Wearable bio-adhesive metal detector array (BioMDA) for spinal implants

Spinal instrumentation is widely used to treat deformities, injuries, and degenerative spine conditions, but postoperative hardware failures remain a significant concern. Current evaluation relies on radiographic imaging (e.g., CT) when symptoms arise, which is costly and exposes patients to ionizing radiation. Some invasive, self-powered sensing approaches have been explored for long-term monitoring but raise risks of infection and inflammation. Wearable technologies offer a promising non-invasive alternative; however, existing wearable imaging modalities (ultrasound or electromagnetic-based) face issues such as interface interference, lack of decoupling models for position inference, and potential heating of implants due to eddy currents at high frequencies. This study proposes a non-invasive, radiation-free, wearable bio-adhesive metal detector array (BioMDA) that uses magnetostatic inductive coupling and an engineered bio-adhesive interface with decoupling models to achieve real-time 3D localization of spinal implants, enabling early detection of hardware failure and monitoring of arthrodesis progression.

The paper contextualizes the need for non-invasive postoperative monitoring by noting limitations of CT and other radiographic methods (cost and radiation). Prior invasive sensor systems for monitoring fusion have been proposed but have limited adoption due to infection and inflammatory risks. Emerging wearable imaging (ultrasound, electromagnetic imaging) can image internal structures in real time but suffer from interference at skin/bone/implant interfaces and lack robust signal-to-position decoupling models. High-frequency electromagnetic methods may also cause eddy-current-induced heating in metal implants. The authors leverage magnetostatic interactions to avoid heating and draw on advances in flexible/wearable electronics and bio-adhesives to achieve stable device-skin coupling. They also recognize shifts in implant materials (from stainless steel to titanium), motivating potential surface magnetic coatings (e.g., nickel) to enhance magnetization for broader applicability.

System architecture: BioMDA comprises a 4×4 array (16 units) of inductive sensing modules connected via ultrathin Cu/PI traces (18/12 μm) on a PDMS elastomer substrate (~200 μm), with a PI isolation layer (5 μm) and PDMS encapsulation. A bio-adhesive hydrogel layer (~500 μm) provides robust, biocompatible bonding between device and skin as well as between hydrogel and silicone encapsulation. Sensing unit design and optimization: Each unit includes a copper toroidal coil (18 mm OD, 2 mm ID, 1 mm thickness), a PDMS support ring (12 mm ID, 18 mm OD), a permanent magnet (8 mm diameter, 1.5 mm thickness), and a PET holding film. The magnet motion under external electromagnetic attraction from metal implants induces emf in the coil. The PET film’s central angle and thickness modulate magnet kinematics and sensor stability/sensitivity. Finite element mechanical simulations (ABAQUS) examined strain and deformation for PET thicknesses (50–250 μm) under ~100 mN loading. Experimental benches measured interaction forces versus implant distance; nickel coatings (100–400 nm) on stainless-steel CPS were sputtered to enhance magnetic response and extend detection distance. Interface hydrogel engineering: The adhesive hydrogel consists of a PAA network crosslinked with GelMA and a gelatin network; PAA-NHS ester enables covalent bonding to amine groups on skin, while APTES functionalization introduces amino groups on PDMS to form covalent bonds with the hydrogel. Mechanical properties (elastic modulus 13–46 kPa) and high resilience (>8× stretch) were characterized. Peel (180°) and lap-shear tests quantified interfacial toughness and shear strength for skin–skin, skin–PDMS, and PDMS–PDMS interfaces. Device fabrication: Laser patterning of Cu/PI traces (ProtoLaser U4), plasma activation and bonding to PDMS, partial encapsulation, and integration of 16 sensing units. A PI isolation layer and PDMS top encapsulation were added; the hydrogel adhesive is modular and replaceable. APTES treatment (after oxygen plasma activation) was used to graft amines on PDMS prior to bonding with hydrogel. Electrical and mechanical characterization: Signals were acquired via a multichannel DAQ (6510, Keithley) at 1000 Hz. Tests included orientation robustness (−60° to 60°), electromagnetic interference tolerance (permanent magnets, metal board, WiFi router at ≥5 cm), cyclic durability (≥4000 approach–separate cycles), temperature tracking over 1000 cycles, and dynamic motion tests with a CPS at constant distance (~10 mm) across the array to assess crosstalk and uniformity. Decoupling models: A two-step model was developed. (1) Vertical distance decoupling: An electromagnetic model (magnetic field distribution, Faraday’s law for emf) coupled with a phenomenological attenuation through tissue and a kinematic model of magnet motion was calibrated with six factors, allowing distance inference from emf amplitude and velocity. Calibration used controlled implant motions from 15 mm to ~2–6.5 mm distance at 7.5–15 mm/s, repeated 20× per setting; regression determined scaling functions in distance and velocity spaces. (2) Horizontal mapping: A received-signal-strength (RSS)-based approach with radial basis function kernels uses multichannel amplitudes across the 4×4 array to map 2D implant position. In-vitro evaluation: On a cervical spine prosthesis with PDMS artificial skin (6 mm) separating BioMDA and implants, scenarios of screw loosening (pulled out 4 mm; reduced vertical distance to ~2 mm) and screw fracture with 5 mm misalignment were created. Four CPS (4.5 mm diameter, 45 mm length) with two rods (5 mm diameter, 120 mm length) were used. Signals underwent horizontal mapping and vertical decoupling to quantify positional differences between normal and failed states. Simulations: Electromagnetic simulations (Ansys Maxwell) modeled magnet (NdFeB) and stainless-steel CPS (relative permeability 4000). Mechanical simulations (ABAQUS) modeled PDMS (Mooney–Rivlin) and PET film behavior. Ethics approval covered human participant testing of the wearable interface; informed consent obtained. Comparative interface tests: Bio-adhesive hydrogel vs commercial double-sided acrylic adhesive (3M 100MP) during repeated bending of porcine tissue with embedded CPS assessed SNR, uniformity, and stability under perspiration-like conditions.

• BioMDA architecture: A flexible, ultralight (~40.3 g) 4×4 sensor array with a tissue-adhesive hydrogel achieves intimate skin coupling, enabling non-contact, zero-power magnetostatic inductive sensing of spinal implants during neck motion.
• Sensitivity and stability optimization: PET film thickness and central angle strongly affect performance; 150 μm thickness and 60° central angle balanced detection range and signal stability, covering typical skin-to-epidural distances (~5–20 mm).
• Force-distance characterization: Interaction force between CPS and magnet decays quadratically with distance. Nickel coating on CPS (up to 400 nm) increased attractive force by >100% in the 8–12 mm range vs uncoated stainless steel, extending detection capability for lower-magnetization implants.
• Signal robustness: Stable operation over >4000 approach–separation cycles without waveform or amplitude drift; temperature fluctuations <2 °C over 1000 cycles. Minimal crosstalk and good node uniformity in dynamic tests; orientation insensitivity from −60° to 60°. The device tolerated common EM interference sources at ≥5 cm separation; a metal plane at <2 cm caused distortion, implying a practical standoff recommendation.
• Adhesive hydrogel performance: Strong, biocompatible, covalent bonding to skin and amino-grafted PDMS. Interfacial toughness exceeded 1000 J/m² (skin–skin 180° peel). Shear strength: 36 kPa (skin–skin) and 12.5 kPa (skin–PDMS). Adhesion was robust across cycling (50 tensile cycles at 20% strain; ≥5 peel-offs) and over time (≥24 h at room temperature; ≥3 weeks at −20 °C). Hydrogel mitigated sweat-induced failures and skin irritation versus commercial adhesive.
• Electrical improvements with hydrogel interface: Compared to commercial double-sided adhesive, the hydrogel yielded larger, more uniform amplitudes, suppressed channel-to-channel variation, and maintained SNR after 100 bending cycles; commercial adhesive showed severe amplitude/SNR degradation due to interface gaps.
• Decoupling model accuracy: The electromagnetic–kinematic distance decoder with six calibration factors reproduced experimental emf amplitudes with an error of 1.5 ± 1.3 μV over 1525 samples. Tissue-induced magnetic attenuation was negligible for the intended body sites, supporting generalizability across body types.
• 3D localization and failure detection: RSS-based horizontal mapping plus vertical distance decoupling identified failure modes. For screw loosening (top-left CPS), decoupled distances shifted from 5.604–6.146 mm (normal) to 1.934–2.237 mm (loosened). For screw fracture with 5 mm misalignment, horizontal mapping identified the affected unit with reduced amplitude; decoupled ranges distinguished normal (5.507–6.037 mm) from fractured (5.901–6.481 mm) states. Overall positional precision reported at <0.5 mm (abstract).

The BioMDA system directly addresses the unmet need for continuous, non-invasive, radiation-free monitoring of spinal implants by converting multichannel inductive responses into precise 3D positional information. The flexible device design and robust hydrogel interface stabilize the sensor–skin coupling, reducing motion artifacts and improving SNR compared with commercial adhesives. The magnetostatic approach avoids eddy-current heating risks associated with high-frequency methods. The calibrated electromagnetic–kinematic distance model, together with RSS-based horizontal mapping, enables accurate detection of clinically relevant failure modes (e.g., screw loosening, screw/rod fracture/migration), thereby allowing early interventions that could mitigate complications and improve outcomes. Bench, simulation, and phantom experiments demonstrate durability, interference tolerance, and accuracy across clinically relevant distances and motions. The system’s modular calibration (six parameters) facilitates scalable manufacturing and straightforward recalibration, promising consistency in real-world deployment. Collectively, these findings indicate that BioMDA could support real-time, at-home postoperative monitoring, potentially reducing imaging burden and enabling proactive patient management.

BioMDA integrates a flexible magnetostatic sensing array, a biocompatible bio-adhesive interface, and calibrated decoupling models to achieve precise, real-time, non-invasive localization of spinal implants without radiation. The system reliably detects hardware positional changes associated with common failure modes and maintains robust performance under repeated motion, varying orientations, and everyday electromagnetic environments. Beyond spinal fusion, the approach is extensible to other orthopedic implants with appropriate sensor geometry optimization. Future work will target enhanced depth sensitivity (e.g., lumbar applications), broadened material compatibility (e.g., titanium via magnetic coatings or composite caps), and AI-driven decoupling models to improve accuracy across diverse implant types and anatomies. The demonstrated feasibility and robustness suggest a translational pathway toward decentralized postoperative monitoring and improved outcomes.

• Material sensitivity: Performance is strongest with ferromagnetic implants (e.g., stainless steel). Many modern implants are titanium, necessitating surface magnetic coatings or composite caps to enhance magnetization for effective sensing.
• Depth of detection: Current detection range targets cervical applications with typical skin-to-epidural distances; deeper implants (e.g., lumbar spine, higher BMI) may challenge sensitivity and precision.
• Environmental interference proximity: Although tolerant to common sources at ≥5 cm, very close proximity (<2 cm) to large metal planes can distort signals, requiring user guidance.
• Study scope: Primary validation was benchtop/simulation/phantom; comprehensive in vivo clinical validation remains to be performed.
• Motion dependence: Optimal signal capture uses prescribed movements; user compliance and motion variability may affect localization accuracy, though model calibration and qualitative velocity inputs can mitigate this.