Design & Validation of a Temperature-Feedback Controlled Automated Magnetic Hyperthermia Therapy Device

Image showing three-piece induction coil set, heat stations, oscilloscope, magnetic field probe and water jacket.

Objective

Magnetic hyperthermia therapy (MHT) is a potent cancer therapy that employs heat generated by magnetic nanoparticles (MNPs) embedded within the target tissue when they are exposed to an alternating magnetic field (AMF), at low radiofrequency (RF), typically <300 kHz. To date, no clinical trial has been conducted in the United States with automated feedback temperature or thermal dose control for MHT.

MHT requires delivery of MNPs to the tumor and application of AMF to generate local hyperthermia (41 °C–46 °C) via magnetic hysteresis loss. For many MNPs, the heat generated, often expressed as specific loss power (SLP), exhibits a linear response with AMF frequency and non-linear with AMF amplitude. This implies that by controlling AMF amplitude at fixed frequency, power deposition and hence temperature within the tumor near the MNPs, can be controlled.

On the other hand, MHT delivery can be challenging because MNP distribution within the tumor is typically variable and heterogeneous, leading to unpredictable temperature variations within the tumor and at tumor margins. New approaches are needed to achieve target hyperthermic temperatures within the tumor microenvironment while minimizing locally under-treated and ablative tumor zones, under-treated margins, and inadvertent heating of healthy tissue.

Computational methods predict improved spatiotemporal control of treatment temperature with AMF amplitude and power modulation. Success with these approaches requires real-time temperature monitoring to provide the needed input into a temperature controller device. However, automated temperature control in clinical hyperthermia systems, particularly for MHT, remains underdeveloped.

Here, we document, verify, and validate the performance of an experimental PID-controlled automated MHT device that uses fiber-optic temperature data as input to manage the power delivered to a custom-designed 20 cm diameter RF coil connected to a 120 kW induction heating power supply. The intended use is treating canine glioblastoma in a future pilot study.

Design

(A) Schematic block diagram of the PID controller depicting the error signal (e(t)) input and control signal (uctrl) generation to minimize the error e(t) using the PID algorithm. (B) Regions in the s-plane, delineated by transient specifications on controller design, including rise time (ωn, red), overshoot (ζ, green), and settling time (σ, blue). The agarose gel + Cu wire system (inset schematic) constitutes a transfer function with poles in the shaded region to meet design specifications. The 3D temperature color graph indicates the temperature distribution in the sample after 10 min of AMF treatment at 123 Oe (peak), 160 kHz.

The design requirements for a temperature feedback controller for MHT include satisfying general hardware and software compatibility, safety requirements, and MHT treatment-specific performance requirements. For the present case, these were:

MHT treatment and controller performance criteria:
- Capability to maintain temperature at a target set point temperature (Tref) for 15–30 min at a single probe location.
- Capability to achieve hyperthermic temperatures (43 °C–45 °C), and maintain the temperature at setpoint (Tref) to attain thermal dose defined by the metric cumulative equivalent minutes at 43 °C (CEM43) of 60 ± 5 min within clinically relevant treatment times (15–30 min).
- Rise time (tr) to target temperature (Tref) < 60 s.
- Overshoot (Mp) = < 5%.
- Settling time (tss) within ±0.5 °C in <5 min
Safety criteria:
- A defined maximum temperature (e.g., 50 °C) within the treatment region to limit power to prevent runaway heating at the feedback sensor location.
- A safety temperature threshold at a distant location for additional safety monitoring, for example, a temperature representing body core temperature.
Hardware and software requirements:
- Integration with a multi-sensor temperature probe to provide temperature readout capability with appropriate sampling interval.
- Prevent power fluctuations that might arise from higher order harmonics generated by the RF power supply that can damage electronics in the 120 kW AMF

(A) 3D mesh model of the RF coil and gel + Cu wire experimental setup, used to simulate open-loop and closed-loop temperature vs. time responses. (B) Zoomed-in view of the 3D mesh model of the gel + Cu wire setup shown in (A). (C) Snapshot of the temperature distribution in the agarose gel + Cu wire system at 30 min, with a setpoint temperature, Tref , at the probe location of 25 °C. (D) Simulated temperature vs. time responses (black curve) for setup in (C), for input gains (Kp Ki, Kd) of (0.26, 0.001, 3). (E) Experimentally measured temperature vs. time response (black curve) in the agarose gel + Cu wire system, during a 30 min PID temperature-control experiment, where the setpoint temperature, Tref , was input as 25 °C. Final gains that resulted in a stable compromise between rise time, settling time, and minimizing power oscillations, were (0.23, 0.0001, 2.84). (F) Experimentally measured temperature vs. time response (black curve) in the “tuned” agarose gel + Cu wire system, during a 30 min PID temperature-control experiment, where the setpoint temperature, Tref , was input as 25 °C and gains of (0.23, 0.0001, 2.84) were used. Inset shows the initial overshoot <5%. Settling time (tss ) of 39.7 s and rise time (tr ) of 8.35 s were observed.

LabVIEW® controller code was developed to convert measured analog temperatures to digital signals, T(t), that were then used to compute error, e(t), as the difference between user-defined set-point temperature, Tref, and T(t), and then to calculate the new control signal, uctrl(t), using the digital LabVIEW® PID algorithm and user inputs of proportional, integral and derivative gains Kp Ki, Kd, respectively. The frequency of these operations was determined by the sampling interval used to digitize the temperature signal, FPGA clock rate (40 MHz), and frequency with which the power supply could respond to (0–5 V) analog signals without generating a fault.

Initial PID gains (Kp Ki, Kd) for closed-loop feedback control were derived from computer simulations of an agarose gel + Cu (wire) and these were verified experimentally. In vitro validation experiments for temperature controller response were conducted three times (N = 3). For in vivo validation, a healthy, adult male beagle (~10 kg) with a custom catheter for MNP delivery in the left frontal lobe of the brain was studied. All studies were performed under general anesthesia.

(A) Schematic depicting setup for in vivo validation of MHT controller in a canine research subject. Magnetic nanoparticles (MNPs) were infused into the canine brain through a custom catheter. One temperature probe was inserted through the catheter’s MNP port into the brain tissue (MHT, black), for feedback temperature control at that location. A second temperature probe was placed in the rectum to monitor core body temperature (R, orange), and a third probe was inserted subcutaneously in the canine head (S, green) to monitor temperature increase from eddy current heating during the treatment. (B) Cone-beam computed tomography (CT) image of the canine brain showing the catheter with infused MNPs localized within and around the catheter. (C) Finite element-based coupled electromagnetic and heat transfer simulations performed on the CT image-segmented brain and MNP heat sources showed that at 20 kA/m peak, 160 kHz, heating occurred mostly at the catheter where the MNPs were located. Temperature increase, ∆T, away from the MNP heat sources (>3 mm from the catheter walls) was negligible due to the heat sink effects of perfusion. (D) Temperature vs. time response at the brain tissue adjoining the MNP catheter tip, for a setpoint of 39 °C, showed that setpoint was achieved within 200 s, however small temperature fluctuations (<0.4 °C) were observed. Rectal temperature remained stable at ~37 °C throughout this test. (E) Temperature vs. time response at the brain tissue adjoining the catheter‘s MNP port-tip, for a hyperthermic setpoint of 45 °C, showed that the setpoint was not achieved in this 10 min test. The temperature at this location increased to ~44 °C and fluctuated between 43–44 °C. (F) Temperature vs. time response at the brain tissue adjoining the catheter’s MNP port-tip, after optimal tuning of the controller PID gains, showed that the controller achieved the setpoint of 45 °C in ~120 s and remained stable (∆T < 0.1 °C) throughout the remainder of the 15 min treatment. Final PID gains of (2, 1×10-4, 2.5) were used. CEM43 isoeffect thermal dose for the treatment was 56 min. Rectal temperature remained stable at ~38.5–39 °C during the entire treatment.

Application

Here, we document the successful design and validation of a PID-based, multi-sensor temperature feedback controller and its integration with a 20 cm diameter RF coil intended for MHT treatments of gliomas in large animals. We verified controller performance in vitro in agarose gel + Cu wire model and validated its performance by achieving CEM43 of 60 ± 5 min with steady temperature within a clinically relevant time (15–30 min) ex vivo and (15 min) in vivo MNPs as heat sources in a live research canine. Careful consideration of safety and performance criteria during device design and development enabled us to minimize operational risk. Specifically, our process placed emphasis on the design inputs, design outputs, and verification process as recommended within the FDA’s Design Control Guidance waterfall method to facilitate device evaluation.

Our results demonstrated in gel phantom experiments that the device can automatically adjust the AMF amplitude to maintain the temperature within the target range despite significant perturbations. However, for complex clinical scenarios, PID gains may require dynamic adjustment by the operator based on an initial pulse test. Future efforts should explore advanced control strategies such as model predictive control (MPC) for multi-input multioutput (MIMO) systems. Clinical translation of such control systems can improve patient safety and quality assurance.

*Published in Frontiers in Thermal Engineering, “Design of a Temperature-Feedback Controlled Automated Magnetic Hyperthermia Therapy Device."
*Published in Cancers, “Validation of a Temperature-Feedback Controlled Automated Magnetic Hyperthermia Therapy Device.”

Page updated

Google Sites

Report abuse