MRI-compatible electromagnetic servomotor for image-guided medical robotics

The study addresses the challenge of enabling precise robotic actuation within the strong magnetic field of clinical MRI systems. Conventional electromagnetic servomotors contain magnetic and ferromagnetic materials that become hazardous near MRI scanners and interfere with imaging, limiting the development of MRI-compatible robots. Existing alternatives—pneumatic and piezoelectric actuators—suffer from limited accuracy, control challenges, and electromagnetic interference that can degrade image SNR. Prior efforts have leveraged MRI fields for propulsion or actuation of small devices but have not provided traditional servomotor functionality during standard imaging. The purpose of this work is to develop and demonstrate a non-magnetic electromagnetic servomotor that uses the MRI’s superconducting main field for torque generation, supports closed-loop control, and operates concurrently with MRI without degrading image quality, thereby enabling more capable image-guided medical robotics.

The authors review the clinical motivation for intraoperative MRI across neurosurgery, orthopedics, biopsies, and cancer therapies, noting limitations of preoperative images due to intraoperative tissue shift. Closed-bore high-field MRI improves image quality but restricts access, motivating MRI-compatible robotic systems. Traditional electromagnetic motors are incompatible in MRI due to ferromagnetic components and projectile risk. Pneumatic actuators, often with long transmission lines, can exhibit oscillation and overshoot and limited precision. Piezoelectric actuators can introduce electromagnetic noise that reduces MRI SNR by 26–80%, with special controllers needed to reduce degradation below 15%, and achieving smooth closed-loop proportional control is nontrivial. Previous research has exploited MRI gradient or fringe fields for propulsion, cell guidance, or tetherless actuation, but these approaches typically require modified imaging sequences or do not deliver conventional rotary servomotor performance during standard imaging protocols. Thus, a gap remains for a hybrid actuator enabling traditional servomotor functionality that can operate during routine MRI without custom sequences.

Design: A DC motor concept is introduced that replaces permanent magnets and ferromagnetic laminations with non-magnetic components and uses the MRI’s static B0 field (1.5–3 T) for torque generation. The torque on a rotor loop is T = (M × B)·ŝ, with M = nIA; for B = B0 ẑ, the shaft torque follows Tz = (Mx cosθ – My sinθ) B0 sinθ, predicting maximum torque when the motor shaft is perpendicular to B0 (θ = 90°) and zero when parallel (θ = 0°). Servomotor construction: The prototype uses a 2 mm G-10/FR4 non-conductive axle; mechanical commutator and brushes from a toy DC motor; a 3D-printed rotor winding support; three 100-turn coils (30 AWG magnet wire, ~20 mm² loop area, ~1.2 Ω per loop) secured with cyanoacrylate; polycarbonate housing with ABS end rings and Olite bushings; graphite lubrication. An optical encoder comprises a 3D-printed ABS encoder wheel (three leaves, 5 mm thick, 35 mm diameter) and two transmissive optical sensors (Vishay TCST2103) providing quadrature states to subdivide each revolution into 12 increments. Control and electronics: An Arduino Uno implements a PI position controller reading encoder signals; an H-bridge motor driver (2 A) provides PWM-based bidirectional control. Power is supplied by a 7.4 V LiPo battery. Signals run over a double-shielded Cat7 cable (one twisted pair for motor power, one for encoder diode power, two for sensor returns). The servomotor is enclosed in a copper Faraday shield with a 2 mm axle penetration; controller electronics are in a grounded, shielded aluminum box. EMI mitigation: To minimize RF emissions near the proton Larmor frequency (123.23 MHz at 2.89 T), the design includes continuous copper shielding for the motor, shielded cable with shield bonded to motor and controller enclosures, a low-conductivity composite axle to reduce antenna effects, and six floating shield current suppression traps per cable end (spaced 15 cm, tuned to 123.23 MHz; two lengths: 38 mm with ~7.4 dB attenuation and 57 mm with ~11.3 dB attenuation). Performance measurements: Conducted at the patient area of a 3 T Siemens Prisma Fit MRI (B0 ≈ 2.89 T at isocenter). Stall torque and no-load speed measured with motor powered by the 7.4 V LiPo at θ = 90°. Back EMF constant ke (equal to torque constant kt) measured by mechanically coupling an MRI-safe drive motor to spin the servomotor as a generator; EMF and angular speed measured to compute ke = E/ωm. ke characterized versus polar angle θ at isocenter and versus axial distance from isocenter (θ = 90°). Each condition measured nine times across ~300–600 rpm. Imaging/SNR tests: Interactions with MRI assessed by SNR of 2D GRE coronal images using a 2-channel Tx/Rx body coil. GRE parameters: TE/TR = 3.58/200 ms, flip = 60°, FOV = 22 cm, resolution = 1.72 × 1.72 × 5 mm, bandwidth = 260 Hz/pixel, 1 average. Scenarios at motor-phantom separations d = 15, 30, 45 cm: motor off (in bore), motor on (DC supply), motor on (PWM 50% duty). Control (no motor in room) also acquired. SNR maps reconstructed with noise-covariance-weighted SOS; mean SNR over phantom computed for 12 repeats per scenario. Robot construction and experiments: A one-DOF linear biopsy introducer robot was built from non-magnetic components. A modified plastic Vernier caliper provides a linear stage (0.1 mm scale for calibration). A 120:1 plastic gearmotor, with ferromagnetic axles replaced by 316 stainless, is driven by the servomotor via a 3D-printed gearbox holder. A 15-tooth, 15 mm plastic pinion drives a matching rack. Max linear speed 10 mm/s, max insertion force 585 N, travel 10 cm. A sheath holder secures the introducer sheath during cutting stylet removal. For visualization, a mock introducer with an MR-visible tip marker was used. Ex vivo procedure: Porcine loin with an embedded pitted olive as target. Pretreatment 3D VIBE (TE/TR 2.46/7.04 ms, flip 10°, FOV 25.6 cm, 1×1×1 mm, BW 890 Hz/pixel, 3 averages) used to select target coordinates. Controller converted target depth to encoder steps. Real-time single-slice imaging at 5 frames/s monitored motion: TRUFI for phantom tracking (TE/TR 1.94/3.87 ms, flip 45°, matrix 256×62, 1.17×1.25×5 mm, BW 1149 Hz/pixel, PF 5/8, GRAPPA R=2 with 22 refs) and FLASH for ex vivo insertion (TE/TR 2.24/4.9 ms, flip 8°, matrix 256×58, 1.17×1.46×5 mm, BW 1150 Hz/pixel, PF 6/8, GRAPPA R=2 with 24 refs). Post-placement 3D VIBE verified introducer sheath location using a plastic obturator to improve tip visibility. All data reported as mean ± SD; source data available on Figshare (doi:10.6084/m9.figshare.18812555).

• Feasibility: A non-magnetic electromagnetic servomotor leveraging the MRI’s B0 field achieved closed-loop rotary actuation inside a 3 T clinical scanner during imaging.
• Closed-loop control: The encoder and PI controller provided position control robust to external perturbations, maintaining the commanded setpoint after disturbances.
• Orientation and field dependence: The back EMF constant ke (and thus torque constant kt) followed the predicted sin(θ) dependence, maximizing at θ = 90° (axle perpendicular to B0). At θ = 90°, ke remained constant within 60 cm of isocenter and dropped to about 52% at 1 m from isocenter, reflecting the fringe-field profile.
• MRI image quality: With comprehensive EMI mitigation (shielded motor and controller, shielded cabling, composite axle, and tuned cable traps), simultaneous motor operation resulted in SNR changes ≤1.5% compared with control across distances d = 15–45 cm and power/control modes (DC, PWM 50% duty).
• Robot demonstration: A one-DOF MRI-compatible robot driven by the servomotor achieved controlled linear motion visualized at 5 frames/s. The system provided up to 10 mm/s linear speed, 585 N maximum insertion force, and 10 cm travel, and successfully placed a 9-gauge biopsy introducer sheath to a predefined target in ex vivo tissue during continuous MRI, with post-procedure imaging confirming correct placement.

The results demonstrate that conventional electromagnetic actuation principles can be adapted for safe, precise operation within the MRI environment by eliminating magnetic materials and exploiting the scanner’s static field for torque generation. Compared with pneumatic and piezoelectric alternatives, the approach enables standard motor control techniques, potentially improving precision, controllability, and integration with established robotics platforms. EMI mitigation strategies effectively preserved MRI image quality during simultaneous actuation, addressing a key barrier to concurrent imaging and robotics. The servomotor’s performance depends on orientation relative to B0 and position within the bore (sin(θ) behavior and field uniformity). Nonetheless, closed-loop control accommodated these variations, and system sizing can ensure adequate torque across required orientations. Future designs using electrical commutation (e.g., brushless with slip rings) could remove constraints on commutator alignment around the axle, improving flexibility for multi-DOF systems. Leveraging B0 allows potential miniaturization compared with permanent magnet motors, as the in-air flux density in clinical scanners exceeds that of typical rare-earth magnets, which is advantageous in confined MRI bores. Clinically, the capability for concurrent imaging and robotic actuation could reduce targeting errors due to needle bending and tissue deformation by enabling real-time feedback and correction, potentially improving safety and efficacy in procedures such as biopsies and neurosurgical interventions. Beyond surgery, the servomotor can support motion requirements in MRI applications like elastography, MRI-compatible ultrasound positioning, and animated phantoms.

This work introduces and validates an MRI-compatible electromagnetic servomotor built from non-magnetic components that uses the MRI’s main magnetic field for torque generation. The servomotor operates with closed-loop control during concurrent MRI, with negligible impact on image SNR. A proof-of-concept biopsy introducer robot driven by this motor achieved high-force, controlled insertions to a target under real-time imaging. These advances open a path to more capable MRI-embedded robotic systems using familiar electromagnetic motor control paradigms. Future directions include: implementing brushless electrical commutation to remove orientation constraints; optimizing motor geometry for miniaturization within MRI bores; extending to multi-DOF robotic platforms; integrating higher-resolution encoders for finer positional control; and exploring broader clinical and research applications requiring motion in MRI.

• Orientation dependence: Motor torque constant varies with the polar angle relative to B0 (sin(θ) behavior), with zero torque when the axle is parallel to B0, imposing constraints on mechanism layout.
• Spatial dependence: Performance decreases outside the homogeneous field region; ke dropped to ~52% at 1 m from isocenter in the tested scanner, implying site- and scanner-specific performance profiles.
• Mechanical commutation: The prototype relies on brush/commutator alignment; rotational alignment around the axle must be maintained for optimal performance, which can constrain robot design until electrical commutation is implemented.
• EMI mitigation complexity: Effective simultaneous operation required comprehensive shielding and tuned cable traps, adding integration complexity.
• Prototype-level encoder resolution: The encoder provides 12 increments per revolution, limiting angular resolution in this prototype; higher-resolution sensing may be needed for demanding applications.