Synopsis

Keywords: YIA, Heart

High-amplitude gradient systems can surpass the IEC regulatory limit for cardiac stimulation (CS), which is based on animal electrostimulation data and simplified electromagnetic modeling. We assess CS in MRI by performing the first cardiac magnetostimulation threshold measurements in pigs, the primary animal model of the human cardiovascular system. We use the measurements to validate a detailed CS modeling pipeline applicable to both pigs and humans. Experimental thresholds and predictions in porcine-specific models agree within 19% NRMSE. CS modeling in a detailed human body model indicates that CS thresholds of the Connectome gradient are ~25-fold greater than the IEC CS limit.

Purpose:

Time-varying MRI gradient fields induce electric fields (E-fields) in the patient, which can induce peripheral nerve stimulation (PNS) and potentially cardiac stimulation (CS)^[1-4]. Unlike PNS, cardiac magnetostimulation thresholds cannot be safely measured in humans. Instead, the IEC 60601-2-33 standard^[5] imposes limits on the switching speed of gradient systems to prevent CS, based on electrode measurements in different species and on simplistic electromagnetic (EM) field simulations in a homogeneous body model^[2,6]. For long rise-time waveforms and high-amplitude gradient systems like the Connectome^[7], the IEC CS limit can become more constraining than the PNS limit. We assess CS in MRI using a combination of cardiac magnetostimulation measurements in pigs and detailed simulations combining state-of-the-art EM and electrophysiological modeling.

Methods:

Measurement of porcine CS thresholds: We measured cardiac magnetostimulation thresholds in pigs^[8], a widely used animal model of the human cardiovascular system^[9]. All experiments were conducted with the approval and under the advice of the Massachusetts General Hospital Institutional Animal Care and Use Committee. CS thresholds were measured in ten anesthetized pigs by discharging a capacitor bank into a spiral coil close to the torso, resulting in a damped sinusoidal dB/dt waveform (f=715 Hz, Figure 1)^[8]. Capacitor discharges were triggered immediately after the T-wave (early diastole) and before the P-wave (late diastole), based on ECG recordings. The stimulus amplitude was increased by increasing the capacitor voltage until cardiac capture was detected on ECG, blood pressure, or peripheral oximetry traces. The CS thresholds were then determined by fitting sigmoid functions to the binary cardiac responses as a function of the capacitor voltage.
Porcine-specific body models: Immediately after the stimulation experiment, fat/water-separated Dixon (multiple bed positions) and CINE images were acquired in eight of the pigs (#1-#4 and #7-#10) using a 3T MRI scanner (Connectome, Siemens Healthineers, Germany)^[8]. The MRI datasets were segmented manually and interpolated to 1 mm³ isotropic resolution to create individualized body models reflecting the anatomy and posture of the animals. Tissues were assigned conductivity values based on the IT’IS low frequency database^[10]. Finally, cardiac Purkinje and ventricular muscle fiber paths were added to the heart of each porcine body model using rule-based modeling algorithms^[11,12].
Prediction of CS thresholds: We used our EM-electrophysiological CS modeling pipeline^[13] to predict CS thresholds in the individualized porcine body models for the spiral coil, and in a female human body model placed in the Connectome gradient system (head at isocenter). In short, we project the predicted E-field onto 3D paths of cardiac Purkinje and ventricular muscle fibers to calculate the electrical potential along all excitable fibers. The potential is modulated in time by the coil current waveform derivative and input into validated electrophysiological cardiac fiber models to identify the smallest coil current waveform inducing an action potential in any fiber.

Results:

ECG and oximetry signals measured in pig #6 show that a capacitor discharge at U=5.0 kV did not cause cardiac capture, whereas an increase to U=5.25 kV capacitor voltage did (Figures 2A&B). No cardiac arrhythmias were induced in any of the pigs other than a single premature ventricular contraction in pig #4 (Figure 2C). The average CS threshold measured in the ten pigs during early diastole was 6.27±1.33 kV, which corresponds to dB/dt=1.60±0.22 kT/s at heart center and a 95th percentile E-field E95=92.9±13.5 V/m over the myocardium. Thresholds measured during early and late diastole were not significantly different (p<0.05, Wilcoxon rank sum test).
Figure 3 shows a porcine body model (including cardiac fibers) and E-fields simulated for the spiral coil. Figure 4 compares measured and predicted thresholds for the eight pigs with animal-specific body models. Modeling predictions and experimental thresholds agree within 19% (NRMSE), and no significant difference was found between them (p<0.05, paired t-test). In all cases, the predicted Purkinje thresholds were >5-fold higher than ventricular fiber thresholds, thus setting the effective CS threshold. We estimated the CS modeling uncertainty to be 27% by propagating the uncertainties of each of the 25 main model parameters through the modeling pipeline.
Figure 5 shows CS threshold predictions in a female human body model for the Connectome gradient (X+Y+Z axis combination, trapezoidal gradient waveform with 500-µs flat-top duration). The predicted CS thresholds were 25-fold higher than the IEC CS limit and the PNS limit at t_rise=0.5 ms.

Discussion:

We performed the first measurements of cardiac magnetostimulation thresholds in ten healthy, anesthetized pigs^[8]. To analyze this data, we created individualized body models reflecting the anatomy and posture of the animals during stimulation. The average dB/dt in the porcine heart at the measured threshold was 1.60±0.22 kT/s. This value is 11 times higher than the human IEC CS limit at f=715 Hz.The measurements agree with threshold predictions in the porcine-specific models within 19%, supporting the validity of the CS model. Predictions for the Connectome gradient coil loaded with a realistic female human body model suggest that the CS thresholds are substantially greater than either the PNS thresholds measured for that coil or the IEC CS limit. Our CS model will be a useful tool for informing safety limits for MRI gradients that protect patients without overly restricting gradient performance and thus imaging speed and resolution.

Acknowledgements

The authors would like to thank Jo Hyun Kim, Shi Chen, and Robert Alan Eder for their help with the experiments. The research reported in this publication was in part supported by the ISMRM Research Exchange Grant and the German Academic Exchange Service (DAAD) as well as by the National Institutes of Health under award numbers R01EB028250, P41EB030006, R01HL151704, and R01HL159010. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.