WO2024007000A2

WO2024007000A2 - A capacitive sensor for monitoring muscle activation

Info

Publication number: WO2024007000A2
Application number: PCT/US2023/069519
Authority: WO
Inventors: Owen PEARL; Louis Dankovich; Eni HALILAJ; Sarah Bergbreiter
Original assignee: Carnegie Mellon University
Priority date: 2022-06-30
Filing date: 2023-06-30
Publication date: 2024-01-04
Also published as: WO2024007000A3; AU2023299303A1

Abstract

A capacitive sensor system is adapted for use as a wearable sensor and is capable of measuring the muscle activity of a user. The sensor system monitors the change in distance between an electrode and the user's skin, where the change is caused by the changing muscle cross-sectional area or muscle bulging. A musculoskeletal model is used to validate the capacitance measurement with the muscle cross-sectional area from gold-standard marker-based motion capture along with gold-standard measurements of muscle excitation, and muscle contraction force. The sensor can be used in real-world settings to assist with gait rehabilitation, among other uses.

Description

TITLE

A Capacitive Sensor for Monitoring Muscle Activation

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit under 35 U.S.C. § 119 of Provisional Application Serial No. 63/357,219, filed June 30, 2022, which is incorporated herein by reference.

STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH

[0002] This invention was made with United States government support under DGE1745016 and DGE2140739 awarded by the National Science Foundation. The U.S. government has certain rights in the invention.

BACKGROUND OF THE INVENTION

[0003] The invention relates to wearable sensors. More specifically, the disclosure relates to capacitive sensors capable of identifying muscle activation and is useful in a variety of biomechanical applications.

[0004] Wearable sensors can be used to monitor movement and are critical in the treatment of mobility impairments and the development of assistive wearables, among other uses. Motion capture equipment remains the preeminent technology for studying movement, yet it is rare to find such equipment outside of biomechanics laboratories preventing wide-scale adoption. Monitoring a patient’s movement during everyday life could transform diagnosis and prognosis and inspire the design of new smart rehabilitation practices with the patient in the loop. Portable and inexpensive wearable sensors have been explored to fill needs unmet by motion capture equipment.

[0005] In this context, wearable sensors should provide reliable data over long periods of time, be robust to placement errors by non-experts, and show tangible translational potential. Inertial sensing has been studied as a wearable candidate, but it remains sensitive to both drift and placement error. Besides being limited by drift, inertial sensors do not capture muscle activity, which is relevant to both mobility-limiting conditions and mobility-enhancing technologies. Wearable technologies that focus on surrogates of muscle activity, such as surface electromyography (EMG), suffer similar drawbacks.

[0006] Capacitive sensing is a diverse class of sensing whereby the natural capacitive coupling between objects is measured to infer information about relative motion, position, material composition, and proximity for a multitude of interaction techniques and applications. Previous use of capacitive sensing for motion sensing focused primarily on activity recognition and gesture recognition. For activity recognition, capacitive sensing provides only coarse classification of gait phase or locomotion mode using machine learning models and is not generalizable. And although useful for human-robot interaction, gesture recognition is not a primary goal in rehabilitation engineering.

[0007] Further, real-time estimates of muscle activity could complement existing wearable technologies and boost their impact across a myriad of clinical applications. These measurements could be informative on their own or as part of multi-modal sensing systems that enable comprehensive kinematic and kinetic analysis of movement. Intelligent fusion of data from multimodal wearables has until now remained elusive, despite its promise to revolutionize human motion tracking in the wild. For example, sensor fusion algorithms applied to data from inertial measurement units (IMUs) are unable to reach high accuracy in human applications, despite their success in robotics, due to the unique challenges posed by sensor-to-body calibration and soft-tissue-motion uncertainties. Therefore, it would be advantageous to develop a wearable sensor capable of providing data related to muscle activity, where the sensor is low-cost and useable outside of a controlled environment.

BRIEF SUMMARY

[0008] According to embodiments of the present invention is a capacitive sensor capable of providing capacitance measurements related to a cross-sectional area of a muscle. Using biomechanical models, the cross-sectional area can be correlated to muscle excitation and muscle force. In one embodiment, the sensor comprises an electrode disposed on a substrate, where the substrate is adapted to contract the skin of a user and provide a gap between the electrode and the user’s skin. An attachment mechanism, such as a hook-and-loop fastener, maintains the position of the electrode and substrate against the skin. As a user activates a muscle, the substrate compresses and the distance between the skin and the electrode decreases, changing the measured capacitance.

[0009] More specifically, a capacitive sensing sleeve, or sensor, measures circumferential capacitive profiles of the shank and thigh of a user, which are primarily governed by bulging of the shank and thigh muscles as they contract to flex the ankle and the knee. The expansion in cross-sectional area decreases the distance between four copper tape electrodes in the capacitive sensor and the subject’s skin, which then results in an increase in their capacitive coupling. [0010] The capacitive sensing can be used as a gait rehabilitation monitoring technology as it can accurately track changes in muscle bulging and fiber length with the fidelity of laboratory tools across people of different age groups and body compositions. Capacitive sensing complements inertial sensing better than electromyography in multimodal wearable sensing systems, achieving state-of-the-art accuracy in natural-environment motion tracking. Further, capacitive sensing as a biomechanics-monitoring technology can operate on its own or complement inertial sensing better than electromyography in multimodal wearable sensing systems. The sensor can be used to detect asymmetric gait modifications that commonly occur after knee surgery and adherence to gait retraining prescriptions that prevent osteoarthritis.

BRIEF SUMMARY OF THE SEVERAL VIEWS OF THE DRAWINGS [0011] Fig. 1 is a diagram of the sensor, according to one embodiment. [0012] Fig. 2 shows the sensor positioned on the shank and thigh of a user.

[0013] Fig. 3 depicts a musculoskeletal model of a thigh and shank.

[0014] Fig. 4 is a set of graphs depicting the performance of the sensor.

[0015] Fig. 5 is a set of graphs showing performance in various settings.

[0016] Fig. 6 compares the sensor compared to lab-based measurement systems.

[0017] Fig. 7 shows the improved performance sensor for monitoring rehabilitation.

[0018] Fig. 8 is a graph showing the capacitive sensor data merged with other sensor data.

DETAILED DESCRIPTION

[0019] Fig. 1 shows a sensor 100 according to one embodiment. As shown in Fig. 1, the sensor system 100 comprises an electrode 110 disposed adjacent to a substrate 120 and an attachment mechanism 130. The attachment mechanism 130 may comprise a hook-and-loop fastener, a buckle, an elastic sleeve adapted to be positioned around a limb of the user, or any other similar devices. In one example embodiment, multiple electrodes 110 are used to provide multiple data channels in order to provide more coverage of the measured body segment and higher resolution of the underlying muscle cross-sectional areas. In this embodiment, the electrodes 110 comprise copper-tape sandwiched between a felt substrate 120 that rests on the user’s skin and an adjustable hook-and-loop layer 130, which enables the sensor 100 to be adjusted to fit different users. Snap connectors 111 are used to connect the electrodes 110 to a processing unit 140. The processing unit 140 may comprise a controller, a microcomputer, a processor, an application specific integrated circuit, a programmable logic array, a logic device, an arithmetic logic unit, a digital signal processor, or another data processor and supporting electronic hardware and software. The processing unit 140 can measure the capacitance between the electrodes 110 and the user’s skin, perform further analysis on the data measurement or communicate the measured data to another processing unit 140.

[0020] Each capacitive electrode 110 in the sensor 100 is capacitively coupled to the human body, which acts as the ground. The capacitance between each electrode 110 and the body can be approximated as a parallel plate capacitor, although fringing fields are also non- negligible:

EQ. (1) C= e(A/d) where ‘s’ represents the dielectric constant of the substrate 120 material between the electrode 110 and the body (i.e. felt in one example embodiment), ‘A’ represents the area of the electrode 110, and ‘d’ the distance between the electrode 110 and the body, which is equivalent to the thickness of the substrate 120. As suggested by the consideration of a dielectric constant, the substrate comprises a dielectric material. In the example embodiment, felt is used but other dielectric materials may be used. Natural and synthetic fiber textiles, neoprene, silicones, resin- coated materials, and combinations of the foregoing are examples of materials suitable for use as the substrate 120. As will be discussed, the substrate 120 is deformable to permit a change in the distance ‘d’ as the muscles activate.

[0021] Fig. 2 is a cross-sectional view of a user’s leg with a sensor 100 fitted, with an example of the relative distance between the skin and electrode 110 shown for an active and inactive muscle. As shown in Fig. 2, an active muscle decreases the distance ‘d’ by compressing the substrate 120. The attachment mechanism 130 firmly encircles the user’s leg, preventing the sensor 100 from simply moving outward from the center with the expanding cross-sectional area of the muscles. Stated differently, the attachment mechanism has a fixed circumference and any change in cross-sectional area is manifested as a decrease in the distance between the electrode 110 and the user’s skin. The capacitive sensor 100 monitors proximity between the muscles and the electrodes 110, which for the shank primarily vary due to bulging of the gastrocnemius, soleus, and tibialis anterior muscles as they are activated to produce ankle plantarflexion and dorsiflexion.

[0022] The capacitive touch sensor 100 measures the change in capacitance as muscles contract, shorten their length, and expand their cross-sectional area. The capacitance of the interaction between the user’s skin and the electrode 110 is measured via the processing unit 140, such as a Teensy 3.2 microcontroller with capacitive touch pins, which uses a dual oscillator method to determine the difference in the charging and discharging times between the capacitor of interest and a reference capacitor on the microcontroller 140. Alternatively, other methods of measuring capacitance known in the art, which leverage different operating principles such as the double differential principal, RC phase delay, and phase angle conversion, may be used. This difference in time is reported in counts, a measure that is directly proportional to the capacitance, and analyzed on the processing unit 140. Alternatively, the measure is sent via wireless communication to a secondary processing unit 140 separate from the sensor 100 for post-processing.

[0023] The raw capacitance measurement has never before been shown to be directly indicative of muscle activation and changes in muscle cross-sectional area. To prove that the sensor directly measures muscle activation and cross-sectional area change, a musculoskeletal model is used. In one embodiment, an open-source platform for simulating and analyzing musculoskeletal systems is used to scale a musculoskeletal model and perform constrained inverse kinematics with marker data to estimate muscle fiber lengths and cross-sectional areas. Marker data is used to validate that the sensor 100 is working accurately across a pool of users. However, marker data is not required outside of validation of the sensor 100. Alternatively, mapping muscle fiber length changes to muscle cross-sectional area changes or mapping individual muscle bulging profiles to the resulting capacitance profile can be used. The former approach produces reasonable results as the fiber length change most prominently controls the resulting anatomical cross-sectional area (ACSA) of each muscle, which depends on the total muscle volume ‘V’, muscle fiber-length T, and the muscle’s nominal pennation angle. The musculoskeletal model can also be used to simulate muscle force production. Fig. 3 shows a data acquisition set-up with markers present on the user’s leg and the resulting musculoskeletal simulation to comprehensibly validate the capacitive sensors measurements against gold- standard measurements in a research level gait lab.

[0024] The validation shows that the capacitance sensor 100 profiles are highly correlated with changes in the cross-sectional areas of the shank and thigh muscles, resulting in physiologically interpretable measurements (see Fig. 4) directly using the raw capacitance measurements. Capacitance profiles at the shank generally rise throughout the stance phase — peaking near 40-45% stride as the posterior shank muscles plantarflex the ankle — then drop off shortly after toe-off near 65-70% stride, and rise again from around 80% stride to 105% stride as the anterior shank muscles dorsiflex the ankle to prepare for heel strike. Capacitance profiles at the thigh generally peak during early stance (near 10% stride) as the quadriceps muscles control the amount of knee flexion and assist in the extension of the knee in midstance, then drop off during late stance (near 40% stride), and begin climbing again after toe-off (60% to 100% stride) as the quadriceps and hamstrings assist in hip flexion by stabilizing, accelerating, and decelerating the leg during swing. A multilinear regression reveals that the cross-sectional areas of the muscles in each body segment are remarkably strong predictors of the corresponding capacitance measurement (adjusted r² = 0.94, p < 0.0001 at the shank; adj. r² = 0.99; p < 0.0001 at the thigh).

[0025] The association between muscle areas and capacitive measurements remains consistently strong over a range of conditions. This association is demonstrated in Fig. 5, which shows trials using both shank and thigh sleeve-type sensors 100 and participants ranging from 19 to 70 years old, as well as on a 76-y ear-old total knee replacement patient. Each unimpaired participant performed a baseline walking gait as well as a simulated impaired- walking gait, where they were directed to keep their nondominant knee straight during the entire walking trial, in a similar fashion to post-surgical stiff-knee gait. During the trials, capacitive sensing (CS), electromyography (EMG), and musculoskeletal simulations derived from motion capture data were captured while subjects walked on an instrumented laboratory treadmill with baseline, stiff-knee, and toe-in gaits. Marker-based motion capture was used with a musculoskeletal model to estimate muscle cross-sectional areas of the lower limbs.

[0026] In each condition, the cross-sectional areas remained a strong predictor of the capacitance signal (all adj . r²’ s > 0.88) with no statistical differences detected between the body segment, walking gait, and participant age groups (all Tukey’s HSD p’s > 0.1392). This relationship remained strong for the total knee replacement patient as well (adj . r² = 0.90 at the shank; adj. r² = 0.88 at the thigh). The trials also measured each subject’s muscle volumes, resting muscle fiber lengths, percent fatty infiltration, and muscle development score via magnetic resonance imaging (MRI) and then tested whether the relationship between cross- sectional area and capacitance became weaker with varying body compositions. The trials showed that there was no significant degradation in the fiber length to capacitance relationship with respect to these factors or the factors of age and body mass index (BMI) (all abs(r)’s < 0.36; all p’s > 0.3475). The steady association between muscle area and capacitance indicates that capacitance is capable of consistently capturing muscle bulging profiles and is not limited in its usefulness to specific body locations, steady-state symmetric gaits, or patients of a particular age, physiology, or musculoskeletal health.

[0027] Sensor 100 measurements are also correlated with electromyography and simulated muscle forces. To show this association, the capacitive profiles from capacitive sensing were compared with aggregate electromyography and simulation profiles of the tibialis anterior, soleus, and gastrocnemius for each gait, showing that muscle bulging measured with the sensor 100 is strongly correlated with muscle excitation estimated from electromyography and muscle force estimated from simulation (see Fig. 6). Capacitive sensors 100 exhibited strong correlations with electromyography (r = 0.87; p < 0.0001) and simulation (r = 0.86; p < 0.0001) across the average of the baseline trials, as well as strong correlations with both EMG (median r = 0.69; p < 0.0001) and SIM (median r = 0.84; p < 0.0001) for individual gait cycles. Capacitive sensing was characterized by lower standard deviations than electromyography across gait cycles within a subject (mean of 8.1% of the maximum for CS vs 10.5% for EMG; p = 0.0033) and across subjects (mean of 11.4% of the maximum for CS vs 13.4% for EMG; p = 0.0010). The same analysis conducted at the thigh yielded similar results, with capacitive profiles exhibiting strong correlations with EMG (r = 0.63; p < 0.0001) and SIM (r = 0.81; p < 0.0001).

[0028] By estimating composite muscle bulging, which results from a combination of contractile force and connective tissue constraints that collectively modulate overall force production, capacitive profiles are inherently better correlated with muscle force (SIM) than muscle fiber excitation (EMG).

[0029] Capacitive Sensing in Clinical Applications

[0030] The shank and thigh sleeve-type sensors 100 show clinical utility in two applications: (1) monitoring of a therapeutic gait-retraining prescription and (2) classification of an asymmetric stiff-knee gait, similar to what patients exhibit after knee surgery.

[0031] Training patients with knee osteoarthritis to walk with their toes in, or out, has been shown to reduce the knee adduction moment — a surrogate of medial compartment knee loading — decrease knee pain, and slow the microstructural progression of the disease. Capacitive sensing is able to more precisely detect gait deviations resulting from participants switching from a baseline to toe-in gait in both a laboratory and an outdoor environment. To demonstrate this application, 17 unimpaired individuals were recorded walking on an instrumented treadmill at their preferred walking speed, while wearing a sleeve-type sensor 100, electromyography sensors, and retro-reflective markers. Outside of the laboratory, capacitance and electromyography measurements were recorded for seven subjects. Subjects were recorded in each environment while performing both a baseline walking gait and a 5° toe- in walking gait. The sensor 100 distinguished toe-in from baseline gait as well as, or better than, existing laboratory benchmarks (see Fig. 7).

[0032] In addition to monitoring adherence to therapeutic gait modifications, it may also be beneficial to monitor negative gait adaptations that result due to injury or surgical interventions. For example, stiff-knee gait, occurring after an anterior cruciate ligament tear or a knee-replacement surgery, can increase the risk of either degenerative diseases or revision surgeries. Binary classifiers using capacitance data (Fl = 0.94) are more accurate than those using inertial measurement units (Fl = 0.81) and as accurate as those using electromyography data (Fl = 0.94) at detecting stiff-knee gait, in both unimpaired individuals and a patient recovering from total knee replacement surgery.

[0033] Capacitive Sensing as an Integral Component of Multimodal Wearables

[0034] Fusion of inertial and capacitive sensing data can increase accuracy in human motion tracking. Also, capacitive sensing can be more complimentary to inertial sensing than electromyography, since capacitive sensing estimates muscle fiber length, which is dependent on the joint angles of the body. Electromyography, on the other hand, estimates neural drive via a time delayed measure of the action potential across the muscles, which is difficult to track with a musculoskeletal model and does not typically improve estimates of body kinematics, but is rather utilized to correct estimates of muscle tendon unit, joint moments, and muscle excitations while holding constant kinematics estimated from marker data alone.

[0035] Three classes of musculoskeletal simulation were performed and compared to ground truth kinematics using marker-based motion capture data (mocap): (1) sensor fusion with inertial measurement units using a musculoskeletal model (IMU), (2) simultaneous trajectory tracking of the kinematic outputs from simulation 1 and electromyography data with a musculoskeletal model (IMU+EMG), and (3) simultaneous trajectory tracking of the kinematic outputs from simulation 1 and capacitive sensing data with a musculoskeletal model (IMU+CS). The inertial measurement unit only simulation utilized eight IMUs placed on the body and commercial sensor fusion algorithms to predict segment orientations from the experimental accelerometer, gyroscope, and magnetometer data, then constrained orientation kinematics with a musculoskeletal model via an open-source platform. The IMU+EMG simulation fused kinematics from the IMU-only simulation with experimental EMG data in an optimal control tracking problem solved with direct collocation methods in a motion control library. The IMU+CS simulation also fused kinematics from the IMU-only simulation with fiber lengths predicted with capacitance data, solved with direct collocation methods in the motion control library.

[0036] The results shown in Fig. 8 (shown for hip flexion; dashed line is ground truth from marker-based motion capture) suggest that capacitance data is not only a better complement to IMU data than traditional methods for measuring muscle activity, but that it is also a powerful enough measure of muscle behavior to support portable, multimodal simulations with accuracies comparable to simulations run with data from a traditional gait laboratory.

[0037] Sensitivity to Sensor Design Parameters and Placement

[0038] The circumferential electrodes 110 in the sensor 100, along with the felt dielectric substrate 120, can be customized to different patients and applications with the expectation that sensitivity to gait changes may also change. Signal quality depends in-part on two factors — the electrode area ‘A’ and the substrate 120 thickness ‘d’. The sensor 100 can perform well even over layers of fabric and thick clothing. Yet, reductions in electrode area (to a fourth of the original area) and increases in substrate thickness (to four times the substrate thickness) decrease the ability of the sensor 100 to detect therapeutically relevant changes in gait from a baseline to toe-in RMSD of 17.0% to an RMSD of 5.5% near the ankle and from an RMSD of 13.3% to an RMSD of 11.5% at the gastrocnemius. However, even at these conditions, the electrode 110 was still capable of distinguishing between the baseline and toe- in gaits, suggesting a remarkable robustness to design parameters.

[0039] When used in this specification and claims, the terms "comprises" and "comprising" and variations thereof mean that the specified features, steps, or integers are included. The terms are not to be interpreted to exclude the presence of other features, steps or components.

[0040] The invention may also broadly consist in the parts, elements, steps, examples and/or features referred to or indicated in the specification individually or collectively in any and all combinations of two or more said parts, elements, steps, examples and/or features. In particular, one or more features in any of the embodiments described herein may be combined with one or more features from any other embodiment(s) described herein.

[0041] Protection may be sought for any features disclosed in any one or more published documents referenced herein in combination with the present disclosure. Although certain example embodiments of the invention have been described, the scope of the appended claims is not intended to be limited solely to these embodiments. The claims are to be construed literally, purposively, and/or to encompass equivalents.

Claims

CLAIMS What is claimed is:

1. A capacitive sensor system for estimating muscle activity of a user comprising: a capacitive sensor comprising an electrode and a substrate separating the electrode from skin of a user, wherein the substrate comprises a dielectric; and a processor adapted to perform the following: measure a capacitance between the electrode and the user’s skin; associate a change in the capacitance with a change in a cross-sectional area of a muscle; and identify an activity of the muscle based on the change in cross-sectional area.

2. The capacitive sensor system of claim 1, wherein the muscle activity is correlated with muscle excitation and muscle force.

3. The capacitive sensor system of claim 1, wherein the capacitance is a circumferential capacitive profile of the user.

4. The capacitive sensor system of claim 1, further comprising: an additional sensor, wherein the processor fuses data from the additional sensor and the capacitance.

5. The capacitive sensor system of claim 4, wherein the additional sensor comprises an inertial measurement unit or an electromyography sensor.

6. The capacitive sensor system of claim 4, wherein the processor fuses data from the additional sensor and the capacitance using a musculoskeletal model.

7. The capacitive sensor system of claim 1, further comprising: an attachment mechanism holding the substrate in contact with the user’s skin, wherein the attachment mechanism circumferentially surrounds the muscle.

8. The capacitive sensor system of claim 1, wherein the substrate is deformable.

9. The capacitive sensor system of claim 7, wherein the substrate is deformable.

10. The capacitive sensor system of claim 1, further comprising: an additional electrode providing a separate capacitance measurement to improve a resolution of the cross-sectional area.

11. The capacitive sensor system of claim 1, wherein associating a change in the capacitance with a change in a cross-sectional area of a muscle comprises: associating a measurement of capacitance with a cross-sectional area derived from a musculoskeletal model.

12. A method of measuring a cross-sectional area of a muscle comprising: providing a capacitive sensor comprising an electrode and a substrate separating the electrode from a user’s skin, wherein the substrate comprises a dielectric material; measuring a capacitance between the electrode and the user’s skin; associating a change in the capacitance with a change in a cross-sectional area of a muscle; and identifying an activity of the muscle based on the change in cross-sectional area.

13. The method of claim 12, further comprising: correlating the muscle activation with muscle excitation and muscle force.

14. The method of claim 13, wherein the capacitance is a circumferential capacitive profile of the user.

15. The method of claim 12, further comprising: providing an additional sensor, and fusing data from the additional sensor with the capacitance.

16. The method of claim 15, wherein the additional sensor comprises an inertial measurement unit or an electromyography sensor.

17. The method of claim 12, further comprising: circumferentially attaching the sensor around at least one muscle of the user. method of claim 12, wherein the substrate is deformable. method of claim 12, further comprising: creating a musculoskeletal model; and associating a capacitance with a cross-sectional area of the muscle based on the musculoskeletal model.