HIGH RESOLUTION COMPUTATIONS AND MEASUREMENTS OF POTENTIAL EM1 WITH MODELS MEDICAL IMPLANTS AND RADIATING SOURCES

Transcription

1 HIGH RESOLUTION COMPUTATIONS AND MEASUREMENTS OF POTENTIAL EM1 WITH MODELS MEDICAL IMPLANTS AND RADIATING SOURCES Howard Bassen and Jon Casamento Center for Devices and Radiological Health Food and Drug Administration Rockviile MD, U.S. OF Abstract - We used high resolution finite difference time domain (FDTD) computer modeling to evaluate the potential of cellular phones to cause interference in an implanted cardiac pacemaker-defibrillator. The model consisted of a comhination of an 835 MHz sonrce which was either (1) a special boatmounted cellular phone antenna (a sleeve dipole fed by a 3W amplifier) or (2) a standard EM1 test model ofa cellular phone handset (a dipole antenna placed 8 mm from the victim). The victim was a simplified model of a pacemaker-defibrillator with one insulated sensinglstimulation lead in a torso simnlator. The handset model and the victim conformed to the AAMI PC69 standard for pacemakerdefibrillator EM1 testing. We compared the EMI-induced voltage in the victim from the sleeve dipole source with the induced voltage from a cellular phone handset model. We performed experimental comparisons of the computed results using the actual antennas. We used a fiber-optically linked pacemakerdefibrillator simulator to measure induced voltage at the junction between the lead and the box of the pacemakerdefibrillator model. Our conclusions were that the sleeve dipole at 15 em, driven with a 3 watt amplifier, is less capable of causing potential interference than the handset The sleeve dipole, with its companion 3-W amplifier, produced 12.7 db less induced voltage in the victim than the phone handset model driven with 0.6 W. Our results were validated since measured values differed by 1.7 db or less from computed values. This is well within the uncertainties of the measurements and computations. Kqwnrds (medico1 device, compufatinn.finite difference rime domain, FDTD, measuremenr, pacemaker, defibrillafor, cardiac, cnmpurer modeling) I. INTRODUCTION This project was initiated in response to concerns expressed by an EM1 standards-setting group dealing with implanted cardiac pacemakerdefibrillators. This group was apprehensive about interference from a three-wan cellular phone power amplifier driving a boat-mounted antenna. The overall goal of this project was to compare the RF voltage induced in a model of a pacemaker-defibrillator (victim) from two sources. I US. Government work not protected by U.S. copyright I One source was a boat-mounted cellular phone antenna (sleeve dipole) and its companion amplifier sold by the antenna manufacturer. The amplifier provides an input to the antenna of 3 Wafts. The antenna was placed at both 15 and 30 cm fiom the surface of a torso simulator prescribed in the AAMI standard PC69 [I]. The torso contained a model of a pacemaker-defibrillator. This was done to simulate a person with a pacemaker-defibrillator standing at the closest reasonable distances from a large, visible whip antenna. The second source was a standard EM antenna for testing pacemaker/defibrillators for interference from wireless handsets in accordance with the AAMI standard. This second antenna was a half-wave dipole spaced 0.8 cm from the surface of the torso simulator. This source was driven with 0.6 Watts to represent the maximum power from a cellular handset sold in the US. The 0.8 cm distance is intended to represent the worst case where a cellphone user places the handset directly on the surface of their chest, adjacent to their implanted pacemaker/defibrillator. In the past we demonstrated the effectiveness of this test method (with half-wave dipole and torso simulator) for experimental determination of EM in pacemakers and in implantable defibrillators 121. To meet our goal, we developed relatively realistic computer models to evaluate interference voltages induced in an implanted cardiac pacemaker-defibrillator by cellular phones operating at 835 MHZ. We planned to use experimental measurements with physical replicas of the computer models to verify the results of the computations. n. Background A number of researchers performed computer modeling of microwave EM1 of medical implants. Most biomedical EM researchers believe the finite difference time domain (FDTD) analysis provides the best combination of spatial resolution and computational efficiency for near-field modeling on modern personal computers. Prior computer modeling of EM1 of medical implants includes work by Wang et al [3]. They studied a simple model of a pacemaker with a bare-wire lead implanted in both simple and realistic models of a human. The 900 MHz exposure source in this paper was a half-wavelength dipole antenna. A comparison was made of the RF voltage induced at the open circuit input port of the pacemaker/defibrillator model while implanted in a highly 75

2 simplified homogeneous cuboid box model of the human body. This was compared to implantation in a highly realistic, multiple-tissue model of a human. The EMI results d e rived from the homogeneous cuboid model were found to be comparable to the results for the human model only when the antenna was more than 60 m from the body. We also used FDTD computations to determine the RF voltage induced at the input of a model of a pacemakerdefibrillator when exposed to emissions from a special microwave antenna (sleeve dipole). The EM interference induced in a cardiac pacemaker-defibrillator is proportional to this voltage. Our study differed from Wang s because we used a detailed, high-spatial-resolution (0.4 m) model of a specific source, a sleeve dipole, driven by a high-power cellular phone antenna. Ill Methods and materials A. Overview We computed the induced voltage in a model of a cardiac implant while exposed to emissions from both the sleeve dipole and to the half-wave dipole that is specified in the AAMI standard. These two antennas were used in conjunction with a simplified replica of the AMMI EMC standard s torso simulator. The AAMI EMC standard is used by most cardiac device manufacturers in the United States to evaluate EMC at microwave frequencies. In addition to computations, we performed experimental validation of the computed results of the exposures using a pacemaker-defibrillator model containing a diode detector and a fiber-optic analog data link: B. Models Computer and experimental models of the EM1 sources We used two sources of radiated EMI. Source I was a boat-mounted cellular phone antenna. This device is sold for use with a three-watt power booster amplifier. A standard cell phone is used to feed the input of the booster amplifier. We used reverse engineering to construct a CAD model of this antenna We took x-ray images of the antenna to generate all of its dimensions using physical landmarks on the antenna. The antenna was a sleeve dipole with a coaxial structure and some very unique, proprietary features. It consisted of a thin outer conductor (1.8 nun wall) and a very thin (0.7 nun diameter) inner conductor. In addition, the antenna used a short section of coaxial cable as a matching network. The cable also was pa~i of the radiating structure (figure 1). The manufacturer s name and the exact details of the antenna will not be presented because of its proprietary status. We developed three dimensional computer models of this antenna in standard AClS format. In order to solve the problem with this high resolution, complex antenna, our software required computer memory of at least 1.5 GB. All of these features required the antenna model to be composed of 3.8 million voxels for dielectric objects and 0.97 million edges for conducting objects (figure 2). This antenna was fed with a sinusoidal 835 MHz signal with an amplitude of 1 volt RMS. For the experimental study we used the actual antenna as the radiating source. Saline Figure 1 Torso simulator showing both sleeve dipole and AAMI dipole (NOTE: AAMI and sleeve dipole are never present in the same setup). Source 2 was a half-wave dipole antenna that followed the AAMI pacemakeddefibrillator EMC-test standard s specifications for the frequency of interest (835 MHz). This antenna s manufacturer (SPEAG, Zurich) provided CAD data and drawings that allowed direct importation into the FDTD program. This source was placed 8 m from the surface of a torso simulator in which a pacemakerdefibrillator model was placed. For the experimental model of the antenna, we used the actual sleeve dipole. The antenna dimensions were 161 mm from tip to tip, and a dipole diameter of 3.5 mm. The center gap between the two dipole arms was centered adjacent to the point on the simulated pacemakerdefibrillator where the lead entered the pacemakerdefibrillator box. This was done for the following reason. The highest H field occurs at the dipole center gap. The E field is highest at the tips of the dipole, far from the gap. When this dipole is placed very close (less than a centimeter) to a torso simulator filled with a lossy liquid, the induced E field inside the liquid is highest adjacent to the gap of the dipole, not near the tips of the dipole [4]. 16

3 16 nun diam. antenna I 5 nun diam. insulated lead Figure 2. Front and top views of voxels and conducting edges of dielectric and conducting objects of a sleeve dipole source (left) and pacemaker victim (right). Computer and experimental models of the EM1 victim. For both experimental and computational models: The victim consisted of three components: a pacemakerdefibrillator "box", a pacemaker-defibrillator lead, and a torso simulator. The box represented the metallic case of a real pacemakerdefibrillator device. The case houses a pulse generator and cardiac electrical sensing circuitry. The box was a relatively large object (1 1.7 x 5.7 x 1.8 cm) of the same size as the case of first-generation implantable defibrillators. The simplified model of the lead had only one conductor instead of commonly used multiple conductors. The lead had a thin (1 mm wall) insulation layer (E, = 2.0) surrounding a perfect conductor (2.7 nun diameter). The torso-simulator in the experiment was completely filled with salt water (saline) 0.18% NaCl by weight, as specified in the AAMI standard. The computer model of the torso simulator was filled with this saline's electrical properties (E, = 79, a= 0.46). The box was 5.5 nun from the front surface of the saline inside the torso simulator. The lead was parallel to the front surface of the saline, and 10.9 mm away from this surface. The plastic front wall of the torso simulator was 10 mm thick and was between the dipole and the saline. The lead had a rectangular configuration and a "loop area" of 225 cm2. The loop area is illustrated in figure 1. The patient torso simulator was constructed as a simplified version of the unit specified in the AAMI pacemaker-defibrillator EMC standard (Figure 3). This box-shaped model had walls of plastic (E, = 2.0) with dimensions of 38 x 10 x 55 cm and a wall thickness of 10 mm Source-victim coupgurations - For both computational and experimental systems that were studied, the following configurations were used. One source-victim configuration consisted of the sleeve dipole placed 15 or 30 cm from the torso simulator. The other (second) source-victim configuration consisted of the AAhfl half-wave dipole 0.8 nun from the plastic front of the torso simulator. In both cases, the pacemaker-defibrillator box was 5 nun behind the EM-exposed front surface of the saline of the torso simulator. Computer model of the victim - For the victim we used a high resolution FDTD model (0.4 nun non-homogeneous meshing, i.e. a graded mesh). This allowed us to create a relatively realistic pacemaker-defibrillator lead, with round wires and thin cylindrical insulation. Figure 2 illustrates the high resolution utilized in our model as indicated in screen shots of the victim's voxels and conducting objects. For the computer model of the torso simulator only the ffont surface had a plastic wall (IO mm thick) with the rest of the sides being open. The torso simulator in the AAMI standard has no front surface material, since it lies down. Our model had a plastic front surface so it could be positioned vertically in our anechoic chamber and not spill the liquid saline. It had its broad side facing the sleeve dipole. The effects of the plastic on the induced EMI voltage in the pacemakerdefibrillator were computed and the difference is reported in section IV. C. Methods Cnmputational methods We used commercial FDTD software (SEMCAD 1.8, SPEAG, Zurich) that incorporated native drawing and importing of 3-D CAD object models in ACIS (.sat) format. The sofhwre had a variable grid (graded-mesh) generator. This software was used on a personal computer with a 2.4 GHz, 32-bit processor, 2.5 GB of RAM, and a Windows XP operating system. The 3D models that were used included sources (sleeve and half-wave dipole) and the victim device (pacemaker-defibrillator in a saline-filled box). This FDTD software required no equations or computer code. Input data are provided by creating 3-D drawings of the physical structures involved and the input of detailed EM specifications for the simulation parameters (e.g. signal frequency, voltage amplitude at the antenna input, and grid dimensions). However, for EMI problems, reasonable knowledge of electromagnetic field theory and measurements is needed (e.g. impedance matching, VSWR, complex (vector) voltage and power, etc.) to use this software and obtain proper results. Special care was taken to inspect all voxels and the electrical parameters used in the simulation. For the configuration with the sleeve dipole, the 30 cm distance between it and the torso simulator required large volumes of vacuum which are represented by a large number of "empty" voxels. Our graded mesh sofhwre used an estimated 36 times smaller number ofvoxels than would be required without the graded mesh feature. Without graded meshing, this problem could not be solved on a 32-bit personal computer with an operating system limit of 2 GB of usable RAM. In addition, we reduced the number of voxels significantly by manual optimization of the grid. This involved inspecting the model and locating areas with no materials in them. Then, in these areas, we increased the grid size by manual adjustment of the grid step value. For the other configuration with the source being the AAMl standard dipole placed 8 mm away from the kont face of the torso simulator, a reduced number of conducting object edges (0.49 million) were required compared to the sleeve dipole with its 0.97 million edges. 77

4 Figure 3. Patient torso simulator and pacemakerdefibrillator simulator (aluminum box x 5.7 x 1.8 nun with a single insulated stimulatingkensing lead). Experimental methods The physical model of the pacemakerdefibrillator was constructed as an aluminum box containing a diode detector and photonic sensor (described below). This box was submerged in the saline of the torso simulator (figure 3). We measured the RF voltage induced between the pacemakerdefibrillator box and the lead as follows. Inside the box we mounted a coaxial diode detector with 50 ohm input impedance. The center conductor of a coaxial connector from the input port of this detector was fed to the outside of the box through a small hole. The sensing lead of the pacemaker was attached to this connector outside the box and sealed with silicone. Inside the box, the output port of the detector fed a DC voltage to a fiber optically-linked photonic sensor (SRICO Inc). The fiber optic link was used to avoid problems from metallic wires: pickup of voltage from data collection leads and distortion of the electromagnetic exposure fields. We performed a comparison of the voltage induced in the model pacemaker/defibrillator from each of two sources in an anechoic chamber. The AAMI dipole source was placed 0.8 mm in front surface of the saline of the torso simulator. Later, the sleeve dipole was placed at both 15 and 30 cm from the saline surface. The power into each of the source antennas under test was increased and recorded when the voltage that was detected by the victim was exactly 0.1 mv. This eliminated effects of any non-linexities and calibration errors of the diode detector and photonic sensor. IV. Results A. Computations Source 1 - We computed the distribution of the electric and magnetic field strengths throughout a volume of space surrounding the antennq without the torso simulator present. We also computed the input impedance. Even using a high degree of physical accuracy to model the antenna it was difficult to obtain an input impedance value that was close to the physical antenna s value of 50 ohms. We believe our problem was due to the fact that the physical antenna was well matched at two, non-harmonically related frequencies, probably with some hidden mechanical or electrical components. We did not model these. The best impedance match for the computer model of the sleeve dipole antenna at 835 MHz resulted in a return loss of 3 db. Computation of the electric fields in free space out to a distance of 30 cm from the antenna required about 18 hours to solve on a 2.4 GHz P~ISOMI computer with 2.5 GB of memory (RAM). These resources were needed for our model with a 0.4 mm grid step and total of 14.1 million mesh cells. Source-victim configuration 1 - The sleeve dipole was placed at 15 and then at 30 cm from the saline surface of the victim. The pacemakerdefibrillator body was placed in the saline at a height where the highest electric field strength was observed (via computations) at the inner surface of the torso simulator. A plot of the electric field strength inside the liquid of the torso simulator is shown in figure 4. Source-victim configuration 2 - The half-wave dipole was placed 8 mm away from the front plastic surface of the torso simulator. We determined the optimum location in our experimental system, and then placed the dipole there in the computational model. B. Interference potential of each E M source The RF net power into each of the two source antennas that was needed to induce 0.1 mv in the victim is shown in Table 1. The received power in the victim and the resulting voltage do not represent all factors that produce potential interference in a pacemaker-defibrillator. Other Factors include the frequency of the RF signal, and the waveform and amount of amplitude modulation imposed on the RF signal. Table 2 presents more data on the interference potential (power into the source and the resulting voltage induced in the victim). These data are presented for the sleeve dipole driven by a 3-Watt booster amplifier and for the AAMI dipole driven by 0.6 watts from a hqpothetical cellular phone handset 78

5 >lax Contour 0.4 vim - (linear scale) calculated the effect of driving the sleeve dipole with more power (3 Watts) as if it were driven by the power amplifier that is sold for it. This was five times greater power that the maximum power (0.6 W) delivered by a cellular phone handset (as modeled by the AAMI half wave dipole). Using a model of the lower-powered handset placed very close to the torso simulator, we induced more voltage into the victim than the sleeve dipole driven by a 3-Wan booster amplifier!ea4 1.0 E&. net power into source Figure 4. Computed electric field induced by the sleeve dipole. Field strength isocontours are shown inside the torso simulator liquid at the plane containing the fiont surface of the pacemaker. The fiont of saline of torso simulator is 15 cm from the antenna. Lightest shade = maximum field. Source Power into source Induced voltage ratio (db) C. Interference potential of additional configurations of the victim In addition to the data above, we computed data for several more realistic pacemaker-defibrillator configurations. Fist, we changed the input impedance. of the pacemakerdefibrillator to infinity (open circuit). The voltage induced in the pacemaker-defibrillator with an open circuit input impedance was 2.0 times higher than the voltage induced in a 50-ohm input impedance. Next, we produced a computer model of a smaller, realistically-sized pacemaker (figure 5). The voltage induced in the smaller model of the pacemaker with a 50-ohm input impedance was 1.9 times less than the voltage induced in the larger box model of a pacemakerdefibrillator with the same input impedance. Finally, we computed effects of the plastic on the front of the torso simulator. For the sleeve dipole as the source, the magnitude of the complex voltage induced in the victim decreased 1.7 db when there was no plastic. This is due to the matching effect of the plastic between the airhaline interface. V. Conclusions Our computed results indicate that the sleeve dipole at separation distances of 15 and 30 cm required significantly more power (19.7 and 25.8 db respectively) to induce the same voltage into the victim (pacemakerdefibrillator) than the AAMI dipole 8 nun from the torso simulator. Next, we Sleeve dipole (15cmfrom saline) Sleeve dipole I 3.0 I saline) AAMIdipole I Table 2 -Induced voltage ratio in the victim (relative to the AAMI dipole) when the actual source net power levels are used. I Agreement between computed and measured values differed by at most 1.7 db). This is a very close agreement (*0.85 df3) given the fact that there are uncertainties in experimental measurements, instrument calibrations, and computational modeling. In addition, during experimentation we observed significant variations (about 3 db) in the measured voltages versus slight changes in distance (less than 1 cm) between the AAMI dipole and the simulated pacemakerdefibrillator model. This is due to large spatial gradients in the E and H fields that exist in the near field where the separation between source and victim is only a few centimeters. For this type of high resolution FDTD computation it is desirable to!q many positions and configurations of the source and victim to identify the worst case EM situation. The time required to compute multiple configurations is somewhat prohibitive, since each computation takes

6 hours. Fortunately, faster FDTD algorithms (such as the ADI-FDTD method) for high spatial resolution computations are or will be incorporated into commercial software. In addition, multiprocessing versions of FDTD software are now available that use relatively inexpensive personal computers (PCs) with up to four processors. Also, this software runs on PC clusters with 30 or more processors. This enables a great decrease in computational times, making interactive, what if analysis feasible. References [IIAAMI, Active Implantable Medical Devices, Electromagnetic Compatibility (EMC) Test Protocols for Implantable Cardiac PacemakeriDefibrillators and Implantable Cardioverters, ANSYAAMI PC69:2000, Association for the Advancement of Medical Instrumentation. [2] H. Bassen, H., Moore, and P. Ruggera, Cellular Phone Interference Testing of Implantable Cardiac Defibrillators, In vitro, Pacing and Clinical Electrophysiology, Vol. 21, no. 9, Sept. 1998, pp [3] J. Wang, T. Ohshima, E. Takahash., G. Showaku, Verification for EM1 Test of Cardiac Pacemaker-defibrillator by Portable Telephones with an Anatomically Based Human Model, IEEE EMC Society Symposium Proc., Turkey, [4] Kuster, N. and Balzano, Q., Energy Absorption Mecbanism by Biological Bodies in the Near Field of Dipole Antennas Above 300 MHz, IEEE Transactions on Vehicular Technology, Vol. 41, no. 1, pp , Feb Figure 5. Box model of pacemaker (lei?) and mode a smaller, realistically-sized pacemaker (right) Disclaimer The mention of commercial products, their sources, or their use in connection with material reported herein is not to be construed as either an actual or implied endorsement of such products by the US. Food and Drug Administration. 80