Wirelessly interfacing sensor‐equipped implants and MR scanners for improved safety and imaging

To investigate a novel reduced RF heating method for imaging in the presence of active implanted medical devices (AIMDs) which employs a sensor‐equipped implant that provides wireless feedback.


INTRODUCTION
Millions of patients worldwide have already been implanted with devices such as deep brain stimulators (DBS), cardiac pacemakers, or spinal cord stimulators, so called active implantable medical devices (AIMDs).Further advancements in device development aim to enhance the quality of life for patients with conditions such as amyotrophic lateral sclerosis, 1,2 vision impairments, 3 speech synthesis, 4 or psychiatric disorders. 5,6lthough MRI is the preferred imaging modality for diagnosing various pathologies, patients with AIMDs may be directed toward other imaging modalities due to the potential risk of RF-induced tissue heating near the implant tip. 7,8MR conditional implants that comply with established standards do exist. 9,10However, scanning these implants often requires significant limitations on RF transmission power, leading to reduced imaging performance and necessitating modifications to imaging sequences and clinical protocols. 11,12Additionally, clinical personnel must adhere to complex guidelines provided by the manufacturers for imaging these implants, which is time consuming, error prone, and also imposes liability onto the radiologist and/or technologist/radiographer. [13][14][15][16] Despite the complexity involved, incidents of burns in MR conditional AIMDs continue to be reported. 17][20][21][22][23] However, the extent of RF-induced heating experienced by an AIMD depends not only on the device itself but also on patient-specific tissue parameters and the distribution of the electric field (E-field) generated by the RF transmitter. 24For instance, the absorbed power around the tip of a DBS electrode can vary by 88% when comparing the same device implanted in different patients, 25 and surgical management of the lead trajectory can significantly impact RF-induced tissue heating. 26,27 promising method to improve implant safety without the need for modifications of the implant itself is the utilization of parallel transmission (pTx) technology.][34][35][36][37][38] The translation of simulation-based methods into clinical practices poses several challenges.One of the main obstacles is the lack of patient-specific simulation models.Even if such models were available, even a slight change of the patient's position within the RF coil could result in strong changes in RF-induced tip heating. 39o overcome these difficulties, measurement-based techniques offer a solution by evaluating the safety risk in vivo and in situ.Sensor measurements of current 40,41 and/or E-field 39 amplitudes and phases within or around an implanted lead have proven to be effective strategies for mitigating tip heating and even enabling control over imaging quality. 39,40,42AIMDs are typically completely implanted inside the body and consist of intricate electronic circuitry.This characteristic makes it challenging to apply phase-sensitive, bulky time-domain sensors, except in certain interventional applications. 39,404][45] Despite the absence of phase information in RMS sensors, studies have demonstrated that a so-called sensor Q-matrix (Q S ) methodology can still be applied to address tip heating and maintain the quality of MR images. 42][48][49][50][51][52] The utilization of embedded sensor data from an implant, wirelessly transmitted to a pTx capable MRI system, holds tremendous potential for enhancing implant safety.This "smart" implant safety concept would define clear responsibilities for both the implant manufacturer and the MRI system manufacturer, thus relieving the radiologists and technologists from this burden in clinical settings.Bluetooth low energy (BLE), already widely used in wireless body area networks found in implants, can be employed for establishing such communication. 53,54BLE operates at GHz frequencies, which do not interfere with the 100 MHz range used in MRI.][58] In this study, we introduce a wireless reference implant that incorporates embedded RMS sensors capable of communicating with a pTx capable MRI system.Phantom experiments were conducted using a 300 MHz (7T) testbed 39 and a 3T scanner, considering various implant positions and configurations including six realistic DBS lead trajectories derived from patient data.It is investigated whether these implant prototypes can detect RF-induced tip heating based on a small foot-print circuitry placed within the generator case that detects signal from the tip.The captured signal was then transmitted to the MR scanner using BLE communication.Subsequently, an excitation mode was selected based on the sensor signal alone, aiming to minimize tip heating while ensuring minimal impact on image quality.

Wireless reference implant
The reference implant consists of a casing with embedded electronics, which is connected to an implant lead (Figure 1).

Hardware
The schematic of the electronic circuit of the reference implant is shown in Figure 1A.The BLE v5.signals coming from the implant lead tip.A 7.8 μs sampling period was programmed, where 5.3 μs was used as on-chip averaging time.In addition, an RF switch module was implemented based on a PIN diode (D1).D1 is biased and controlled by the SoC via the digitally adjustable resistance R1 and enables sensor signal measurements to assess implant heating while simultaneously protecting the ADC.L1 serves as an RF choke to attenuate RF-induced voltages at the digital port, protecting the port from high peak power RF pulses during MR experiments while D1 is biased.L2 is used as the DC return path for the PIN diode bias, and C1 and C3 are used to block unwanted DC currents.An SMA connector was used as an interface to the implant lead.The outer shield of the connector is connected over the feedthrough capacitor C2 to prevent spurious RF signals from the outer shield corrupting measured signals from the implant lead tip.
The electronics circuit was enclosed within a metallic box (1455D601RD, Hammond Manufacturing, Guelph, Canada) acting as a Faraday cage with a 2.4 GHz Bluetooth antenna (1003893FT-AA10L0025, AVX Corporation, USA) located outside (Figure 1B).The antenna was connected through a high-pass filter 46 adjusted to pass the Bluetooth signal, while blocking MR frequencies.A photo of all connected components as used in an experimental setup is shown in Figure 1C.

Firmware
0][61] To be able to measure the sensor Q-matrix (Q S ) 42 with the implant hardware and transfer this data wirelessly, the maximum transmission unit size was set to 251 packets and the physical link layer was adjusted to 2 Mbps.In addition, the connection interval was set to 7.5 ms without any slave latency between the master and slave.These settings provide the maximum throughput to transfer the measured Q S data, including a fast, wireless trigger detection.The ADC measurements were controlled by an independent microcontroller embedded in the SoC and programmed via its own integrated development environment. 62After saving the data to the internal RAM of the SoC, they were transferred wirelessly to the PC.The total measurement time of the Q S was 58 ms.

Software
The communication between reference implant and pTx console is controlled by a self-written Python software running on the pTx console.

Implant leads
Six models of realistic DBS lead trajectories extracted from patient data 63,64 were converted into CAD models and 3D printed (S5, Ultimaker BV, Utrecht, Netherlands) (Figure 2).Afterward, a semi-rigid coaxial wire (2 mm diameter, 15 mm uninsulated tip) was bent along the trajectories to mimic complex and realistic implant settings.The end of the cable was soldered to an SMA connector and connected to the implant casing (Figure 1C).Please note that the implant leads in this work do not have any sensors embedded at the implant tip.Lead tip signals are deduced from induced voltages between the inner and outer conductor of the coaxial wire and measured in the generator casing.

Wireless communication workflow
The communication workflow between implant and pTx console is shown in Figure 3. Reference implant and BLE server establish a connection before any RF transmission.The connection parameters, implant power settings, and digital control switches are initially set by the pTx console before any trigger event.Afterward, the BLE server waits for a trigger from the pTx console indicating the start of RF transmission.The following RF-induced signal is recorded Implemented implant leads derived from realistic DBS implant lead trajectories. 63The top black structures are 3D printed models.The bottom series are bent semi flexible coaxial wires with 2 mm diameter.The end of the coaxial wire was soldered to an SMA connector that can be connected to the implant casing or the extension wire.Fifteen millimeter insulation was removed from the implant tip to mimic a DBS electrode.

F I G U R E 3
The wireless communication workflow for the AIMD interface with the pTx console is shown.In the initialization phase (black dashed box), a BLE connection is made between implant and pTx console, and the connection parameters are adjusted.In the data acquisition phase (green dashed box), the implant records the RF-induced signals from the implant tip after an initial trigger signal indicating the start of an RF transmission and transmits the measured data to the pTx console via BLE server.After detection of the induced signal or calculation of the pTx mitigation pulses by calculating Q S , the implant disconnects from the console.
by the implant and wirelessly transmitted to the pTx console.The pTx console computes Q S to derive mitigating RF settings or simply reports the amount of RF detected by the implant.

Testbed experiments
Benchtop experiments have been performed in an implant safety testbed 39 using an eight-channel pTx system at 300 MHz and an 8Tx/8Rx 7T head coil (Figure 4A). 65nsor calibration The sensor readings that are performed in the implant generator case of the wireless implant were compared to external reference E-field and temperature measurements at the tip of the electrode as shown in Figure 4B.
The wireless reference implant with extension wire and Lead 3 was placed into a cylindrical PVP phantom. 42The lead was fully immersed in the phantom with its uninsulated tip 15 mm below the liquid's surface.A time-domain E-field probe (E1TDSz SNI Speag, Zurich, Switzerland) was positioned closely to the implant lead tip to measure the induced radial E-field component concurrent with the implant sensor recordings.

F I G U R E 4
Photographs of the experimental setups.(A, B) The experimental setup for the testbed experiments at 300 MHz.The implant lead with extension wire was fully immersed into a cylindrical phantom.A pTx console and a 7T eight-channel pTx RF coil were used for RF excitation. 39External fiber-optic temperature probes and a time-domain E-field probe were positioned at the implant lead tip to correlate implant sensor signals with measured E-fields and temperatures.Implant sensor readings were wirelessly transferred to the pTx testbed console.(C) depicts the MR experimental setup of an ASTM phantom with the reference implant case positioned in the upper chest area of the phantom, while the implant lead tip was in the head part.The latter was inserted into an eight-channel 3T pTx coil.The sensor measurements in the implant were wirelessly triggered by the MRI system and transferred to an external computer for monitoring and calculations of the pTx transmission modes.
A total of 1000 random excitation vectors were transmitted for E-field measurements.Out of these, 24 excitation vectors were selected for RF heating experiments (60 s continuous RF heating per excitation) to measure the tip temperatures using external fiber-optic temperature probes (CANSAS FBG-T8, Imc Test Measurement GmbH, Berlin, Germany).
The correlations of the measured data from implant, E-field, or temperature probe were analyzed by Pearson correlation coefficient (R) for a linear model without intercept: f (x) = a * x.

Detection and pTx mitigation
The sensor Q-matrix (Q S ) was measured by the implant to characterize the induced E-fields from the eight-channel RF coil. 42Based on measured Q S , the worst-case (WC) excitation vector was computed, here defined as the eigenvector to the largest eigenvalue of Q S .The orthogonal projection method (OP) was applied for heating mitigation by projecting the circular polarized (CP) mode on the orthogonal subspace of WC. 39,42 In addition, the null mode (NM) method 40 was computed, by selecting the vector of the minimum absolute value of the singular value decomposition of Q S .Only a single NM was used for the experiments.All four modes were normalized to the same total transmit power.

Experimental setup
MRI experiments were performed at a 3T scanner (Verio, Siemens Healthineers, Erlangen, Germany) with eight-channel pTx array.The implant casing was connected over a 500 mm extension to the extracted DBS lead trajectories, giving a total wire length ∼1200 mm to mimic realistic DBS scenarios 66 (Figure 1C).The implant was then immersed in an ASTM like body phantom filled (h = 14 cm) with a PVP solution ( r = 50,  = 0.33 S/m) (Figure 4C).The implant casing was fixed using a 3D printed Polylactic acid holder 7.5 cm away from the "shoulder" and ∼3 cm away from the phantom wall outside the RF coil.The implant tip was positioned within the ASTM phantom head section using a 3D printed holder from Polylactic acid.The head section of the phantom was inserted into an eight-channel pTx coil (RAPID Biomedical, Rimpar, Germany) and connected to pTx system of the MR scanner.The trigger signal, indicating the start of the pulse sequence on the scanner, was fed to the BLE server to activate the receive switch on the implant during MR transmission.
The trigger of an RF-only FID pulse sequence of the pTx system was set to 100 ms before the RF pulse to accommodate for the internal processing time of the SoC.Amplitude and phase of the RF pulses were determined from the text files stored in the pTx console of the MR scanner.The step size of the pulses was 10 μs and the maximum RF pulse length was 7.8 ms at the scanner.Therefore, the duration for the Q S pulses was adjusted to 60 μs which is followed by 50 μs idle time.The eight-channel Q S acquisitions thus could be performed with a single pulse sequence of 7.5 ms (50 V peak , P total = 48 mW, measured at the output of the amplifiers).For the assessment of the RF-induced signal on the implant for different pTx excitations, a single rectangular RF pulse was transmitted (50 V peak , P total = 35 mW).The amplitude and phases of the pTx pulses were adjusted using the pTx console's GUI.

BLE timing uncertainty and repeatability
BLE is a discrete communication scheme where master and slave devices maintain the connection for a limited time.Then, they reestablish the connection after a predetermined time period, which is so-called connection interval. 61The minimum connection interval 7.5 ms is used in all experiments.However, this creates a 7.5 ms deadtime between two devices meaning that an RF trigger may fall within this timeframe.In addition, there is a probability of having undetectable RF-induced signals on the implant for some channel combinations during a Q S acquisition.Therefore, four additional RF pulses (forward and reverse circularly polarized) were inserted at the start and at the end of the Q S pulsetrain.These pulses were used to detect the start of the RF pulse to mitigate the timing uncertainty of the BLE communication.This method was investigated in an MRI setting using 10 consecutive Q S acquisitions.

RF heating
RF heating experiments were performed for all implant lead trajectories.The same external fiber-optic temperature probes were used to measure RF-induced heating for the above-mentioned (WC, CP, OP, and NM) pTx excitation scenarios.The same FID sequence (TR = 200 ms) was adjusted to a total time of 67 s.The total transmit power for all pTx excitation settings was 22 ± 0.5 W (measured at the output of the amplifiers).Please note that RF coil and cable losses are substantial and further decrease the effective transmitted power.RF heating was performed additionally for 400 s for Lead 3 to investigate temperature increase and mitigation performance for a longer period of RF heating.

Imaging evaluation of sensor based pTx mitigation modes
A GRE sequence (TE/TR: 2.41/221 ms; resolution: 1.0 × 1.0 ×5.0 mm 3 ; matrix size: 512 × 512; multislice (20): interleaved; gradient bandwidth: 980 Hz/Px; averages, two; measured power at the output of the amplifiers: 3.75 ± 0.5 W) was used to evaluate the imaging performance of the pTx mitigation modes (OP, NM) against WC and CP mode transmission.To evaluate the imaging performance of a pTx excitation mode based solely on sensor measurements, an imaging score (S image ) is introduced, which is defined as: where complex-valued u im describes the pTx excitation vector for the target imaging mode, for example, CP in this study; u opt is the optimized excitation for implant imaging for example, OP or NM. is the angle between the two excitation vectors and can be calculated by the vector dot product.S image indicates the geometrical similarity between a calculated pTx excitation and the reference imaging mode.Consequently, S image = 1 indicates that the imaging performance is on par with the CP mode.

Testbed experiments at 300 MHz and sensor calibration
The results from the simultaneously acquired implant sensors signals and time-domain E-field probe recordings are shown in Figure 5A with a good correlation of R 2 = 0.929.The correlation between implant sensor readings acquired in the implant case and fiber-optic temperature probes at the lead tip is R 2 = 0.954 (Figure 5B).For small RF-induced sensor signals (<0.1 V), the readings are close to the noise floor due to the limited resolution of the 12-bit ADC in the SoC, while for pTx pulses with larger induced implant signals (>0.26 V), a saturation of the time-domain E-field probe (>2300 a.u.) was observed.This is also visible (Figure 5C) from correlating the E-field probe measurements at the implant tip with the temperature measurements at the implant tip (R 2 = 0.824).

3.2
Wireless Q S acquisition The correlation between implant sensor measurements and the time-domain E-field probe measurements of the radial E-field components at the implant tip using 1000 random pTx excitation pulses is shown.For the low induced voltages, E-field probe sensitivity is higher compared to the implant sensor, whereas E-field probe saturation is observed for higher recorded E-field values.A good correlation (R 2 = 0.93) is found between implant sensor and E-field probe readings.(B) A total of 24 pTx pulses from (A) were used to perform RF heating measurements with fiber-optic temperature probes as a reference.Fiber-optic tip temperature rises show a good correlation (R 2 = 0.95) with implant sensor measurements from the wireless implant.(C) The correlation (R 2 = 0.82) between the fiber-optic tip temperature rises and the measured E-fields at the lead tip using the time-domain E-field probe.

F I G U R E 6
Repeatability results for wireless Q S acquisitions inside the MR scanner.(A) Ten consecutive 7.5 ms long raw acquisitions of the Q S pulsetrain illustrating the timing uncertainty (7.5 ms) of the BLE connection interval parameter.(B) Time corrected Q S data.The first and last two pulses are used as timestamps for synchronization.Violin plots 100 depicting SD in the wirelessly acquired OP, which were less than 1% for both (C) amplitudes and (D) phases.pulses (amplitudes in Figure 6C and phases in Figure 6D) of the OP method was less than 1% for all channels.

MR experiments
Input voltage transmitted by the MR system versus measured sensor signal in the reference implant is shown in Figure 7.Over the whole investigated range, the simulated non-linearity correction works well as indicated by a strong linearity (R 2 ≥ 0.99 for all eight channels), despite the hostile MRI environment (B 0 , high peak RF power).Similar to the testbed experiments, the implant sensor readings become unreliable below 0.1 V and saturation is observed above 1.6 V (cf.The results of the MRI experiments using all six DBS lead configurations are depicted in Figure 8.The normalized amplitudes and phases of the Q S acquisition using the wireless implant were utilized to identify the worst-case RF excitation and calculate the OP mode (Figure 8A). Figure 8B displays the RF-induced implant sensor signals for the WC, CP, and OP excitation scenarios.Across all lead configurations, the WC mode generated the highest signals, ranging from 0.572 V to 1.841 V.The CP mode produced measured voltages ranging from 0.223 V to 1.335 V.The OP and NM mitigation modes effectively reduced the induced RF signals for all lead configurations, resulting in readings within the implant noise floor.Specifically, for the OP mode, the measured voltages were below 0.137 V, while for the NM mode they were below 0.125 V.The corresponding temperatures recorded with the fiber-optic temperature probes are presented in Figure 8C for all lead configurations and RF excitation modes.The WC mode resulted in the highest RF heating, ranging from 0.52 K to 3.33 K, followed by the CP mode with temperature rises ranging from, 0.07 K to 1.28 K.The OP mode successfully reduced RF heating for all lead trajectories, with temperature rises between 0.03 K and 0.14 K. Last, the NM excitation exhibited the lowest temperature rises at the tip, ranging from 0.00 to 0.07 K.
RF heating experiments of longer duration (400 s) were conducted on Lead 3, as shown in Figure 9.The recordings of two fiber-optic temperature probes (positioned ∼1 mm apart at the lead tip) for WC, CP, OP, and NM modes as used in Figure 8.In the pTx experiments, the WC mode resulted in a maximum temperature rises of 3.14 K (probe 1) and 2.24 K (probe 2), stabilizing after ∼150 s, even though RF transmission was still ongoing.For the CP mode, RF-induced heating reached 0.81 K (probe 1) and 0.67 K (probe 2) after 400 s.The OP mode recorded temperature rises of 0.05 K (probe 1) and 0.02 K (probe 2), while the NM mode showed 0.12 K (probe 1) and 0.10 K (probe 2).
Coronal MR imagines for two lead configurations and excitation modes are shown in Figure 10.MR images of all six leads are provided in Supporting Information Figure S1, while transversal images can be found in Supporting Information Figure S2.Image quality of the OP mode was often (S image ≥ 0.8) but not always (S image = 0.7 for Lead 5) comparable to the CP mode.For a single NM mode S image was between 0.17 and 0.56.

DISCUSSION
This study introduces implant hardware with embedded sensors capable of wirelessly transmitting safety-related data to an MR scanner.It was demonstrated that these data strongly correlate with the induced E-fields and temperatures near the tip of the implant lead, allowing the MR scanner to automatically assess the safety status of the implant in situ.Moreover, it was shown that utilizing the implant sensor readings can substantially reduce tip heating while preserving image quality when using pTx.These results were achieved through experimental investigations on various complex and realistic DBS lead configurations, without relying on prior simulations, solely based on the sensor readings.The findings indicate the feasibility of a "smart" implant safety concept, where the implant communicates with the MR scanner to significantly enhance implant safety.

Comprehensive implant safety concept
Based on the demonstrated feasibility in this study, it is possible to establish a comprehensive safety strategy that includes native pTx safety (without implant). 42,67Native pTx safety can be ensured through state-of-the-art electromagnetic field simulations conducted on human voxel models to calculate Q-matrices in advance, which enable the generation of safe excitation vectors.In the presence of an implant in a patient, the Q S of the implant can be measured in situ, transmitted to the MRI system, and combined with the native safety Q-matrices.In addition to ensuring RF safety, optimizations can be performed to improve image quality when an implant is present.Simulations have shown that using this strategy with an eight-channel body coil at 3T, mean B + 1 can be increased three-fold compared to CP excitation, while still maintaining a maximum temperature increase of the implant tip of ≤1 K. 67 Although for 7T eight-channel pTx systems and potentially eight-channel body coils might become available RF heating experiment performed using Lead 3 for a duration of 400 s using two fiber-optic probe locations (∼1 mm deviation).The WC mode shows the maximum temperature rise at probe 1 that is about 3.15 K after 200 s.For probe 2, the maximum temperature rise was at 2.24 K about ∼1 K lower.For the CP mode, temperature rises were 0.81 K and 0.67 K for the probes 1 and 2, respectively.For the OP mode, the temperature rises 0.05 K (probe 1) and 0.02 K (probe 2), for the NM transmission 0.12 K (probe 1) and 0.10 K (probe 2) were measured.
in the future, 68 current clinical 3T systems are limited only to two transmission channels.Studies have shown that even with these two-channel systems, it is possible to mitigate RF-induced heating from the implant while simultaneously preserve imaging performance compared to single-channel transmission. 37,42,69onetheless, further evaluation is required to determine the specific conditions under which the limited degree of freedom offered by two-channel systems is able to maintain image quality while ensuring safe RF excitations.
Another advantage of the proposed concept is its scalability to multiple implants or an implant with multiple leads. 27,36,70The MR system could easily consider Q S information from multiple sources to calculate pTx excitation vectors under RF safety constraints.
The wireless communication between an implant and an MR scanner, once established and standardized, is an interesting asset not only for RF safety of implants but also for many other novel applications that rely on in vivo sensor information during MRI scans.For example, investigations of neuromodulatory effects of DBS using fMRI. 71Similarly, applications with dynamic modifications of implant location and imaging parameters, or the insertion of external devices under MR guidance, such as in interventional MRI, would also benefit from an established infrastructure.In the case of interventional MRI, the Q S information could be easily utilized by the MR system to transmit high-or low-intensity tip signal excitation modes specifically targeted to visualize the tip of the interventional device. 41

Hardware and wireless communication
The reference implant was designed to measure the RF-induced signal at the tip using a rectification circuit in the generator casing.Small RMS sensors in SMD size can be embedded in realistic implants for this purpose.Similar components that measure the neural activity from the target tissue at the electrode are already present in existing closed-loop stimulation devices, which could also be applied to RF safety. 72,73In addition, alternative designs for RF detectors can be used in conjunction with the methodology presented here. 74,75e used off-the-shelf components for the implant casing, which needed additional copper foil to improve RF shielding of the implant electronics when exposed to the high peak powers of the MRI system.A new, better shielded design of the implant casing may improve the results further.In addition, for the antenna, an inverted-f antenna design can be used with the antenna and lead connections embedded in an insulated header. 47ccording to the current BLE standard, 55 the connection interval cannot be reduced to less than 7.5 ms, which would set the theoretical limit to acquire an eight-channel Q S to 15 ms (7.5 ms Q S +7.5 ms BLE timing uncertainty).A fixed 100 ms time delay between the RF trigger from the scanner and the start of the RF pulse is furthermore present due to internal delays in the SoC, leading to the total time of 115 ms to acquire Q S signals for an eight-channel pTx system.This delay could be reduced by application-specific circuits and optimized firmware.
Another timing limitation arises from the amount of data that is being transferred between the implant and the scanner.In our implementation, the complete Q S raw data measurements were transmitted to the pTx console, where all subsequent calculations were performed (extraction of Q S matrix from raw data, calculation of eigenvectors and eigenvalues, and calculation of pTx excitation vectors).Some of these computations could, in principle, also be performed on the implant to transfer a reduced dataset, for example, only the calculated Q S and not the fully sampled raw data.It should also be noted that there can be MR-specific timing limitations, such as RF pulse TR or RF coil ringing, which may impact Q S acquisition and need to be considered.

Lead design
In this study, realistic DBS lead trajectories were derived from patient data. 63,64For DBS leads of different lengths, modified lead trajectories or a different structure of the lead tip, distinct absolute RF heating values are expected. 23,76,77In our investigated configuration, a voltage is measured between inner and outer conductor of the implant lead.The inner conductor is shielded along the implant lead path, with its main contribution to the measured signal coming from the implant lead tip.Similar shielded DBS lead designs exist, 78 and for DBS implant leads with multiple electrodes, differential signals could be used.A limitation of the current study is the limited number of lead trajectories and real implant leads used.Therefore, further investigations are necessary to explore the robustness and uncertainties of the presented methodology for a larger set of realistic implant lead configurations. 67,79hile it is appealing to position the readout electronics in the implant casing to avoid modifications of the implant lead design, sensors embedded directly at the implant's lead tip could also be used.For example, it has been previously demonstrated that temperature sensors (e.g., thermistors) embedded at the implant tip can also be used to measure Q S and mitigate RF heating. 42uch sensors are already embedded, for example, in RF ablation catheters. 43,441][82] Temperature measurements can be transmitted to the scanner using a wireless link and they can provide an additional watchdog functionality during the MR scan, which improves patient safety. 46

Sensor calibration and reference E-field and temperature measurements
The implant sensor in the generator case was calibrated against E-field and temperature reference measurements at the lead tip (Figure 5B).E-field probe orientation and positioning with respect to the implant lead tip is crucial and small variations will impact absolute E-field readings due to strong E-field gradients close to the lead tip.Temperature sensors might provide a more robust calibration setting, and the temperature readings can be also converted to lead tip SAR values. 67,79These calibrations are in principle frequency-dependent and need to be performed at each field strengths, independently.Similar to the transfer function assessment of implants, the electromagnetic and thermal properties of the investigated tissue will alter lead tip heating and need to be accounted for as an uncertainty of this approach (Supporting Information Figures S3 and S4). 10,79,83,84Insulated leads have the hot-spot location typically focused around their uninsulated lead tip. 8,85In a more general case, the hot-spot location should be determined beforehand to assess the effectiveness of sensors to suppress unwanted RF heating.Validated simulations can help here to investigate multiple configurations. 79he SoC used in this study has a 12-bit ADC, which has a maximum of 11.6 effective bits 86 translating to <4096 quantization or approximately to 1 mV resolution (without diode non-linearity correction).The Schottky diode has a gating threshold; therefore, very low RF-induced currents may not be detected by the implant (Figure 5A).In a watchdog functionality, this is less critical since low sensor values effectively mean low tip heating of the implant (Figure 5B).For the acquisition of Q S , however, a sufficient signal strength is required as any uncertainty of Q S would affect all subsequent results.On the higher end of the induced signals, it was seen that a saturation is observed above 1.6 V (Figure 7), which is based on limitations of the SoC.For realistic implementations in implants, this would need to be considered and adjusted.
Even if no or only negligible RF signals were detected by the AIMD, there could still be a small temperature rise in the order of 100 mK around the tip for the OP or NM method (see Lead 2 in Figure 8).Again, this might reflect the thresholding behavior of the Schottky diode for the small signal regime (cf. Figure 5A).An alternative explanation is the temperature rise due to the "native," not implant related, SAR deposition. 67[89]

pTx imaging
In general, there are three possible imaging scenarios for AIMDs: (A) when the AIMD is located outside the target FOV for imaging, (B) when the AIMD is positioned inside the FOV and the target imaging location is far away from the implant, and (C) when the AIMD is positioned inside the FOV and the target imaging location is close to the implant.For (A) and (B), the OP method can be utilized, which demonstrated comparable image quality to the standard imaging mode, as long as the imaging mode is not close to a worst-case RF excitation.For scenario (C), more advanced imaging techniques might be required to improve contrast-to-noise ratio (CNR) around the implant. 90In this particular scenario, the implant sensor signal may provide a useful indicator of RF-safe excitations, while image quality is being optimized.2][93] These methods are limited to imaging scenarios (B) and (C), and it still has to be demonstrated that these methods work robustly in vivo.The sensor approach described here is much faster and straightforward to calibrate.It can also complement image-based methods by providing the reference values for a quantitative assessment and further development of these methods.
A simple metric for basic imaging quality prediction of a pTx excitation vector was introduced.The S image determines the similarity of the excitation vector with the reference imaging mode CP and is calculated based on the implant's sensor measurements without the need to acquire additional B + 1 maps. 94Whenever the CP mode is similar to the WC excitation vector (see, e.g., Lead 5 in Figures 8B,C and 10), the projected OP vector will automatically deviate more strongly from CP, which is also reflected in a lower S image (i.e., S image = 0.7 for Lead 5 compared to S image ≥ 0.8 for all other leads, Figure 10, Supporting Information Figures S1 and S2).Here, we reach an inherent limitation of the OP method and static RF shimming in general: if homogeneous excitation happens to heat an implant the most, some image quality must be traded in for safety.S image can be utilized for a rapid evaluation of the imaging information contained in the OP mode. 946][97][98][99] Linear combinations of other NMs could also be used to improve the imaging performance of a single NM.This optimization could be performed based on acquired B + 1 -maps 40 or the sensor signal alone by using the S image of each NM for optimization.In general, the image quality of the proposed approach needs further investigation in realistic DBS leads and comparison to single-channel transmission under current MR-safe conditions of ISO 10974.Potential performance benefits (image quality, acquisition time, etc.) of sensor-embedded implants need to be quantified for example, by maintaining a <1 K or <2 K lead tip heating and optimizing image quality using pTx.In this study, all pTx imaging experiments were performed in a homogeneous ASTM phantom, which has limitations for evaluating imaging performance.Therefore, it has not yet been demonstrated whether sufficient CNR could be obtained with the lower heating modes for clinically relevant situations.Anthropomorphic phantoms or ex vivo 37,46,66 imaging studies are needed to assess CNR including all uncertainties (e.g., from sensor calibration or Q S acquisition) of the proposed technique.

CONCLUSIONS
It was demonstrated that a "smart" implant safety concept, where an AIMD communicates wirelessly with an MR scanner is feasible and capable to substantially improve implant safety for patients.Such strategy would define clear and distinct responsibilities: the implant manufacturer provides the sensor data and calibration; the MR manufacturer translates this information into safe excitation vectors.Simultaneously, the clinical personnel would be relieved from the current burden and liability to ensure implant safety.In combination with pTx, a substantial reduction of implant heating is possible, at a very moderate, in most cases, loss of image quality.
. The results at 3 T of the electromagnetic (A-D) and thermal (E-H) simulations of the implant tip for varying background materials (PVP phantom liquid [ r = 45 and  = 0.2 S/m @123 MHz], white matter, gray matter and cerebrospinal fluid 101 ).Simulation power levels were adjusted to the same implant sensor voltage reading.The thermal maps depict the heating at 60 s and (I) the temperature evolutions are plotted using a virtual sensor, indicated with "x."More information on the simulation setup can be found in the reference below.
2 system-on-chip (SoC) module (cc2652RB, Texas Instruments, Texas, USA) was powered by using a coin cell Li-ion battery (CP 1654A3, Varta Microbattery GmbH, Ellwangen, Germany) including a battery protection circuit (AP9211SA, Diodes Incorporated, Plano, Texas, USA).A Schottky barrier diode (D2) rectifier module with an RLC filter (R2, L3, C4) was embedded in the implant casing and connected to the 12-bit ADC module (200 kHz sampling rate) of the SoC in order to measure rectified RF-inducedF I G U R E 1The wireless reference implant hardware.(A) The block diagram of the implant electronics inside the implant case.A battery powered BLE SoC module was utilized for the experiments.D2 is the Schottky diode that rectifies the RF-induced signals.R2, C4, and L3 are the elements for the Schottky rectifier.D3 is the Zener diode for the overvoltage protection.D1 is the PIN diode which is biased by using SoC module during wireless trigger events.C1 and C3 are the DC block capacitors.L2 is used for the DC return path for the PIN diode bias.L1 is the RF-choke (RFC) (B) 3D illustration of the implant with a connected implant lead.(C) Photograph of all components used in an experimental setup.The lids of the casing are insulated with a copper tape and glue for improved RF shielding in the MR experiments.The temperature rise at the tip was measured with external temperature probes (a zoomed inset is shown at the top right).

Figure 6
Figure 6 shows the wireless Q S acquisition using the 3T scanner with eight-channel pTx coil.The received raw data for 10 repeated measurements span a 15 ms time interval of (Figure 6A) even though Q S is acquired in 7.5 ms.The additional 7.5 ms dead time indicates the connection interval timing uncertainty of the BLE protocol.The same 10 consecutive Q S acquisitions with synchronized timing of the pulses are shown in Figure 6B.The first and last two RF-induced induced signals in the Q S pulsetrain are the CP pulses used for synchronization.The repeatability of the wireless pTx mitigations was shown by calculating the OP mode from the acquired Q S .The SD in the calculated pTx

Figure 7 :
Figure 7.Over the whole investigated range, the simulated non-linearity correction works well as indicated by a strong linearity (R 2 ≥ 0.99 for all eight channels), despite the hostile MRI environment (B 0 , high peak RF power).Similar to the testbed experiments, the implant sensor readings become unreliable below 0.1 V and saturation is observed above 1.6 V (cf.Figure 7: Channel-6) which is the upper limitation of the SoC.

F I G U R E 7
Induced implant sensor readings versus MR scanner output voltage at a single pTx channel.Overall, a good linear correlation is observed with an R 2 ≥ 0.99.For very low induced signals (<0.1 V, noise floor) and for very high induced sensor signals (>1.6 V, saturation) stronger deviations are observed.

F I G U R E 8
Wireless implant sensor readings and corresponding tip temperature measurements for varying RF excitation modes and implant lead configurations on a 3T MR scanner.(A) The amplitudes (normalized) of Q S for each lead configuration.(B) Measured implant sensor signals for WC, CP, OP, and NM.RF pulses for WC and OP were calculated based on the implant sensor measurements.WC and CP modes induce high induced signals, which can be suppressed by applying the OP and NM RF excitation modes.(C) Corresponding RF-induced heating measured by fiber-optic temperature probes.The RF-induced signals correlate well with the measured temperature rises on the implant tip.OP and NM successfully suppress RF-induced heating at the implant tip for all implant lead configurations.

F
I G U R E 10 MRI of two different DBS lead trajectories for different RF excitations modes (WC, CP, OP, and NM) as determined by the implant sensor readings and their calculated imaging scores (S image ).Single coronal images of the implant tip are shown.The total transmit power and the scaling factor are the same for all imaging modes.Lead 1 has the best imaging score (S image = 0.97) in OP mode, while Lead 5 showed the worst (S image = 0.7) among all investigates lead configurations.Images of all lead trajectories and scores can be found in Supporting Information Figures S1 and S2 .

Figure S4 .
Figure S3.The results at 3 T of the electromagnetic (A-D) and thermal (E-H) simulations of the implant tip for varying background materials (PVP phantom liquid [ r = 45 and  = 0.2 S/m @123 MHz], white matter, gray matter and cerebrospinal fluid101 ).Simulation power levels were adjusted to the same implant sensor voltage reading.The thermal maps depict the heating at 60 s and (I) the temperature evolutions are plotted using a virtual sensor, indicated with "x."More information on the simulation setup can be found in the reference below.FigureS4.The results at 7 T of the electromagnetic (A-D) and thermal (E-H) simulations of the implant tip for varying background materials (PVP phantom liquid, white matter, gray matter and cerebrospinal fluid101 ).Simulation power levels were adjusted to the same implant sensor voltage reading.The thermal maps depict the ). Simulation power levels were adjusted to the same implant sensor voltage reading.The thermal maps depict the heating at 60 s and (I) the temperature evolutions are plotted using a virtual sensor, indicated with "x."More information on the simulation setup can be found in the reference below.PetzoldJ, Silemek B, Winter L, Ittermann B, Seifert F. Experimental and numerical calibration procedure for RF safety evaluation of implant-embedded sensors.In: Intl.Soc.Mag.Reson.Med.Vol 31.; 2023:0752.How to cite this article: Silemek B, Seifert F, Petzold J, Brühl R, Ittermann B, Winter L. Wirelessly interfacing sensor-equipped implants and MR scanners for improved safety and imaging.Magn Reson Med.2023;90:2608-2626.doi: 10.1002/mrm.29818