RF-induced Heating Estimation of a Stent in a 3T MRI using Transfer Function Approach with a Tabletop E-field Generator

Abstract

The use of a transfer function (TF) method enables a conservative estimation for radio frequency (RF) safety assessment of active implantable devices (AIMDs). The TF approach can be applied to various scan conditions, patient populations, and device trajectories inside the human body, reducing the computational burden of full-wave electromagnetic (EM) simulation. The in vitro TF model validation process is time-consuming, requiring tests in various sample trajectories that collectively exceed eight hours. Here, we demonstrated reducing the burden of the TF approach using a low-power tabletop E-field generator. We measured the TF of the stent via the piecewise excitation method at 128 MHz and validated it by exposing the device under diverse test exposure fields using a tabletop E-field generator that requires less phantom material, lower cost than whole-body coil or MRI scanner, and with reduced experimental safety hazards or shielded room requirements. The TF approach was used to predict radio frequency (RF)-induced power near the stent tip at 128 MHz and predicted values were then compared against measured values. We also used a body transmit coil to compare the conventional in vitro TF model validation approach and tabletop E-field generator. With the tabletop E-field generator, the equivalent absolute normalized error was (0.37 ± 0.31 dB) compared to the body transmit coil tests (0.43 ± 0.15 dB), and the required test time decreased from eight to three hours. In summary, we showed how a low-power compact E-field generator can be used for in vitro TF model validation with reduced testing time and cost without using a shielded room.

I. INTRODUCTION

MAGNETIC Resonance Imaging (MRI) is one of the most versatile diagnostic imaging techniques employing a strong magnetic field and non-ionizing radiation to non-invasively image soft tissues. Nearly 40 million patients undergo MRI scans annually in the United States [1], and the likelihood of patients with implantable devices accessing MRI increases over time [2]. Patient risk assessment includes magnetically-induced torque and force, radio frequency (RF) or gradient-induced heating, vibration, and unintended stimulation to tissue, and device malfunction [3].

One key safety concern is electromagnetic (EM) interactions with metallic devices in patients’ bodies [4]. The RF wave transmitted to the patient’s body for imaging could interact with metallic components of implantable medical devices, acting as antennae that enable the induced current flow to tissues (referred to as the “antenna effect”), potentially causing severe thermal damage. However, estimating RF-induced heating is not a trivial task due to variations in device sizes and trajectories inside complex body tissues, which also vary from patient to patient.

To tackle this challenge, a transfer function (TF)-based RF-induced heating estimation approach has been introduced to predict the induced heating or voltage through the elongated metallic lead, which produces a local hot spot or voltage at the end of the device [5]. The TF approach simplifies the problem by characterizing the metallic leads in miniature configurations separately from the relatively large and heterogenous human body interactions with the EM fields. It assesses the induced heating or voltage with reasonable overestimation compared to the concurrent EM modeling and simulation [3].

Device TF models can be experimentally measured or determined using computational simulation, but require in vitro model validation experiments before being applied to the Tier 3 method defined in International Organization for Standardization (ISO) Technical Specification (TS) 10974 [3] that are used to estimate induced-power or voltage in human body models. The lowest numbered Tier method (e.g., Tier 1) represents the least complex test method, produces the least accurate results, but is conservative to ensure safety. As the tier number increases (up to Tier 4), the test method becomes more complex, but provides a more accurate results [3]. For Tier 3, the most common approach is to measure the RF-induced heating in various device trajectories inside a tissue-mimicking gel phantom [6] within a whole-body-sized RF-transmit coil or 1.5T and 3T MRI scanners for TF model validation [7]. This in vitro model validation process is essential to confirming TF suitability for computational modeling studies but is expensive and time-consuming.

To mitigate this burden, a test field diversity (tfd) method [8] was proposed. This method produces and varies the incident RF field on the device under test (DUT), reduces the number of DUT test positions needed to validate the TF by increasing the test field variety. An alternative option is to use a high E-field generator dedicated to testing and validating measured TF [9] within confined box-shape test equipment. Compared to traditional methods described above, this approach requires less gel phantom material, lower cost than whole-body coils or MRI scanners, smaller form factor, reduced safety hazards or shielded room requirements with a simpler geometry of transmit source to model. Thus, it facilitates conducting more straightforward measurements within confined workspace and accurately positioning the DUT. While tabletop-sized high E-field generators have several advantages, testing still involves repositioning the DUT at various locations, adding potentially burdensome experimental time.

The first goal of this study was to use an alternative versatile, low-power tabletop E-field generator for RF field exposure to validate TF model by predicting RF-induced power near the stent tip using the Tier 3 method as defined in ISO/TS 10974 [3]. This was compared with the measured RF-induced power value near the elongated passive device, stent tip at 3T. The second goal was to compare TF model validation results from the tabletop E-field generator with the conventional in vitro TF model validation method using a whole-body size quadrature birdcage coil (QBC) tuned at 128 MHz. Towards these goals, we measured the TF of a stent to predict the RF-induced heating at the stent tip at 128 MHz (typical operating frequency of 3T MRI) and tested this prediction by comparing measured RF-induced power near the stent tip with various incident E-field scenarios inside tissue simulating liquid phantom using a tabletop E-field generator.

II. METHODS

A. TRANSFER FUNCTION MEASUREMENT

The piecewise excitation system (piX, Zurich MedTech AG (ZMT), Switzerland), was used to measure TF at 128 MHz (see Fig. 1a). A stent sample (L ≈ 100 mm, O.D.=7 mm, thickness= 0.2 mm) made with Nitinol (Fig. 2a) was positioned on a straight racetrack for the TF measurement, submerged under conductive tissue simulating liquid (σ: 0.47 S/m, ɛ_r: 78; TLe78c0.47, ZMT) and the E-field was measured near the DUT tip using a time-domain active optical E-field probe (E1TDSz, Schmid & Partner Engineering AG (SPEAG), Switzerland). The piecewise excitation coil (piX excitor coil) tuned at 128 MHz was attached to the robot arm (Dosimetric Assessment System (DASY52), SPEAG), which was swept along the device trajectory every 10 mm between −20 mm from the tip of the DUT (y: −20 mm) and 10 mm after the end of the DUT (y: 110 mm). TF measurement was repeated three times, and the data range outside the device was trimmed after the measurement. Next, the background response was measured without the DUT by sweeping the excitor coil along the same TF measurement trajectory to subtract excitor response. The measured TF was then normalized such that the integral of TF magnitude over the DUT length equaled 1 [8].

FIGURE 1.

Transfer function measurement system. a) Experimental set up for the TF measurement; b) zoomed view of the DUT inside tissue simulating liquid and the excitor coil.

FIGURE 2.

Tabletop E-field generator for in vitro TF model validation. a) dimension of the stent sample; b) Experimental set up with the RF-induced power measurement near DUT tip using a dosimetric probe; c) Top-view of the DUT test positions relative to the phantom.

B. IN VITRO TRANSFER FUNCTION MODEL VALIDATION USING TABLETOP E-FIELD GENERATOR

Experiment:

We validated the measured TF model by exposing the DUT under diverse test fields using a commercial tabletop E-field generator, Medical Implant Test System TableTop (MITS-TT, ZMT) (Fig. 2b). Sixteen different incident test fields were chosen to produce diverse E-field exposures by changing the relative gain (0 to 4 dB) and phase (0 to 315°) between two transmit sources. The dosimetric E-field probe (EX3DV4, SPEAG) was positioned near the stent tip to measure the specific absorption rate (SAR) in W/kg. The deposited power at the tip was estimated using (1),

$$P_{\mathit{\text{tip}}}=A{\left|{\int }_0^{L}S\left(l\right)E_{\mathit{\text{tan}}}\left(l\right)\mathit{\text{dl}}\right|}^2$$ (1)

where P_tip is the estimated local power deposition near the stent tip in (mW); A is the scaling factor (μW/(V/m)²) to estimate deposited power near the DUT; S(l) is the normalized transfer function (1/m); E_tan(l) is the simulated incident electric field over the DUT pathway (V/m); and l is the specific point along the lead trajectory (m).

We measured the transient point SAR near the stent tip at 16 test fields. A bamboo stick was guided through the inside of DUT, cut slightly longer to allow for firm securing the DUT on the oval-shaped FR-4 glass reinforced epoxy racetrack (ZMT) using Kapton tape (Fig. 2). The volume scan was performed above the stent tip (e.g., 10 mm × 10 mm × 3 mm), which required co-registering and scaling to estimate deposited power in −30 dB contour [10]. After the volume scan, the DUT was removed from the tabletop E-field generator, and background E-field scans were performed at 16 test fields for a 200 mm straight line at every 10 mm distance, which is required to scale the simulation results. The same procedure was repeated after the DUT was positioned at the second TF model validation location (Position 2, Fig. 2c) using an e-shaped FR-4 racetrack (prepared to guide the elongated device inside the MITS-TT at a distance of >2cm away from the wall in a vertically centered position).

Simulation and Post-processing:

We used MRIxLAB, a subtool of IMAnalyticsSuite (ZMT) containing the pre-computed EM field of a tabletop E-field generator, to extract the tangential E-field (E_tan) at each test field at 128 MHz along the input device test trajectories. MRIxLAB requires three experimental input data: i) scaling factor, A, calculated from Position 3, ii) deposited power (mW) from the volume scan, and iii) DUT trajectories within MITS-TT. The scaling factor (A) was calculated from an experiment conducted at an additional DUT position (Position 3) minimizing the deviation between measured power and predicted power (see Supplemental Materials S1 for details). We chose the position producing a different incident E-field compared to Position 1, as the trajectory variation was limited due to the short DUT length. We calculated the −30 dB deposited power near the stent tip at each DUT position using a “Touchless” Python notebook with volume scan measurement co-register with EM simulation of DUT at 128 MHz [10] (see Supplemental Materials S2). The “Touchless” process removes the measurement location error by co-registering the measured volumetric E-field into the simulation and translates the measured E-field (V/m) into −30 dB deposited power (mW) near the DUT tip. Finally, DUT trajectories were prepared as 3D models (.sat), storing the three-dimensional geometry information of the DUT inside MITS-TT. The E_tan over 16 test fields were extracted at each DUT position, then combined with the normalized TF and scaling factor using (1) to predict the deposited power near the DUT tip.

C. TRANSFER FUNCTION MODEL VALIDATION USING WHOLE-BODY TRANSMIT COIL

Experiment:

In addition to the tabletop E-field generator method, we also performed in vitro TF model validation using a traditional method based on a quadrature 16-leg band-pass 3T body transmit coil (i.e., QBC) with a length of 480 mm operating at 128 MHz (MITS-3.0; ZMT) operated inside a shielded room (Fig. 3a). The gelled tissue-mimicking phantom was fabricated according to the American Society for Testing and Materials (ASTM) standard method F2182–19^e2 [6] (σ: 0.47 S/m, ɛ_r: 78) which was positioned at the center of the QBC and allowed to equilibrate to room temperature (over 2 hours). A DUT was positioned near the inner wall of the gel phantom where the highest E-field is expected. A fiber optic temperature probe (T1S-02-PT05, Neoptix, Canada) was positioned near the stent tip to measure the temperature changes (Fig. 3b). RF field was generated using the QBC with a rectangular pulse at 55 dBm input power at 40% duty cycle in circularly polarized (CP) mode for 15 minutes. We repeated the tests at the seven different DUT positions by shifting DUT 5 and 10 cm from the wall in x-, y- directions (Fig. 3b). After each experiment, the gel phantom was manually stirred and cooled down more than 30 minutes before the following experiment to establish a homogeneous background temperature distribution [6]. We performed an additional test to independently calculate the scaling factor, A_QBC, at the DUT position of 2 cm away from the inner phantom wall for the QBC experiment, because this position produced the highest signal-to-noise ratio (SNR; ΔT).

FIGURE 3.

Experimental setup using the whole-body sized birdcage coil and gel phantom. a) whole-body birdcage coil with fiber optic temperature probes; b) Positions of the DUT inside a gel phantom (orange dots indicate the temperature measurement location).

Simulation and Post-processing:

For the QBC simulation, Sim4Life software (ZMT) was used to calculate the incident E-field along the device trajectory in the QBC. The inner diameter of the QBC was 700 mm, and the length (L) was 480 mm, with a copper strip width of 40 mm for the end ring and 25 for each rod. The RF shield (ID: 830 mm, outer diameter (OD): 837 mm, and L: 855 mm) was also modeled outside the QBC, and the phantom (L=650 mm, W=420 mm, height(H)=90 mm) was placed at the center of the QBC following the experimental condition. The log files containing forward and reflected power at each port was used to scale the simulation results. The E_tan was extracted over the DUT location in simulation, and MATLAB (MathWorks, MA) was used to combine scaled E_tan(l) with the normalized TF to predict the RF-induced heating using (1) and the following assumptions: $\Delta T\approx \tau \frac{\mathit{\text{SAR}}}{C_p}$, where ΔT=temperature change (°C), τ=duration of heating (s), SAR=locally deposited power absorbed in the medium (W/kg), C_p=specific heat capacity of the medium (J/kg·°C), when there is no internal heat source other than SAR, thermal conduction and convection are relatively small, so the change in temperature is assumed linear with time [11].

D. TECHNICAL COMPARISON OF TWO IN VITRO TF MODEL VALIDATION METHODS: TABLETOP E-FIELD GENERATOR AND QBC

We conducted the technical comparison between the two approaches regarding: i) total required test time, ii) Total standard uncertainties in the incident field exposure system, iii) error comparison in predicted versus measured induced power, and iv) the number of linearly independent E_tan fields. The normalized error, N.E. [dB], was calculated using (2),

$$N.E.=10{\mathit{\text{log}}}_{10}\left(1+\frac{P_{\mathit{\text{Measured}}}-P_{\mathit{\text{Predicted}}}}{{\mathit{\text{Max.Power}}}_{\mathit{\text{Predicted}}}}\right)$$ (2)

where P_Measured is the −30 dB deposited power from measurement (mW), P_Predicted is predicted power (mW) using Tier 3 method defined in ISO/TS 10974, and Max.Power_Predicted is the maximum predicted power (mW) near the stent tip. The uncertainty of the TF modeling, and TF model validation with tfd exposure system and QBC system was estimated as reported in previous publications [10], [12], [13] following the methodology of the Guide to the Expression of Uncertainty [14]. Total standard uncertainty (k=1) was calculated by root-sum-square of each term of standard uncertainty. Finally, the number of linearly independent E_tan matrices was assessed using the rank function in MATLAB (MathWorks) along the 10 cm DUT trajectories for all tested conditions (see Supplemental Materials S3 for an additional analysis performed on the 20 cm lead trajectories).

E. EXAMPLE TIER 3 ANALYSIS FOR RF-INDUCED HEATING IN HUMAN MODELS

IMAnaytics (ZMT) with under-sampled tutorial example database was used to estimate the RF-induced power deposition near the DUT tip in the human body as defined in ISO/TS 10974 [3]. High-pass QBCs with a diameter of 600 mm in three lengths (500mm, 600mm, 700mm) were simulated without the presence of the DUT while operating in CP mode at 128 MHz. We utilized the Duke human model (34-year-old male) [15] at nine different imaging positions (or landmarks); position z=0 corresponds to a location inside the model’s head at the eyebrow level centered in QBC, then moved every 200 mm distance to 1600 mm towards the feet. The E_tan was extracted from the representative generic implant lead routing for cardiovascular devices, then combined with normalized TF, S(l), in two polarities (0 of TF at DUT tip measured E-field response, and 0 of TF at opposite end of DUT) to account for induced heating at both ends. We used the scaling factor, A (uW/(V_rms/m)²), calculated from the TF model validation process to assess the RF-induced heating. The corresponding values in Fig. 10c were used to scale up the power level of the MR imaging system until one of quantities is reached to the Normal Operating Mode (NOM) limit defined in IEC 60601-2-33 [16].

III. RESULTS

A comparison of TF model validation using a tabletop E-field generator and QBC is presented in terms of measured RF-induced power (mW) and RF-induced heating (°C).

A. TRANSFER FUNCTION MEASUREMENT

The normalized TF of the stent at 128 MHz is shown in Fig. 4. Peak TF magnitude was found 10 mm from the tip and decreased over the DUT pathway, while the phase continued to decrease when the piX excitor coil position moved away from the stent tip. The maximum standard deviation of the TF measurements was 0.09 dB in magnitude, and 0.01 dB in phase.

FIGURE 4.

Measured transfer function model of DUT at 128MHz. a) TF of the stent along the DUT trajectory in magnitude, and b) phase; c) magnified view of the TF shows the three measurement data range at 20 mm distance from the stent tip in magnitude, and d) phase.

B. IN VITRO TRANSFER FUNCTION MODEL VALIDATION USING TABLETOP E-FIELD GENERATOR

We calculated the RF-induced power in the −30 dB contour near the DUT scaled to volume scan as 30.8 mW and 10.6 mW for DUT positions 1 and 2, respectively (Fig. 2c). The scaling factor, A, was calculated as 33 μW/(V_rms/m)² at the independent DUT location (Position 3), which was then used to predict RF-induced power using (1) (see Supplemental Materials S1). Absolute mean normalized error was 0.12 ± 0.11 dB at Position 1 (Fig. 5–6) and 0.61 ± 0.25 dB at Position 2.

FIGURE 5.

Comparison of predicted vs. deposited power at the stent tip at sample Position 1 (Blue arrow at top left indicates the corresponding power calculated from the “Touchless” process; see Supplemental Materials for further details)

FIGURE 6.

Examples of tested exposure fields shown in simulation (Normalized to total input power: 10 W). EM simulation was performed without the device present, while Etan was extracted along the pathway of the tested device at each position.

C. IN VITRO TRANSFER FUNCTION MODEL VALIDATION USING WHOLE-BODY TRANSMIT COIL

Results of the in vitro TF model validation test using QBC are shown in Fig. 7. The maximum temperature rise was 4.6°C when the DUT was tested at Position 1. Predicted temperature rise near the stent tip in other DUT positions was then compared with the measured temperature from the QBC experiment in Table I. The error between measured and predicted values was calculated using (2) and shown in Fig 8. Absolute normalized error from the QBC experiment was 0.43 ± 0.18 dB (Fig. 8).

FIGURE 7.

Measured RF-induced heating near the stent tip using the QBC during a 15-minute RF exposure at seven DUT positions.

TABLE 1.

MEASURED AND PREDICTED TEMPERATURE RISES NEAR THE STENT TIP AT POSITIONS SHOWN IN FIG. 3B.

Probe location (x,y) [cm]	Measured [°C]	Predicted [°C]
Position 1 (3.5, center)	4.60	3.98
Position 2 (3.5, 5.0)	1.40	0.80
Position 3 (3.5, 10.0)	2.10	1.73
Position 4 (5.0, center)	3.80	3.48
Position 5 (5.0, 5.0)	1.10	0.69
Position 6 (5.0, 10.0)	1.60	1.51
Position 7 (10.0, center)	1.50	1.96

FIGURE 8.

Normalized error between estimated power vs. measured power near the DUT tip

D. TECHNICAL COMPARISON OF IN VITRO TF MODEL VALIDATION METHODS: TABLETOP E-FIELD GENERATOR AND QBC

Table II shows the uncertainty budget for TF modeling of the DUT using the piecewise excitation method. Tables III and IV show the uncertainty budgets for the tabletop E-field generator and the QBC experiment, respectively. Fig. 9 shows the in vitro TF model validation results using the tabletop E-field generator and QBC. The circle points inside the blue dotted lines represent the predicted and measured deposited power using the tabletop E-field generator (Fig. 9a) and QBC (Fig. 9b). The σ value is the combined standard uncertainty of measurement and modeling at 14.1% ± 1.3 mW for the tabletop E-field generator, and 26.1% ± 0.1°C for QBC [9] following the procedure described in ISO 10974 [3]. The σ value includes the uncertainty from the TF model measurement (Table II) and uncertainty from the in vitro TF model validation using the tabletop E-field generator (Table III) and QBC (Table IV). The in vitro TF model validation results were within the 2 σ boundary with 95% or more confidence in prediction. Finally, Table V summarizes the technical comparison between the proposed and traditional QBC methods.

TABLE 2.

UNCERTAINTY OF THE NORMAIZED TRANSFER FUNCTION MODEL OF THE TEST ITEM [12].

Source of Uncertainty	Std. Uncertainty (dB)
Analyzer drift (20 min)	0.01
Cable movement	0.08
TDS E-field sensor linearity	0.25
piX Excitor positioning (x,y 1 mm)	0.05
piX Excitor positioning (z 5–20 mm)	0.25
piX Phantom boundaries (1 mm)	0.20
Reflected power	0.06
Tissue Simulating Liquid properties (5.5%)	0.13
Device definition (5 mm)	0.05
Post-processing (DoF >5)	0.07
Combined Standard Uncertainty of transfer function model (k=1);	0.45

TABLE 3.

UNCERTAINTY OF THE INCIDENT FIELD EXPOSURE DUE TO THE TABLETOP E-FIELD GENERATOR AND PHANTOM [10], [12].

Source of Uncertainty	Std. Uncertainty (dB)
Tabletop tfd exposure deviation	0.15
Test item position	0.58
Phantom fill-height	0.12
Tissue Simulating Liquid properties	0.25
Estimating power deposition using ‘Touchless’ [10]	0.82
Readout electronics (SAR acquisition system) [12]	± 1.34 mW
Combined Standard Uncertainty of tfd experiment (k=1)	1.05 dB ± 1.34 mW

TABLE 4.

UNCERTAINTY OF THE INCIDENT FIELD EXPOSURE DUE TO THE QBC AND PHANTOM [10], [12].

Source of Uncertainty	Std. Uncertainty (dB)
Phantom position	0.41
Phantom fill-height	0.12
Test item position	0.58
Tissue Simulating Liquid properties	0.25
Polarization	0.16
Temperature probe position (mm) [10]	1.80
Readout electronics	± 0.10°C
Combined Standard Uncertainty of tfd experiment (k=1)	1.96 dB ± 0.10°C

FIGURE 9.

Tabletop E-field generator and QBC comparison through the predicted deposited power on the DUT tip versus measured deposited power. a) in vitro TF model validation results using the tabletop E-field generator; b) in vitro TF model validation results using the QBC.

TABLE 5.

TECHNICAL COMPARISON BETWEEN IN VITRO TRANSFER FUNCTION VALIDATION METHODS USING A TABLETOP E-FIELD GENERATOR AND QBC.

performance	Tabletop E-field Generator	QBC
Phantom equilibrating time to room temperature	N/A	2 hours
Total measurement time during RF exposure	10 mins (5 s per test field)	2 hours (15 mins per test)
Volume scan	1 hour 30 mins (30 mins per position)	N/A
Background scan	48 mins (1 min per test field)	(Optional: 15 min)
Total cooling time (DUT and probe positioning time)	N/A (10 mins per position)	4 hours (30 mins per position)
Number of test conditions (+ scaling factor)	32 (+16)	7 (+1)
Total required test time for in vitro TF model validation	≅ 2.5 hours	≅ 8 hours
Number of linearly independent incident Etan	21	7
Total std. combined uncertainty for exposure system experiment (dB)	1.05 ± 1.34 mW	1.96 ± 0.10°C
Errors in comparing predicted vs. measured RF-induced power (dB)	0.37 ± 0.31	0.43 ± 0.15

E. EXAMPLE TIER 3 ANALYSIS FOR RF-INDUCED HEATING IN THE HUMAN TISSUE

Results of Tier 3 analysis example are shown in Fig. 10. We observed elevated power deposition in the tissue at the imaging position between 200 and 800 mm when the imaging landmark was chosen close to the generic implant routing for cardiovascular devices. Maximum estimated power deposition near the DUT tip was calculated to be 48.3 mW (comparable to 48.3 W/kg SAR over one gram of tissue; 1gSAR) at the imaging position z = 600 mm (Fig. 10b) at the maximum allowed power level of the MR imaging system in Normal Operating Mode, as defined in IEC 60601-2-33 [16]. Fig. 10c shows the corresponding volume-averaged SAR value used to scale up the power level in the MR imaging system, as a function of the position of a Duke model inside QBC (Diameter: 60 cm, Length: 70 cm).

FIGURE 10.

Examples of Tier 3 approach defined in ISO/TS 10974 [3]. a) Duke model with representative generic implant lead routing for cardiovascular devices in red; b) Tier 3 deposited power as function of the position (x) of a Duke anatomical model inside all simulations; c) corresponding volume-averaged SAR values as function of the position (x) for the Duke model inside a high-pass birdcage coil (diameter: 60 cm, Length: 70 cm).

IV. DISCUSSION

In this study, we investigated a novel in vitro TF model validation method using a tabletop E-field generator with a stent in two device positions. We measured TF using the piecewise excitation method and predicted RF-induced power at the stent tip as a part of the TF model validation procedure. Also, we measured the temperature rise near the stent tip with the traditional method of a QBC in seven device positions to compare the new TF model validation method with the conventional one. The induced heating at the stent tip was relatively low in both cases due to the low conductivity of Nitinol from which it was made, and potentially due to the effects of the coating applied on the stent. Measured DC resistance of the stent was 13.4 Ω, > 30 times higher compared to the 0.4 Ω measured from the 10 cm stainless steel rod. This could be due to the meshed structure. Stents made with different designs and materials may produce higher induced heating [17], therefore it is important to consider the specific stent designs and materials for the potential RF-induced heating in MR environments. We found that there was phase reversal of the TF model at the end of the DUT sample (Fig. 4). This is due to the open-ended condition of the DUT. We note that Fig. 4 shows the highest potential heating from the piece-wise TF in magnitude at one end only. Meanwhile, we investigated the induced heating at both ends of the stent by reversing the TF orientation in Tier 3 analysis, assuming that the other end would generate an equivalent heating pattern (see Supplemental Fig. 2).

Consistent experimental condition control is crucial to validating the measured TF model with reduced uncertainties. Thus, the inaccurate positioning of the device and a sensor probe (e.g., point SAR or temperature), in the testing medium may introduce unintentional deviation. It is reported that both temperature and SAR rapidly drop near the hot spot [10], where the submillimeter positioning accuracy is required to achieve robust validation, and where the probe positioning error in SAR measurement was corrected with the co-register of volume scan using the “Touchless” tool in the proposed method [10] (see Supplemental Materials S2 for details). The tfd method enabled the reduced necessary test sample positions to validate the TF by manually alternating induced E-field exposed to the device [8] while reducing the probe position error by co-registering the probe position using volume scan to calculate RF-induced power at the device tip [10].

The relative probe measurement position to the tabletop E-field generator can be accurately controlled using the robot arm. In contrast, the conventional TF validation using a whole-body birdcage coil required more effort and time to position the sample inside the phantom. The normalized error from the TF validation using the conventional birdcage coil was 0.43 ± 0.18 dB compared to the 0.37 ± 0.31 dB in the tabletop E-field generator. The DUT at Position 2 generated relatively high errors compared to Position 1 results, producing a maximum error slightly higher than that of the traditional TF model validation method. This was due to the low SNR (deposited power near the DUT tip) as the incident E-field near the center of the tabletop E-field generator is small. Position 2 provided easy control of DUT positioning without complexity, but one may choose the curvature trajectory near the E-field hot-spot on the side, especially when the device is short (Fig. 6). Validating TF model in various test fields with reduced necessary device positioning demonstrated the equivalent overall measurement accuracy. Sample positioning was relatively easy; the sample was fixed on two types of racetracks; an additional test was required for scaling factor calculation. At the same time, the transparent liquid that filled inside the tabletop E-field generator was also beneficial to accurately positioning the test sample and the dosimetric probe.

Furthermore, the incident E_tan required for the in vitro TF model validation was assessed using the 10 cm DUT trajectories used in this study (Table V), and was further evaluated with 20 cm lead pathways using a tabletop E-field generator compared to the QBC (Supplemental Fig. 4). The tabletop E-field generator could produce the linearly independent incident E_tan by alternating the relative gain and phase of the transmit sources with a minimum of two necessary DUT positions. Meanwhile, the traditional QBC experiment required pre-adjusted routing shapes and test positions to create various incident field conditions.

The primary advantage of using a tabletop E-field generator is a dramatic reduction in the total required time from more than eight hours to less than three hours compared to the traditional in vitro TF model validation process using a birdcage coil. This is mainly due to the reduced number of necessary sample positions required for the test; two test positions were enough when combined with 16 test field condition resulting in 32 exposure scenarios and an additional test for the scaling factor calculation. The point SAR measurement with a dosimetric probe only required less than five seconds per test field, while volume scan at the representative test field and reference background scans for 16 test fields could take 15 minutes for each task (Table V). Direct SAR measurement in a liquid medium does not require cooling time between measurements; thus, the entire test could be efficiently managed.

Meanwhile, the birdcage coil test required device and temperature sensor positioning in a torso-sized gel phantom. Each experiment needed 20 minutes preparation time for the device and additional probe positioning which could vary depending on the tested device. In addition, cooling time was required to achieve the thermal equilibrium between the gelled saline and surroundings before the next test with new DUT position; achieved by manually stirring the gel and letting it cool until background temperature was stabilized. This cooling required at least 30 minutes, sometimes more than one hour depending on the amount of heating that occurred during the previous test. Also, changing the positions of the sample and probe is cumbersome; requiring one to remove gel phantom outside the birdcage coil, and rearrange the sample and probe positions by either elevating the baseline grid to the top of the phantom box or draining the portion of gel and refilling it after the device positioning. Overall, the tests at each sample position required at least one hour, multiplied by the number of test positions.

In addition, the tabletop E-field generator did not require a shielded room, and the 12 W amplifier and control system required only a small portion of the laboratory. Our laboratory had an industrial-grade precision robot arm (DASY52, SPEAG), which was the primary occupier of physical space. Although the proposed process does not require a large working area, the alternative smaller robot-arm system may be compatible with the proposed task. The birdcage coil experiment was performed in a dedicated shielded room to avoid electromagnetic interaction with the surrounding environment and to maintain a controlled environment for the stability of the ambient temperature. Additionally, a high-power amplifier (output: 10 kW) is required to operate a QBC resulting in loud sounds during the experiments. Thus, soundproofing was necessary to protect the operator's hearing during the internal calibration and the RF exposure run.

Alternatively, the proposed method can be implemented in a gel-type phantom with a fiber optic temperature probe, which still has the benefit of reducing the necessary sample positions by controlling the RF exposure by relative phase and amplitude of the transmit sources. However, unlike direct E-field measurements, volume scans of temperature cannot be performed. This is due to thermal convection, which adds uncertainty from probe positioning. Also, sufficient RF-induced heating at the sample tip may require a high signal-to-noise temperature ratio to counteract the inherent precision of the fiber optic probe (e.g., ΔT ≥ 1°C is recommended when using the probe with the precision of 0.1°C [6]). This may only be applicable to test devices producing large amounts of RF-induced heating, or those which use a temperature probe with a precision greater than 0.1°C.

This study focused on the practicality and validity of TF model validation using a tabletop E-field generator. However, the ultimate goal of the Tier 3 method is to assess the MRI safety of implantable devices with elongated leads in various scenarios, especially active implantable devices, in clinically relevant device trajectories. These clinically relevant trajectories are often complex – including a couple of loops to allow patient movement and differ from individual to individual. MRI safety assessment can be reasonably conservative with reduced overestimation when various patient populations (various ages, sexes, and body mass indices) are considered in various landmark positions with several whole-body coil sizes and transmitting polarizations. This allows safety assessment of >10,000 scenarios without computational burden using the validated TF model.

Limitation:

The tabletop E-field generator may produce slightly different relative phase and gain sets from target values during deposited power measurement near the DUT tip, volume scan, and background scan. These differences were included in the uncertainty analysis. Stent geometry is stretchable in both length and shape; combined uncertainties may be higher when it is tested in a trajectory with curvature. The Tier 3 analysis with the Duke human model shown in Fig. 10 was conducted using a significantly under-sampled MRIxViP database to provide insight into the example application of the Tier 3 method defined in TS/ISO 10974. Thus, Tier 3 analysis results may be different from the results generated by full MRIxViP database. Also, the test sample was a Z-shaped stent consisting of 18 struts in one segment with irregular physical links between segments in the longitudinal direction. We assumed that the stent had a symmetrical wire mesh geometry when calculating the deposited power in −30 dB contour near the stent tip (see Supplemental Fig. 2).

V. CONCLUSION

This study showed that, compared to traditional methods, a tabletop E-field generator used for in vitro TF model validation of RF-induced heating assessment resulted in a 50% reduction in measurement time and equivalent agreement between RF deposited power estimation and measurement. The traditional in vitro TF model validation process requires test measurements using multiple device positions in the ASTM phantom. This is time-consuming work; whereas using the tabletop E-field generator that produces multiple incident E-field exposures enabled the completion of 16 test measurements in approximately one hour for a given sample position. Here, we demonstrated the validity of a proposed novel approach for assessing the RF-induced heating of a stent, an elongated passive implantable device.

We employed an alternative compact E-field generator for the in vitro TF model validation and the assessment of the RF-induced heating. The tabletop TF model validation method was simpler, faster, and cheaper than the current standard approach. Its advantages were that it did not require a shielded room, testing took less than three hours vs. more than eight hours, measurement uncertainties were reduced, and agreement in predicted RF power deposition improved. This tool may be useful in supporting MR safety evaluation of implanted medical devices.

Supplementary Material

APPENDIX

This article has supplementary materials provided by the authors and color versions of one or more figures available at 10.1109/ACCESS.2024.DOI_Number.