HYPER: pre-clinical device for spatially-confined magnetic particle hyperthermia

Abstract Purpose Magnetic particle hyperthermia is an approved cancer treatment that harnesses thermal energy generated by magnetic nanoparticles when they are exposed to an alternating magnetic field (AMF). Thermal stress is either directly cytotoxic or increases the susceptibility of cancer cells to standard therapies, such as radiation. As with other thermal therapies, the challenge with nanoparticle hyperthermia is controlling energy delivery. Here, we describe the design and implementation of a prototype pre-clinical device, called HYPER, that achieves spatially confined nanoparticle heating within a user-selected volume and location. Design Spatial control of nanoparticle heating was achieved by placing an AMF generating coil (340 kHz, 0–15 mT), between two opposing permanent magnets. The relative positions between the magnets determined the magnetic field gradient (0.7 T/m–2.3 T/m), which in turn governed the volume of the field free region (FFR) between them (0.8–35 cm3). Both the gradient value and position of the FFR within the AMF ([−14, 14]x, [−18, 18]y, [−30, 30]z) mm are values selected by the user via the graphical user interface (GUI). The software then controls linear actuators that move the static magnets to adjust the position of the FFR in 3D space based on user input. Within the FFR, the nanoparticles generate hysteresis heating; however, outside the FFR where the static field is non-negligible, the nanoparticles are unable to generate hysteresis loss power. Verification We verified the performance of the HYPER to design specifications by independently heating two nanoparticle-rich areas of a phantom placed within the volume occupied by the AMF heating coil.


Introduction
Magnetic particle hyperthermia (MPH) relies on the delivery of magnetic nanoparticles (MNPs) to tumors followed by the application of an alternating magnetic field (AMF), causing local heating of tissue to kill tumor cells, either directly or by enhancing the cytotoxic effects of radio/chemotherapy [1,2].A key advantage of MPH is its versatility with remote activation of MNPs.AMFs penetrate tissue with minimal attenuation to heat the MNPs, which generate heat localized to the target area.Early clinical trials demonstrated the benefits of MPH to treat prostate cancer [3], and the European Medicines Agency approved MPH for treating recurrent glioblastoma with radiation [4].As with other forms of hyperthermia, MPH has four main effects on tumors: induction of tumor cell apoptosis (inhibition or abrogation of DNA damage repair) [5], upregulation of tumor heat shock proteins [6], increased blood perfusion that improves oxygenation to enhance radiotherapy [7], and more recently, it has been suggested that MPH can also enhance cancer immunotherapies [8,9] as well as induce immunogenic cell death [10].
For MPH to improve, it must offer precision heating to control intratumor temperature and limit thermal damage to the surrounding normal tissue.Current MPH implementations use AMF coils that illuminate large volumes of the patient and rely on MNP delivery methods that produce heterogeneous intratumoral MNP distributions [11][12][13].Achieving spatial control of MNP heating within the tumor requires new technology that inhibits the activation of MNPs outside the defined treatment volume.
The challenge of confining AMF heating to a specific area has been traditionally addressed by modifying RF coil geometry (e.g., coil diameter, coil length, orientation, etc.) [14].Changing the shape of the coil creates subsequent changes in magnetic flux density and field shape, which affects the temperature distribution within the sample [15,16].
In addition, the degree of heating could also be varied by adjusting the sample position within the coil; however, changing the coil requires iterative tuning with the inductive heating power supply, design validation, and a potentially large financial cost if multiple coil designs are used.
As an alternative, our prior work demonstrated localization of MNP heating using a strong magnetic field gradient to create a field free region (FFR), as shown in Figure 1 [17,18].To provide background, MNPs possess inherent magnetic anisotropy, which defines an energy barrier to achieve time-dependent moment reversal (relaxation) at a given temperature [19][20][21].An AMF exerts work on a particle system in an effort to reverse the magnetic moments at the same frequency.At low amplitude, the work expended by the external magnetic field does not exceed the anisotropic energy barrier and results in a lag (or hysteresis) in the magnetic moments of the particles as the field alternates.Per the first law of thermodynamics in magnetic systems, the generated thermal energy (via loss) is equivalent to the net work expended on the particle system by the AMF, or the area encompassed by the hysteresis curve [19].As shown in Figure 1, denoted by region 'A', the magnetic moments of the MNPs undergo relaxation in response to the AMF (light blue arrows) and generate thermal energy.For the nanoparticles outside the FFR, the higher intensity static field (gray arrows) fixes the magnetic moments of the MNPs to where moment reversal, and thus thermal energy generation does not occur.This concept allows for the 3-D confinement and control of MNP heating.
In Hensley et al. the FFR was fixed with respect to size and position, and spatially-confined heating was shown along one axis [18].In Tay et al. the same technology (spatially fixed FFR) is applied in 2D and validated in vivo, where the researchers selectively heated a tumor without damaging the MNP-loaded liver [17].While these works successfully demonstrated spatially-confined MNP heating, the utility of this technology can be improved by enabling greater spatial control and ease of use through automation.Herein, we describe the development, building, and testing of a prototype device, designated HYPER, that creates a FFR within a static magnetic field to enable spatially confined MPH.The device enables the user to designate regions of interest (ROIs) for heating in three-dimensional space, select the size of the heated region with adjustable permanent magnets, and create a custom treatment plan, which the automated prototype then implements.We will showcase the constructed prototype, as well as experimentally verify the design.

Prototype design
Our goal for the HYPER system was to enable spatially-confined MPH in small animal models.Intended for in vivo experiments with mice, the system needed to produce both an AMF for heating MNPs (within the Hergt-Dutz biological limit, 5 � 10 9 A/m/s) [22] and produce an adjustable magnetic field gradient to control the volume and position of the FFR used in heating confinement.We wanted to enable the end user to selectively heat volumes as small as 1 cm 3 to hyperthermia-relevant temperature ranges.
Our final design comprised several subsystems housed in a portable, aluminum frame encased with plastic paneling, as shown in Figure 2.
The circuit that generates the AMF consists primarily of Subsystems 5 and 6, which function as the resonance tank and tuning assembly, respectively.Here, the 'resonance tank' (Subsystem 5) refers to the RF coil (8 turn Cu coil, 55 mm diameter), in parallel with capacitors (2 � 200 nF in series), which resonates at a fixed frequency of 340 kHz with a Figure 1.Opposing magnets create a strong magnetic field gradient containing a FFR that provides the basis for spatially-confined heating with the HYPER prototype.The volume of the FFR depends on gradient strength, where a stronger gradient creates a smaller FFR.When and AMF is applied, the nanoparticles within the FFR (region A) are free to generate thermal energy due to hysteresis losses; however, outside the FFR (region B), the magnetic moments of the nanoparticles are fixed and contribute little to heating.
variable peak magnetic field amplitude between 0 and 15 mT (0-12 kA/m).The resulting field/frequency product is 4.1 � 10 9 A/m/s, which is within the biological safety limit.Water circulates through the capacitors and hollow RF coil to manage resistive thermal loads.The tuning assembly (Subsystem 6) consists of adjustable inductors and capacitors in order to match the impedance of the power supply and RF coil, which ensures efficient power transfer.
A series of four mechanical actuators adjust the sample and FFR simultaneously to coincide with a specific ROI.Two linear actuators (Subsystem 4) independently adjust the distance between the two opposing permanent magnets (Subsystem 2) in the x-direction, where the separation of the magnets relates inversely to the magnitude of the magnetic field gradient observed at the FFR (0.7-2.3 T/m).This actuation determines both the volume of the FFR and the position of the FFR along the x-axis.To adjust the FFR in the y-direction, a lift platform (Subsystem 7) moves the entire gradient magnet assembly vertically, independently from the resonance tank.The FFR location is not adjusted along the z-axis, but rather a mechanical actuator moves the sample stage (Subsystem 3) to coincide with the FFR position.
Prescribing a heating pulse sequence is done via a configuration file.The user can select ROIs for heating using a graphic user interface (GUI), shown in Figure 3, which is aligned with bi-planar cameras (Subsystem 1) that provide an anteroposterior (AP) and lateral (LAT) view of the sample.In this case, the ROIs 0 and 1 correspond to red and yellow crossed lines, respectively.
The user ROI selection within the GUI provides commands to the various actuators to accurately position the FFR with respect to the heating target.Upon selection of the desired number of ROIs, the software generates a configuration file template using the chosen locations.The user can then edit this file in the configuration editor.In addition to sequential heating of the selected regions, the user can also implement pulsed heating strategies that vary both duty cycle and sample position.After creating the desired treatment plan, the software provides an internal validation to ensure the plan does not exceed various hardware and software thresholds.After selecting 'Start' within the GUI, the HYPER executes the heating pulse sequence, while actively gathering temperature data from any connected thermal probes Figure 3. System health, hardware commands, and errors are monitored during the test in a Status Messages window.

Construction
Upon receipt of the disassembled HYPER unit from the manufacturer, Magnetic Insight, Inc., the HYPER system was assembled and calibrated at Johns Hopkins University School of Medicine, as shown in Figure 4.The magnet and RF assembly coincide with the labeled CAD model shown in Figure 2.An electronics rack contains the HYPER control systems, RF amplifier, two fiber optic thermal probes (Micronor Sensors, Inc, Ventura CA), and a desktop computer, from which the user can operate the HYPER.An external 2.5 kW chiller circulates water through the resonance tank at a temperature range of 20-40 � C in order to manage thermal loads from the coil.
The HYPER was designed for in vivo experiments with small animals; one of the challenges we wanted to address was limiting nonspecific tissue heating from the coil [23] and maintaining constant body temperature of anesthetized mice.As such, circulating water from the external chiller at 40 � C promotes higher environmental temperatures within the RF coil, which is useful for maintaining the core body temperature of anesthetized mice.Additionally, we used a custom 3D-printed sample holder to circulate heated water in close proximity to the mice.

Materials and methods
We performed two nanoparticle heating experiments to verify that the HYPER meets design goals.To verify the adjustable FFR conformation to the design criteria, a small aliquot of iron oxide nanoparticles was heated at varying positions along the x, y, and z-axes at several user-selected gradient fields within the design range.Second, we demonstrated user-selected spatially-confined heating using a mouse phantom.For all tests, we heated the nanoparticles with an AMF amplitude of 11 kA/ m at a frequency of 340 kHz.

Magnetic nanoparticles selection and characterization
All HYPER verification experiments were conducted with Synomag-D70V R nanoparticles (micromod Partikeltechnologie, GmbH, Rostock, Germany; Lot#: 09122104-02) in aqueous suspensions at a concentration of 1 mgFe/mL.The concentration was high enough to produce a moderate temperature rise as to adequately characterize the FFR, while limiting water vaporization during testing.Additionally, the selected concentration was slightly lower, but generally consistent, with those used in some of our previous works to achieve hyperthermia temperatures in vivo [7,8].Concentration verification was performed using ferene-s assay [24].After a 100fold dilution, the samples were digested in the ferene-s working solution (acetate buffer and ascorbic acid) for 20 h before reading with UV/vis spectrophotometer.The SLP for these particles at the selected field and frequency was determined to be 233 ± 30 W/gFe using the previously reported pulsed Box-Lucas (PBL) method [25].A graph of the temperature change vs time graph can be found in Figure S1 in the Supplementary Information.

Verification of adjustable FFR
A 100 mL aliquot of nanoparticles in a 1.5 mL Eppendorf tube was placed on the HYPER sample holder.Using the GUI, the FFR was centered on the sample starting position; the sample was then moved in 2 mm increments along the bore using the linear stage over a 60 mm range [−30 mm, 30 mm] relative to the fixed FFR (z-axis).At each interval, we heated the sample (applied an AMF) for 20 s, followed by a 120 s cool down period before shifting to the next position.Sample temperature was recorded using a fiber optic temperature probe throughout the test; the total change in temperature between the start and end of the 20 s heating pulse, DT, was recorded within each interval.We repeated this for both the x-axis and y-axis as well over the achievable range.Measurements were performed in triplicate, and the calculated DT values were assumed to be normally distributed.The average DT at each position increment was recorded with the standard deviation of the three replicates as the error.We performed the measurements for five gradient field values, 0.7, 1.1, 1.5, 1.9, and 2.3 T/m, which were varied by controlling the distance between the two gradient field magnets.The resulting plots were fit to the derivative of the Langevin equation, to determine the full width at half maximum (FWHM), which can be seen in the Supplementary Information.

Demonstration of spatially-confined heating with mouse phantom
A fillable mouse phantom (BioEmission Technology Solutions, Athens, Greece) was utilized to mimic an in vivo spatially confined MPH treatment, with particles in both the tumor and liver; this type of phantom has been used in previous imaging studies [26].The main benefit of the mouse phantom is the relevant geometry, which is similar to that of an in vivo mouse model.We filled two compartments within the phantom with the 1 mgFe/mL MNP solution: 'lower tumor' (204 mm 3 ) and 'liver' (1164 mm 3 ), which are approximately 3 cm apart.This type of MNP distribution is what would be expected from a mouse model given both an intratumor and intravenous dose of MNPs.The two compartments were labeled as ROI 0 and ROI 1, respectively, on the GUI.After selecting a 1.8 T/m gradient, the FFR was centered on ROI 0 and heated for 300 s; afterward, the FFR was moved a distance of to ROI 1 and heated for an additional 300 s.Overall DT was recorded in each ROI.

Verification of adjustable FFR
After heating the samples incrementally along each axis and at each gradient field, the impact of spatially confined heating became apparent.Figure 5 shows the measured temperature change along the x-axis (Figure 5a), y-axis (Figure 5b), and z-axis (Figure 5c) graphically, with tabulated data provided in the Supplementary Information in Tables S1-S3.Also called the thermal point spread function (PSF), the shape of these curves demonstrates how adjusting the gradient field affects the MNP heating profile.
The FWHM of the thermal PSF provides information regarding spatial resolution, specifically, the volume of the region undergoing AMF heating.We used the derivative of the Langevin function to model the PSF, which enabled us to fit each thermal PSF to analytically determine the FWHM (Figure 5d).Further information regarding these fits can be found in Supplementary Information in Table S4.The HYPER design only consists of one pair of opposing magnets in the x-dimension, which accounts for the FWHM being smaller (smaller FFR) along the x-axis.Without two additional pairs of magnets to further constrain the static field in the other orthogonal dimensions, the resulting FFR is asymmetrical, i.e., larger in the y-and z-axes.This analysis provides verification that the size of the effective heating region can be tuned automatically by the user by varying the distance between the permanent magnets with the x-axis actuators.

Demonstration of spatially-confined heating with mouse phantom
In conjunction with measuring the thermal PSF, we demonstrated spatially confined heating in a mouse phantom, as shown in Figure 6.Initially, the FFR was focused on ROI 0 (tumor), which showed a 4.9 � C rise in temperature during the 300 s heating pulse.Only a 0.6 � C rise in temperature was observed in ROI 1 (liver) during the same time.Following the first heating pulse, the FFR automatically aligned with ROI 1 and a second 300 s heating pulse began.During this period, ROI 0 began to cool due to atmospheric convection, while ROI 1 showed a 7.2 � C rise in temperature.Given that the volume of the liver compartment exceeded that of the tumor by a factor of 5.7, it is unsurprising that a higher temperature rise was observed.The consecutive heating of each of these regions within the RF coil using the gradient magnets demonstrates the ability of the HYPER prototype to perform the heating of a phantom tumor with minimal damage to other phantom organs that may contain MNPs.

Limitations
The development of the HYPER represents an advancement in spatially-confined AMF heating; however, we acknowledge design limitations: 1. Presence of Joule Heating: Confinement of MNP heating to a selected ROI eliminates peripheral hysteresis heating of nanoparticles; however, it does not eliminate off-target tissue Joule heating from generated eddy currents.Using a FFR to spatially confine heating relies on the concept of the particle magnetic moments being fixed by the gradient field, thus prohibiting relaxation and thermal energy generation.Eddy currents arise from interactions between AMFs and depend on the conductivity of the material (biological tissue), radius of eddy current path (related to volume of tissue exposed), as well as AMF frequency and amplitude.Though weakly (electric) conductive tissue is diamagnetic, it does not magnetically couple with the gradient field.2. Unguided ROI Selection: In its current state, the selection of the ROI depends solely on the visual alignment of the FFR using external cameras.The user selects the ROI in the GUI, which is based on their best judgment of the MNP localization; however, the selected ROI may not encompass the optimal volume for tumor heating.Only when coupled with magnetic particle imaging (MPI) will the user be able to optimally center the FFR.As the name implies, MPI is an emerging imaging modality, which solely images the MNPs [27][28][29][30].When co-registered with anatomical imaging (e.g., CT, MRI), the true MNP distribution in tissue can be observed and used to aid the HYPER user in the anatomical placement of the FFR. 3. MNP Selection: The responsiveness of a particle suspension to an applied magnetic field depends largely on material properties.As such, choosing a different MNP heating agent would result in a different SLP and thermal PSF, which the user would need to consider when performing spatially-confined heating on HYPER.4. Thermometry: The current setup consists of only singlepoint thermometry, which limits the system capabilities.Without multi-point thermometry, the user cannot gather spatially-dependent temperature data or properly calculate deposited thermal dose.Remote thermometry using MPI has been reported in the literature as a way to extract multi-dimensional temperature profiles, but a combined MPI/MPH system would be necessary [31].

Benefit to MPH
Aside from the limitations, spatially confined heating has the potential to provide clinical benefit.Results obtained using the mouse phantom shown in Figure 6a and b represent a conceptual in vivo MPH treatment in which both intravenous and intratumor injections of MNPs may be used.Following intravenous injection, the liver typically sequesters the majority of the MNPs which presents challenges for MPH.If the entire mouse body is encompassed by a homogenous flux density AMF (without any applied gradient field), both the tumor and the liver heat commensurately, with a higher thermal dose deposited into the tissue containing more MNPs (i.e., liver and/or spleen), leading to potentially severe off-target organ damage [32,33].Using spatially-confined heating, as demonstrated with HYPER, would confine heating to the tissues of interest (tumor) and significantly reduce unwanted heating along the periphery.

Conclusions
We designed, constructed, and verified a pre-clinical MPH device with spatially confined MNP heating capability.Verification consisted of two primary experiments that analyzed the impact of the FFR on heating.Based on the results of the experiments, we conclude: 1. Spatial resolution of the heated region can be effectively controlled by adjusting the magnetic gradient field strength.Within the HYPER system, the separation of the permanent magnets defines the gradient field magnitude, and thus the volume and position of the FFR.Based on the results, the FFR volume can be adjusted between (0.8-35 cm 3 ) within a 28 mm � 36 mm � 60 mm space.2. The HYPER prototype can achieve selective heating of different nanoparticle regions within the same RF coil by moving the FFR and the sample.This offers a unique capability for pre-clinical MPH, where off-target in vivo MNP heating can produce substantial tissue damage.
Our overall goal is to encourage research in this field to improve the clinical viability of MPH treatment.With the basic design operational, we are now investigating improvements to functionality.This includes adding temperature feedback control to allow for better control of time at temperature and improve the in vivo sample chamber to better maintain mouse body temperature while anesthetized.

Figure 2 .
Figure 2. Three-dimensional computer-aided design (CAD) models of the magnet and RF assembly of the HYPER system.(a) Isometric view of the magnet assembly; (b) top view (normal to y-axis) of the HYPER without exterior paneling.This view shows the permanent magnets along the x-axis, the RF coil assembly, the tuning circuitry, and the sample platform.(c) Front view (normal to z-axis) of the HYPER without exterior paneling.From this angle, the y-axis stage is visible.

Figure 3 .
Figure 3. Heating sequence programming using the graphic user interface (GUI).(a) the GUI enables the user to execute a spatially-confined heating plan entered as command lines.The user first selects the ROIs in 3D space on the sample using the bi-planar camera views: AP and LAT, and the program generates a preset plan in the configuration editor; (b) Flow chart outlining the processes that occur during a measurement with the HYPER.

Figure 4 .
Figure 4. Assembled hyper prototype.(a) In addition to the magnet and AMF assembly, the system includes a rack containing control systems, an RF amplifier, a water chiller, and animal heater.The chiller circulates water through the RF coil, with temperature that can be varied between 20 � C and 40 � C. The animal heater circulates water through the sample stage during in vivo experiments to maintain the core body temperature of the mouse models.(b) Diagram of the sample stage orientation within the HYPER, denoted by the dashed box.

Figure 5 .
Figure 5. Thermal PSF at five gradient strengths along all three axes.A nanoparticle sample was heated at varying positions along the x, y, and z-axis relative to the FFR, located at the isocenter.Thermal PSF at varying gradients along (a) x-axis; (b) y-axis; (c) z-axis.(d) Upon fitting the thermal PSFs to the derivative of the Langevin function, we calculated the FWHM in each dimension for the tested particles.

Figure 6 .
Figure 6.Verification of spatially confined heating with mouse phantom.Regions of interest ROI 0 (red) and ROI 1 (yellow) were selected in the HYPER GUI from both the top (a) and side (b) views.(c) Sequential heating of ROI 0 and ROI 1 is demonstrated.Despite both ROIs existing within the RF coil, the gradient field limits observed heating to the selected ROI.