Dr. Michael C. Kohler

Surface Acoustic Wave (SAW) Force Myography Sensor for Comprehensive Joint Torque Evaluation

Accurate quantification of skeletal muscle forces and resulting joint torque is essential for preventing fatigue-related injuries, monitoring performance changes, and managing neuromuscular conditions such as Parkinson’s disease, multiple sclerosis, and muscular dystrophy. Conventional torque evaluation using electromechanical dynamometers provides high-precision isometric and isokinetic measurements and is regarded as the clinical and research gold standard. However, their large size, high cost, and need for controlled environments limit accessibility outside laboratory settings. Wearable systems such as electromyography (EMG), inertial measurement units (IMUs), and ultrasound offer real-time motion monitoring but face challenges including skin-electrode sensitivity, motion artifacts, and calibration drift. Force myography (FMG), which detects muscle deformation via volumetric expansion rather than bioelectric activity, mitigates several of these issues by offering signal stability and minimal interference from motion artifacts. Surface acoustic wave (SAW) sensors further extend these advantages by providing passive, wireless, and battery-free operation with high strain sensitivity.

This study investigates the feasibility of a SAW-based FMG (SAW-FMG) system for estimating elbow joint torque by directly comparing its performance against a reference electromechanical dynamometer (Biodex System 4 Pro) under isometric and isokinetic conditions. A custom one-port SAW delay-line sensor fabricated on 128° YX-cut lithium niobate (LiNbO₃) was integrated into an armband positioned over the biceps brachii. The device consisted of interdigitated transducers and reflectors forming a delay path sensitive to surface strain. Muscle contraction modulated the acoustic wave phase, which was measured in real time through a network analyzer. Consistent armband tension across participants was ensured using a reference strain gauge.

Two experiments were conducted. In the isometric protocol, participants performed maximum voluntary contractions at 15°, 30°, 45°, 60°, 75°, and 90° elbow flexion. Normalized phase and torque data were time-synchronized and modeled using a second-order polynomial capturing the relationship among torque, angle, and phase shift. In the isokinetic protocol, concentric flexion from 0° to 90° was executed at angular velocities of 10°/s and 20°/s, and the calibrated polynomial was applied to predict torque from phase and angle. Model accuracy was evaluated using normalized root-mean-square error (NRMSE) and Pearson correlation coefficients, and statistical comparisons employed repeated-measures ANOVA and Wilcoxon signed-rank tests.

Under isometric conditions, the subject-specific polynomial achieved a mean NRMSE of 13.6% and R² = 0.834, slightly outperforming the group-level model (NRMSE = 14.4%, R² = 0.808). Although differences were statistically significant, they were modest in magnitude. These results compare favorably to previously reported muscle-force modeling studies, considering variations

in methodologies and sensors. The SAW-FMG system exhibited a slightly longer time-to-peak torque than the dynamometer, attributed to indirect deformation sensing and acquisition latency, yet maintained strong temporal consistency. During isokinetic trials, the model yielded mean NRMSE values of 24–25% and correlation coefficients of 0.65–0.76 across all cycles, improving to 18–20% NRMSE for the best cycles. No significant performance difference was observed between 10°/s and 20°/s conditions, indicating model stability across velocities.

These findings validate the feasibility of SAW-FMG for noninvasive torque estimation. Subject-specific calibration enhances accuracy by accounting for inter-individual variations in muscle geometry and placement, while the compact, battery-free configuration demonstrates promise for continuous, wearable joint-torque monitoring. Future work will focus on reducing latency, refining calibration procedures, and integrating advanced modeling approaches to improve prediction accuracy, paving the way toward untethered, real-time applications in sports medicine, rehabilitation, and neuromuscular diagnostics.