A wireless implantable sensory ring for continuous airway stent migration tracking

Abstract

Airway stents play a vital role in managing central airway obstruction (CAO) caused by lung cancer and other pulmonary diseases by providing structural support to collapsed airways and restoring airflow. However, complications such as stent migration often require urgent medical intervention while early monitoring is essential to reduce the risk. Regular monitoring through bronchoscopy requires anesthesia in the hospital, which causes pain and an economic burden on patients. Computed tomography involves risky radiation and lacks the ability to provide continuous, real-time feedback outside of hospital settings. Here we report a fundamental mechanism of wireless tracking based on magnetic field in a wirelessly powered sensory ring integrated on an airway stent. The sensory ring is designed for continuous, real-time monitoring of stent position and orientation. This sensory ring, integrating an on-board magnetic sensor, and a wearable magnetic field generation system, enable accurate localization by detecting the magnetic field generated externally. The sensory ring is powered wirelessly via a charging coil, ensuring long-term operation. Our system achieves tracking accuracy of 0.5 mm and 2.2 degrees, with a temporal resolution of 0.2 Hz. Beyond migration monitoring, the sensor also detects airway deformation, offering the potential to sense pathological changes associated with lung cancer and other pulmonary conditions. By eliminating the need for radiation-based imaging or bronchoscopy, this approach enables safe, long-term surveillance of stent patency and surrounding tissue conditions. The proposed sensing mechanism and platform are also adaptable in other organs, such as the esophagus, for monitoring stent migration and deformation.

Introduction

Central airway obstruction (CAO) is a condition characterized by the narrowing or blockage of the trachea and mainstem bronchi, often caused by malignant tumors¹, inflammatory diseases², or trauma³. It leads to significant breathing difficulties, wheezing, coughing, and recurrent infections due to impaired airflow and mucus clearance. Left untreated, CAO can result in severe respiratory distress and diminished quality of life. Airway stents are hollow tubes to support the airway^4,5 which are critical in managing CAO as they provide structural support to maintain airway patency, alleviate symptoms, and improve airflow^6–8. They are particularly beneficial for patients with recurrent obstructions or those who are not candidates for surgery, offering a minimally invasive solution to restore respiratory function. Despite the wide utilization of airway stents, existing designs have common complications such as stent migration^9–11 which requires monitoring. Stent migration occurs when the stent shifts from its intended position due to improper sizing, insufficient anchoring, or external forces like coughing or breathing. Migration can lead to recurrent airway obstruction, ineffective symptom relief, and potential injury to the airway or surrounding tissues. To minimize this risk, accurate stent sizing, careful placement, and advanced designs like covered or self-expanding stents¹² are employed, but stent migration remains an issue that must be closely monitored. Early detection of airway stent migration allows swift intervention to prevent serious complications, optimize treatment outcomes, and minimize the risk of invasive procedures.

Existing methods for airway stent monitoring are through a combination of medical imaging and bronchoscopic techniques⁹. Flexible bronchoscopy is the gold standard for assisting stent delivery¹³ and directly visualizing stent patency¹⁴ in complications like mucus plugs or tissue overgrowth⁹ thanks to its steerability compared with rigid bronchoscopy. However, bronchoscopic methods for monitoring stent patency require frequent hospital visits and recurrent anesthesia, which carries both risk and economic burden to patients. In addition, Multidetector CT (MDCT)¹⁵ evaluation of airway stents is widely used because it provides high-resolution, 3D visualization of stents and surrounding airway anatomy. For example, symptoms like shortness of breath or increased mucus due to stent complications are evaluated by checking the stent position and conditions of obstructions¹⁶. MDCT provides excellent structural imaging but cannot directly assess stent function, such as mucociliary clearance, dynamic airway collapse, or subtle airway physiology¹⁷. A major drawback is the requirement for ionizing radiation, which can become significant with repeated scans—particularly concerning for younger patients or those requiring lifelong surveillance. Additional disadvantages include radiation exposure, imaging artifacts, high cost, and limited functional assessment.

Emerging technologies, such as sensor-integrated stents¹⁸ for sensing pressure^19–21, flow^22,23, and biofluid properties²⁴, offer real-time data for early detection of issues, but mostly focus on cardiovascular diseases^25–27 and related stent complications such as in-stent restenosis^21,27. They are not yet available for airway stent migration sensing due to the lack of a wireless sensing mechanism for stent migration tracking, the large distance from the central airway to the skin, as well as the short duration due to the limited battery life. Therefore, it remains challenging to monitor airway stent migration continuously with existing methods, especially out of the hospital.

To tackle this challenge, we propose a stent migration sensing mechanism coupled with a sensory ring integrated into an airway stent and a magnetic field generator to monitor for stent migration. The tracking is based on an on-board magnetic sensor, which can sense the magnetic field generated by a wearable magnetic field generator. The calibration model of the magnetic actuation module is established by modeling the magnets and performing the calibration. The translational accuracy is 0.51 mm. The rotational accuracy is 2.2 degrees. The sensing bandwidth is 0.2 Hz. We demonstrate the delivery of the sensory ring and its remote charging in phantoms and sheep models ex vivo. The sensory ring is fully remotely charged using induction coupling at a distance of more than 3.5 cm, enabling long-term operation. The sensing speed and accuracy are proved to be capable of being further improved with a more sensitive sensor and a rotating magnet with a higher speed. Our proposed sensory ring and wearable unit thus are promising for the next-generation smart airway stent for continuous monitoring of stent migration and other complications.

Results

Concept of monitoring airway stent migration by an implantable sensory ring

The concept of monitoring the airway stent migration by a sensory ring is illustrated in Fig. 1A and Movie S1. As shown in Fig. 1A, B, we propose a sensory ring that can detect its migration both translationally and rotationally. With a known external magnet on a wearable magnetic field generator (WMFG), the sensory ring, equipped with an onboard magnetic sensor, can pick up the magnetic field and then send it to a user device via Bluetooth Low Energy (BLE). The data are then analyzed with a model-based method to estimate the location and the orientation of the sensory ring for a potential stent migration. The sensory ring is fully powered by inductive coupling from an external transmitter (TX) coil and an implanted receiver (RX) coil, eliminating the need to integrate batteries. The sensory ring is composed of a flexible circuit and a soft back layer to withstand harsh mechanical squeezing during the stent deployment process, as shown in Fig. 1C, whose fabrication process can be found in Materials and Methods.

Fig. 1. Concept of monitoring airway stent migration using a sensory ring and a wearable magnetic field generator.

A Schematic illustration of a sensory ring bonded to an airway stent, a wearable magnetic field generator mounted on a human chest for monitoring stent migration within the human trachea. B More detailed illustration of the wearable magnetic field generator mounted on the human body, showing the external magnet as the magnetic field source, the remote charging unit, the sensor-integrated airway stent inside the trachea, the global coordinate, and the local coordinate of the system. C Optical images of the sensory ring before and after mechanical deformation. D Simulated magnetic field distribution on the plane z = 0 mm. Colored arrows indicate magnetic flux direction; the red dashed line marks $x$ = 47.5 mm. E Simulated magnetic field strength along the line $x$ = 47.5 mm, generated by two N42 magnets (25 mm × 25 mm ×25 mm) positioned 3.5 cm away. F Schematic illustration of a sensory ring bonded to an airway stent, a wearable magnetic field generator, and a mobile device for monitoring stent migration within the human trachea. Created with BioRender.com.

The sensing mechanism is briefly illustrated in Fig. 1D, E and Movie S1. A magnetic field model is developed to calculate the spatial magnetic field distribution, with which the magnetic sensor can be located based on its readout. For example, the magnetic fields on the x-y plane when z = 0 cm are modeled and plotted in Fig. 1D when the external magnet is at a location and orientation as specified in Fig. 1B. The magnetic fields on the line of x = 47.5 cm in the x-y plane are further plotted in Fig. 1E, showing that the three magnetic field components $B_x$, $B_y$, $B_z$ are varying when changing the y position of the sensor along the line. Intuitively, based on the change of the three magnetic field components, the location and orientation of the sensory ring can be estimated. As multiple locations can share the same magnetic field in space, it causes a computational singularity. Rotating the external magnet allows more sensed magnetic fields to be collected to avoid the singularity. Lastly, Fig. 1F shows that the stent migration information can be sent out to a mobile phone via Bluetooth signals.

Compared with passive tracking approaches that rely on magnetic markers used in capsule robots and related devices^28–31, the proposed method offers a significantly smaller footprint. In addition, unlike millimeter-scale magnetic beads, which provide relatively short sensing distances³² (typically <2 cm) and are therefore unsuitable for airway tracking, our method achieves a sensing distance of more than 3.5 cm, which can be further extended by incorporating more sensitive magnetic sensors. Table S1 provides a detailed comparison and summary of existing tracking methods.

Design and characterization of the flexible sensory ring

The components of the sensory ring include an nRF52 System-on-a-Chip as the computation and wireless communication unit, a remote charging circuit with an induction coil, and a magnetic sensor, as shown in Fig. 2A. The schematic of the circuit and the fabrication process are shown in Fig. S1 and Fig. S2, respectively. First, the magnetic sensor integrated on this circuit board is able to sense a magnetic field at a resolution of 0.1 mT. Other sensors could be further integrated for sensing other properties. Second, the long-term operation of the sensory ring is crucial to allow monitoring the airway stent migration. Using batteries can only support a short duration of the electronics. To address this issue, we propose to integrate a flexible receiving coil on the sensory ring, which could be inductively coupled with an external transmitting coil for remote charging (Fig. S3).

Fig. 2. Design and characterization of the sensor ring integrating a magnetic sensor, Bluetooth chip, and energy harvesting units.

A Schematic illustration and optical images of the system components and their interconnections. B Optical image of the electronic components assembled in the sensor ring with and without encapsulation. C Data flow diagram of the electronic system and illustration of the mobile user interface. D Top and side view optical images of the experimental setup, showing Bluetooth signal reception from a mobile phone during wireless energy transfer testing. The sensory ring integrates a receiving coil and is tested with a transmitting coil placed 3.5 cm away. E Output voltage of the remote charging coil during wireless charging under a 24 V DC power supply, plotted as a function of charging distance. An illustration of the operational range and voltage output is also shown. F Output voltage of the remote charging unit over time during the wireless charging process under a 24 V DC power supply, with an illustration of operational time and voltage trend. G Time-varying magnetic field detected via Bluetooth communication for two different positions and orientations of the sensor ring, while the magnet in the wearable magnetic field generator (actuation) remains fixed. Illustrations indicate the locations of the magnetic field source and sensors.

Figure 2B shows the optical image of the fully integrated sensory ring embedded into a back layer made of Thermoplastic Polyurethane (TPU). The electronic components are further encapsulated using polyimide tapes to ensure sealing (Fig. S4). We conduct soaking tests in phosphate-buffered saline (PBS) and a mucus surrogate, confirming that: (1) no visible delamination of the polyimide layer occurred, (2) the sensor exhibited no baseline drift, and (3) no leakage was detected based on dye penetration and impedance monitoring. These results indicate that polyimide sealing is sufficient for the short-term functional duration of airway stents. For longer-term applications with an operation time of weeks, additional encapsulation, such as Parylene-C or epoxy coatings, may be incorporated.

To further illustrate the system structure and data flow, in Fig. 2C, we show the flow chart of the three system units, including the wearable actuation and charging unit, the sensory ring, and the mobile phone with a user interface. The wearable magnetic field generator and charging system will be introduced in the next section, which integrates a motorized permanent magnet and induction coil with a resonance frequency of 110 kHz (Fig. S5). The user interface on the phone is used to receive the sensor readings and potentially perform the decoding of the information.

We systematically investigate whether the curved receiving (RX) coil and the remote charging unit could allow sufficient voltage for powering the sensor ring. As shown in Fig. 2D, a commercial transmitting (TX) unit with a resonance frequency of 110 kHz is placed above a phantom. A sensory ring is deployed inside a trachea phantom and placed on a holder. The sensed signal on the phone clearly shows that the system allows powering the sensory ring at a distance of about 3.5 cm, which is sufficient for a typical adult. In addition, Fig. 2E shows that the powering voltage drops when the distance between the TX and RX coils increases. Furthermore, Fig. 2F shows that the voltage of the RX circuit is more than 2.7 V during the charging process at a distance of 3.5 cm from the TX coil to the top surface of the sensory ring. With the electronics components and remote charging function, the device can send the magnetic field data in real time, as shown in Fig. 2G.

Lastly, as shown in Fig. S6, we quantify the device’s power consumption. The magnetic sensor consumes an average of 21 mW, while the BLE-SoC operates at approximately 48 mW. The device is designed for typical usage of three activations per day. These measurements indicate that the system operates at a low average power.

Characterization and calibration of the wearable magnetic field generator

The system components, function, and modeling of the WMFG are introduced in Fig. 3. Figure 3A shows the overall configuration of the wearable magnetic field generator when placed on a human chest phantom. The WMFG is integrated with two cube permanent magnets in addition to the remote charging unit. As shown in Fig. 3A, the WMFG provides time-varying magnetic fields using two cube magnets (25 mm by 25 mm by 25 mm, N42, NdFeB). This is controlled by an embedded controller (Fig. S7 and Fig. S8) by regulating the rotating speed and the location of the magnets using two miniaturized motors. The two permanent magnets on the WMFG are fixed in a fixture which can rotate at a maximum rotation speed of up to 120 rpm. As shown in Movie S2, their rotating angle is controlled by a step motor. A DC motor further controls the translation of the magnet to allow magnetic field magnitude control when needed. Subsequently, the magnetic field sensor on the sensor ring could record the magnetic field for tracking.

Fig. 3. Characterization of the wearable magnetic field generator (WMFG).

A Optical images of the wearable magnetic field generator mounted on a human chest phantom, including a zoomed-in view of the device. B Illustration of the cubic permanent magnets and their generated magnetic field distribution. C Spatial characterization of the magnetic field on a plane ($y$ = 0), as shown in the accompanying schematic on the right, along with the stepper motor and magnet. D Spatial characterization of the magnetic field on a plane ($z$ = 0), with the corresponding schematic indicating the relative positions of the stepper motor and magnet. E Comparison of predicted and measured magnetic field values over time at a selected location ($x$ = 40.1 mm, $y$ = 18.0 mm, $z$ = -34.5 mm) as indicated in the schematic. F Distribution of magnetic field prediction errors at sampled locations. Plane 1: $y$ = 0 mm, $z$ ∈ [–40, 40] mm, $x$ ∈ [60, 100] mm; Plane 2: $z$ = 0 mm, $x$∈ [60, 100] mm, $y$ ∈ [0, 40] mm.

A critical step is to calibrate a magnetic field prediction model such that the magnetic field can be used for comparing with the sensor ring readings. System calibration is performed when the magnet is rotating at different angles. At each rotating angle, the magnets generate a spatially distributed magnetic field. Figure 3B shows the modeled magnetic field distribution on a given plane by a discretized magnetic field model. The permanent magnet is discretized into $n$ = 1863 elements and approximated by magnetic dipoles at the location of the center of mass of each element. The position and orientation of the magnet, as well as its magnetic moment, are further calibrated using a magnetic field sensor at multiple locations. To build the calibrated model, Fig. 3C, D shows the magnetic field generated by the WMFG on two given planes measured by a magnetic field sensor. These measured data in a coordinate associated with the wearable magnetic field generator are used to correct the magnetic moment of the permanent magnets to minimize the prediction error.

To assess the accuracy of the calibrated magnetic field model, we compare the predicted magnetic field using the calibrated magnetic field model and the magnetic field measured by a magnetic field sensor at different time stamps and at different locations. Figure 3E shows the predicted and the experimentally measured time-varying magnetic field at a given location ($x$ = 40.1 mm, $y$ = 18.0 mm, $z$ = -34.5 mm) as marked by a yellow dot. The experimental data match the predicted data well in several cycles, with a Root-Mean-Square error of 1.57 mT. In addition, we also validate the magnetic field predicting model at points across space. Figure 3F shows the error distribution for the predicted and experimentally measured magnetic field at randomly selected positions in the workspace. A total of 86 locations are selected randomly across an area of 80 mm by 40 mm in the $x-z$ plane ($y$ = 0 mm) and an area of 40 mm by 40 mm in the plane ($z$ = 0 mm). Four different magnetic fields are generated from each location for experimental data, while the external magnet provides four different orientations for each location. The predicted spatially varying magnetic field matches the experimentally measured magnetic field with a Root-Mean-Square error of 1.64 mT, while the maximum normalized error is below 2%.

Compared with electromagnet actuation systems, the permanent-magnet–based magnetic field generation unit is more compact and energy-efficient than electromagnet-based systems, making it well suited for portable or wearable use (Fig. S9). Overall, the proposed sensory ring and magnetic field generation system enables at-home monitoring of stent migration with a compact stent design and a practical, low-power magnetic field generator.

Mechanism of tracking the sensor ring position and orientation

With the distributed magnetic field model, we introduce the tracking mechanism for stent migration sensing. First, Fig. 4A shows the relative coordinates of the external magnet, the sensory ring, and the global coordinate attached to the human chest. The magnetic field generator has a magnetic field ${\mathit{B}}_{\mathit{ak}}\left({\mathit{r}}_s\right)$ generated at a location ${\mathit{r}}_s$ at the $k$-th time stamp (k = 1, 2, ..., P), which is denoted in the global frame $\left\{{\mathit{O}}_g\right\}$. By discretizing the permanent magnet into $N$ small magnets, ${\mathit{B}}_{\mathit{ak}}$ could be precisely modeled using a magnetic dipole model.

$$\begin{array}{c}{\mathit{B}}_{\mathit{ak}}\left({\mathit{r}}_s\right)=\sum _{i=1}^N{\mathit{B}}_{\mathit{dipole}}({\mathit{r}}_s;{\mathbf{R}}_{\mathit{ai},k}{\mathit{M}}_{\mathit{ai}},{\mathit{r}}_{\mathit{ai},k})\\ =\sum _{i=1}^N\frac{{\mu }_0}{4\pi }[\frac{3({\mathit{r}}_{\mathit{s}}-{\mathit{r}}_{\mathit{ai},\mathit{k}})\left({\mathbf{R}}_{\mathit{ak}}{\mathit{M}}_{\mathit{ai}}\cdot \left({\mathit{r}}_{\mathit{s}}-{\mathit{r}}_{\mathit{ai},\mathit{k}}\right)\right)}{{\left|{\mathit{r}}_{\mathit{s}}-{\mathit{r}}_{\mathit{ai},\mathit{k}}\right|}^5}-\frac{{\mathbf{R}}_{\mathit{ak}}{\mathit{M}}_{\mathit{ai}}}{{\left|{\mathit{r}}_{\mathit{s}}-{\mathit{r}}_{\mathit{ai},\mathit{k}}\right|}^3}]\end{array}$$ 1

Fig. 4. Mechanism and characterization of tracking the sensor ring position and orientation.

A Coordinate definitions of the magnetic tracking system, showing the relative positions of the external magnet, sensor ring, and global frame. B Experimental setup for position and orientation tracking. (i) Overview of the setup, including WMFG, motion stage, and sensor ring. (ii) Close-up of the inertia measurement unit (IMU)-integrated sensor ring. C Sensor ring positions at two time points during magnet rotation. (i) Tracked positions. (ii) Optical images of corresponding locations. Time-varying magnetic field signals from the sensor ring at locations 1 (D) and 2 (E). F Video (Movie S3) frames and tracking accuracy during translational motion through three positions spaced 5 mm apart, with comparison to ground truth. G Video (Movie S3) frames and tracking accuracy during rotational motion with 45° intervals. (i) Optical images and estimated poses. (ii) Comparison of predicted and ground-truth orientations in Euler angles. H (i) Position tracking error across 50 random points ($x$: [68 78] mm, $y$: [–40 0] mm, $z$: [–12 –2] mm), with (ii) illustration of sampling distribution. I (i) Orientation tracking error across 30 random poses (roll: –2 to 4 rad, pitch: –0.5 to 0 rad, yaw: -4 to 4 rad), with (ii) illustration of sampled orientations.

The external magnet has a local coordinate system $\left\{{\mathit{O}}_{\mathit{mag}}\right\}$. ${\mathbf{R}}_{\mathrm{ai},\mathrm{k}}$ is the rotational matrix of the magnet, and ${\mathit{M}}_{\mathit{ai}}$ is the initial magnetic moment of the magnet. ${\mathit{r}}_{\mathit{ai},k}$ is the position of the $i$-th element of the magnet at the $k$-th time stamp. At the same time, the sensor has a local coordinate system $\left\{{\mathit{O}}_{\mathit{sensor}}\right\}$ defined by its position ${\mathit{r}}_s$ and rotation matrix ${\mathbf{R}}_s$. The sensed magnetic field at the $k$-th time stamp by the magnetic sensor is ${\mathit{B}}_{\mathit{sk}}$. The sensor location and orientation could be obtained by solving the following optimization problem, given by

$$\underset{{\mathbf{R}}_{\mathbf{s}},{\mathit{r}}_{\mathit{s}}}{\min }\sum _{k=1}^P\left|{{\mathbf{R}}_s^{-1}\mathit{B}}_{\mathit{ak}}\left({\mathit{r}}_{\mathit{s}}\right)-{\mathit{B}}_{\mathit{sk}}\right|$$ 2

Figure 4B shows the optical image of the magnetic field generator and the relative position and orientation of the external magnet and the on-board magnetic sensor. A motion stage carries the sensory ring to simulate the motion and orientation of the sensory ring. The translation and rotation are obtained from the motion stage and an inertia measurement unit.

Spatially varying locations of the sensor are estimated using the proposed method. Figure 4C shows the two locations of the sensory ring in the workspace and the relative displacement relative to the external magnet. When the magnet is rotating, the magnetic fields recorded by the sensor ring are different at the two locations. Figure 4D shows the time-varying magnetic field sensed by the sensory ring at $P_1$, while Fig. 4E shows time-varying magnetic fields sensed by the sensory ring at $P_2$. The different waveforms of the magnetic fields in the $x,\mathit{y},\mathit{z}$ axis are the key information to further recover the position and orientation of the sensory ring. By solving the optimization problem in Eq. (2), the position and orientation of the sensory ring are estimated.

We further evaluate the tracking accuracy using the proposed method. The experimental setup for calibrating the system and validating the tracking performance is shown in Fig. S10. The tracking and validation process follows the steps shown in Fig. S11. The ground truth data of the sensory ring location are obtained using a set of stereo cameras. Figure 4F to I show the results of tracking sensor position using the proposed method. First, Fig. 4F shows the tracking process when the sensory ring is only translating. A two-step translation motion (0.5 cm for each step) along the negative direction of the $z$ axis of the global coordinate is performed by an approximate uniform linear vertical translational movement of the motion stage. $P$ = 8 rotational angles of the external magnet are used in this optimization, leading to a relative position estimation error of less than 2%. Subsequently, the sensory ring is rotated, and its orientation in terms of Euler angles is estimated using the same method. A two-step clockwise rotation motion with +45 degrees for each step about the y axis of the global coordinate is performed by an approximate uniform linear rotation movement of the motion stage. The relative orientation estimation error is also less than 2% for the Euler angles, as shown in Fig. 4G. Finally, in Fig. 4H, I, we plot the error distribution for both the translation and rotation tests and illustration of sampling distribution, which quantitatively shows the relatively accurate tracking performance of the proposed method.

To further investigate the potential of increasing the tracking speed, accuracy, and reducing the size of the WMFG when needed, we demonstrate the use of a smaller external magnet paired with a high-resolution magnetic sensor. First, this magnetic field sensor (LIS3MDL, STMicroelectronics), with a 16-bit resolution and a sensing range of ±16 Gauss, is used in the same tracking framework shown in Fig. 4. Figure S12A, B present the experimental setup for validating the translation and rotation tracking using a smaller magnet (8 mm × 8 mm × 8 mm) actuated by a DC motor. The discretized magnetic field model in the plane $x$ = 40 mm is shown in Fig. S12C. Second, the tracking results for translational and rotational motion are presented in Fig. S13, showing a tracking accuracy of less than 0.5 mm in translation and less than 0.1 rad in orientation, as well as a tracking rate of 0.85 Hz. In addition, to further validate the magnetic field prediction and sensor tracking accuracy, additional data are collected from multiple motion trajectories of the sensor. Figure S14A shows an example of the magnetic field signal over time during translation along the $y$-axis, while Fig. S14B demonstrates the accuracy of magnetic field prediction. By optimizing the system with $P$ = 8 discrete rotational angles of the external magnet, we achieve tracking errors below 2% across randomly sampled points on a target plane, as shown in Fig. S14C.

To further demonstrate the effectiveness of the proposed method for potential applications in the human lung, Fig. 5A presents a representative scenario in which the sensory ring and airway stent are implanted inside a trachea phantom, mounted within a human chest phantom. The WMFG is secured to the chest using belts, maintaining a realistic distance from the sensory ring and tracheal structure. An experimental setup mimicking this scenario is shown in Fig. 5B, where the WMFG is positioned 3.5 cm above the trachea phantom, approximating the skin-to-airway distance, to evaluate the tracking performance of the sensory ring during both translational motion and deformation.

Fig. 5. Quantification of tracking accuracy inside a trachea phantom.

A Demonstration of the WMFG and sensory ring placed inside a trachea phantom mounted on a human chest phantom. (i) Front view. (ii) Back view. B Experimental setup for tracking the sensory stent at a distance of 3.5 cm. (i) Side view. (ii) Top view. C Video (Movie S3) frames showing translational motion of the sensory ring at three locations. D Estimated positions and orientations of the sensory ring during translation. E Comparison of (i) tracked and ground-truth positions, (ii) tracked and ground-truth orientations in Euler angles during translation. F Histogram of the relative position (i) and angle (ii) tracking errors for 15 trials in the translation tests. G Video (Movie S3) frames showing deformation of the sensory ring within the phantom. H Estimated positions and orientations during deformation. I Comparison of (i) tracked and ground-truth positions, (ii) tracked and ground-truth orientations in Euler angles during deformation. J Histogram of the relative position (i) and angle (ii) tracking errors for 15 trials in the deformation tests.

Online tracking of the sensory ring locations in a phantom is further demonstrated. As shown in Fig. 5C–F, an airway stent bonded to a sensory ring was manually translated along the trachea phantom. The subtle translational motion is captured in Fig. 5C using a camera as a comparison. A two-step movement is performed by gently pushing the sensory ring with a finger, resulting in measured translation distances of 7.19 mm and 4.16 mm, respectively. Figure 5D shows the detected motion, while Fig. 5E compares the tracked positions and orientations with ground-truth data obtained from the images. Figure 5F further quantifies the relative tracking error when deforming the sensory ring in 15 trials. The results confirm that the translational motion is accurately detected and quantified, and even the slight rotation induced during movement is captured by the system.

The sensor ring also allows sensing airway deformation using a phantom model. Figure 5G–J illustrate the system’s ability to detect deformation of the sensory ring, which can occur when airway tissue compresses the stent and sensory ring. Figure 5G presents a sequence of images capturing the deformation process under compression along the positive $\mathit{x}$ axis, resulting in translations along the $x$ axis. Figure 5H shows that the deformation is clearly detected by tracking changes in sensor positions. As shown in Fig. S15, using a single magnetic sensor as a tracker, sensory ring deformation can be distinguished from translation and rotation by leveraging the geometric constraints of the trachea. During pure deformation, the distance along the tracheal long axis changes minimally. In contrast, during translation or rotation, the distance from the sensor to the tracheal long axis remains nearly constant. Figure 5I compares the measured displacements and rotation with ground-truth data, showing that the system detects 2.6 and 1.1 mm displacements along the $x$- and $z$-axes, respectively, and a rotation of 0.305 radians, all closely matching the ground truth. Figure 5J further quantifies the relative tracking error when deforming the sensory ring in 15 trials. These results demonstrate the system’s strong potential for accurate deformation sensing.

In addition to the system shown in Fig. 5, we also demonstrate position and orientation tracking using a compact WMFG and a sensory ring integrated with a magnetic field sensor of a higher resolution with Least Significant Bit (LSB) = 0.98 Gauss. Specifically, we embed the high-resolution magnetic field sensor (LSB = 0.49 mGauss) into a flexible BLE SoC circuit (Fig. S16) and prepare a sensory ring to showcase the reduced size of the WMFG. As shown in Fig. S17 and Movie S4, the sensory ring is powered by the remote charging unit and enables real-time tracking of its position and orientation within a trachea phantom using the compact WMFG. For practical applications, the selection of sensor sensitivity, sensing range, and external WMFG design will depend on the requirements for magnetic actuation, sensing distance, and constraints on system size and weight, which will be explored in future work. In addition, decoupling stent migration from deformation can be further improved by integrating multiple magnetic sensors on the sensory ring. Although this increases device complexity, it enables more accurate separation of deformation from translation and rotation, as demonstrated in the two-sensor test in Fig. S18 and the three-sensor test in Fig. S19.

Demonstration of stent tracking in a sheep trachea ex vivo

To provide a proof-of-concept demonstration, we validate the delivery of the sensory ring and airway stent, followed by successful tracking of stent migration in both a phantom and an ex vivo sheep trachea. As shown in Fig. 6A, a customized delivery tool is developed by modifying a flexible delivery tool originally designed for self-expandable airway stents. The redesigned cone-shaped tip, with a maximum diameter of 18 mm and an effective length of 100 mm, accommodates the foldable sensory stent. Importantly, the sensory ring retains its functionality throughout the folding and unfolding process. Figure 6B illustrates the delivery procedure in a trachea phantom. The sensory ring, bonded to the airway stent, is deployed by advancing the stent through the delivery tool. First, the cone tip is inserted to the desired position within the trachea phantom. The stent is then gradually released by pushing the handler into the outer tube. As most of the stent is deployed, the tool is slowly withdrawn while continuing to push the handler. Once fully advanced, the stent is successfully delivered and fixed in place within the phantom, confirming effective deployment.

Fig. 6. Validation of the tracking performance of an airway stent in a phantom and ex vivo organ.

A Optical images of a customized delivery tool, a sensory ring integrating with a hybrid airway stent, as well as the squeezed sensory ring and stent inside the delivery tool. B Video (Movie S5) frames showing the delivery process for the sensory ring and the airway stent. C Optical images of the top view and side view of the experimental setup for testing the sensory ring and airway stent in a sheep lung with a chest phantom. D Bronchoscope video (Movie S5) frames of the delivered sensory ring and the airway stent. E Bronchoscope video frames (Movie S5) of the sensory ring and stent inside the trachea when being translated by external forces. F Tracked position during translation and deformation of the stent and the sensory ring. G Bronchoscope video (Movie S5) frames of the sensory ring and stent inside the trachea when being deformed by external forces. H Tracked rotational motion during translation and deformation of the stent and the sensory ring. I Biocompatibility test of the encapsulated sensory ring. (i) Before soaking in Phosphate-Buffered Saline (PBS). (ii) After submerged in PBS. (iii) Fluorescence images of the cell viability tests with the control and experimental groups. Scale bar, 100 μm. J X-ray images of the sensory ring and the stent inside the trachea (i) and a more detailed X-ray image of the sensory ring (ii). K Histogram of the relative error of $n$ = 9 trials of tracking stent translation. L Histogram of the relative error of $n$ = 8 trials of tracking stent rotation (rotation angles 7–12 degrees). M Validation of stent translation tracking under X-ray imaging. N Validation of stent rotation under X-ray imaging.

To further validate the delivery system and stent monitoring functionality, we integrate an ex vivo sheep lung into a chest phantom and mount the magnetic actuation and wireless charging system, as shown in Fig. 6C. Using a fixed bronchoscope, we successfully visualize the sensors on the sensory ring inside the sheep trachea (Fig. 6D), confirming compatibility with endoscopic imaging. We then demonstrate the system’s ability to sense and track stent migration by inducing translational and deformational movements of the sensory stent. As shown in Fig. 6E, G, bronchoscope images capture motion of the sensory ring inside the trachea, resulting in detectable translation and rotation of the embedded magnetic sensor. Specifically, around 20 s, a pushing motion produces a translation of 10.9 mm along the positive y-axis. At approximately 48 s, deformation caused additional movement: translation in the positive $x$ and negative $z$ axes, and rotations of 0.23 rad about the x axis, 0.05 rad about the $y$ axis, and 0.45 rad about the $\mathit{z}$ axis. The initial push also induces slight rotations across all axes. The tracked translational motion (Fig. 6F) closely aligns with the externally applied displacement, while the rotational data (Fig. 6H) further confirms deformation of the sensory ring during stent migration. In addition, we also validate the encapsulation of the device with a biocompatibility test for 72 h after being soaked in PBS. The results in Fig. 6I show that the device maintains a good sealing to allow biocompatibility after a 72-h test.

In addition to wireless signal-based monitoring, the position and condition of the sensory ring and airway stent can be assessed using X-ray imaging. For this demonstration, a sheep trachea containing the sensory stent is placed in an X-ray cabinet (Faxitron MX-20, Hologic Inc.) with a voltage of 30 kV and an Automatic Exposure Control period of 7.36 s. The resulting X-ray images clearly reveal the sensors and electronic components in Fig. 6J, even in the presence of the remote charging coils, offering an alternative method for device localization and condition assessment. To further validate the stent tracking accuracy inside the sheep lung, we used the X-ray images as comparison. Histogram of the relative error of $n$ = 9 trials of tracking stent translation is reported in Fig. 6K with relative error of less than 2.5%. Similarly, a histogram of the relative error of $n$ = 8 trials of tracking stent rotation is reported in Fig. 6L, demonstrating the accurate tracking with relative error less than 4%. These experiments are shown in Fig. 6M, N, where nine stent translational displacements (5–10 mm) are induced along the long axis of the stent and nine stent rotational displacements are induced about the long axis of the stent (7–12 degrees).

Discussion

In summary, we have presented a wirelessly powered sensory ring integrated with an airway stent and a magnetic field generator for continuous monitoring of stent migration. The system leverages an on-board magnetic sensor to detect magnetic fields generated by a wearable magnetic field generator. We established a calibration model through analytical modeling and experimental calibration of the magnetic field, achieving a translational sensing accuracy of 0.5 mm, a rotational accuracy of 2.2°, and a sensing bandwidth of 0.2 Hz. We have demonstrated the delivery of the sensory ring and its wireless charging in both phantom models and ex vivo sheep trachea. The sensory ring is fully charged remotely via inductive coupling at distances exceeding 3.5 cm, enabling long-term, untethered operation. Furthermore, the sensing speed and accuracy can be further enhanced by incorporating a more sensitive magnetic sensor and a faster rotating magnet. Overall, the proposed sensory ring and wearable magnetic field generator provide a promising platform for next-generation smart airway stents, offering real-time, continuous monitoring of stent migration and potentially other stent-related complications.

To further enhance the performance of the sensory ring for potential clinical applications, several aspects can be improved. First, the tracking speed can be significantly increased by using a higher-speed motor and reducing the magnet size, allowing for faster magnetic field generation. Second, the mechanical resilience and durability of the sensory ring can be improved by optimizing the electronic circuits, including tighter bonding at the electrode interfaces, and incorporating serpentine structures^33,34 to enhance stretchability and mechanical robustness. Third, the system will be further validated through in vivo testing in a large animal model to assess the delivery performance, sensing capability, and long-term biocompatibility of the sensory stent in addition to the in vitro biocompatibility test demonstrated in this work. The sensing functions for in vivo applications may also need to consider the motion artifact due to breathing and body motions. The motion artifact could be reduced by performing filtering the periodic and vibrational signals through effective algorithmic intervention³⁵. Lastly, we mainly focus on stent migration detection in this work and would like to explore integrating multiple magnetic sensors for stent deformation sensing for monitoring tumor regrowth after stent deployment.

Magnetic field–based tracking has been explored for wireless localization of medical devices such as catheters³⁶ and capsule endoscopes^29,31,37. Two primary approaches have been developed including tracking permanent magnets using external sensor arrays^{31,32,38–40}, and using on-board magnetic sensors with external electromagnetic field generators. However, both methods face significant limitations. In the first approach, tracking range is restricted due to the rapid decay of magnetic fields generated by small embedded permanent magnets, especially as the distance between the magnet and the sensor array increases^32,39,41. In the second approach, while on-board magnetic sensors offer direct localization, most existing implementations are tethered⁴², limiting their applicability for long-term, wireless operation. These sensors are typically embedded in catheters, endoscopes, or needle tips and are not designed for continuous wireless use. In contrast, our method integrating magnetic-based tracking with wireless charging capabilities in airway stents overcomes these limitations, which can potentially enable long-term, wireless monitoring of stent position and orientation, offering a promising solution for tracking stent patency without relying on tethered systems or limited-range permanent magnets as on-board tracker.

In future work, the wearable magnetic field generation system may be compared with electromagnetic field generators⁴³, which would enable further improvement of the tracking speed while the issue of heat generation needs to be considered. However, for applications focused on stent patency monitoring several times a day, the current sensing speed is already sufficient. Additionally, multiple magnetic sensors could be integrated on the sensory ring to allow sensing stent deformation more precisely. Moreover, the sensory airway stent can be expanded to integrate other types of sensors, such as flow and pressure sensors²⁰, to enable comprehensive physiological data collection for disease diagnosis. Furthermore, the sensory stent could be combined with a mucus-clearing mechanism, as demonstrated in our previous work^44,45, to provide on-demand and closed-loop mucus removal for patients with Chronic Obstructive Pulmonary Diseases, Cystic Fibrosis, and other lung diseases associated with excessive mucus accumulation. In summary, the proposed sensory ring, together with the wearable and remote charging system, holds great potential for at-home, long-term monitoring of airway conditions and stent performance.

Methods

Components of the flexible sensory ring

The sensory ring was prepared by wrapping a flexible sensory board using a multi-step procedure. The flexible sensory board has a total length of 31.95 mm and a width of 26.64 mm. The flexible board hosted a BLE System-on-a-Chip (MDBT42V-P512KV2, Raytac Corporation), a three-axis Hall-effect sensor (TLV493D-A1B6, Infineon Technologies AG), and a remote charging circuit. The remote charging unit includes a low VF Schottky diode array (BAS4002ARPPE6327HTSA1, Infineon Technologies), a 6 V Zener diode (MM5Z6V2T1G, Onsemi), a 3 V Zener diode (MM5Z3V3T1G, Onsemi), a low-noise, low-dropout regulator with shutdown (LP2985 150-mA, Texas Instruments), and a rechargeable battery (Coin, 6.8 mm 3 V Lithium Battery Rechargeable, 5.5 mAh). See Table S2 for the component details.

Prepare the flexible PCBs on the sensory ring

To prepare the flexible PCBs, the following procedures were used. Pyralux with Polyimide (PI) tapes as masks were prepared using a laser cutting machine (LPKF U4 protolaser, LPKF AG). A single-sided flexible copper sheet (Pyralux) was attached to a rigid substrate (1.5 mm in thickness) using PI tapes to fix its four edges. Another layer of PI tape was attached to the flexible copper sheet to cover the whole Pyralux sheet. The total thickness of the rigid substrate, the Pyralux sheet, and the PI tape was measured. The laser power and speed were tuned to cut the PI tape only. After the cutting, the PI tapes covering the unwanted areas were carefully removed. Subsequently, the Pyralux sheet with selectively covered areas was etched in a ferric chloride solution. Ferric chloride solution was prepared in a glass petri-dish (90 mm in diameter, 15 mm in depth) for etching. Pyralux sheets were placed into the ferric chloride and sank to the bottom of the container. The etching process took about 20–30 min. A plastic dropper was used to remove the black residues on the Pyralux surface. The Pyralux sheet was checked every 2 min to prevent the sheet from being over-etched or damaging the conductive traces. Next, the PI tapes covering the traces were peeled off after taken out from the etching solution. Finally, electronic components were soldered to the Pyralux sheets based on the design. The flexible board was bonded to a TPU tube as a back layer, which was printed by a 3D printer (Bambu Lab X1C 3D Printer, Bambu Lab, Inc.).

Wearable magnetic field generator and remote charging unit

The external magnetic field generation system was composed of two cube magnets, which were placed inside a 3D printed fixture (see Table S3 for the component details). The magnets were further mounted on a step motor such that the rotation angle of the magnets could be controlled. The translations of the two magnets were further controlled by a servo motor with a sliding-bar linkage mechanism. The two motors were driven by L298N mini motor drivers, which were controlled by an embedded controller (Arduino Nano Sense BLE 33). The controller was communicated via Bluetooth with a MATLAB program on a computer. A battery or a DC power supply was used to power the embedded controller and the motor drivers. To remotely power the sensory ring, a transmission coil with a transmitter board module was a commercial device (XKT-801-49, Taidacent Inc., Shenzhen). The transmission coil has an inner diameter of 60 mm, an outer diameter of 84 mm, and a thickness of 1.3 mm. The transmitter board module has a size of 50.14 mm by 16.35 mm. Its power is about 0.72 W when working.

Preparing the RX coil and charging unit for the sensory ring

The customized RX coil was a square planar coil made by a single-sided flexible copper sheet (Pyralux) using a laser cutting machine (LPKF U4 protolaser). The single-sided flexible copper sheet (Pyralux) was attached to a rigid substrate (1.5 mm in thickness) using PI tapes to fix its four edges. The laser power and speed were tuned to cut the surface of the single-sided flexible copper sheet (Pyralux). The RX coil has three layers, each with a length of 30 mm for the side length and 70 turns. For a maximum charging efficiency, the RX unit’s resonance frequency was designed to match that of the TX output signal. The capacitor connected to the charging board was used to tune the frequency of the RX unit to maximize the energy harvesting performance. The RX coil and the connected capacitor were considered as an RLC series circuit, of which the resonance frequency could be found out using a frequency sweeping method.

Test the remote charging performance

To test the remote charging performance, the TX unit, the RX unit, a DC power supply, and an oscilloscope were used. The TX unit was connected to the DC power supply. An Arduino was connected to the RX unit to measure the output voltage continuously. The TX coil and RX coil were placed concentric with a given distance. To test the maximum functional charging distance, the distance between the two coils was steadily increased until the output voltage of the RX unit was about 2.5 V. To test the charging duration at the maximum functional charging distance, the rechargeable battery was first discharged and then charged with the RX unit at a charging distance of 3.5 mm. The voltage of the rechargeable battery was measured continuously. The time for the rechargeable battery to reach 3 V was recorded. After charging, the time for the voltage of the rechargeable battery to drop to 2.5 V was recorded.

Power consumption measurement

We characterized the power consumption of the device using the Power Management IC Development Tools Devkit for wireless protocol and ICs (Nordic Semiconductor). System power consumption was measured step by step. First, with the magnetic sensor and LDO (Low Dropout) regulator turned off and the RX coil removed, we measured the current drawn by the nRF52832 BLE System-on-a-Chip (BLE-SoC) under a regulated 3.3 V supply. We then sequentially activated the magnetic sensor and LDO to quantify their contributions to the total power consumption. To evaluate the RX coil energy consumption, we measured the voltage and current at the RX coil circuit output, which includes the power delivered to the BLE-SoC and peripherals, as well as heat generation. The average power consumption of each component is summarized in Fig. S6. We measured an average operating power of 21 mW for the magnetic sensor and 48 mW for the BLE-SoC. These results confirm that the device operates at a low average power. Finally, it is important to note that the device is intended for typical daily use of approximately three activations per day.

Procedure of testing the biocompatibility of the sensory ring

Human fibroblast cells (CRL-2522) were obtained from the American Type Culture Collection (ATCC) and cultured in Minimum Essential Medium (MEM; Gibco) supplemented with 10% (v/v) fetal bovine serum (FBS; Gibco). Fibroblasts were seeded into 6-well plates at a density of 5 × 10⁵ cells per 500 µL of medium to facilitate initial attachment and spreading under standard incubation conditions (37 °C, 5% CO₂, humidified atmosphere). After 24 h, the encapsulated sensory ring was placed into the wells and co-cultured with the cells for an additional 72 h. Cell viability was assessed using the LIVE/DEAD Viability/Cytotoxicity Kit (Invitrogen, Thermo Fisher Scientific), and fluorescence signals were observed using a fluorescence microscope (Eclipse Ti-E, Nikon, Japan).

Calibration of the magnetic field generation system

A commercial magnetic field sensor board (TLE493D-A2B6 3D-MS2GO, Infineon Technologies AG) was used to measure the magnetic field at a given location for calibrating the magnetic field generation system. The positions of the measured points were measured using two cameras. The magnetic field was modeled by discretizing the magnet geometry into n = 1863 elements using the Geometry and Mesh functions. A discretized magnetic dipole model by discretizing two cube magnets into small elements was used to predict the magnetic field at a given location using customized codes.

Camera-based tracking for obtaining ground truth

Two cameras (Flir USB3 camera) were used to obtain the position and orientation of the sensory ring and the commercial hall effect sensor board to be used as the ground truth data. Stereo Camera Calibrator App in MATLAB 2024b (MathWorks Inc.) was used for calibrating the two cameras with a prepared checkerboard.

Prepare the phantom and the sheep organ

The phantom used to test the tracking accuracy of the sensory ring is based on a lung model obtained from NIH 3D (https://3d.nih.gov/entries/3DPX-021148). Subsequently, the model’s size and shape were adjusted for printing in AutoDesk MeshMixer (AutoDesk Inc.). With this processed model, the phantom was directly printed in Elastic 50 A (Formlabs Inc.) in a stereo photolithography (SLA) 3D printer (Form 3+, Formlabs). The sheep trachea was obtained from Vanderbilt University Medical Center and stored in a freezer before the experiments. During the experiments, the sheep trachea was unfrozen and placed under a chest phantom for testing the sensory ring delivery and sensing functions with the magnetic field generation and remote charging system.

Statistics

Standard deviations of at least n = 3 trials were indicated in the figures. The data were processed in MATLAB 2024b.

Author contributions

Conceptualization: X.D. Methodology: X.D., R.G., and Y.W. Investigation: R.G., Y.W., C.N., and H.F. Visualization: R.G., Y.W., and X.D. Supervision: X.D. Writing—original draft: X.D., R.G. Writing—review and editing: X.D., R.G., Y.W., C.D., F.M., V.S., and Y.Z. All authors have read and approved the manuscript. R.G. and Y.W. are equally contributed co-first authors.

Data availability

All data is contained within the manuscript and supplementary files.

Code availability

The C++ and MATLAB codes for tracking can be accessed via the link: https://github.com/dong-mrlab/stent_migration.

Competing interests

Vanderbilt University has filed a provisional patent application related to this work. The authors declare that they have no other competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41528-025-00526-0.