define testing procedures more relevant to the actual clinical conditions at which such implants are scanned. The roles of gradient and RF as implant heating sources are investigated in silico; recommendations for combining results of RF and gradient testing are provided in view of standardized procedures. The result is a robust methodology to streamline the in vitro evaluation of bulky passive implants and translate into actual clinical MRI exposures with suitable confidence levels, which is proposed in Deliverable 7 “Standard Test Method for Measurement of Gradient Magnetic Field Induced Heating On or Near Nonactive Implants During Magnetic Resonance Imaging” of the STASIS project as a draft ASTM test standard. Key outcomes of the work described in this deliverable report include: demonstration of significant superposition of RF and gradient heating of bulky passive implants under clinically relevant conditions o65% in phantom, 10% in vivo worst conditions and 0.35K in realistic condition refined RF E-field metrics developed with lower overestimation than Tier 2 validated representative confidence interval of in silico RF and gradient heating assessments These outcomes are summarized in this report, and most are also published in the following open-access peer-reviewed scientific journal articles: Zanovello Umberto, Fuss Carina, Arduino Alessandro, Bottauscio Oriano. Efficient prediction of MRI gradient-induced heating for guiding safety testing of conductive implants. Magnetic Resonance in Medicine 2023, doi:10.1002/mrm.29787 Zanovello Umberto, Arduino Alessandro, Fuss Carina, Goren Tolga, , Bottauscio Oriano, Impact of simultaneous exposure to RF and gradient electromagnetic fields on implant MR safety labeling. Magnetic Resonance in Medicine 2025, doi:10.1002/mrm.70059 Arduino Alessandro, Bottauscio Oriano, Grappein Denise, Scialó Stefano, Vicini Fabio, Zanovello Umberto, Zilberti Luca, 3D–1D modelling of cranial mesh heating induced by low or medium frequency magnetic fields, Computer Methods and Programs in Biomedicine 2025, doi: 10.1016/j.cmpb.2025.109009 4 of 46 1Arduino Alessandro, Zanovello Umberto, Hand Jeff, et al. Heating of hip joint implants in MRI: The combined effect of RF and switched-gradient fields. Magnetic Resonance in Medicine. 2021;85(6):3447 2Clementi Valeria, Zanovello Umberto, Arduino Alessandro, et al. Classification Scheme of Heating Risk during MRI Scans on Patients with Orthopaedic Prostheses, Diagnostics. 2022;12(8):1873. 2 Motivation The heating of implants during an MR examination due to the radiofrequency (RF) field and the gradient field are potentially additive12, yet the two phenomena are typically assessed independently; their superposition is rarely considered in safety assessments, as it is not directly required in any standard or regulatory guidance. There are several reasons for this: 1. RF fields deposit energy directly in the biological tissues surrounding the implant, whereas gradient-induced fields deposit energy in the metallic implant. The spatial distributions of the resulting temperature increases, as well as the hotspot locations around a given implant, are significantly different. 2. GC heating is usually stronger when the implant is relatively far from the isocenter, whereas RF heating is often stronger when the implant is close to the isocenter or to the end-ring of the RF coil; 3. MR sequences which are the worst-case for GC-heating are also not typically the worst-case for RF-heating, and vice versa; 4. The test equipment of benchtop heating assessment, as well as the selected MR protocols for heating assessments in clinical scanners and the standards specifying the test protocols, are different and independently developed, encouraging their independent treatment. These methods are necessarily summarized down to heating rates, worst-case temperatures, or thermal doses of local hotspots, for implants under unrelated test exposure scenarios. Therefore, no information about the 3D spatial/temporal distributions of heat is typically available; 5. The MR sequence parameters which can limit RF-heating (SAR, B1+rms), GCheating (dB/dt rms), or both (scan duration), are independently specified, thus RFGC heating superposition can be mitigated by adjusting any of three parameters, creating difficulties for developing guidance. These reasons can be subdivided into risk-based considerations (items 1–3) and practical considerations (4–5). The increasing use of high-fidelity numerical modeling of patient exposures using human body models and implant surrogates, in lieu of or in supplement to phantom-based physical testing, is relevant to several of these points. The most significant practical consideration, item 4, is surmounted by access to the full 3D spatial distribution of local hotspots granted by numerical methods, increasing the feasibility of efficiently assessing combined RF-GC heating. Numerical methods now allow potential safety risks to be reliably identified ahead of time through multi-physics in vivo simulations, rather than waiting for a clinical report of an injury directly attributed to the phenomenon in question. Under these circumstances, the question is whether the (now feasible) effort of considering combined RF-GC heating is justified by real potential risks, i.e., considering points 1–3, do realistic MR examinations produce simultaneous RF and GC heating which superposes 5 of 46 sufficiently at the worst-case locations to justify their consideration in implant safety labeling? The first objective of this report is to identify general conditions under which the heating of GC and RF can be considered as independent or interacting, causing highest-heating scenarios. The entire analysis is performed in silico, simulating the exposure of anatomical human models modified by inserting the implant through virtual surgery. For the situations where superposition is found to be significant, recommendations for combining results of RF and GC testing were provided in view of standardized procedures. The second goal of this report is to define a robust methodology to streamline the in vitro evaluation of bulky passive implants, to translate in vitro analysis into clinical MRI exposure and to provide suitable confidence levels. The analysis will include development of reference approaches for RF heating estimation of bulky passive implant, bridging the gap between ISO 10974 Tier 2 and Tier 3-type approaches to RF heating estimation; analysis of the presence of perfusion in RF and GC heating; possibility to define optimal phantom thermal properties for GC heating to match in vivo results; possibility to combine tests performed on single implant components to get the final behavior of the entire implant for GC heating. The results represent a justification to streamline the in vitro evaluation of bulky passive implants and translate into actual clinical MRI exposures with suitable confidence levels, as well as the validated tools and metrics required to do so; this approach is proposed in Deliverable 7 “Standard Test Method for Measurement of Gradient Magnetic Field Induced Heating On or Near Non-active Implants During Magnetic Resonance Imaging” submitted to ASTM F04.15 as a standard draft item. 3 Activity A3.3.1 INRIM, ITIS and MRC will define a set of realistic exposure scenarios for performing the analysis of superposition of RF and GC heating. The considered scenarios will include different passive and active implants (at least two bulky orthopaedic implants, two orthopaedic plates, one active implant), a selection of MRI sequences (at least two classes), two hardware configurations (RF coil and GC), implant positions within the scanner (at least six positions). The following implants were used: Hip implant: Implant manufactured by Adler Ortho SpA, Italy (model APTA-FIX + FIXA Ti-Por) with CoCrMo sphere, Ti+HA stem, Ti6Al4V cup and UHMW-PE liner (Figure 1a). Knee implant: Implant manufactured by Adler Ortho SpA, Italy (model GENUS MB) with CoCrMo femoral and tibial components and UHMW-PE liner (Figure 1b). Shoulder implant: Implant manufactured by LimaCorporate SpA, Italy (model TRAUMA SMR Anatomic) made in Ti6Al4V alloy (Figure 1c). Cranial plate: Implant manufactured by Medartis AG, Switzerland (model M2-7093S) made in titanium (ASTM F67), semi-rigid. The plate is 100mm × 100mm and 0.6mm thick (Figure 1d). Ankle plate: Implant manufactured by Medartis AG, Switzerland (model 4954 25S (left)) made in titanium (ASTM F136), alloy. Together with the ankle plate, two 6 of 46 titanium (ASTM F136) screws are considered (models 5901 50 and 5800 60) (Figure 1e). Active implant: The SAIMD-U implant manufactured by ITIS is considered. SAIMDU is a verification and validation device for evaluations according to ISO/TS 10974 (Figure 2). The electric and thermal properties of the material composing the passive above implants are listed in Table 1; the materials of the SAIMD-U are given in Annex U of that standard. Table 1: Material electric and thermal properties (a) Adler Ortho SpA hip (b) Adler Ortho SpA knee (c) LimaCorporate SpA shoulder Implant CAD implant CAD implant CAD Material El. Cond. (MSm−1) El. Rel. Perm. Th. Cond. (Wm−1 K−1) Sp. Heat Cap. (Jkg−1 K−1) Mass Dens. kg/m3 Titanium 0.58 - 7.2 520 4420 CoCrMo 1.16 - 14 450 8445 UHMW-PE 0 2.3 0.41 1840 930 7 of 46 (d) Medartis AG cranial plate (e) Medartis AG ankle plate CAD CAD Figure 1: CAD models of the implants Figure 2: ZMT SAIMD-U AIMD verification device The following MRI sequences were considered along the next project activities: Echo-planar imaging (EPI): Selected to stress the heating effects due to switched gradient fields (Figure 3) Balanced steady-state (TrueFISP): Selected to stress both the heating due to radiofrequency and to switched gradient fields (Figure 4) Turbo Spin Echo (TSE): Selected to stress the heating due to radiofrequency fields (Figure 5) 8 of 46 As required by the protocol, six imaging positions were analyzed: Femur Knee Pelvis Abdomen Thorax Head For each imaging position multiple implant positions were simulated according to a fixed 7cm spatial step along the z-axis (see Figure 6). Finally, a three-axial actively shielded gradient coil with whole-body access manufactured by Nanjing Cichen Medical Technology Co., Ltd, Nanjing, China (model: Solaris-R) were used for the analyses of switched gradient field heating. 16-leg high-pass birdcage body coils were used to investigate the effects of radiofrequency heating both at 1.5T and 3T. Figure 3: Echo-planar imaging (EPI) 9 of 46 Figure 4: Balanced steady-state (TrueFISP) Figure 5: Turbo Spin Echo (TSE) 16 of 46 Table 3: Target deposited powers at SAIMD-U lead tip (in W). Position / Frequency 64 MHz 128 MHz 30 0.003413 0.006014 100 0.007563 0.009639 170 0.011134 0.012712 240 0.01513 0.023597 310 0.016356 0.032438 380 0.01643 0.03659 450 0.015529 0.030868 520 0.012891 0.019755 590 0.008001 0.009391 660 0.003252 0.003414 730 0.000808 0.000712 800 0.000237 3.78E-05 Position / Frequency 64 MHz 128 MHz 30 0.003413 0.006014 100 0.007563 0.009639 170 0.011134 0.012712 240 0.01513 0.023597 310 0.016356 0.032438 380 0.01643 0.03659 450 0.015529 0.030868 520 0.012891 0.019755 590 0.008001 0.009391 660 0.003252 0.003414 730 0.000808 0.000712 800 0.000237 3.78E-05 Figure 11: Target deposited powers at SAIMD-U lead tip at 64 MHz. 17 of 46 3Arduino Alessandro, Bottauscio Oriano, Brühl Rüdiger, Chiampi Mario, Zilberti Luca. In silico evaluation of the thermal stress induced by MRI switched gradient fields in patients with metallic hip implant. Physics in Medicine & Biology, 2019, doi: 10.1088/1361-6560/ab5428 Figure 12: Target deposited powers at SAIMD-U lead tip at 128 MHz. The distribution around the SAIMD-U lead tip has been extracted on a grid of 2mm in all directions and placed in a box with a grid corresponding to the grid used for the results for the passive implants so that the center of the distribution lies at the position of the lead tip of a generic pacemaker routing. The distribution has been scaled so that the deposited power extracted within the -30dB contour corresponds to the target powers shown in the above table. Additionally, simulations with the SAIMD-U IPG only have been done in the same way as described above for the passive implants. 6 Activity A3.3.4 INRIM will perform in silico GC simulations for the exposure scenarios identified in A3.3.1, using the digital models developed in A3.3.2. All GC simulations have been performed complying with the positions, implants and sequences selected in activity A3.3.1. Each simulations has been repeated twice to compare the effect of the reaction (skin effect) on the power deposition within each implant. Due to the negligible role played by tissues in GC heating, all the simulations have been performed in air in a spatially homogeneous magnetic field corresponding to that generated by the GCs in the relevant implant position. Gradient simulations were performed in Sim4Life and with a home-made validated code3 when simulations involved the ASTM phantom and anatomical body models, respectively. A 2mm voxel resolution was seen to be sufficient to achieve numerically stable results, and 18 of 46 4Bottauscio Oriano, Chiampi Mario, Hand Jeff, Zilberti Luca. A GPU Computational Code for Eddy-Current Problems in Voxel-Based Anatomy. IEEE Transactions on Magnetics, 2015, doi: 10.1109/TMAG.2014.2363140 the simulation domain was restricted to the region of the metallic implants since the power deposited by the GCs in the phantom and body tissues is negligible. In simulations involving the ASTM phantom, the implants were exposed to a harmonic, spatially homogeneous, magnetic field directed along the direction which maximizes the induced heating. An algorithm implemented in Sim4Life determined the worst exposure direction, and simulations were performed at 1750 Hz as suggested by the ISO 10974 draft IS1 standard. These simulations exposed the implants to an arbitrary power, as the temperature increases were scaled to a reference peak value at a later stage. Simulations involving MR pulse sequences and realistic body models followed the structure discussed in Arduino et al.,3and complied with the implementation described by Bottauscio et al.4As a consequence of the frequency content of the gradient waveform harmonic spectrum in the considered MR sequences, electromagnetic (EM) simulations modeled the skin effect in order to achieve more reliable results. 7 Activity A3.3.5 Starting from the data collected in A3.3.3 and A3.3.4, INRIM will perform thermal simulations either comparing the results obtained with RF and GC acting separately or superposing their contributions. Thermal simulations have been performed in three different scenarios: 1. Only power from RF 2. Only power from GC 3. Both RF and GC powers Given a specific sequence, the RF power have been scaled to the average power associated to the required to obtain the flip angle αineeded by the sequence. The refers to the spatial average within the slice at the isocenter of the coils and it is obtained as: where γis the gyromagnetic ratio and τiis the i-th pulse duration. The RF power PRF associated to the considered scenario is therefore obtained as: PRF where P1µTis the power associated to a of 1T averaged on the isocenter slice and TR is the repetition time. Since the thermal regulation has not been accounted for in 19 of 46 5Arduino Alessandro, Bottauscio Oriano, Chiampi Mario, Zilberti Luca. Douglas–Gunn Method Applied to Dosimetric Assessment in Magnetic Resonance Imaging. IEEE Transactions on Magnetics. 2017, doi: 10.1109/TMAG.2017.2658021 simulations, the RF temperature distributions associated to the same implant position and different sequences simply scales with the power value PRF. The power density distributions resulting from GC and RF simulations were extracted on a regular 2mm grid. During GC exposure, the tissues heat up indirectly, due to thermal diffusion from the metallic components of the implant, whereas during exposure to RF, the heat diffuses from the tissues to the implant, where the direct thermal effect of RF fields is negligible. Pennes' bioheat equation was expressed in terms of temperature elevation with respect to the temperature at rest and solved, by means of a home-made code3,5. In the case of the anatomical body models, the bioheat equation also modeled the blood perfusion in the tissues, according to the parameters reported in the database of the IT'IS Foundation. This led to two temperature increase distributions, one due to GC alone and another to RF alone. Finally, the temperature increase due to the simultaneous presence of gradient and RF effects was obtained as summation of the previous GC and RF temperature distributions. Thanks to the linearity of the Pennes' equation, this corresponded to solve the problem using the sum of the two power distributions as the forcing term. The analysis with the ASTM phantom combined, point by point, the spatial distributions of the temperature increases produced by the GC and RF fields, scaled to the same peak value within the phantom. Simulations involving the MR pulse sequences and realistic body models scaled the gradient power according to the running sequence waveform and the RF power according to the required flip angle, RF pulse duration and modulation envelope, i.e. a sinc function with 0.5 apodization coefficient. Both powers were averaged over the repetition time (TR) of the sequence and used as a source of the thermal problem. The GC and RF temperature increases were combined in two ways: the first by summing GC and RF temperature increases scaled to the same maximum value across the landmarks, and the second by summing GC and RF temperature increases as obtained from the actual powers generated by the relevant MR pulse sequence. All temperature increases are provided as peak values across the computation domain, for a 30 min exposure to the different power sources. 8 Activity A3.3.6 INRIM and ITIS will define correction factors to account for the effect of superposition of RF and GC heating to be considered for an extension of the ASTMF2182 Standard related to RF heating of passive devices. 20 of 46 Table 4 collects the relative enhancements of the peak temperature increases in the ASTM phantom due to the simultaneous exposure to GC and RF fields, with respect to the peak temperature increase obtained with GC or RF alone scaled to the same reference value. Relative enhancements range from 17-67% at 1.5T and from 24-51% at 3T, depending on the implant. Table 4: Relative enhancements of the maximum temperature increase due to the simultaneous application of GC and RF fields with the anatomical body models, with respect to the temperature increase due to GC or RF alone. For each implant and MR pulse sequence, RF and GC temperature increases were scaled to the same maximum value across the landmarks. The exposure conditions of the implants inside the ASTM phantom are quite different from those in a clinical setup. This reflects on the temperature increase enhancements obtained when the analysis involved MR pulse sequences and realistic body models, and the GC and RF temperature increases were scaled to the same maximum values across all landmarks before being combined (Figure 13 and Figure 14) . In this regard, Table 5 collects the relative enhancements of the peak temperature increases in the cases of realistic MR pulse sequences (Table 6) and anatomical human body models. The results show that the maximum relative enhancements never exceed 9% and 11% at 1.5T and 3T, respectively. It is interesting to notice that, most of the time, the maximum temperature increase due to the simultaneous exposure to GC and RF occurs at the same landmark as in the case of exposure to the GC field alone. 21 of 46 Figure 13: Peak temperature increase, in kelvin, on the six implants exposed to different MR pulse sequences as a function of multiple axial landmarks. Results refer to 1.5T exposure to the RF EM field alone (red line), GC EM field alone (green line) and simultaneous application of RF and GC EM fields (blue line). The gray area highlights the enhancement of the peak temperature increase due to the simultaneous exposure to RF and GC EM fields with respect to the maximum between the peak temperature increase due to RF or GC alone. 22 of 46 Figure 14: Peak temperature increase, in kelvin, on the six implants exposed to different MR pulse sequences as a function of multiple axial landmarks. Results refer to 3T exposure to the RF EM field alone (red line), GC EM field alone (green line) and simultaneous application of RF and GC EM fields (blue line). The gray area highlights the enhancement of the peak temperature increase due to the simultaneous exposure to RF and GC EM fields with respect to the maximum between the peak temperature increase due to RF or GC alone. 23 of 46 Table 5: Relative enhancements of the maximum temperature increase due to the simultaneous application of GC and RF fields with the anatomical body models, with respect to the temperature increase due to GC or RF alone. For each implant and MR pulse sequence, RF and GC temperature increases were scaled to the same maximum value across the landmarks. 24 of 46 Table 6: Main parameters of the sequences considered for the analysis. Maximum Whole-Body (WB) averaged SAR refers to the maximum including all implants and exposure landmarks for the specific sequence. The three dB/dt rms values reported for the EPI sequence refer to the X, Y and Z readout directions, respectively. Figures 13 and 14 show the peak temperature increase for 1.5Tand 3% scenarios, respectively, resulting from the exposure to the specific MR pulse sequences whose parameters complied with those listed in Table 6. The figures report the temperature increases as a function of the axial position of the implant barycenter with respect to the scanner isocenter. The selection of the reported results was performed on a case-by-case basis, according to the combination of implant and sequence leading to the highest enhancements of the temperature increase due to the simultaneous effect of RF and GC EM fields with respect to RF and GC alone. This choice did not necessarily lead to show the cases where the absolute maximum temperature was reached. Nevertheless, the maximum peak temperature increases also played a role in excluding those scenarios where the heating due to both RF and GC were less prominent. The gray areas in the figures highlight the enhancement of the peak temperature increase when the RF and GC EM fields are applied simultaneously with respect to the maximum between the peak temperature increase due to the application of RF or GC EM field alone. Scenarios with the greatest distance between the blue and the dashed lines represent those where the addition of the two effects has the largest effect on the combined temperature increase. In this regard, Table 7 collects the maximum peak temperature increase enhancements for each exposure scenario. The table reports the enhancements as the peak temperature increase following the combined application of GC and RF, reported next to the maximum between the peak temperature increase due to RF or GC alone. 25 of 46 Table 7: Maximum peak temperature enhancements, among all the landmarks for each specific scenario. Each table entry reports the temperature increase due to RF or GC alone (the highest of the two as reported in the brackets) on the left of the temperature increase due to the combined effect of RF and GC. The temperature enhancement is added in square brackets near the temperature increase values. All temperature increases are reported in kelvin. For each implant, the first row refers to results at 1.5T and the second row at 3T. When the temperature increase enhancement resulted to be negligible for all the landmarks, a hyphen is reported in the corresponding table entry. Results shown in Figures 13 and 14 are reported in bold. The U.S. Food & Drug Administration (FDA) suggests to include, along with the MR Conditional labeling of a Medical Device, the maximum allowable gradient slew rate per axis, the maximum permitted Whole-Body (WB) averaged SAR and/or maximum permitted head averaged SAR and the maximum B1+ rms admissible value. Constraining the maximum gradient slew rate reduces excessive GC heating, whereas limiting the maximum SAR and/or B1+ is useful to limit RF heating. Therefore, the previous quantities are restricted on the basis of testing results and a safety threshold in terms of maximum acceptable temperature increase. Independently of the actual value of the temperature increase threshold, when GC and RF testing are performed separately, applying independent limits is appropriate as long as the two heating effects do not combine significantly. When this holds true, if an MRI scan is performed e.g., at the maximum B1+ value, the peak of the RFinduced temperature increase may be close to the established threshold, if the clinical exposure is close to worst-case. Of course, if at the same maximum B1+ value, the gradient also contributes significantly to the heating, the actual peak temperature increase may be higher than the safety threshold. As a consequence, if the interaction of RF and GC heating is not properly accounted for, maximum temperature increases could reach unexpected and unsafe values, exposing the patient to a potential hazard. The analyzed implants were not labeled as MR conditional, therefore the above information was not available; nor is there any generally accepted published value of maximum acceptable temperature increase. For these reasons, the study abstracted from these inputs, first by scaling the peak temperature increases obtained from GC and RF exposure to the same reference value, and then by considering an exposure to the EM field generated by MR sequences complying with normal operating mode and hardware watchdog. 32 of 46 While the conservativeness of Tier 4 is limited by the access of the CAD model to the site of the worst-case incident field, the proposed metric can achieve full coverage of the target surface or volume and generate the desired statistics using the same tools and effort as Tiers 2 / 3. 11 Activity A3.4.2 ITIS will perform validation SAR/temperature RF measurements on selected generic implants from A3.4.1 with defined conditions to assess the accuracy and conservativeness of the prediction approaches. 11.1 Validation and Confidence Interval of RF-Heating The validation of the accuracy and conservativeness of the “Tier 2+” approach defined in A3.4.1 extends the confidence interval of the Tier 3 library and toolset, IMAnalytics with MRIxViP, from where the incident fields are derived. The confidence interval of the use of IMAnalytics with the MRIxViP library for a generic elongated implant, which was estimated by following the standard guidelines defined by the Guide to the Measurement of Uncertainty (GUM) and by the ASME V&V 10, is shown in Table 8. The procedure is based on describing the total uncertainty of the evaluation as the convolution of independent uncertainty sources using Type B evaluation. Each uncertainty term was evaluated in terms of its contribution to variation of the final result (power deposition at a hotspot or induced voltage at lead terminal). Therefore, the uncertainty contribution of each term comprises both the intrinsic variation of the source of uncertainty, and the local sensitivity of the final result to that variation − thus the uncertainties are presented as unitless factors, in decibels (as they are applicable for deposited power and for voltage). The uncertainty associated with other IMAnalytics inputs, such as uncertainty of the transfer function models or routings/regions of interest, must be assessed separately and included in the final confidence interval. 33 of 46 Table 8: Confidence interval of IMAnalytics + MRIxViP output (deposited power or induced voltage) due to uncertainty of in vivo exposure conditions. 34 of 46 The confidence interval of ISO 10974 Tier 3 evaluations of deposited power around SAR hotspots in tissue, or induced voltage in a channel, was assessed for tests performed by the IT'IS Foundation on AIMD in previous EMPIR project (MIMAS 17IND01). In that report, the calculation of temperature rise, thermal dose, and tissue damage corresponding to the assessed in vivo deposited power were not discussed as being outside the scope of the ISO 10974 approach. Unlike deposited power, the local temperature rise depends strongly on the shape of the lead tip as well as the tissue environment around the hotspot (tissue properties and distribution). In this work, the MIMAS report was extended to cover in vitro temperature rise uncertainty around strong gradients such as found near screw tips or AIMD electrodes. To appropriately measure the temperature rise, a system check has to be conducted to guarantee proper functionality of the equipment. If possible, the racetrack in the vicinity of the electrode should be cut away to minimize its contribution to the measurement. The system check consists of the following steps: measure the electrical and thermal properties of the TSM together with the corresponding temperature; verify the correct operation of the measurement system, by conducting a background temperature rise measurement and comparing it to the expected value from the measured E-field. In order to minimize errors from disturbing the test item and slurry, the system check is performed with the test item in place. The background temperature rise measurement location should be selected to have the least enhancement due to the implant (<-30dB) and be at least 2cm away from any interface. 1. Fill up the desired phantom with fresh TSM in slurry form. TSM should not contain air bubbles. 2. Measure the electrical and thermal conductivity of the TSM using a thermometer, the DAK, and the thermal conductivity meter. All values should be close to nominal, i.e. within the range defined by the uncertainty budget for your test protocol. 3. Let the covered phantom rest (e.g. over-night) to equilibrate and for air bubbles to dissolve. 4. Before starting the measurement, check that TSM properties have not drifted. 5. Mount the phantom in the exposure system, and configure it to expose the test item to a uniform incident E-field 7. Validate the incident B-field as appropriate. 9. Mount the temperature probe and position it at the defined location for background temperature rise measurement. 10. Record the baseline temperature; temperature drift should be ≥10dB below the measured temperature rise of the sample. 35 of 46 12. Engage the exposure. The measurement protocol should record a temperature rise of at least 5 × sensor resolution; the total temperature rise should be <3K to avoid change of the TSM properties. 13. Mount the SAR probe and measure the total E-field at the same location as the temperature, rotating the probe 360◦ and averaging total E. 14. The measured ∆T has to be equal to σE2/(ρc)×tmeas within the confidence interval defined in Table 9. If this is the case then the measurement system is validated. 15. Move to the first measurement point defined for the test item (e.g., dark blue in Figure 17). Repeat 10 - 12. Repeat for the desired number of measurement points. The background temperature rise should be validated against the analytical solution (SAR= C dT/dt). The SAIMD-1-200mm temperature rise can be validated against numerical simulations, as shown in Figures 18 and 19. Target values can also be obtained from Annex I of ISO 10974. Figure 17: Schematic of SAIMD-1-200mm with the locations of interest (dark and light blue) for temperature measurement. Figure 18: Numerical benchmark for p1 and p2, normalized to background incident field. The background temperature rise has been removed from the temperature rise due to SAIMD-1-200mm. The numerical data has been averaged over the dimensions of the tip of the probe (d = 0.88mm). 36 of 46 A correction factor has to be applied to compensate for the physical size of the temperature probe and the temperature gradient in z-direction (gradient compensation section in Table 9). Figure 19 shows the correction factor, derived from a linear fit of SAIMD-1 measurements with two different models of temperature probe. Table 9: Example confidence interval of temperature rise validation at the highest heating point (point 1, dark blue) of SAIMD-1-200mm. References marked with [?] are not generalized and should be updated as appropriate for the TSM selected. 37 of 46 Figure 19: Correlation between measurement deviation and the temperature gradient in the probe direction at both 64 (big symbols) and 128 MHz (small symbols). The outliers (marked with black circles) are excluded from the linear fit. 38 of 46 11.2 Validation and Confidence Interval of GC Heating In this activity, the work was extended to include this uncertainty for in vitro RF and GC exposures. The measurements were conducted on a semi-rigid titanium (ASTM F67) cranial mesh manufactured by Medartis AG (Basel, Switzerland). The implant was 93 mm×93 mm and its thickness was 0.6 mm. The thickness of the implant corresponds to the diameter of its one-dimensional structure, so that 𝑟= 0.3 mm. It was equipped with three optical fibre temperature probes, two positioned in the peripheral regions, where the largest temperature increase was expected to be induced, and one in the central region, as shown in Figure 20. The probes were connected to an eight-channel AccuSens signal conditioner produced by Opsens (Québec, Canada) sampling at 20 Hz. The manufacturer declared an overall accuracy of ±0.30 ◦C and a resolution of 0.01 ◦C. 39 of 46 Figure 20: Experimental setup and virtual modelling. (a) Semi-rigid cranial mesh in titanium equipped with the optical fibre temperature probes. (b) System of GCs for cylindrical MRI scanners. (c) Phantom filled with expanded polystyrene grains. (d) Computational model of the cranial mesh with depicted the boxes of the active parts of the temperature probes. (e) Computational domain including the phantom with the cranial mesh and the model of the GCs. 40 of 46 In a first experiment, simulating the heat transfer from the cranial mesh towards soft tissues, the implant with the positioned probes were plunged into a cuboid container (base of 13 cm × 44 cm, height of 20 cm) filled with gel simulating average tissue parameters produced by Zurich MedTech AG (Zurich, Switzerland) in accordance with ISO/TS 10974. In a second experiment, simulating an almost adiabatic condition, the same cuboid container was filled with expanded polystyrene grains whose diameter varies between 1 mm and 3 mm. The base of the implant was fixed in a slot cut in a spongy support to keep it in the correct position within the container. In particular, the implant was positioned at 45◦ with respect to the container walls to have it perpendicular to the generated magnetic field, which is the configuration expected to maximize the induced temperature increase. The magnetic field was generated by the system of actively shielded GCs with whole-body access for cylindrical MRI scanners. Precisely, it is the model Solaris-R manufactured by Nanjing Cichen Medical Technology Co., Ltd (Nanjing, China). The system, featuring three coils to generate gradient magnetic fields along three orthogonal directions, has an internal diameter of about 67 cm and a total length of about 150 cm. The gradient directions determine the used reference system, with 𝑧directed longitudinally with respect to the coils, 𝑦directed vertically and 𝑥perpendicular to the other two directions. The GCs were supplied by a NG500 1.3 gradient amplifier built by Prodrive Technologies (Eindhoven, The Netherlands) able to drive each coil independently and to provide a peak current of 1000 A and a maximum voltage of 940 V. To dissipate the loss power in the coil conductors due to the high current values, a water cooling system was connected to the coils. The cooling system made negligible the impact of the heating dissipated by the GCs’ conductors on the temperature values measured on the implant, as verified through a fourth optical fibre temperature probe positioned at the boundary of the phantom, where no heating due to the currents induced on the cranial mesh was expected. In the experiment, the coils generating gradients directed along 𝑥and 𝑧were driven with two in-phase sinusoidal currents at the frequency of 2 kHz. According to the numerical model described in Section 2.2.2, the peak current intensities of 150 A were such that a peak magnetic field of 3.5 mT was generated at the implant barycentre, located at 𝑥= 14 cm, 𝑦= 0 and 𝑧= 30 cm with respect to the coil isocentre (i.e., the central point where all the coils generate a null magnetic field). The temperature detected by the optical fibre temperature probes were recorded every 2 s for 900 s of continuous exposure of the cranial mesh to the harmonic magnetic field generated by the GCs. Temperature recording started 30 s before switching on the power amplifier to acquire the value at which the system and the probes were thermalized. Since the actual measurand to be compared with the simulation results is the temperature increase, an offset equal to the average of the temperature values recorded in the preliminary 30 s were applied to the measurement results. The measurements were repeated two times a day apart without moving neither the probe nor the phantom, to assess the repeatability of the experiment. Figure 21 depicts the measurement pipeline. The temperature probes attached to the implant were connected to the temperature acquisition. 41 of 46 Figure 21: Experiment pipeline. The experiment was monitored by a controller PC that collected the acquired temperature data through a serial connection with the temperature acquisition system. The same controller PC controlled the gradient amplifier through a LAN connection and monitored the generated currents through a Thunderbolt connection and an acquisition board connected through coaxial cables to the gradient amplifier. The gradient amplifier supplied the GCs through high current cables and the cranial mesh was exposed to the generated magnetic field. Figure 22: Comparison between measurements and simulations. The noisy pale orange and pale red lines report the measurement results in the first and second repetition, respectively. The solid green and blue lines report the simulated results with the thermal seed model (purely 3D FEM) and the 3D–1D coupling,