Finite volume analysis of temperature effects induced by active MRI implants with cylindrical symmetry: 1. Properly working devices
© Busch et al; licensee BioMed Central Ltd. 2005
Received: 16 November 2004
Accepted: 08 April 2005
Published: 08 April 2005
Active Magnetic Resonance Imaging implants are constructed as resonators tuned to the Larmor frequency of a magnetic resonance system with a specific field strength. The resonating circuit may be embedded into or added to the normal metallic implant structure. The resonators build inductively coupled wireless transmit and receive coils and can amplify the signal, normally decreased by eddy currents, inside metallic structures without affecting the rest of the spin ensemble. During magnetic resonance imaging the resonators generate heat, which is additional to the usual one described by the specific absorption rate. This induces temperature increases of the tissue around the circuit paths and inside the lumen of an active implant and may negatively influence patient safety.
This investigation provides an overview of the supplementary power absorbed by active implants with a cylindrical geometry, corresponding to vessel implants such as stents, stent grafts or vena cava filters. The knowledge of the overall absorbed power is used in a finite volume analysis to estimate temperature maps around different implant structures inside homogeneous tissue under worst-case assumptions. The "worst-case scenario" assumes thermal heat conduction without blood perfusion inside the tissue around the implant and mostly without any cooling due to blood flow inside vessels.
The additional power loss of a resonator is proportional to the volume and the quality factor, as well as the field strength of the MRI system and the specific absorption rate of the applied sequence. For properly working devices the finite volume analysis showed only tolerable heating during MRI investigations in most cases. Only resonators transforming a few hundred mW into heat may reach temperature increases over 5 K. This requires resonators with volumes of several ten cubic centimeters, short inductor circuit paths with only a few 10 cm and a quality factor above ten. Using MR sequences, for which the MRI system manufacturer declares the highest specific absorption rate of 4 W/kg, vascular implants with a realistic construction, size and quality factor do not show temperature increases over a critical value of 5 K.
The results show dangerous heating for the assumed "worst-case scenario" only for constructions not acceptable for vascular implants. Realistic devices are safe with respect to temperature increases. However, this investigation discusses only properly working devices. Ruptures or partial ruptures of the wires carrying the electric current of the resonance circuits or other defects can set up a power source inside an extremely small volume. The temperature maps around such possible "hot spots" should be analyzed in an additional investigation.
The amplification is adjustable by the quality factor Q of the resonance circuit and allows the imaging of the inner volume of such metallic implants as well as gathering functional information from this location, for example, velocity or flow. For these special investigations inside the implant volume the quality factor should be as high as possible to increase the signal to noise ratio for this region and to get as low signal as possible from spins outside the volume of interest. For an overall overview of the implant inside the anatomic surrounding it is only necessary to raise the decreased signal inside the Faraday cage to the normal level, which is possible with a lower quality factor. The active implant locally increases the absorbed power during MR investigations. The temperature increase induced by this extra power loss around vessel implants with a cylindrical geometry is calculated with a finite volume analysis for a "worst-case scenario". For most of the simulations only thermal heat conduction is assumed as "worst case" without cooling due to blood perfusion and without cooling due to blood flow inside the active implant. Partly also a blood flow is considered carrying energy out of the simulation volume. The calculations are carried out with respect to power loss, time of rf-exposure and size of the aMRIi. The aim of this investigation was to perform a detailed assessment of the safety for such implants under "worst case" assumptions.
Specific Absorption Rate (SAR)
where σ [S/m] is the electric conductivity, ρ [kg/m3] is the mass density, E [V/m] is the amplitude of a sinusoidal time dependent electric field and r is the position vector. Assuming that the amplitude of the electric field only arises by induction from a uniform, linearly polarized magnetic field with amplitude Bm [T] and angular frequency ω [rad/s] in a homogeneous body of rotational symmetry, the following equation holds 
with r as component of cylindrical coordinates and r = 0 on the rotational symmetry axis aligned parallel to the magnetic field Bm. Combining Eq. (1,2) the spatially averaged SAR over a cylinder with radius Rcyl and using dV = 2 π r h dr yields
For pulsed MR sequences two correction factors are necessary,
The duty cycle factor cdc equals N* τ /TR, where TR [s] is the MR-sequence repetition time, N is the number of identical excitations during TR and τ [s] is the duration of one excitation pulse. cdc equals to the time ratio of "rf-on" to "rf-on + rf-off" during an MR scan. cpwm corrects for rf-pulses which are not rectangular in shape and is the ratio between the energy of the MR excitation pulse and the energy of a rectangular shaped pulse with identical amplitude and identical duration.
Principles of B1-field amplification inside a resonator
A circularly polarized magnetic field with magnitude B1 or a linearly polarized magnetic field with amplitude Bm for excitation is amplified inside an aMRIi with respect to the used resonance circuit. Because of the shape of vascular implants and to simplify calculations, a solenoid is chosen as inductor of the resonator and is investigated as a magnetic antenna. For the theoretical estimation, the axis of the solenoid resonator is assumed to be parallel to a sinusoidal linearly polarized magnetic field with amplitude Bm, or equivalently to be in the plane of a circularly polarized magnetic field with magnitude B1. For the resonance case (2πν0 = ω0 = (LC)-1/2), where the overall resistance of the resonator is just R, the following equation can be derived from the law of induction, Ohms law and the definition of the quality factor Q of a resonator,
where Vind [V] and Vself [V] are the induced and self-induced voltage respectively, Z [Ω] is the impedance, R [Ω] is the resistance, L [H] is the inductance of the inductor and αres and α are the flip angles inside the resonator inductance and outside the resonator respectively. The total magnetic field inside the resonator arises from both components Bm and Bres. For large quality factors (Q>>1) Bres dominates Bm. With Eqs(3, 4) it is possible to calculate the magnetically induced SAR and with the SAR the corresponding power loss inside a resonator. But this takes into account only power losses due to eddy currents and not the total power loss of a resonator.
Total power loss of a resonator
The overall power loss P [W] also includes the electric losses on and around the resonator. It is given by the following basic equation ,
where Wtotal [J] is the energy stored inside the resonator.
where V, A, ℓ with index imp are the volume, cross section and length of the implant inductor. For a pulsed MR sequence with excitation magnitude B1 the total power loss can be calculated by combining Eqs. (4, 5, 6) to
A comparison of Eqs. (3b, 7) shows that for a certain resonator tuned to the resonance frequency of a specific MR field strength, the supplementary absorbed power is proportional to the SAR of the MR sequence due to the identical dependence on cdc, cpwm and B12. For a specific sequence on a MRI system with a definite resonance frequency, the extra power is proportional to the quality factor Q and the volume Vimp of the inductance (Eq. (7)). The proportionality to the quality factor Q may be surprising, because it states that for better quality factors the power loss is larger. This is due to the total energy of the resonator, which depends quadratically on the field strength inside the resonator. This field strength is proportional to the exciting field B1 (or Bm) as well as the quality factor Q. Combining Eq. (5) and Eq. (6) alters the inverse proportionality to Q (Eq. (5)) to a proportionality (Eq. (7)) with respect to the field established by the transmit coil. Examining the power loss with respect to the magnetic field inside the resonator confirms the inverse proportionality to Q.
Finite volume analysis
All temperature increases are calculated as temperature difference maps by using a simulation volume divided into many small simulation cells. The energy exchange ΔE [J] between two simulation cells with a specific contact area A [m2], temperature difference ΔT [K] and heat diffusion path length Δx [m] for a time interval Δt [s] is given by the equation for heat conduction as,
where λ [W/(m K)] is the thermal conductivity.
The total energy change ΔEtot of one cell during a time interval Δt is the sum of exchanges of this cell with all adjacent cells with non zero contact area and the energy change due to the power loss pcell inside the cell. From this total energy change the temperature difference ΔT* can be calculated,
where c [J/(kg K)] is the specific thermal capacity of the cell material and Vcell is the cell volume.
This investigation did not use any commercially available finite volume package. The simulation is self-coded for problems with cylindrical geometry in Kylix and Delphi, a software development environment, based on object oriented Pascal. The graphical outputs are mostly generated by an evaluation (registered β-test) version of Teechart 7 used within the Delphi and Kylix environment. The simulation describes the time developing temperature difference maps around an aMRIi with cylindrical geometry for a constant total power loss P.
Al[r] [m2] is the contact area in both cylinder axis directions (from index x to x-1 and to x+1) whereas Ar[r] [m2] is the contact area in radial direction from index r to r+1. The contact area in radial direction from index r to r-1 is identical to the area Ar[r-1]. A last "shell" with cells as heat (energy) sink is placed as one boundary of the simulation volume at r = m + 1 and x = n + 1. These cells always keep a constant temperature (ΔT = 0), even when receiving energy during one simulation step. Partly the simulations use a second heat sink representing a blood flow through the inner volume of the implant. For this second heat sink the temperature differences of all cell elements below a radius rflow*Δr are also kept zero independent of the energy transfer from the cell elements with higher index r. This approximation assumes that the energy applied to the inner cylinder volume with blood flow is completely transported away from the simulation volume during one time step Δt. At r = 1 and x = 1 as well as at r = m + 1 and x = n + 1 the simulated volume has its boundaries. The symmetry and the use of cylinder coordinates imply that cells with index r = 1 or x = 1 can not exchange energy with cells at a lower index. For x = 1 the symmetry plane defines an identical temperature at index x = -1, (Figure 3) with no energy exchange across the symmetry plane. For r = 1 no cells with lower index r exist and therefore also no energy exchange is possible.
Physical constants of tissue, titanium, iron and tantalum
specific heat c
thermal conduct. λ
electric conduct. σ
The simulation starts with a temperature field T [r,x] = 0 for all r and x. For each iteration step with duration Δt the total energy change for each cell element is calculated according to Eq. (9a). This change ΔEtot[r,x] of one cell is the sum of five different energies. One is due to the power loss pcell = p[r,x] defined for a cell element at index r and x. The four others are energy exchanges (Eq. (8), depending on the contact areas (Eq. (10a, b)) and the temperature differences between adjacent cells with respect to the equivalent heat diffusion length through the material of the examined region. From ΔEtot [r,x] the temperature change ΔT*[r,x] is calculated according to Eq. (9b). This value is added to the prior value (Tnew [r,x] := Told[r,x] + ΔT*[r,x)) for each cell during the whole iteration process. The entire simulated time tsim consists of q iterations with time interval Δt (tsim = q * Δt).
Analytical model for a heat generating linear wire as test of the simulation
One test of the principal correctness of the self-coded simulation is possible by a comparison of the simulation results with those of an analytical solution. An easy straightforward analytical solution is available for an endless linear wire with radius rwire, which is heated by a constant power density P* [W/m], defined as power per wire length [W/m]. After reaching the thermal equilibrium, the power associated with P* penetrates through every cylinder surface surrounding the linear power source independent of the radius r (with r>rwire). The temperature difference ΔT between the wire surface and a point at radial position R can be calculated in a homogeneous medium from the equation for heat conduction.
Control of simulation results
The implemented algorithm was controlled with different checks, beside the afore-mentioned comparison with an analytical model. First, the total energy uptake of the heat sink surface is added up during all iteration steps. This energy summed with the energy stored inside the simulation volume must equal the total applied energy. The energy inside the simulation volume is calculated independently from the finite volume algorithm by using the final temperature increase, the specific heat, the density and the volume of each cell.
Second, the algorithm was tested as to whether it provided similar results for identical geometries with different spatial resolution, different time resolution as well as a different size of the simulation volume surrounding the implant.
Heat generation inside a cylinder of length Limp and radius Rimp.
The heat generation is assumed to be uniform over the entire volume of Vimp = π R2impLimp; the entire simulation volume is assumed to be tissue.
Heat generation in nr metallic rings with rectangular cross section of rwire times xwire at radius Rimp.
The rings are equally spaced over the length Limp. These power generating rings are used instead of a solenoid with nr turns to retain the cylindrical symmetry and they stand for a model where the power loss is equally distributed on and nearby the circuit paths forming the inductor. For the rings freely selectable physical parameters c, λ and ρ can be used (Table 1). Partly these simulations assume a flow inside the implant volume not allowing a temperature increase for the region of flow.
Heat generation in a cylinder shell of length Limp and thickness rwire at radius Rimp.
This is an approximation of a modified solenoid structure with a high density of circuit paths on the cylinder surface. In this case, despite the assumed dense wire, most of the shell volume is tissue. Therefore the physical properties of tissue are used for the entire simulation volume.
None of the three cases exactly match the real situation. However, consistent results for all cases, which also are checked to the order of magnitude with the linear wire model of Eq. (11), could certify a realistic estimation.
Test of Simulations: Endless wire model
Test of Simulations: Temporal and spatial resolution
For sufficiently short time steps Δt, the simulations show the expected behavior that the temperature increases monotonically in each simulation cell. For overly large time steps the energy flow out of a cell during a time step can become larger than the energy loss inside it. This leads to a physical incorrect situation with decreasing temperatures inside a cell. For this case the simulation produces huge deviations and the temperature differences oscillate in an unpredictable fashion during the temporal development.
Finite volume analysis
Data for calculation of power density PV according to Eq. (7) for an MRI sequence with an SAR of 4 W/kg (manufacturer declaration)
permeability of vacuum
[V s / (A m)]
magnitude of magnetic excitation field
repetition time of MRI sequence
duration of one excitation
number of identical excitations during TR
duty cycle cdc = N τ/TR
pulse waveform modulation factor cpwm
Larmor frequency = resonance frequency of LC circuit
power loss density PV = P/(Q Vimp) (Eq. (7)
Geometrical data and power losses for example resonators
Radius of ring resonator
Length of ring resonator
Dimension of ring wire axial
Dimension of ring wire radial
Number of rings (turns)
Volume of implant
Total power loss of resonator
SAR inside resonator (Eq. (3b))
Power loss of resonator due to SAR
Heat generation inside a cylinder
Heat generation in nr metallic rings with rectangular cross section
Heat generation in a cylinder shell
Test of simulations: endless wire model
The simulation of an endless wire uses the first of the three above mentioned models. Especially, the radius Rimp is taken as equal to rwire. The condition that the wire is endless can be realized in two ways. The first way is to choose a very long Limp and to evaluate the temperature increases only at positions on the center plane (index x = 1) and at radial distances r small compared to Limp. The second way is to modify the simulation software in such a way that the energy transfer in axial direction is excluded. This can be done by considering only volume elements of the center plane (index x = 1) and setting the heat sink shell temperature (index x = 2) after each iteration not to zero, but to the same value as at index x = 1 for all r respectively.
The second way has been realized. The common and expected result was that during the temporal development the simulated radial temperature differences increasingly approach the analytical solution for the thermal equilibrium (Eq. (11)). One comparison between a simulated and the analytical result is shown in Figure 6. Although a test on agreement of both results is independent of the specific choice of the power loss density P* [W/m], we have chosen a realistic value of P* in order to obtain realistic temperature increases. The power loss density P* was calculated with Eq. (7) for the MRI sequence of Table 2 for the example resonators (Table 2) to be between 0.24 W/m and 4.4 W/m. The simulation uses value of 1 W/m.
Test of simulations
A correct simulation of an endless wire should deliver results similar to the analytical curve of Eq. (11) using an implementation of the algorithm with no energy exchange in longitudinal (x) direction. The physical model of such a linear wire claims that the temperature difference between the wire and a cell at a certain distance increases monotonically with that distance. The theoretically calculated curve is valid after reaching the thermal equilibrium. With increasing simulated time the calculated curves show an increasingly close coincidence with the prediction of Eq.(11) (Figure 5), indicating the correctness of the algorithm.
The analytical linear wire model can also indicate which power per unit of wire length P* is safe for active magnetic implants (Figure 4). The theoretical estimation according to Eq.(11) shows for a power density larger than 2 W/m a risk of reaching temperatures increases greater than 5 K. The results do not vary with the time resolution Δt as long as the value is small enough to prevent the simulation from oscillating (Figure 6c,d). Also the spatial resolution (Δr, Δx) does not affect the results very much as long as the temperature maps appears smooth (Figure 6d,e,f). On the other hand, the calculation time explodes with a finer resolution. An increase of the spatial resolution always requires a better temporal resolution. Especially if a metallic cell completely surrounded by tissue is divided into a few metallic cells, a superior temporal resolution is necessary. For a metallic cell completely surrounded by tissue the shortest diffusion length equals roughly half of the spatial resolution with the diffusion path through tissue. For adjacent metallic cells the diffusion length equals the spatial resolution, but the diffusion path is completely through metal with a much larger thermal conductivity. The energy transfer for an identical temperature difference through this path is at least an order of magnitude larger than the one between a metal cell and a tissue cell. This higher energy transfer requires the drastically reduced time step Δt for a stable temporal development of the simulation. The calculation time for an identical simulated time increases by several orders of magnitude with metallic rings and a high spatial resolution and can easily reach computing times of several days on a normal desk top computer. Some examples for the calculation time is given in the second Table below Figure 6.
Power loss of resonator
Achievable quality factors were derived from the construction of experimental solenoid resonator prototypes (Table 3). The quality factor for a given geometry is strongly dependent on the environment and the thickness of the wire insulation. It is reduced substantially after changing the environment from air or distilled water to saline solution. Also Q decreases with a thinner insulation of the wire. Using this information it is possible to conclude that the main power losses are electric losses. Losses due to eddy currents induced by the magnetic field would not change with the thickness of the electric insulation. The dominance of the electric losses is also verified by comparing the power loss from eddy currents inside the active implant (Eq. (3)) with the total power loss (Eq. (7)). For all example prototypes in Table 3, the power loss calculated with the SAR inside the resonator (Eq. (3)) is lower by several orders of magnitude than the total power loss according to Eq. (7) (Table 3).
Realistic values for the quality factor of a resonator placed in ionic surroundings such as tissue are below 5 for all example resonators of Table 3. The calculated examples are worst case assumptions as they assume the maximum achievable Q of 5 for resonator no 1 and therefore the maximum power loss. For resonator no 2 (Q = 12.5) and no 3 (Q = 80) the chosen value is far beyond the reachable inside an ionic surrounding. These high values for resonator 2 and 3 are chosen to pronounce the safety statement for smaller resonators with quality factor values well above achievable values. The unrealistic high quality factors do not show physiologically critical temperature increases. For resonator number 1 (with a large volume) a realistic value was chosen, because in this case the temperature increase reaches physiologically dangerous values.
Part of the simulation results could be confirmed on stents implanted inside the aorta of rabbits [9, 10]. Small coated resonators with a volume below 0.5 cm3 and quality factors of 3 to 4 were used. After excision and histopathologic examination, the tissue did not show any indications of heating after several MRI investigations. This is in coincidence with the simulation results, because even for small resonators with a much higher quality factor no dangerous heating was calculated.
Ten minutes of simulated time is 5 minutes less than the critical time according to the FDA regulations for imaging of the trunk assuming a SAR of 4 W/kg. But all simulations show only small changes after 10 minutes. Therefore it is unnecessary to increase the simulation time further. Nevertheless, one example of 15 minutes simulated time is given in the movie (see additional file 1) and shows only small changes in the last five minutes.
The weak influence on the distance between power generating cells and heat sink (the outer most layer) is also shown in the movie (see additional file 1). The difference between two almost identical calculations is shown. Starting from a first calculation with simulation parameters Lsim and Rsim, the second calculation uses doubled values of Lsim and Rsim without changing the spatial resolution or the other simulation parameters. This approach shifts the heat sink to a larger distance from the heat generating cell elements. As expected the heat sink layer drags down the temperature increase in its vicinity, but the effect is small if the volume is adequately chosen. The differences for identical cells between both simulations are shown by alternating both views a few times at the end of the movie (see additional file 1).
Finite volume analysis
For all three example resonators (Table 3) the first simulation model assumes a uniform heat generation over the whole volume of the inductor. Figures 7,8,9 show the results of the simulations. All temperature increases are below the critical value of 5 K. The maximum temperature is at the cylinder axis, corresponding to the center of the vessel for a vessel implant. The uniform heat generation inside homogeneous tissue is – as discussed above – not the best approximation. The ring model is a more appropriate match for realistic conditions.
Therefore resonator 1 is assumed as a solenoid with 2 turns (Figure 10). The simulation assumes these turns as metallic rings, which should be similar to a solenoid for the question of temperature increases. For resonator no 2, six rings are calculated (Figure 12) and resonator 3 is simulated with twelve rings (Figure 13). Only the simulation for resonator no 1 shows temperature increases over 5 K, which are located on and directly adjacent to the power generating rings. For resonator 2 the temperature increase with 6 rings is slightly above 1 K (Figure 12). The same is true for resonator no 3 with twelve rings (Figure 13).
For 6 or 12 rings and after 600 s calculated simulation time, the peak values near the location of the power generating rings inside the tissue become more and more negligible. Therefore a model using a power generating cylinder shell is appropriate for resonators with a reasonable number of turns.
For resonators no 2 and no 3, the cylinder shell simulation did not show any unsafe heating and is similar to the ring calculations. Neglecting the peaks of the ring simulation the maximum temperature increase is nearly identical.
Only the simulation for resonator no 1 with two rings reaches a critical temperature value above 5 K, all others stay below this value. For a large number of rings or, stated more precisely, a high wire density at the cylinder surface, it is possible to use the cylinder shell model (compare Figures 12, 14 and Figures 13, 15). As expected, the rings cause higher maximum temperature increases according to the higher local power density. The difference between the shell generating the entire power (Figures 14, 15) and a uniform power generating volume (Figures 8, 9) is mainly a different temperature distribution inside the inductor. The former simulation first shows a moderate increase in radial direction up to the power generating shell inside the volume, followed by a decrease. The latter has a larger peak value at the axis of the cylinder and falls off continuously with greater distance from the axis. The ring model is the most preferable because of the dominance of the electric losses on and near the wire of the inductor. Using different metals without altering the other simulation parameters changes the results only marginally (Figures 10b,c; 11 b-d; 12 a, b).
The simulation worked on a "worst case" basis, neglecting in most cases all cooling effects except the energy transport due to the thermal conductivity. Especially for vascular implants in the "normal case", some blood flow and blood perfusion, causing a much faster energy transport, will reduce the temperature increases. The effect of blood flow is simulated for the example of resonator 1 with the two rings. This was the only simulation reaching physiologically critical temperature increases. Blood flow is simulated by keeping the temperature difference at zero for a central part of the cylindrical implant. Changing the dimension of the central part can simulate an implant, which may be more or less infiltrated to the vessel wall or covered by tissue (thrombosis, calcification or intimal hyperplasia). The reduced temperature increase for resonator no 1 and for various sizes of the cooling blood flow is shown in Figure 11. For an implant placed directly inside the wall, with none or only a small amount of tissue between wire and blood flow the temperature increases drop to a physiologically tolerable value (Figure 11). Nevertheless for implants with a tissue coverage of a few mm, which can be plaque or a thrombosis, critical temperatures can be reached.
With respect to the overall power absorption greater than 100 Watts inside the human body during MRI investigations with a maximum SAR, one additional Watt inside the entire body is negligible.
This investigation assumes a "worst-case scenario" in different ways. Firstly, the resonator is assumed being perfectly aligned within the plane of the excitation field B1. Secondly, to some extent no blood flow inside the inductor (vessel) of the implant is included. Thirdly, no blood perfusion inside the tissue around the resonator is taken into account. Lastly, too large quality factors are used.
For most "normal" cases, intact active implants will therefore be less critical, but it is not possible to exclude the worst case conditions. Pathologies (thrombosis, intimal hyperplasia, plaque) may alter blood flow and perfusion in the area directly adjacent to the current paths of the resonator and certainly the resonator can be perfectly aligned to the plane of the exciting rf-field B1.
Especially for resonators with a large volume (such as resonator no 1) and with a small number of rings it is possible to reach critical temperature increases above 5 K (Figures 10, 11). For peripheral and especially cardiac vessels, even with an unrealistically high quality factor, the power loss is too low for dangerous heating. On the other hand, stent grafts or vena cava filters built as active implants can reach volumes of a few ten cm3, which may be dangerous, if the tissue around the wires of the implant is not exposed to blood flow or sufficient blood perfusion. To reduce the risk for such active implants, the quality factor has to be low. This reduces also the amplification of the MR signal the resonator is made for and can thus be dispensed with altogether.
As the above simulations assume properly working resonance circuits without any failures on the electric paths of the system, one worst case scenario was not presented. Defects such as ruptures or partial ruptures may generate a relatively high resistance over very short distances. The current flow through this resistance can produce a large power loss inside an extremely small volume. This can generate a very high power density which, even for small implants, may induce physiologically critical temperature increases for a small volume. In fact the analysis of these "hot spots" is important, because ruptures of stent struts are likely and a high power density can occur also for smaller implants. An additional investigation estimating the maximum possible power loss inside such "hot spots" and the resulting temperature maps around them is necessary to check the safety of active implants under these circumstances. This is being prepared for future publication.
The study protocols for the cited animal experiments were approved by the responsible authority (Landesamt für Arbeitsschutz, Gesundheitsschutz und technische Sicherheit, Berlin, G 0142/99).
Many thanks to Peter van Leeuwen, a bilingual native English speaker who has reviewed and corrected the entire "German English" manuscript.
- Bartels LW, Smits HF, Bakker CJ, Viergever MA: MR imaging of vascular stents: effects of susceptibility, flow, and radiofrequency eddy currents. J Vasc Interv Radiol 2001, 12(3):365–371.View ArticleGoogle Scholar
- Hug J, Nagel E, Bornstedt A, Schnackenburg B, Oswald H, Fleck E: Coronary arterial stents: safety and artifacts during MR imaging. Radiology 2000, 216(3):781–787.View ArticleGoogle Scholar
- Maintz D, Kugel H, Schellhammer F, Landwehr P: In vitro evaluation of intravascular stent artifacts in three-dimensional MR angiography. Invest Radiol 2001, 36(4):218–224. 10.1097/00004424-200104000-00004View ArticleGoogle Scholar
- Meyer JM, Buecker A, Schuermann K, Ruebben A, Guenther RW: MR evaluation of stent patency: in vitro test of 22 metallic stents and the possibility of determining their patency by MR angiography. Invest Radiol 2000, 35(12):739–746. 10.1097/00004424-200012000-00008View ArticleGoogle Scholar
- Schnall MD, Barlow C, Hirihara Subrmaniam V, et al.: Wireless implanted magnetic resonance probes for in vivo NMR. J Magn Reson 1986, 68: 161–167.Google Scholar
- Farmer TH, Cofer GP, Johnson GA: Maximizing contrast to noise with inductively coupled implanted coils. Invest Radiol 1990, 25(5):552–8.View ArticleGoogle Scholar
- Patent of Simag GmbH, Teltowkanalstr.2, D-12247 Berlin, Germany, EP1092985A3Google Scholar
- Quick HH, Kuehl H, Kaiser G, Bosk S, Debatin JF, Ladd ME: Inductively coupled stent antennas in MRI. Magn Reson Med 2002, 48(5):781–90. 10.1002/mrm.10269View ArticleGoogle Scholar
- Kivelitz D, Wagner S, Schnorr J, Wetzler R, Busch M, Melzer A, Taupitz M, Hamm B: A Vascular Stent as an Active Component for Locally Enhanced Magnetic Resonance Imaging: Initial In Vivo Imaging Results After Catheter-guided Placement in Rabbits. Invest Radiol 2003, 38(3):147–152. 10.1097/00004424-200303000-00002Google Scholar
- Kivelitz D, Wagner S, Hansel J, Schnorr J, Wetzler R, Busch M, Melzer A, Taupitz M, Hamm B: The active magnetic resonance imaging stent (AMRIS): initial experimental in vivo results with locally amplified MR angiography and flow measurements. Invest Radiol 2001, 36(11):625–631. 10.1097/00004424-200111000-00001View ArticleGoogle Scholar
- Athey TW, Ross RJ, Ruggera PS: Magnetic Fields Associated with Nuclear Magnetic Resonance Medical Imaging System. Mag Res Imaging 1982, 1: 149–154. 10.1016/0730-725X(82)90206-5View ArticleGoogle Scholar
- Margulis AR, Higgins CB, Kaufman L, Crooks LE: Clinical Magnetic Resonance Imaging. In Radiology Research and Education Foundation. Library of Congress Catalog Card number 83–81823; 1983.Google Scholar
- Johnson CC, Guy AW: Nonionizing Electromagnetic Wave Effects in Biological Materials and Systems. Proc of the IEEE 1972., 60(6):Google Scholar
- Yeung CJ, Susil RC, Atalar R: RF Safety of Wires in Interventional MRI: Using a Safety Index. Magnetic Resonance in Medicine 2002, 47: 187–193. 10.1002/mrm.10037View ArticleGoogle Scholar
- Hand JW, Lau RW, Lagendijk JJW, Ling J, Burl M, Young IR: Electromagnetic and Thermal Modeling of SAR and Temperature Fields in Tissue Due to an RF Decoupling Coil. Magnetic Resonance in Medicine 1999, 42: 183–192. 10.1002/(SICI)1522-2594(199907)42:1<183::AID-MRM24>3.0.CO;2-IView ArticleGoogle Scholar
- Liao SamuelY: Microwave Devices and Circuits. Prentice-Hall, Englewood Cliffs, New Jersey; 1980:134.Google Scholar
- Sminia P, Haveman J, Wondergem J, van Dijk JD, Lebesque JV: Effects of 434 MHz microwave hyperthermia applied to the rat in the region of the cervical spinal cord. Int J Hyperthermia 1987, 3(5):441–52.View ArticleGoogle Scholar
- Haveman J, Van Der Zee J, Wondergem J, Hoogeveen JF, Hulshof MC: Effects of hyperthermia on the peripheral nervous system: a review. Int J Hyperthermia 2004, 20(4):371–91. 10.1080/02656730310001637631View ArticleGoogle Scholar
- Khoei S, Goliaei B, Neshasteh-Riz A, Deizadji A: The role of heat shock protein 70 in the thermoresistance of prostate cancer cell line spheroids. FEBS Lett 2004, 561(1–3):144–8. 10.1016/S0014-5793(04)00158-9View ArticleGoogle Scholar
- Poland J, Schadendorf D, Lage H, Schnolzer M, Celis JE, Sinha P: Study of therapy resistance in cancer cells with functional proteome analysis. Clin Chem Lab Med 2002, 40(3):221–34. 10.1515/CCLM.2002.037View ArticleGoogle Scholar
- Sinha P, Poland J, Schnolzer M, Celis JE, Lage H: Characterization of the differential protein expression associated with thermoresistance in human gastric carcinoma cell lines. Electrophoresis 2001, 22(14):2990–3000. 10.1002/1522-2683(200108)22:14<2990::AID-ELPS2990>3.0.CO;2-FView ArticleGoogle Scholar
- Gowrishankar TR, Stewart DA, Martin GT, Weaver JC: Transport lattice models of heat transport in skin with spatially heterogeneous, temperature-dependent perfusion. BioMedical Engineering OnLine 2004, 3(1):42. 10.1186/1475-925X-3-42View ArticleGoogle Scholar
- Geddes LA, Baker LE: The specific resistance of biological material – a compendium of data for the biomedical engineer and physiologist. Med & Biol Eng 1967, 5: 271–293.View ArticleGoogle Scholar
- Holmes KennethR[http://www.ece.utexas.edu/~valvano/research/Thermal.pdf]
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.