Gradient heating of bulk metallic implants can be a safety concern in MRI

It is generally accepted that potentially hazardous tissue heating can occur when a metallic implant is exposed to strong radiofrequency (RF) fields in MRI 1-4. Although device heating by gradient switching is mentioned in some normative documents 1, 2, but not all 3, 4, the popular opinion is that this effect must be irrelevant for patient safety, and does not warrant concern or deeper investigation. The few existing literature on this subject appeared to confirm this viewpoint 5. This scenario changed when a numerical simulation study by Zilberti et al 6 reported significant gradient-induced temperature rises in and around hip prostheses. Primarily, the implant itself was heated (by eddy currents), and thermal conduction caused the adjacent tissue to follow closely. These findings were not expected by many, as it had often been assumed that the much lower frequencies of gradient switching, compared with the RF case, should result in correspondingly less heating. Here, experimental results are reported that confirm the possibility of substantial gradient heating of an orthopedic implant. The test object was the acetabular cup (diameter of 54 mm, mass of 70.7 g) of an excised hip prosthesis (Plasmafit Plus-3, Aesculap, Center Valley, PA, USA). The device, untested for MR compatibility by the manufacturer 7, is made of Ti-6Al-4V alloy 8, whereas CoCrMo showed even larger temperature effects in 6. With four fiber-optical temperature sensors (Luxtron, LumaSense Technologies, Santa Clara, CA, USA) attached, the implant was placed in a clinical 3 tesla scanner (Verio, Siemens Healthcare, Munich, Germany) at position with respect to the iso-center in gradient coordinates. Anatomically, this corresponds to a male adult, lying supine, and laterally off-center by . Based on manufacturer-provided field maps, and were chosen for maximum expected effect while keeping reasonable. This represents a bad example—one that is not typical, but possible. A self-written echo planar imaging like sequence with continuous, trapezoidal z-gradients normal to the implant face was used. The gradient strength ( ), slew rate ( ), and frequency ( ) were set to the scanner limits when the sequence was run in "normal operating mode." All safety features (eg, stimulation monitor, temperature supervision) were in place; RF was off. Scanning a subject with this sequence would have been permitted. Experiments were run with both the implant thermally "insulated" (in polystyrene) or "embedded" (in gelatin gel). Within 10 min, maximum temperature rises of for the insulated and for the embedded implant were observed. If the adjacent tissue follows the implant temperature, as it was found in 6, we can estimate a specific gradient heating power using , where is that tissue's specific heat capacity and is the initial slope of the temperature curve (identical, within errors, for the insulated or embedded case) (Supporting Information Fig. S1). For example, for muscle 9; hence, . Applying, for comparison, the product echo planar imaging sequence, we obtained 40%, with Graf's 5 sequence and implant position 2.6% of this value. The resulting tissue temperatures can only be determined by simulations that take into account heat dissipation, such as by perfusion 6, 10, 11, whereas here an independent experimental test of such simulations is intended. An RF heating test performed for comparison (Supporting Information Fig. S2) showed no detectable temperature effect. The relationship of to the local specific absorption rate (SAR) for RF heating, for which—based on an implicit model of tissue cooling—a limit of 20 W/kg is imposed (1), remains to be clarified. Both quantities are different in their physical nature and expected effect, so their values are not directly comparable, and established local SAR exposure limits cannot immediately be applied to . Averaging over 10 g 1 or 1 g 4 of tissue, for example, is an integral part of the local SAR rationale, whereas as derived previously applies only to a possibly very thin layer surrounding the implant. The inductive heating power of a gradient sequence depends only on its root mean square–averaged 2. Our self-written sequence generated close to the bore, and at the implant (ie, not vastly exceeding the FPO:B limit of 56 of the new fixed-parameter option 1, which is currently the only existing limitation in a standard with respect to scanning patients with MR-conditional implants). From the scarce data presented here, it cannot be concluded that gradient heating of bulk metallic implants is automatically dangerous. However, the opposing opinion that it can safely be ignored, can no longer be maintained. The low frequencies are no impediment, as gradients primarily heat the implant; the effectiveness of this mechanism is demonstrated in Figure 1. These results imply the existence of an additional critical region for implants at and beyond the ends of the gradient coil (ie, in a region far outside the field of view, the footprint of the RF transmit coil, and also possibly the operator's attention). Left: Prosthesis cup embedded in gelatin with four thermo-sensors attached. The top-most sensor (arrow) showed the fastest temperature response. Right: Temperature increase above ambient temperature at the topmost sensor for the insulated implant under 10 min of gradient heating. The curve has an initial slope of 0.11 K/s. Inset: Corresponding curve for the embedded implant. We are grateful to Jochen Rohrbeck, MD, from Park-Klinik Weissensee, Berlin, for providing the excised hip joint, to Patrick Waxmann for programming the gradient sequence, and to Siemens Healthcare for providing the gradient field maps. Additional Supporting Information may be found in the online version of this article Fig. S1. Absolute temperature curves for the insulated (left) and embedded (right) implant under gradient heating (same measurements as in Fig. 1, but with unprocessed raw data). The blue lines are least-squares fits of empirical model functions (power laws). Both of the depicted fit curves have initial slopes of 110 mK/s, but for the case of the embedded implant, a reliable determination of this quantity is not possible from the available data. Fig. S2. Setup and results of an RF heating test. The implant was mounted in the same way and in the same container as for gradient heating, but because RF heating is caused by electrical currents in tissue, the nonconductive gelatin was replaced by a viscous, tissue mimicking liquid (hydroxyethyl cellulose doped with NaCl) with measured conductivity σ=0.45 S/m and permittivity ε=80 at 123.2 MHz. The photo on the left shows the arrangement with one thermo-sensor measuring directly in the liquid. The container was placed in the scanner bore with the implant at , as the strongest E fields from the RF body coil are expected at and laterally as far off-axis as possible. The curves on the right show the temperature curves (raw data, vertically offset for clarity) for 6 min of RF heating at nominally 143.6 W RF power without gradients. The RF on and off time points are indicated by vertical lines. The figure demonstrates that for this particular experimental arrangement, RF heating is negligible compared with gradient heating (Fig. 1), but no further conclusions should be drawn. Compared with gradient heating, RF heating depends much more on composition and size of the phantom or the loading of the RF coil. This was not comprehensively investigated, as RF heating was not the primary subject of the present study. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.