One limitation on the ability to monitor health in older adults using Magnetic Resonance (MR) imaging is the presence of implants, where the prevalence of implantable devices (orthopedic, cardiac, neuromodulation) increases in the population, as does the pervasiveness of conditions requiring MRI studies for diagnosis (musculoskeletal diseases, infections, or cancer). The present study describes a novel multiphysics implant modeling testbed using the following approaches with two examples:
- an in-silico human model based on the widely available Visible Human Project (VHP) cryo- section dataset;in-silico human model based on the widely available Visible Human Project (VHP) cryo- section dataset;
- a finite element method (FEM) modeling software workbench from Ansys (Electronics Desktop/Mechanical) to model MR radio frequency (RF) coils and the temperature rise modeling in heterogeneous media.
The in-silico VHP Female model (250 parts with an additional 40 components specifically characterizing embedded implants and resultant surrounding tissues) corresponds to a 60-year-old female with a body mass index (BMI) of 36. The testbed includes the FEM-compatible in-silico human model, an implant embedding procedure, a generic parameterizable MRI RF birdcage two-port coil model, a workflow for computing heat sources on the implant surface and in adjacent tissues, and a thermal FEM solver directly linked to the MR coil simulator to determine implant heating based on an MR imaging study protocol. The primary target is MR labeling of large orthopaedic implants. The testbed has very recently been approved by the US Food and Drug Administration (FDA) as a medical device development tool (MDDT) for 1.5 T orthopaedic implant examinations.
This manuscript will provide a valuable method to evaluate the safety of MR in patients with orthopaedic implants, which is required in clinics. A strength of the work is that the in-silicon testbed is solid, based on the widely available human project, and validated. In addition, the toolbox will be open for clinical practice.
One limitation on the ability to monitor health in older adults using Magnetic Resonance (MR) imaging studies is the presence of implants, where the prevalence of implantable devices (orthopedic, cardiac, neuromodulation) increases in the population, as does the pervasiveness of conditions requiring MRI studies for diagnosis (musculoskeletal conditions, infections, or cancer). In 2020, 26% of the US population over 65 was estimated to carry a large joint or spinal implant . Simultaneously, 12.6 million patients over 65 who carry orthopedic or cardiac implants will need an MR study within 10 years, according to an estimate in 2020  with this number expected to rise. Similarly, over 70% of the estimated 3 million pacemakers in the US are implanted in patients older than 65 ,, where approximately 20% of these patients will need an MR study within 12 months of device implantation .
Terms to be used to label MR information for medical devices – implants – include MR safe, MR conditional, and MR unsafe ,,,. MR safe items are nonconducting, nonmetallic, and nonmagnetic items, such as a plastic Petri dish . MR unsafe items include in particular ferromagnetic materials ; they should not enter the MR scanner room . All other devices that contain any metallic components, such as titanium, (regardless of ferromagnetism) are MR conditional and will need to be evaluated and labeled for RF-induced heating, image artifact, force, and torque . For the corresponding labeling icons, see .
MR conditional implants may safely enter the MR scanner room only under the very specific conditions provided in the labeling. Patients should not be scanned unless the device can be positively identified as MR conditional and the conditions for safe use are met . When present, information about expected temperature rise and artifact extent may inform the risk/benefit decision of whether or not a patient should undergo an MR examination .
Given the large numbers of implants subject to conditional labeling, the number of cleared FDA 510(k) submissions for orthopedic implantable devices with MR labeling has been growing exponentially since 2014 , approaching 100 in 2019 . However, practical testing is limited by constraints related to cost and resources, including testing tools . As a result, a number of implants have been labeled “MR Not Evaluated”, which precludes patients’ access to MR imaging procedures . Other implants may be labeled too restrictively , limiting patient access to MR imaging . When estimating combined data from ,,,, up to two million elderly patients in the US are potentially affected by MR labeling uncertainty.
Presumably, the most important consequence of this uncertainty is restricting general access to MR imaging studies for patients with implants. This also prevents the use of MR imaging for better soft tissue monitoring in the vicinity of implants. A prime example of the latter is periprosthetic joint infection following total hip replacement surgery, which occurs in only 1%–2% of primary arthroplasties , but in up to 30% of revision arthroplasties . This form of infection occurs due to mechanical loosening and dislocation, currently the most common causes for revision of total hip arthroplasty in the United States . Periprosthetic joint infection related mortality is approaching 5-8% at one year . Presently, X-rays and other methods are used for diagnosis , but results of MR imaging with metal artifact reduction were recently shown to be the most accurate tool in the diagnosis of several biomarkers of periprosthetic hip joint infection ,.
One major safety concern, relevant to both passive and active implants, is implant heating within MR radiofrequency and gradient coils . Along with the required yet not entirely anatomical ASTM phantom test ,, numerical simulations with virtual human models generate accurate predictions of temperature rise ,,,,,, accepted by the FDA . The electromagnetic and thermal simulation algorithms based on finite element, finite difference, and boundary element methods are reasonably well developed ,,,. However, accessible, full body, detailed anatomical virtual human models reflecting major age, sex, race, and obesity variations are severely lacking. Their creation is a long, tedious, and labor-intensive process. Even today, it requires manual and semiautomatic supervised segmentation of full body MR images, surface mesh reconstruction, mesh intersection resolution, software compatibility and robustness testing, and finally examination of hundreds of different body compartments by anatomical experts.
An excellent collection of in-silico human body models intended for this purpose is the Virtual Population, a product of the IT’IS Foundation , widely used in both industrial and academic applications. While this population has many highly detailed body models, it is relatively homogeneous: reasonably fit, younger Caucasian European subjects, representing a comparatively limited subsection of human anatomy and physiology. Although three obese models were added in 2023 , two of the three are not truly anatomical and were obtained via morphing (the ‘Fats’ model being the exception). Also, while models are available for purchase and for research purposes, no background MRI data enabling independent tissue structure verification have been made publicly available for this population set.
The present study describes a complete ready-to-use implant modeling testbed for RF heating based on
a FEM modeling software workbench from Ansys HFSS (Electronics Desktop) to model the physical phenomena of a MR RF coil and corresponding temperature rise in heterogeneous media.
The in-silico VHP Female model (250 anatomical structures with additional 40 components specifically characterizing embedded implants ) characterizes a 60-year-old female subject with a body mass index (BMI) of 36. The open-source version of this model  has over 600 registered users from both industry and academia worldwide. The testbed includes the in-silico model, an implant embedding procedure, a generic parameterizable MR RF birdcage two-port coil model, a workflow for computing heat sources on the implant surface and neighboring tissues, and a thermal FEM solver directly linked to the MR coil simulator to estimate implant heating based on an MR imaging protocol. The primary target is MR labeling of large orthopaedic implants. The testbed has recently been approved by the US Food and Drug Administration (FDA) as a medical device development tool (MDDT)  for 1.5 T orthopaedic implant examinations. We also present two simple application examples pertinent to choosing an appropriate MR imaging protocol for a particular orthopaedic implant as well as validation against measurements of the heading of an ablation needle in bovine liver.
2. Materials AND Methods
2.1 In silico-human model and implant embedding procedure
Fig. 1 shows surface CAD meshes for the VHP-Female model , (with some muscles removed for clarity) and examples of passive femoral implants embedded into the model. The corresponding physical femoral implants are shown on the top right of the figure.
The implant registration enforces an anatomically correct implant position, and a certain part of the bone matter (cortical and/or trabecular) to be removed as necessary. A semiautomatic implant registration algorithm requiring limited user intervention has been employed based on the principal idea to use at least two anchor nodes per implant: a fixed node and a floating node. The floating anchor node is a vertex of the implant mesh belonging to a certain curve, say, the long axis of the bone. The fixed anchor node is a joint coincident vertex of the femur mesh and the implant mesh. These nodes define the proper implant position given the bone model and a cost function, with a “best fit” based on a mesh intersection check and the signed normal distances between implant/bone boundaries. An additional criterion involves the minimum required thickness of the cortical bone matter with an embedded implant.
2.2 Computation of heat sources due to microwave absorption in MR RF coils
A generic, parameterized, and tunable MR RF birdcage two-port coil model (high-, low-, or bandpass), at 64 MHz (1.5 T) with a variable number of rungs was implemented in Ansys Electronics Desktop (Ansys HFSS, Fig. 2a). This model is used to compute heat sources – either specific absorption rate (SAR) in W/kg, or power loss density (PLD) in W/m3 at any point in the body, including on the surface of the implants (Figs. 2b and c ). The in-silico model with the implant(s) can be positioned at any appropriate landmark.
2.3 Determination of implant temperature rise as a function of scan time
An Ansys FEM transient thermal solver was employed to determine tissue temperature rise close to the implant caused by the heat sources. It requires knowing the relevant thermal properties of the tissues. The solver may approximately model blood perfusion, which is less important for bone, but is important for cardiac implants and other soft tissue implants .
The entire testbed has been integrated into Ansys Workbench, which allows the combination of different multiphysics modules within a single environment, as shown in Fig. 3.
In this way, the output of the electromagnetic solver (Figs. 3A and B) is the input to the thermal solver (Fig. 3D). Both accurate FEM solvers utilize the same human model geometry throughout, but with different material properties (thermal vs electromagnetic).
3.1 What is heating-related MR labeling of implants?
For the implantable devices that are categorized as MR Conditional, the labeling includes scan and rest times at a given whole-body SAR or (“root mean square” value of averaged over a period of 10 seconds). This is described in the FDA’s guidance document  (cf. also examples in ) where devices are to be labeled for a 1-hour MR session, including both scan and rest times. The guidance states a certain interleaving combination of scan (e.g., 5 min) and rest (e.g., 15 min) times that guarantees implant heating is less than 5 °C or another specified number . The FDA-required procedure is a measurement test in an ASTM gel-based homogeneous phantom ,. When performing relevant numerical modeling, the pulse sequences and scan times should be converted to equivalent CW (continuous wave) operation, which is easier to model.
However, the response of the ASTM phantom is quite different from that of a real body, which includes bones and other tissues of varying electrical and thermal conductivities, as well as blood circulation and perfusion. In several test cases, our testbed prototype predicted a higher maximum temperature rise (up to 40% higher) at the implant tips versus in-vitro experiments with a simplified gel phantom. In other cases, and for other implants, however, the heating was substantially lower (by 50% or so). Therefore, the in-silico testbed will augment the ASTM measurements with accurate multiphysics modeling. Additionally, this modeling can assist with implant design in an efficient manner.
We note that this work is solely focused on the RF safety aspects of MR labeling. The results expressed herein should be considered supplemental to existing published guidelines.
3.2 Example: labeling long femoral titanium nail
Fig. 4 shows an example of testing results for a long titanium femoral nail subject to three cycles of 15 min with a 2.3 W/kg average equivalent SAR exposure followed by 5 minutes of rest, resulting in a 1-hour total exposure in a 1.5 T MRI coil. The model predicts that the temperature near the implant reaches 41 °C after the first exposure with its final value approaching 45 °C, a total increase of about 10 °C which is clearly unacceptable! Further simulations show that 4 min exposures followed by 16 min of rest would be a safe solution. In the testbed, the exposure time is arbitrary and can be rapidly tested and adjusted (within 5-7 min) to meet the FDA requirements  and construct the proper MR exposure protocol.
3.3 Comparison with experiment for a long resonant metal conductor with a sharp tip
The most challenging and important cases correspond to testing large and potentially resonant metal implants  with relatively sharp tips or terminations since detecting resonance requires accurate high-frequency modeling. One extreme example was studied in Ref. , where a simulated percutaneous RF ablation surgical procedure using MR heating was performed in ex-vivo bovine liver in a 1.5 T scanner. The device under study was a bent long wire ‘antenna’ made resonant at the scanner Larmor frequency with an adjustable series capacitor. The antenna, a 26 AWG (0.40 mm) Teflon-insulated silver-plated copper wire taped around the edge of the patient table, was terminated in a simulated RF ablation needle (a 15 cm long 16 AWG/1.30 mm diameter bare copper wire), the tip of which was embedded into the liver to simulate the percutaneous ablation of a solid hepatic tumor. The parametrized testbed coil model was used to replicate the RF antenna and needle geometry of Ref. , using standard electrical properties of human liver tissue. The peak tissue temperature increase imaged in Ref.  by the proton resonance frequency shift (PRFS) method (20 °C) and a 22 °C increase recorded by a fiber optic temperature sensor at the needle tip agreed well with our modeled prediction of a 23 °C increase using the modeling testbed.
A very fine FEM mesh resolution is required to accurately resolve temperature rise close to a sharp lead tip with a diameter of 1.3 mm – cf. Fig. 5. This is achieved using local automated adaptive FEM mesh refinement, which is a unique property of the present testbed. Fig. 5 shows the corresponding testbed setup along with the lesion. A few relevant field movies are available as supplementary materials .
4. Discussion AND Conclusion
The temperature rise in the surrounding tissues of a large orthopedic metallic implant subject to MR imaging is a significant point of concern and a potential barrier for the development of better implants. Numerical electromagnetic and thermal modeling offers a way to solve this complex problem with a sufficient degree of accuracy. We developed a complete testbed for realistic implant modeling, which includes a detailed FEM-compatible obese human female model, a parameterized tunable generic MR coil model, a method for implant embedding, and an accurate radio frequency solver directly coupled with a transient thermal solver.
In the testbed, the MR exposure time is arbitrary and can be readily adjusted and rapidly tested (typically within ∼5-7 min). This enables the user to meet regulatory requirements  and to construct a proper MR exposure protocol. A cross-platform compatibility of the in-silico model has been established previously . Further validation of cross-platform electromagnetic and thermal coupling performance is currently underway with Dassault Systèmes CST Studio Suite software package. The testbed along with the open-source version of the human model VHP-Female College  is available online .
This work was supported by the National Institutes of Health grants R01AR075077 and R01EB029818.
- Scanning patients with MR conditional implants. Philips Professional Healthcare
- Implantable Cardiac Electronic Devices in the Elderly PopulationArrhythm Electrophysiol Rev 8:143–146https://doi.org/10.15420/aer.2019.3.4
- Puette JA, Malek R, Ellison MB. Pacemaker. [Updated 2022 Sep 12]. In: StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing Co; 2022 Jan-. Available from: https://www.ncbi.nlm.nih.gov/books/NBK526001/
- Calculation of MRI RF-Induced Voltages for Implanted Medical Devices Using Computational Human ModelsBrain and Human Body Modeling: Computational Human Modeling at EMBC 2018 [Internet] https://doi.org/10.1007/978-3-030-21293-3_1
- Testing and Labeling Medical Devices for Safety in the Magnetic Resonance (MR) Environment. U.S. Department of Health and Human Services, Food and Drug Administration, Center for Devices and Radiological Health. May 20th 2021. Online: https://www.fda.gov/media/74201/download.
- Understanding MRI Safety Labeling. U.S. Food and Drug Administration. Summary, online: https://www.fda.gov/media/101221/download
- ACR Manual on MR Safety
- MR labeling information for implants and devices: explanation of terminologyRadiology 253:26–30https://doi.org/10.1148/radiol.2531091030
- Retrospective Analysis of Radio-frequency Safety of Orthopedic Passive Implantable Device
- Medical Device Development Tool MDDT Q170004 Communications with FDA Center for Devices and Radiological Health, US Food and Drug Administration (FDA)
- FDA MDDT: Computational Tool Comprising Visible Human Project® Based Anatomical Female CAD Model and Ansys HFSS/Mechanical® FEM Software for Temperature Rise Prediction near an Orthopedic Femoral Nail Implant during a 1.5 T MRI Scan. 03/30/32. Neva Electromagnetics, LLC. Online: https://www.fda.gov/media/166724/download. xU.S. Food and Drug Administration Bulletin of 03/30/2023. Online: https://content.govdelivery.com/accounts/USFDA/bulletins/351ef58?reqfrom=share
- FDA 510(k)/De Novo Clearance and Premarket Approval Applications: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/search/default.cfm, accessed Jan. 2023.
- NEVA Electromagnetics, LLC. VHP-Female 2.2 (College). Online download page: https://www.nevaelectromagnetics.com/vhp-female-2-2
- Prosthetic-joint Infections: Mortality Over The Last 10 YearsJ Bone Jt Infect 4:198–202https://doi.org/10.7150/jbji.35428
- The Expanding Role of MRI in the Evaluation of Periprosthetic Hip Joint InfectionRadiology 296:109–110https://doi.org/10.1148/radiol.2020201419
- What’s New in Hip ReplacementJ Bone Joint Surg Am 100:1616–1624https://doi.org/10.2106/JBJS.18.00583
- Joint Replacement Infection
- Diagnosis of Periprosthetic Hip Joint Infection Using MRI with Metal Artifact Reduction at 1.5 TRadiology 296:98–108https://doi.org/10.1148/radiol.2020191901
- Retrospective analysis of RF heating measurements of passive medical implantsMagn Reson Med 80:2726–2730https://doi.org/10.1002/mrm.27346
- Practical Aspects of MR Imaging Safety Test Methods for MR Conditional Active Implantable Medical DevicesMagn Reson Imaging Clin N Am 28:559–571https://doi.org/10.1016/j.mric.2020.07.008
- Computational dosimetry in MRI in presence of hip, knee or shoulder implants: do we need accurate surgery models?Phys Med Biol 15https://doi.org/10.1088/1361-6560/aca5e6
- Heating of hip joint implants in MRI: The combined effect of RF and switched-gradient fieldsMagn Reson Med 85:3447–3462https://doi.org/10.1002/mrm.28666
- Parameters Affecting Worst-Case Gradient-Field Heating of Passive Conductive ImplantsJ Magn Reson Imaging 56:1197–1204https://doi.org/10.1002/jmri.28321
- Gradient coil and radiofrequency induced heating of orthopaedic implants in MRI: influencing factorsPhys Med Biol 23https://doi.org/10.1088/1361-6560/ac3eab
- MRI-Related Heating of Implants and Devices: A ReviewJ Magn Reson Imaging 53:1646–1665https://doi.org/10.1002/jmri.27194
- MDDT Summary of Evidence and Basis of Qualification Decision for IMANALYTICS with MRIXVIP1.5T/3.0T and BCLIB
- Virtual Human Models for Electromagnetic Studies and Their ApplicationsIEEE Rev Biomed Eng 10:95–121https://doi.org/10.1109/RBME.2017.2722420
- Computational Tool Comprising Visible Human Project® Based Anatomical Female CAD Model and Ansys HFSS/Mechanical® FEM Software for Temperature Rise Prediction Near an Orthopedic Femoral Nail Implant During a 1.5T MRI ScanBrain and Human Body Modelling 2021 https://doi.org/10.1007/978-3-031-15451-5_9
- Development of a new generation of high-resolution anatomical models for medical device evaluation: the Virtual Population 3.0Phys Med Biol 59:5287–303https://doi.org/10.1088/0031-9155/59/18/5287
- IT’IS Foundation. Three Class III Obese ViP Models for Improved Patient Coverage. Jan. 31st 2023. Online: https://itis.swiss/news-events/news/virtual-population/class-iii-obese-models/
- Visible Human Project® female surface based computational phantom (Nelly) for radio-frequency safety evaluation in MRI coilsPLoS One 16https://doi.org/10.1371/journal.pone.0260922
- The CAD-Compatible VHP-Male Computational PhantomBrain and Human Body Modeling 2020 https://doi.org/10.1007/978-3-030-45623-8_19
- Comparative analysis of different hip implants within a realistic human model located inside a 1.5T MRI whole body RF coilAnnu Int Conf IEEE Eng Med Biol Soc :7913–6https://doi.org/10.1109/EMBC.2015.7320227
- “The visible human male: A technical reportJ. Amer. Medical Informatics Assoc 3:118–130
- The Visible Human ProjectProc :504–511
- MRI-Related Heating of Implants and Devices: A ReviewJ Magn Reson Imaging 53:1646–1665https://doi.org/10.1002/jmri.27194
- Magnetic Resonance Mediated Radiofrequency AblationIEEE Trans Med Imaging 37:417–427https://doi.org/10.1109/TMI.2017.2753739
- Multi-Purpose VHP-Female Version 3.0 Cross-Platform Computational Human ModelProceedings of the 10th European Conference on Antennas and Propagation :10–15
- Online Dropbox link: In-Silico Testbed for MR Labeling of Orthopaedic Implants based on open-source VHP-Female 2.2 College. June 2023.