SELF-POWERED LOAD SENSING CIRCUITRY FOR TOTAL KNEE REPLACEMENT

STATEMENT OF GOVERNMENT RIGHTS

This invention was made with U.S. Government support under R21 AR068572 awarded by the National Institutes of Health. The U.S. Government has certain rights in the invention.

CROSS REFERENCE TO RELATED APPLICATIONS

The present application claims benefit of priority from U.S. Provisional Patent Application No. 63/403,638, filed September 2, 2022, the entirety of which is expressly incorporated herein by reference.

FIELD OF THE INVENTION’

The present invention relates to implantable energy harvesting sensor devices, and more particularly to an orthopedic implant employing a triboelectric harvester as both sensor and power source.

BACKGROUND OF THE INVENTION

A Total Knee Replacement (TKR) alleviates pain and restores mobility to patients suffering from knee arthritis. en.wikipedia.org/wiki/Knee_replacement However, instability and premature prosthesis wear continue to be major causes of failure. Early detection of incipient failure and excessive wear is essential to avoid revision surgery and reduce medical costs. Self-powered, embedded sensors could provide a new in vivo capability to detect mechanical issues sooner and address them using more targeted treatments. The loads transmitted through the knee during activities of daily living may be used as a source of energy to power a load sensor. Self-powered embedded sensors perform measurements, process and store results, and communicate them noninvasively to the patient and/or alert a health care provider. Aggregate data can be used to better understand and improve the function of commercially available implants and minimize costs and harm to patients.

Continuously monitoring the magnitude and distribution of loads on total knee replacement (TKR) implants permits detection of unsafe loads, improvement of implant designs, and response to the patient before the implant failure. A monitoring system can also help post-operatively to alert patients and surgeons about problems that could potentially be addressed early to avoid a more complex and expensive surgery later on. There has been significant research in the field of smart knee implants. A prior smart knee implant powered by the energy harvested from an electromagnetic generator was developed. This harvester uses a combination of several magnets inserted between the coil, which consumes a large area and consumes power in the range of milliwatts. Other knee implant systems harvest energy using piezoelectric elements. Since the energy is harvested through deformations, service life of the implant is low. TKR implants were embedded in the knee of a senior patient, consisting of four load cells with wireless micro-transmitters. A major limitation of these designs was that the load measurement could only be performed in a medical clinic, as remote powering using magnetic near-field coupling was required for operation. While the instrumented system provided a detailed description of the forces and moments acting on the knee joints, the range of input loads that could be monitored by this implant is highly limited. Yet, they provide data only during the surgery and should be removed by the end of the surgery. A scalable, wearable e-textile triboelectric energy harvesting (WearETE) system has been developed. The WearETE system features low cost material and manufacturing methods, high accessibility, and high feasibility for powering wearable sensors and electronics.

Triboelectric energy harvesting relies on contact electrification and electrostatic induction. A triboelectric generator (TEG) typically consists of two materials (tribolayers) with different polarities. At least one of the materials should be dielectric with a conductor on both sides of the tribolayers. When the two tribolayers are pressed or rubbed against each other, opposite charges are produced, referred to as contact electrification. Due to these charges, the conductor layers are electrostatically induced and can create a flow of electrons, if the tribolayers are periodically moved back and forth. This energy harvesting technology has shown promising results in a wide range of sensor applications, including biomedical systems. In recent work, it was demonstrated that TEG can be used to harvest reasonable power from natural knee motions. Hence, triboelectric harvesters can be incorporated in designing a self- powered electronic system, which can continuously monitor the knee load. Furthermore, triboelectric generators offer several advantages over piezoelectric or electromagnetic based energy harvesting such as higher energy density (313W/m ² ), simple fabrication at low cost, excellent reliability, high efficiency, and greater sensitivity. Furthermore, since triboelectric transducers generate voltage from pressure (force), they do not require large deformations. As such, they do not have the size limitation of piezoelectric devices. Since knee is one of the body joints that takes 3 to 6 times the body weight, triboelectric energy harvesting could serve as a promising candidate for scavenging energy from the loads at the knee joint

There have been instrumented knees developed for research. They have been pivotal to better understand knee loading during activities of daily living (ADL), to propose standardized loading protocols for pre-clinical testing, and to provide broadly beneficial model validation data. These previous systems, however, are not well suited to the type of remote monitoring and telemedicine described above, primarily due to their need for external powering. These systems typically employ metallic strain gauge or piezoresistive types of load cells embedded in the tibial tray to measure tibiofemoral forces and moments during ADL. Their energy demand is high and must be provided through an inductive link. In an inductive link, an external primary coil transmits electromagnetic power to a secondary coil integrated embedded in the implant. The external coil is constantly worn to power the implant during data collection; this requirement restricts the utility of these devices for continuous load measurement and may interfere with natural patient behavior. To avoid the internal inductive link, a rechargeable battery could be integrated within the implant. This battery would be periodically recharged through an external inductive link. This option, however, is not feasible because there is inadequate space available for a suitably sized battery.

Recent studies have developed self-powered instrumented knee implants for postoperative load detection using electromagnetic or piezoelectric mechanisms. However, these mechanisms have severe drawbacks that make them impractical. Electromagnetic devices require major design changes that may weaken the implant structure. The most common piezoelectric material is perovskite lead zirconate titanate (PZT) which is a brittle ceramic and contains lead. Lead-free piezoelectric material has been developed, but it has much lower power density. A relatively new technology, triboelectric energy harvesting, has the greatest power density of 313 W/m ² among all transducers. Triboelectric harvesters generate electricity using contact electrification and electrostatic induction. The contact electrification occurs from rubbing two materials against each other or sliding at micro-patterned textured contacting surfaces. The triboelectric transducer is ideal for implant applications because of its lower cost, easier fabrication, lower detection limits, and greater power density.

Piezoelectric ceramics generate a voltage from deformations. The use of piezoelectric ceramic within TKRs was first proposed in 2005. Later developments include embedding four piezoelectric load sensors sandwiched between the tibial tray and a polyethylene bearing. The piezoelectric units may also be integrated inside the polyethylene bearing, in a design that produced 13 microwatts of power under gait cycles. However, the piezoelectric ceramics including the common PZT-5A contains lead that causes health risks and has a lower power density compared to triboelectric energy harvesting. Other proposed mechanisms employ electromagnetic devices, which require incorporation of coils and permanent magnets within the implant that may weaken the structure. T riboelectric generators that convert motion have shown promise for biomedical devices because of their high power density and biocompatibility.

SUMMARY OF THE INVENTION

There has been a significant increase in the number of total knee replacement (TKR) surgeries over the past few years, particularly among active young and elderly people suffering from knee pain. Continuous and optimal monitoring of the load on the knee is highly desirable for designing more reliable knee implants. A smart knee implant is provided consisting of a triboelectric energy harvester and a frontend electronic system to process the harvested signal for monitoring the knee load. The harvester produces an AC signal with peak voltages ranging from 10 V to 150 V at different values of knee cyclic loads. Measurement results of a PCB prototype of the frontend electronic system fabricated to verify the functionality and feasibility of the approach for a small range of cycling load were obtained. The front-end electronic system consists of a voltage processing unit to attenuate high peak voltages, a rectifier and a regulator to convert the input AC signal into a stabilized DC signal. The DC voltage signal provides biasing for the delta-sigma analog-to-d igital converter (ADC). Thus, the output of the triboelectric harvester acts as both the power signal that is rectified/regulated and data signal that is digitized. The power consumption of the proposed PCB design is approximately 5.35 pW. The frontend sensor circuitry is improved to accommodate a wide range of cyclic load. Triboelectric energy harvesting permits self-monitoring of the load inside knee implants.

The present technology incorporates triboelectric energy harvesting to power a smart knee implant and produce a sensor signal. The technology provides compactness, customization for various knee implants, ability to be powered only by the harvester, and continuous force measurement.

The total knee replacement (TKR) system consists of a femoral and tibial tray and ultra-high mechanical polyethylene (UHMWPE) bearing parts as illustrated in Fig. 1 A. The triboelectric harvesters are placed between the tibial tray and the UHMWPE bearing for load monitoring. The frontend electronic system is placed on the tibial tray and powered entirely by the power generated by the designed triboelectric harvesters without external biasing. The harvesters and the electronic system are placed in a 3D package. An enlarged view of the package with the triboelectric generators is shown in Figs. 1 B and 1 C. Fig. 1 D shows a detail of the triboelectric generator. PDMS may be used in fabricating the triboelectric harvester because of its biocompatibility and flexibility. FR4 (flame retardant) epoxy glass PCB substrate may be used for fabrication as it has high dielectric strength, thereby contributing to its electrical insulation properties. It also has high strength-to-weight ratio and is sufficiently lightweight.

Fig. 1 E shows a schematic drawings of a signal processor integrated circuit (IC) for processing sensor signals. The fundamental blocks of the IC are for power management, feature extraction, memory, and an inductive wireless link to transmit data. The ultimate goal for this measurement system is to provide detailed load measurements during rehabilitation clinical visits and coarser aggregated load data outside of the clinic. An Artificial Neural Network (ANN) relates the two different measurement types, enabling accurate reconstruction of the applied loads from the coarse measurement data and identification of irregularities or mechanical changes.

The present technology provides a self-powered frontend electronic circuit to process the output voltage signal of the triboelectric harvester. Both a small and wide range of cyclic loads can be processed with sufficient accuracy.

Triboelectric transducers can use a broad range of lead-free materials; are biocompatible, have minimal costs and the highest energy density. They are ideal for biomedical implants. Compared to piezoelectric and electromagnetic devices, the harvester has greater sensitivity and can fit into any TKR system without modifying their structures, thus enhancing their versatility and reducing the cost. The sensor circuit operates in two modes, providing continuous load monitoring requiring only microwatts of internal power.

The instrumented knee implant operates solely on harvested power and extract statistical measures of loading forces during specific activities. This is achieved through a combination of analog and digital signal processing and efficient power and memory management. Within a specific duration of daily activities, a set of parameters is extracted through low power analog circuit processing for each loading cycle. These parameters are then stored in on-chip volatile memory. The parameters captured from a pulse-like signal include peak value, the time period between events, and the integrated sensory signals from an event The digital data is processed and statistical measures, such as frequency and the mean and standard deviation of the peak loading force, are extracted. This compressed information is moved to non-volatile memory. The data from the non-volatile memory can be read by an external device, where neural network analysis can be performed to extract approximate load profiles. This neural network is trained with raw data obtained during rehabilitation clinic visits where the implant is externally powered. This externally-powered mode permits high frame rate data transmission without compression for training. The low power on-chip signal processing facilitates the use of small capacity non-volatile memory that fits within the limited area while enabling a significantly richer dataset than previously reported for self-powered instrumented knee implants.

The harvester may have a partitioned sensor architecture, to identifying the force distribution across the tibial tray. For example, the sensor may comprise four quadrants. This information is critical to increase the life of the implant because shifting of the center of pressure causes loosening and instability. Some training data may be derived from mechanical testing in cadaveric knees, to optimize ANNs or statistical algorithms for reconstructing loading patterns from minimal extracted features and classifying abnormal joints.

The present technology provides a self-powered load monitoring system that takes advantage of triboelectric energy harvesting for implant applications. A preferred implementation of the self-powered system consists of four triboelectric generators and a read-out circuit for sensory signal processing, data logging into memory, and wireless transmission. Unlike existing approaches that continuously rely on an inductive link to power the implant, the preferred embodiment only requires a continuous inductive link during clinic visits when data is measured with a high sampling rate. Other times (at home), the sensor relies on harvested power and an inductive link is used for only less than a minute, once a day, for data transmission only. The load sensor is operated with power scavenged from loads transmitted through the knee during daily activities and does not suffer from limited charging cycles like batteries. Because the power generated is proportional to the amount of load, the generator is simultaneously used to measure the applied forces. This property enables simultaneous energy generation and sensing. This synergy between generation and sensing enhances overall efficiency compared to strain-gauge and inductive-powering technologies. The entire package is preferably installed between the UHMWPE bearing and the tibial tray of the implant, eventually allowing the device to be incorporated into any TKR, unlike recent piezoelectric mechanisms that require modification of the implant components. The self-powered sensor system has separate quadrants enabling it to detect how load is distributed across the joint; a feature that can be leveraged to detect abnormal load transmission through the joint, which may be used to phenotype patients into different treatment groups (normal, non-surgical, bearing change, etc.). While accelerometers have recently been introduced for this purpose, direct load measurement potentially provides better data. For example, when coupled with computer models, the measured load data can also predict polyethylene wear progression, stress distributions in the implant and the implant-bone interface, and the stress transferred to the underlying bone.

The interface circuit for the triboelectric generator generates sufficient energy for powering an analog-to-d igital converter, and digital processing circuit, for initial data processing and storage. A supercapacitor or other energy storage element is not required for basic functioning, due to the time characteristics of power availability from the triboelectric generator, and a small amount of energy storage within the converter circuit.

The triboelectric mechanism can be customized and embedded in a wide range of commercial implants. The energy harvesting mechanism receives both power and signal from the walking motion and load profiles may be extracted using various algorithms executing on a low power microcontroller, e.g., artificial neural networks. Two modes of operation (high and low resolutions) are available to mitigate the power limitations. In a low resolution mode, the circuitry reliably operates on the harvested power alone. In the high resolution mode, power from inductive coupling may be available to permit higher resolution digitization and signal processing.

Triboelectric energy harvesters are feasible for converting equivalent human body motions to sufficient microwatts to power a voltage regulator and digitization circuit The harvester preferably employed has micro-patterned features on the surface to improve output efficiency. The micro-patterned features are, for example, sawtooth ridges protruding from the surface.

Packages made of titanium and polyethylene were fabricated, and the harvester generated a voltage signal based on the triboelectric effect when it experienced a cyclic loading. The human walking motion was approximated with a 1 Hz sine-wave motion. An axial load was applied using a servo-hydraulic test system to simulate the axial load passing through the knee while walking. The triboelectric energy harvester converted mechanical energy from the applied axial load to electrical energy that powered a digitization electronic circuit 25 pW of power was generated under an equivalent average gait load of 2 kN, as shown in Fig. 21. A model was used to predict the harvester output and the simulated results were in close agreement with the measured results, see Fig. 22. The triboelectric mechanism showed an ability to generate more power at higher frequencies, which can produce more voltages at activities such as jogging and running. For example, running twice as fast as walking yields twice the output power. That means with more energetic activities, more power can be generated, stored, and used to log data for long- duration daily activities. Therefore, an adaptive process may be implemented to limit data acquisition and processing during low activity periods, and permit increased data resolution (samples and/or bit resolution) when more power is available, corresponding to greater activity.

Preliminary results showed the digitization circuit can operate with only the power generated by the triboelectric harvester under physiologically relevant harmonic loading with a power consumption as low as 5.1 pW.

Harvesters made in halves (Fig. 23A) increase the power of harvesters three or four times when they are in parallel. To identify force distribution, the harvester may be divided in quadrants (Fig. 23B). Each harvester comprises two contacting surfaces separated by a gap (Fig. 1 D). The top layer is made of titanium and has triangular micropatterned ridges of 100-300 . The bottom layer is a patterned or flat dielectric layer like Polydimethylsiloxane (PDMS) or Styrene Ethylene Butadiene Copolymer (SEBS) layer that is coated on a titanium electrode. The area of the harvester directly influences the power it produces, so there is a trade-off between size and power. The front-end electronic system can be reliably powered because one half of the harvester at an average load of 2 kN provided 20 p W of power (Fig. 21 ), which is four times what is needed for the preferred circuit. The half area preferably provides at least 10 p W, which is almost double the required power by the circuit of 5.1 p W.

The area of the harvester governs the harvester output. The larger the area, the larger the output power. A design that uses four harvesters (Fig. 23B) instead of two (Fig. 23A) to be able to detect the center of pressure and identify load imbalances. The power conditioning circuit may be placed between them at the center of the tibial tray (Fig. 1 C). The output of each harvester is proportional to the load it receives. Thus, load information for the medial, lateral, front, and back compartments may be obtained. To make the best use of the available area, the harvester is configured to closely follow the shape of the tibial tray. This design permits greater accuracy in load sensing, and the ability to sense a load imbalance between the medial and lateral compartments.

Harmonic loading may be used to test the quadrant design in a package to see the effects on durability and output stability. A prior design used polydimethylsiloxane (PDMS) as an insulator, which did not have a long life because of the repeated deep penetration by the upper electrode. A preferred triboelectric harvester technology may produce power with minimal mechanical contact. More durable dielectric materials include fluorinated ethylene propylene (FEP), and Styrene-ethylene-butylene-styrene (SEBS). They can be coated on acrylic to act as an insulator material. A tape-peeling method is used to increase the surface voltage potential. These polymers have advantageous dielectric properties and the triboelectric charges can be preserved on their surface for a very long time. This design should significantly enhance the durability of the harvesters as the layers have minimal contact and has gaps in the range of 5-100 microns.

The housing includes a spring structure, called a package, separating the two layers to ensure cyclic motion. The triboelectric converter must be durable, yet compliant, with a target lifetime of up to ten million cycles with sustained output stability.

Two packages made of TisAkV and polyethylene were designed and constructed. For the titanium package itself, structural analyses were conducted using finite element simulations. This enabled tuning of the stiffness of the harvester package to provide an optimal amount of compression under different levels of physiological loading. The models permitted analyses of the stresses and strains in the deformable members of the compliant package, and evaluation of the safety and long-term fatigue resistance. These models were compared directly with experimental results obtained by cyclically loading prototypes on a joint motion simulator until fatigue failure.

Preliminary results on cyclic testing of the harvester showed the harvester output varies only within 7.5% after testing at average normal load of 2 kN for 10,000 cycles. The insulator material has a great influence on the output stability of the harvester. T riboelectricity gives freedom in material selections. Any two different materials with opposite tendency to lose or gain electrons during contact could be used in the harvester. Different insulator materials such as ethyl cellulose (EC) and polylactic acid (PLA)/Styrene-ethylene-butylene-styrene (SEBS), which are biocompatible, may be employed.

If the output of the triboelectric harvester varies too much over time, the interface circuit may no longer be optimal. This may be addressed by providing components within the circuit that are adaptive or controlled by the processor. Alternately, an interface circuit with a broader range of acceptable inputs may be provided.

The preferred device has a total thickness limited to 3-4 mm so it can be incorporated in various prostheses.

The sensor circuit may have two modes of operation. The self-monitoring system measures loads, with high- resolution data collected during rehabilitation and clinical visits, low-resolution data collected at home, and their connection analyzed using artificial neural networks (ANNs).

During clinical visits, patterns such as jogging, running, and gait can be recorded by asking the patients to wear a knee brace that provides the required power externally via the inductive link and achieves raw data transmission of at least 100 points in a cycle. At home, the sensing circuit has low resolution because of limited power. In this case, the circuit extracts only certain features of the signal and stores statistical characteristics of this data for each cycle, solely relying on the internal (harvester) power. For a few minutes a day, the patient turns on a battery-powered external reader near the knee to transmit this compressed data via an inductive link (i.e., NFC or RF-ID standard communication, e.g., IS0 18092, IS0 14443, ISO-21481 , IS0 18000 (-1 to -7), ISO-11784, ISO-11785, ISO-15693, ISO-14223, ISO-18092, FeliCa, EMC-340, EMCA-352, etc.). The reader can be a smartphone, or itself then use WiFi or Bluetooth/BLE to transmit the data to a mobile device. The external reader may be a hand-held sampling device that can be tethered to a cell phone. The data may be analyzed locally within an app downloaded to the memory of the smartphone or reader, or relayed to a server or cloud processing center for further analysis, data aggregation, etc. For example, the sensor may be used to determine defects or degradation in an orthopedic implant, and based on population data from a plurality of patients, a prediction of outcome for a respective patient may be made.

If an abnormality is detected in the pattern of statistics, an alert is created and more examinations can be conducted at a clinic. This abnormality can originate from shifting of the center of pressure or very strenuous activities. When the patient visits the clinic, the data may be analyzed via a neural network which was trained earlier with the low- and/or high-resolution data of various activities. Thus, the sampling rate at normal operation is significantly reduced using ANNs without the loss of accuracy. Note that an ANN may not be required, and traditional algorithms may also provide a mapping of the low resolution and high resolution datasets, and/or prediction of underlying clinical condition of the patient or the implant.

The triboelectric generator occupies most of the available area to maximize scavenged power. However, about 4 cm ² is used to integrate the sensory system circuit that extracts important features from the harvested signal, stores this data within the memory, and facilitates the wireless transmission to the external reader. When the cyclic gait load is applied, the harvester generates an electrical signal that is related to the load. This signal behaves as a data signal to be processed and a power supply signal to be rectified. The harvester and the circuit are compatible to have high efficiency and resolution.

A front-end electronic circuit was developed and demonstrated using discrete components and off-the-shelf integrated circuits for power management (including voltage rectification and regulation), signal conditioning, and digitization. The passive components act as impedance matching to maximize the power transferred to the circuitry while attenuating the large peak voltages for reliable operation of the circuitry. The circuit was fabricated on a printed circuit board (PCB) and connected to the harvester for testing under the MTS machine (Fig. 9). The front-end circuit successfully regulated the harvested signal with a ripple voltage of less than 1 % of the regulated supply voltage. The harvested supply voltage then successfully powered an 8-bit successive-approximation-register based analog-to- digital converter (ADC), thereby digitizing the sensed data while consuming approximately 5.1 pW. The measured and simulated ADC input/output waveforms are shown in Fig.13. The circuit was entirely powered via the triboelectric harvester without requiring any external supply or bias voltage.

Rather than digitizing the entire signal, the unit may extract only the critical data of peak voltage, pulse width, and integration from the signal for a specific duration of different daily activities. Certain statistical measures may be determined from this data, thereby significantly lowering the in-knee storage requirements. A four-channel integrated circuit may be provided for the quadrant design for maximum power efficiency. The system may have discrete ADCs for each channel, or a multiplexer and sample hold amplifier (SHA) or sample hold capacitor preceding an oversampled ADC.

When the knee is loaded, the generator produces a pulse-like signal, where the maximum peak voltage can reach 80 V or higher. The internal impedance of the generator is about 100 MQ. This signal carries information about the loading force and supplies energy for the operation of electronic circuits that process this signal. The sensory system for extracting and logging the information contains passive discrete impedance matching circuits at the harvester’s interface with an application-specific integrated circuit (ASIC), which provides signal acquisition and processing, storage, analysis, and communication. Two discrete passive impedance matching circuits produce output signals to generate DC supply voltage (Vdd) and sense loads. Discrete components are used in the prototype design at the interface because of the large generator voltage, though integration is possible. The ASIC is powered by the harvested energy. In addition to the instantaneous generation of the supply voltage for the operation of the ASIC, extra energy is stored in a supercapacitor. ASIC processes the output signal from the harvester to quantify the loading of the knee implant. As the continuous recording of the digitized output signal of the harvester might require prohibitive memory size, the recorded data may be compressed or processed before storage. The digitized data is stored in non-volatile memory and periodically, through the use of the in-package coil, read wirelessly by the external battery- powered reader. ASIC may be fabricated in a 180 nm CMOS process. The ASIC, along with the impedance matching circuits, non-volatile memory, and the supercapacitor may be integrated on a printed circuit board (PCB) or other carrier, which is connected to the in-package coil. The layout of the sensory system and the building blocks of ASIC are shown in Fig. 1 E.

To optimize the efficiency of the energy harvesting circuit, the input impedance of the interface circuit matches the very high internal impedance of the harvester. The impedance matching circuitry provides voltage attenuation that enables the IC to process the harvester output. Two impedance matching circuits provide two types of signals at the output, one for voltage supply generation and one for information processing.

In the voltage supply generation path, the goal is the efficient energy transfer, while in the information processing path, the goal is the maximizing the signal-to-noise ratio. The passive-interface circuit, which transforms the high voltage, was demonstrated via a fabricated PCB. The main design challenge of the impedance matching circuit is the wide range of harvester voltages, which leads to the adaptive passive circuit with variable values of the passive components. These values may be configured by sensing the energy harvesting signal within the ASIC.

The input AC signal to the ASIC along the energy harvesting path provides a regulated DC voltage supply for the ASIC and energy storage. In the information processing path, the ASIC conditions the signal, digitizes, compresses, and stores the information. Additionally, the ASIC controls the wireless transfer from the on-board memory to the external reader. The most stringent design requirements are a small form factor and low power consumption.

The ASIC is powered from the harvested AC signal, which has to be efficiently converted to DC voltage that matches the supply voltage required for the operation of the integrated circuits. The AC signal, after voltage attenuation by the impedance matching circuit, is rectified and regulated to provide the instantaneous supply voltage. Usually, if the generated energy is more than the energy instantaneously consumed by ASIC, the excess energy would be consumed by the voltage regulator and be wasted. To avoid this issue, an energy management circuit may be provided that senses the voltage at the input of the regulator and enables the transfer of the excess energy to a storage device, i.e., a supercapacitor. Supercapacitors have a lower energy storage density than batteries but can have an almost unlimited number of charge and recharge cycles. The energy management circuit can ensure optimal charging and discharging of the supercapacitor. The energy stored in the supercapacitor provides power supply voltage when there is no AC signal from the harvester. This enables data processing by ASIC when an activity is finished, and there is data that needs to be processed and stored.

The harvester output voltage signal depends on the knee loading force and contains important information on the fatigue, lifetime and impact of the implant on the surrounding tissues and bone. Continuous wireless transmission of the sensed data has a prohibitive power cost and data logging is required. The data logging requires the integration of non-volatile memory in the sensory system. The speed and energy cost of writing to non-volatile memory may limit the amount of data that can be stored between readouts to a few thousand data points. Note that evolution of memory devices is ongoing, and therefore this limit is not theoretical or enduring.

The sensor signal is generated with only knee movement, so the system spends significant time in an inactive, sleep mode. Because of the self-powered nature of the sensor, there is no wake-up circuit and the system logs data only while energy is harvested. The entire transient signal waveform cannot be stored in the memory because of limited space. Signal compression should be performed. Because of the pulse-like nature of the signal for a gait cycle, certain features can be extracted from each pulse. The main candidates for the set of the features that would preserve most of the information are the peak voltage of each pulse, the integrated value of the pulse, and the time between peaks.

The extraction of the proposed feature set in the analog circuit requires low amounts of power. The circuits for peak detection and pulse integration, where the pulse duration is measured in milliseconds, consume power on the order of several tens of nW. After extracting the peak value and the integrated value for the pulse-like signal, these voltages may be converted to the digital domain. The power consumption of the analog-to-digital converter (ADC) with the sampling rate of 1 Hz, with a moderate resolution, is below 10 nW. The timers that need to track time between peaks also require very low power. The collected digital values for each pulse are initially stored in small on- chip volatile random access memory (RAM). Note that with analog parameter extraction, and a quadrant-based sensor, a multiplexed ADC is preferred.

The locally stored digital data representing a set of features for each of the loading cycles are processed after the activity to obtain statistical measures, such as mean and standard deviation. The processing is performed using on-chip digital logic, and because of the low-frequency operation, the power consumption on the order of nano Watts. This operation is powered by the supercapacitor. These statistical measures are stored in non-volatile memory. The overall power consumption of the data compression circuitry is typically lower than 1 W, which is much less than the harvested power in a gait cycle (about 20 pW). The digital logic may be permanently defined as structures or readonly memory, or may be updatable under control from a remote device through the inductive coupling communication link.

In some cases, the supercapacitor may be a structural element of the unit, and therefore its volume may be distributed without interfering with the area of the triboelectric harvesters/sensors.

Data telemetry in the implant is performed through the inductive link powered externally by the reader, e.g., in a backscatter communication mode, though the implant may also harvest energy through the inductive coil and operate in an active communication mode. Active reader-based telemetry has been commonly used in various implantable devices. The inductive link comprises an external primary coil and a secondary coil located within the implanted device. The data from the secondary coil is transmitted using the backscattering communication principle. The secondary implant coil changes the loading impedance, thus changing the reflecting magnetic field. The backscattering enables data transmission at sufficiently low energy per bit for the data readout The primary coil detects the changes in the magnetic field and demodulates the data it receives. The logged sensing data in nonvolatile memory of the implant, with the presence of a coil on the implant, can then be read by an external primary coil. An active data communication system may also be employed, in which power is transferred at a first frequency from the reader to the implant, and the implant actively transmits data at a second frequency using the received power. The active transmission may be an arbitrary communication protocol, such as Bluetooth, BLE, NFC, or other low power standard.

The primary coil is embedded in a reader device and the secondary coil located in the package. The coils’ size and shape are optimized and the wireless channel characterized to determine optimum frequency, both analytically and by using a finite-element method-based field simulator (such as Ansoft HFSS). HFSS is used to accurately model the attenuation of signal throughout soft tissues such as muscle, fat, and skin. The effect of 3D package on wireless transmission was also considered and possible misalignments investigated for the robustness of the link. At an optimum frequency of 90 MHz, a coupling coefficient of 0.15 was achieved. Note that this optimum does not correspond to a ubiquitous communication standard, and therefore a custom communication system is preferred. The 90 MHz emission is preferably compliant with FCC regulations, and may be a digital spread spectrum signal (DSSS) emission which avoids narrowband interference to other devices and enhances signal privacy, for example.

The discrete impedance matching circuit and ASIC may be designed and fabricated on separate or a common circuit board. Given the 3-4 mm thickness of the implant, it may be possible to stack two boards to provide an increased effective board area. The circuits are preferably integrated on a single PCB, along with the non-volatile memory and the supercapacitor.

Depending upon the applied force, the peak voltage generated by the harvester varies between 10V and 150V. The resolution of the sensor in measuring the voltage is about 10 bits full range, i.e. , harvester voltages as low as 70 mV will be detected. Each of the four signal processing channels of the circuit consumes approximately 1 pW. The overall power consumption including writing into and reading from nonvolatile memory is approximately 8 pW, meaning continuous load monitoring is possible because each quadrant of the harvester provides about 10 pW. The available power from each quadrant may be combined. The externally powered inductive link may achieve a data transmission rate of approximately 50 kb per second. Thus, the load data can be transmitted in less than a minute using an external reader and the internal memory is reset to store new load data. The ASIC implementation may be large voltage CMOS fabrication process, which would help limit the number of discrete components and save area on the PCB.

The output of the sensor under simulated activities of daily living when installed in a knee replacement implanted into a cadaver knee may be used to generate test data fortuning and training algorithms and the ANN. A resulting library of sensor data collected across a large sample of loading and soft tissue balancing scenarios within a large sample of cadaveric specimens may be used to assess sensor accuracy. Furthermore, derived overall statistical measures may be used to train ANN models to reconstruct the original load-time curves and flag abnormal joint mechanics and classify knees by their soft tissue balance.

When external power is supplied in a clinical or lab environment, the sensor is capable of collecting joint loading data at a sampling rate suitable for biomechanical analysis. This enables analysis of joint loading and direct comparison with other techniques, such as inverse dynamic analysis and musculoskeletal modelling. The calibrations performed on isolated implants using simple cyclical loading may not be accurate for sensors implanted into the knee as part of a total knee replacement, and therefore in situ measurements in patients or cadavers can provide validation.

When operating in “autonomous” mode, the sensor does not typically continuously record joint loads, but rather extracted features (peak voltages, period between peaks, integrated voltage over a cycle). A desired application of these sampled data is to use them to reconstruct the original joint loading patterns (the time-dependent magnitude of force and center of force on the proximal tibia). This ability is attractive because it could obviate the need to use the “high resolution” sensing mode which required external power. Artificial neural networks (ANNs) may be used to reconstruct joint forces from the extracted features. The ANN implementation is advantageous because it can transparently correct for non-linear effects and interaction between parameters, which might require significant processing and empirical data analysis using traditional algorithms.

Although a minimal clinically important difference (MCID) for tibiofemoral contact forces has not been agreed upon, computational models have reported minimum detectable changes (MDCs) ranging from 0.245 to 0.66 BW; therefore, RMS errors in reconstructed medial and lateral joint force less than approximately 0.2 BW would be considered sufficiently accurate.

An alternative application of the extracted features would be for classifying the measured joint loads as normal or abnormal, with further classification by the type of aberration detected (e.g., “medial compartment is too lax in extension”). Such classification could be performed by processing the detailed force versus time data but may also be possible by training an ANN to recognize trends in the extracted feature sets. Therefore, additional ANNs may be developed for the this type of classification. This type of classification will depend less on absolute sensor accuracy than the joint load reconstruction would, while still providing important clues for healthcare providers to consider when planning treatments (operative or non-operative) for patients who are dissatisfied with the performance of their replaced knee. This approach, however, first requires a library of extracted feature data obtained from knees with these abnormal loading patterns intentionally introduced.

Characterizing the performance of the integrated sensor requires in vitro experimentation using physiological loading. Experiments are performed using a state-of-the-art six-degrees-of-freedom (6-Dof) joint motion simulator (AMTI VIVO). Protocols were developed for measuring the knee’s neutral path of motion, laxity and motion during simulated activities of daily living, all of which are function of joint contact and soft tissue contributions. Using this testing platform, the alterations of the mechanics of the knee were measured with respect to altered ligament balancing philosophies, modelled using “virtual ligaments” around implant components mounted onto the simulator. A similar approach to was used to measure how the UHMWPE bearing thickness influences knee mechanics by over- and under-stuffing the joint. Using cadavers, experiments were performed to examine the influence of implant design and PCL resection during TKR, and philosophical differences in prosthesis alignment. Considering the intact knee joint, similar techniques were used to examine the influence of soft tissues, such as during simulated PCL and medial ligamentous complex injuries. The capabilities of the VIVO joint motion simulator were enhanced to enable gravitydependent muscle-driven squatting motions, such as those of an Oxford rig, for more realistic joint biomechanical analyses.

The sensor may be paired with Triathlon (Stryker) prosthesis components, for example.

In order to obtain testing sensor data, the sensor is powered via an inductive link with an external coil. This permits measurement of joint loading patterns using the higher sampling rate (non-harvesting) model, with load data continuously streamed to an external receiverover a wireless link. Recorded data is synchronized to time-stamped load and motion data recorded by the mechanical testing system (VIVO), and stored on a separate computer for later analysis and ANN training. The long-term (1000 cycle) simulations are then repeated without an external power supply, and the energy harvesting rates measured. The waveform feature data is not recorded, as this can be subsampled from the higher sampling rate data later, avoiding the need for a priori determination of which features to sample.

The joint is unloaded, the medial parapatellar incision re-opened, and the energy-harvesting load sensor removed and replaced with a different load sensor, identical in shape and size, but constructed using traditional straingauge based load cells which have been carefully calibrated. The incision is closed and the loading scenarios described above are repeated while measuring data from this calibrated sensor. This provides another set of directly- measured joint loads which is used to validate the calibration of the energy-harvesting load sensor. The two sets of loading data are then directly compared to measure the in situ accuracy of the energy-harvesting load sensor.

Shims attached to the energy-harvesting load sensor can be removed and modified in order to increase (+2 mm) or decrease (-2 mm) the thickness of the sensor body, which can then be re-implanted. This alters the amount of ligament pre-strain in the joint, simulating an over- or under-stuffed joint scenario. Shims which are tapered in the frontal plane (3° varus and 3° valgus) and sagittal plane (4°, 7° and 10°) are also be used to alter the medial-lateral compartment soft tissue balance posterior tibial slope of the UHMWPE bearing. Shims can be combined, and fractional-factorial design of experiments used to determine which combinations of shims are considered for each specimen. For each shim configuration, the sensor is re-installed and the aforementioned loading scenarios repeated. In a similar fashion to the joint balance parameter study, loading is repeated with simulated soft tissue injuries. In particular, whether or not to resect the PCL and supplement its contribution using a condylar stabilized or cruciate sacrificing implant is a relevant clinical decision which, if chosen incorrectly, may result in post-operative instability. In specimens where the PCL was left intact (surgeons discretion), PCL insufficiency is simulated by complete transection of the PCL. Loading simulations are repeated and resulting sensor data recorded.

From the cyclical loading experiments, thousands of cycles of electrical signals from the energy harvesting load sensor under varying conditions were recorded, along with corresponding loading and kinematics data from the joint motion simulator. Recorded data is labelled according to the conditions under which it was measured. To simulate the actual sampling characteristics of low-power circuitry employed on-chip, key features are extracted from each loading cycle such as peak magnitude, period, integrated voltage over time, number of peaks per cycle and mean. Recorded data is used to develop and test the ANN. 70% of the data is used for training, 20% for validation, and 10% for testing, although other subset of data may be used. Different ANNs may be be constructed. The ANN is used to reconstruct force patterns (corresponding with what was measured by the sensor and the VIVO) in the time domain based on the recorded key feature data extracted from the cyclical loading experiments, see Figs. 3 and 20. In a similar manner, another ANN may be constructed which classifies the joint behavior as normal (based on the loads measured with the appropriately sized and balanced TKR knee, with an intact PCL) versus abnormal (any of the other scenarios). A third ANN attempts to classify joints even further, based on the conditions which have caused abnormal joint loading (such as the bearing being too thick, the PCL being deficient, or combinations thereof).

The strength of the data transmission signal from sensor is preferably sufficient to permit transmission across reasonable distances (2 m); and inductive powering from the external coil through soft tissues is preferably adequate for high power mode sampling.

The RMS error of the sensor is preferably less than 0.2 BW, when compared with a strain-gauge based sensor under identical in situ loading conditions. The accuracy of time domain loads is reconstructed by the first ANN will be worse (greater than 0.2 BW), but qualitatively still useful for observing loading patterns. Key feature data and the ANN make it possible to accurately classify normal versus abnormal loading patterns (>80% classification accuracy), and that prediction accuracy will increase as more key features are added.

A possible concern is that the ANNs may lack adequate predictive power using only key features from the electrical signal. The key feature data may be supplemented with data from an inertial measurement unit (IMU) worn near the knee, (e.g., nP-BLE52, LSM6DSO). This data is used to train the ANN. Because the ANN could ultimately reside outside of the knee sensor (e.g., in the cloud, in a mobile app or on a computer), integration with data from other sources such as this could be possible provided that key feature data be time-stamped. It is unlikely that the small compliance of the sensor package will influence joint mechanics.

It is therefore an object to provide an orthopedic implant, comprising: a compliant housing; a segmented triboelectric generator configured to produce a plurality of outputs; an interface circuit, configured to receive the plurality of outputs of the triboelectric generator, and to produce: a first set of outputs comprising sensor signals which monotonically increase and decrease according to a compression of the compliant housing along an axis; and at least one second output comprising a filtered power output; and an active processing circuit configured to process the first set of outputs using power from the at least one second output

It is also an object to provide a method of collecting data from an orthopedic implant, comprising: applying a force to a compliant housing implanted adjacent to a bone; producing a plurality of outputs with a segmented triboelectric generator; receiving the plurality of outputs of the triboelectric generator with an interface circuit, and producing: a first set of outputs comprising sensor signals which monotonically increase and decrease according to a compression of the compliant housing along an axis; and at least one second output comprising a filtered power output; and processing the first set of outputs in an active processing circuit using power from the at least one second output.

It is another object to provide an orthopedic implant, comprising: a compliant housing; a segmented triboelectric generator configured to produce a plurality of outputs; an interface circuit, configured to receive the plurality of outputs of the segmented triboelectric generator, and to produce: a first set of outputs comprising sensor signals which increase and decrease corresponding to a compression of the compliant housing along an axis; and at least one second output comprising a filtered power output; and an active processing circuit configured to process the first set of outputs using power from the at least one second output. A corresponding method is provided for collecting data from the orthopedic implant, comprising: applying a force to the compliant housing implanted adjacent to a bone; producing the plurality of outputs with the segmented triboelectric generator; receiving the plurality of outputs of the segmented triboelectric generator with the interface circuit, and producing the first set of outputs and the at least one second output; and processing the first set of outputs in an active processing circuit using power from the at least one second output.

It is a further object to provide a non-transitory computer readable medium for acquiring data from the sensor orthopedic implant, comprising: instructions for controlling a microcontroller to estimate a force vector to the compliant housing implanted adjacent to a bone, based on the plurality of outputs from a segmented triboelectric generator; instructions for controlling the microcontroller to operate in a first mode using power solely derived from the segmented triboelectric generator; and instructions for controlling the microcontroller to operate in a second mode using power derived from an inductively coupled backscatter communication antenna.

It is also an object to provide a method of collecting data from an orthopedic implant, comprising: applying a force to a compliant housing implanted adjacent to a bone; producing a plurality of outputs with a segmented triboelectric generator; receiving the plurality of outputs of the triboelectric generator with an interface circuit, and producing: a first set of outputs comprising sensor signals which increase and decrease corresponding to a compression of the compliant housing along an axis; and at least one second output comprising a filtered power output; and processing the first set of outputs in an active processing circuit using power from the at least one second output.

It is a further object to provide a non-transitory computer readable medium for acquiring data from a sensor orthopedic implant having at least one programmable processor, comprising: instructions for estimating a force vector on a compliant housing implanted adjacent to a bone, based on a plurality of outputs from a segmented triboelectric generator; instructions for operating in a first mode using power solely derived from the segmented triboelectric generator; and instructions for operating in a second mode using power derived from an inductively coupled backscatter communication antenna.

It is also an object to provide a non-transitory computer readable medium for acquiring data from a sensor orthopedic implant, comprising: instructions for estimating a force vector to a compliant housing implanted adjacent to a bone, based on a plurality of outputs from a segmented triboelectric generator; instructions for operating in a first mode using power solely derived from the segmented triboelectric generator; and instructions for operating in a second mode using power derived from an inductively coupled backscatter communication antenna.

The orthopedic implant may comprise a total knee replacement The orthopedic implant may be situated in a tibial tray.

The compliant housing may have a height along the axis of less than 6 mm, less than 5 mm, less than 4 mm, or less than 3 mm. The compliant housing may reduce a force along the axis on the segmented triboelectric generator by less than 70%, less than 60%, less than 55%; less than 50%; less than 45%; less than 40%; less than 35%; or less than 30%. That is the force is distributed between the housing and the segmented triboelectric generator.

The compliant housing may have a set of fenestrations acting as leaf springs. The fenestrations may be optimized by finite element analysis for compliance and durability.

The compliant housing may comprise high molecular weight polyethylene.

The compliant housing comprises TisAhV and polyethylene.

The segmented triboelectric generator may have two segments and the first set of outputs comprises two sensor signals, three segments and the first set of outputs comprises three sensor signals, four segments and the first set of outputs comprises four sensor signals, or a larger number of segments and corresponding sensor signals.

The at least one second output may comprise a single output

The sensor signals may have a range of less than ±5 Volts peak, less than ±4 Volts peak, less than ±3 Volts peak, less than ±3.5 Volts peak, less than ±3 Volts peak, or less than ±2.5 Volts peak. The signal may be processed to represent a biased unipolar signal of, e.g., less than 5 VDC.

The at least one second output may comprise a regulated voltage output.

The plurality of outputs of the segmented triboelectric generator have a maximum voltage exceeding 80 V, 90 V, 100 V, 110 V, 120 V, 130 V, 140 V, or 150 V. The at least one second output may be configured to supply at least 5 pW, 10 pW, 15 pW, 20 pW, 22 pW, 23 pW, 24 pW, 25 pW, or 26 pW. The at least one second output may be configured to supply at least 9 pW from a force along the axis of at least 450 N.

The at least one second output may be configured to produce a regulated voltage within ±10% from a 1 Hz sinusoidal 450 N force applied along the axis.

The interface circuit may present an impedance of at least 200 MQ to each of the plurality of outputs.

The segmented triboelectric generator may be configured to provide 15-140 V peak AC voltages and 0.3-26 pW of apparent power (Vrms x Ims) at 100-2000 N sinusoidal force along the axis.

The segmented triboelectric generator may be configured to provide over 50 V peak AC voltage under a load of 500 N, or to provide 60 V peak AC voltage under a load of 600 N. The segmented triboelectric generator may be configured to provide 14 pW apparent power from the first set of outputs under a load of 525 N. The segmented triboelectric generator may be configured to provide 17 pW apparent power from the first set of outputs under a load of 600 N.

The active processing circuit may comprise an analog to digital converter (ADC), a sample hold circuit (e.g., SHA), a peak hold circuit, an integrator, a microprocessor (pP), a microcontroller (pC), an application specific integrated circuit (ASIC), non-volatile memory (e.g., NV-RAM), and/or a timer. The active processing circuit may comprise a statistical signal processor configured to determine a mean value, a standard deviation, a variance, and/or a significant difference. The active processing circuit may comprise a signal processor configured to indicate an alarm state. The active processing circuit may comprise an ARM M0 or MO+core.

The orthopedic implant may further comprise a near field communication (NFC) transceiver, RF-ID circuit, and/or a backscatter communication transmitter. The orthopedic implant may further comprise an inductive coil configured to receive external power and/or perform inductively coupled data communication.

The orthopedic implant may further comprise a supercapacitor and/or rechargeable battery configured to store power from the filtered power output.

The active processing circuit may be configured to jointly process the set of first outputs to analyze changes in the axis over time.

The segmented triboelectric generator may have a normal axis, and the active processing circuit may be configured to jointly process the set of first outputs to dynamically analyze changes in the axis with respect to the normal axis over time.

The active processing circuit may have a first mode of operation powered solely by the segmented triboelectric generator, and a second mode of operation powered externally, e.g., through inductively transferred power.

Each segment of the segmented triboelectric generator may comprise a first surface separated by a gap from a second surface, the first surface comprising a metal and the second surface comprising a dielectric. The first surface may comprise triangular micropatterned ridges of 100-300p, e.g., 10Op. The metal may comprise titanium metal. The metal may comprise titanium having regularly spaced triangular micropatterned ridges of 10Op. The second surface may comprise polydimethylsiloxane (PDMS) or styrene ethylene butadiene copolymer (SEBS). The dielectric may be formed on a titanium surface. The dielectric may be spin-coated on the metal.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 A shows a schematic of the total knee replacement (TKR) system implanted within the knee.

Figs. 1 B and 1 C show an enlarged view of the package with the harvesters.

Fig. 1 D shows a detail view of the triboelectric generator.

Fig. 1 E shows a schematic diagram of a signal processor.

Fig. 2A shows a 3D design view of a prototype of the polyethylene packaged harvesters.

Fig. 2B shows an upper packaged harvester’s components.

Fig. 20 shows a lower upper packaged harvester’s components.

Fig. 3 shows a voltage signal from the harvester at a force of 500 N and 600 N.

Fig. 4 shows a measured peak harvester voltage as a function of applied force.

Fig. 5 shows an architectural block diagram of the prototype electronic system for small range of cyclic loads.

Fig. 6 shows a schematic of attenuator I consisting of a two-stage LC filter to reduce the peak voltages of the harvested signal.

Fig. 7 shows a schematic of attenuator II along the signal path consisting of a single-stage LC filter and diode rectifier.

Fig. 8 shows a PCB prototype of a frontend electronic system for a small range of cyclic loads.

Fig. 9 shows an experimental setup for measurements.

Fig. 10 shows a PCB and ORCAD results for the output of attenuator I.

Fig. 11 shows a PCB and ORCAD results for the regulator output

Fig. 12 shows a PCB and ORCAD results for the ADC input data.

Fig. 13 shows a PCB and ORCAD results for the output of the ADC.

Fig. 14 shows an architectural block diagram of the sensor circuitry for wide range of cyclic loads.

Fig. 15 shows a schematic of the linear and non-linear attenuators within the sensor circuitry.

Fig. 16 shows a peak linear and non-linear attenuator output vs. peak harvester output.

Fig. 17 shows a PCB prototype of frontend electronic system for wide range of cyclic loads.

Fig. 18A shows output waveforms of the harvester output sensor circuitry with 105 V peak voltage Vin.

Fig. 18B shows the amplifier input voltage Vi of the harvester output sensor circuitry.

Fig. 18C shows the nonlinear attenuator output voltage Vout of the harvester output sensor circuitry.

Fig. 18D shows the linear attenuator output voltage of the harvester output sensor circuitry.

Fig. 19 shows the voltage resolution of the proposed sensor circuitry for various intervals of the harvester voltage.

Fig. 20 shows Extraction of knee load profiles through on-chip and cloud processing. Fig. 21 shows the variation of power with load for the left and right energy harvesters and their parallel connections measured across a 220 MQ resistance.

Fig. 22 shows experimental and theoretical voltage output to periodic load of 2000 N at 1 Hz frequency.

Fig. 23A shows a 3D schematic of a triboelectric harvester area shaped like the tibial tray divided into halves.

Fig. 23B shows a 3D schematic of a triboelectric harvester area shaped like the tibial tray divided into quadrants. Fig. 24 shows a schematic drawing of a system employing the triboelectric energy harvester.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

I. HARVESTER DESIGN AND FABRICATION

Vertical contact mode triboelectric energy harvester has been used to generate the AC voltage signal from cyclic contact and separation motions. For providing the necessary contact and separation motions, the parts of the harvester are fixed inside a mechanical spring-controlled housing. A polyethylene packaged triboelectric harvester prototype was used to generate the AC voltage signal from cyclic contact and separation motions. The package serves as housing for the generators and the electronic system together. The prototype’s overall geometry was based on the perimeter shape of a size 7 tibial tray (Triathlon, Stryker, Kalamazoo, Ml) that adds approximately 16 mm to the overall height of the tibial component. The upper tribolayer of the harvester, also a metal electrode, was CNC machined from micro-patterned titanium (100 pm sawtooth ridge) and the lower tribolayer was fabricated by spincoating PDMS mixtures on a back electrode that was machined from a flat titanium.

To make PDMS, first, the titanium electrodes are cleaned with acetone and distilled water in an ultrasonic cleaner. Then, the PDMS elastomer base and the curing agent are mixed in 10: 1 weight ratio. The mixture is stirred thoroughly and degassed in a vacuum chamber. After degassing, the PDMS paste is spin coated at 500 RPM for 36 seconds on the surface of the flat titanium. Finally, the PDMS coated titanium is cured at 90°C for 45 minutes on a hot plate. The upper and the lower titanium parts closely follow the shape of a standard tibial tray. The TEG and its package were tested for 10,000 cycles and maintained stable outputs under equivalent body loads of 1000 N-2000 N. The assembly of the upper and the lower harvester parts inside the package are shown in Fig. 2.

The designed harvester produces an AC signal at approximately 1 Hz with a peak voltage that is proportional to the force applied to the plates of the harvester. In this packaged harvester prototype, a single harvester when connected to its optimum resistance (—220 MQ) can provide 15-140 V peak AC voltages and 0.3-26 pW of apparent power (Vrms x Irms) at 100-2000 N sinusoidal force. For example, as shown in Fig. 3 the harvester produces an AC voltage signal with a peak voltage of 53 V when the packaged prototype is under 500 N of sinusoidal loads. This peak voltage increases to 60 V when the applied force is 600 N. If two harvesters are connected in parallel, 14 pW apparent power is generated at 525 N. This power increases to 17 pW at 600 N. Fig. 4 plots the measured peak harvester voltage for various forces to further demonstrate the dependence of voltage signal on applied force.

II. FRONTEND SENSOR CIRCUITRY

A. Small Range of Cyclic Loads The architectural block diagram of the frontend electronic system for the smart knee implant is shown in Fig. 5.

This system comprises an Attenuator I, a two-stage LC filter for attenuating the high voltages from the harvester, which is common for both the signal and power paths. The signal path has an additional Attenuator II (consisting of a single stage LC filter and a diode) to further condition the data signal, which is digitized by a successive approximation register (SAR) analog-to-digital converter (ADC). The ADC converts the attenuated input analog signal into digital bits for monitoring the knee load. During testing, the digital output from the ADC is read using an STMicroelectronics evaluation board STM32F407. The power path has a rectifier and regulator to extract the supply voltage from the harvested signal. The rectifier converts the output AC signal of the two-stage LC filter into a DC signal, which is further stabilized and regulated by the regulator. The system can be entirely powered through the power generated by the harvesters without any external bias voltage. This proposed system has been fabricated on a printed circuit board (PCB) in order to demonstrate the feasibility.

The high peak voltages in the harvested signal are attenuated into low peak voltages through the Attenuator I, a two-stage LC filter, as shown in Fig. 6. This filter also acts as an impedance matching block to maximize power transfer from the load. The values of the circuit elements are also indicated in the figure. These values are chosen to achieve the desired attenuation (approximately by a factor of 14) while ensuring high power efficiency. The equivalent series resistance (ESR) is considered during the design process for the inductors and capacitors.

A diode rectifier is utilized along the power path to convert the output AC signal from the two-stage LC filter into a DC voltage. This DC voltage is then passed through a linear dropout regulator to produce a stabilized voltage free from the variations in the input voltage. Along the signal path, an Attenuator II (single stage LC filter and a diode) is used at the output of the two-stage LC filter, as shown in Fig. 7. This additional attenuator is needed since the ADC input signal should be in the range of 0.3 V to 2.8 V whereas the output of the two-stage LC filter lies in the range of -6 V to 6 V. The output of this attenuator is the input data to the SAR ADC for load monitoring. As the input data signal frequency is approximately 1 Hz, the clock frequency for the ADC is chosen to be 10 kHz which provides sufficient accuracy.

B. Fabricated PCB Prototype

The frontend electronic system described above is fabricated on a PCB, as shown in Fig. 8. Large form factor passive devices and commercially available rectifier, regulator, ADC and diodes are used for this prototyping. Miniature components may be used in production.

1) An experimental setup was built in order to generate the voltages using the triboelectric generator (TEG) is depicted in Fig. 9. It consists of an MTS 858 Servo Hydraulic Test System for conducting cyclic axial load, and a FlexTest controller for the amplitude tuning. The MTS has a built-in load cell for measuring the applied force. The generated voltage signal is captured using a Keithley M6514 and an ExceLINX program.

2) Measurement Results: Table I lists the PCB experimental results at various stages of the system when tested with the triboelectric harvester. It also compares these test results with the ORCAD simulation results. In this experiment, the harvester output signal has a peak voltage of 53 V. A preliminary package is used for testing the harvester.

TABLE I

COMPARISON OF PCB TEST AND ORCAD SIMULATION RESULTS FOR

53 V PEAK HARVESTED SIGNAL

Fig. 10 compares the measured output of the Attenuator I, two-stage LC filter with the ORCAD simulation result The PCB output signal has a peak of 4 V as compared to the 4.1 V peak signal from the ORCAD simulation.

Fig. 11 compares the measured output of the regulator with the simulation result. The PCB output is a 2.4 V DC signal whereas the ORCAD simulation output is a 2.38 V DC signal.

Fig. 12 compares the measured data input to the ADC (output of the Attenuator II) with the ORCAD simulation result The measured signal has a range of 0.1 V to 1.1 V. The ORCAD simulation also produces a signal in the similar range.

Finally, Fig. 13 compares the measured output of the ADC with the ORCAD simulation result. In this case, the analog input signal to the ADC is approximately 0.9 V and the output digital data is identical in both cases (11111100). Note that the digitized data is wirelessly transmitted to an external reader device via inductive coupling. This data telemetry is powered an external reader, e.g., NFC or RFID, e.g., IS0 10536, IS0 11784; IS0 11785; IS0 14443; IS0 18000; IS0 18001 , ISO 24753, etc.

These test results validate the proposed design as PCB test results and circuit simulation results are sufficiently close. The slight mismatch is primarily due to the disparity between the input signals in each case. Specifically, for simulations, the input signal is produced by the electrical model of the harvester, which approximates the actual signal generated by the harvester.

3) Power Consumption: The overall system consumes approximately 5.35 pW power. The regulator and SAR ADC contribute the most to overall power consumption. Two harvesters connected in parallel can generate 9 pW power at an applied force of 450 N. Thus, the proposed frontend electronic circuitry can be entirely self-powered.

C. Wide Range of Cyclic Loads

The PCB prototype described above can monitor loads in the range of 450-650 N, corresponding to peak voltages of 45 V to 65 V. In practice, the harvester output ranges from 10 V peak to 150 V peak voltage depending upon the applied force. Thus, linear LC filter-based attenuation is not sufficient since the filter output would have a peak voltage of 11.5 V when the harvester output signal has peak voltage of 150 V. Furthermore, LC attenuation requires relatively large passive circuit elements due to low operating frequency of approximately 1 Hz. The architectural block diagram of the sensor circuitry designed for wide range of cyclic loads is shown in Fig.

14. In this design approach, the signal and power paths do not share a passive filter or attenuator. Instead, a nonlinear attenuator is used for the power path, followed with rectification and regulation. For the signal path, a capacitive divider based linear attenuator is used with significantly lower capacitors. These blocks are explained below. ATTENUATOR

1 ) Linear and Non-Linear Attenuator The improved electronic circuitry comprises of two attenuators; a linear attenuator along the signal path and a non-linear attenuator along the power path, as shown in Fig. 15. The input impedance of the signal path is significantly larger to minimize current flow into the signal path. The input impedance of the non-linear attenuator is adjusted to match the input impedance of the harvester (—220 MO) to maximize power transfer. Table II lists the values of all the passive components from these two attenuators. The equivalent series resistance (ESR) is also considered for all the capacitors. Along the power path, the signal from the harvester is attenuated through the capacitive divider, C1 and C2 (attenuation factor is 3.5). This signal is passed through the diode rectifier, D2, and capacitor, C6, to provide the biasing for the amplifier, LM358-N, incorporated in the design. The signal from the capacitive divider is attenuated again through the capacitive divider, C3 and C4 (attenuation factor is 1.5), and diode, D1. This step ensures that the input signal to the opamp, V i, is always less than the bias signal, V+. C5 is a feedback capacitor that provides the desired attenuation between the input and output of the amplifier, as described below. Referring to Fig. 15 and applying KCL, the following expressions are obtained, where Z1 , Z2, Z3, Z4 and Z5 are the impedances of the respective capacitances. Note that the current of the diodes D1 and D2 is neglected in (1) and (2). For the amplifier, the characteristic equation is, i-ii — A ( V7 — Vcw) , (3) where A is the gain of the amplifier. Replacing (3) in (2) and rearranging yields, where k1 and k2 are functions of impedances Z1 to Z5 and are constant at constant frequency. The amplifier, LM358-N, is a voltage controlled current source where the transconductance (and therefore the open loop gain A) changes with the input voltage. Specifically, as the input voltage increases, the gain decreases, thereby achieving non-linear attenuation at the output of the amplifier. The accuracy of (4) is evaluated by comparing the analytic results with the simulated values (see Fig. 16) for different harvester output voltages. The average error is approximately 1.71%.

Along the signal path, linear attenuation is achieved via C7 and C8 (attenuation factor is approximately 62.5), followed with a rectification stage consisting of D3 and C9. This attenuation ensures that the input data to ADC is always less than the supply voltage V dd of the ADC, which is determined by the power path. Note that the supply voltage of the ADC varies in the range of 1.8 V to 2.8 V depending upon the peak voltage of the harvested signal.

2) Measurement Results: The frontend electronic system for wide range of cyclic loads is fabricated on a PCB, as shown in Fig. 17. The voltages at the output of the linear and nonlinear attenuators are illustrated in Fig. 18A for a harvested signal with a peak voltage V in of 105 V. The input voltage of the amplifier, V i has a peak voltage of 3.5 V, as shown in Fig. 18B. The output voltage of the non-linear attenuator, V out has a peak value of 3.1 V, as shown in Fig. 18C for both simulations and measurements. This signal is converted into a DC voltage for the ADC bias through the rectifier and regulator. The output of the linear attenuator (input data signal for the ADC) has a peak voltage of 1.6 V, as shown in Fig. 18D for both simulations and measurements. The measured data is sufficiently close to the simulation results, particularly for the peak voltages.

The peak output voltages of the linear and non-linear attenuator for different harvested voltages are also plotted in Fig. 16. The output peak voltages for the non-linear attenuator range from 2.5 V to 3.5 V, whereas the output peak voltages of the linear attenuator lie in the range of 0.16 V to 2.4 V. The ADC works properly for this range of bias voltage and input data signals, which are within its resolution range. Thus, the overall circuit can work for a wide range of harvester signals (from 10 V to 150 V peak voltage).

Finally, the minimum change in the harvester voltage that can be sensed by the circuit with sufficient accuracy (i.e., voltage resolution) is characterized. This result is shown in Fig. 19 for various intervals of the peak harvester voltage. The resolution ranges from approximately 12 mV to 21 mV. The worst-case resolution corresponds to harvester voltages greater than 111 V. Also, the difference between the digitized ADC output and corresponding harvester voltage is calculated to evaluate accuracy. For this comparison, the ADC output data is multiplied by the overall attenuation factor of the signal path, which is approximately 62.5. The maximum error is 3.48% whereas the average error is 2.83%.

3) Power Consumption: The overall power consumed by the circuit is approximately 5.1 pW. The regulator and SAR ADC contribute the most to overall power consumption. According to this result, the proposed frontend electronic circuitry can be entirely self-powered by a single harvester for cyclic loads 600 N to 2000 N, corresponding to harvester output peak voltages of 60 V to 140 V. Two harvesters connected in parallel can generate 41 -67 W of power at 1000-1500 N of sinusoidal loads. If the load is less than 600 N, the parallel connection of the harvesters can be utilized or supercapacitors can be used to store excess energy. Note that less power is consumed as compared to the previous design due to the much smaller passive devices with significantly lower ESR values.

The system comprises an energy harvesting circuit that supplies voltage to operate the signal processing and data logging circuitry. For example, the data logging circuit may be an ARM MO or M0+ core, such as STM32F0 or STM32L0.

Efficient power management techniques enabled by an embedded supercapacitor or rechargeable battery ensure high conversion efficiency and sustained operation of the system on intermittent power. The statistical measures of the loading during a single activity may be extracted from the sensor signal and logged into non-volatile memory. The power required for feature extraction and data logging is provided solely by the generator (except when external power is available). The logged data is periodically uploaded to an external reading device through an inductive link between a coil located in the implant and a coil in the external device. During these transmissions, the inductive link is powered by an external device, e.g., a smartphone, RF-ID or NFC interrogator, or a dedicated reader device. The accuracy of the sensor may be verified through comparison against a gold standard, in this case strain gauge load cells, exposed to mechanical loading by a materials testing machine. Underspecified conditions, the load on the tibial plateau or other surface may be calibrated with an external load cell such as in footwear.

An artificial neural network (ANN) may be trained with sensor data to reproduce measured joint loads from simpler measured features, classify joints as normal or abnormal, and further classify by type of irregularity. The feature data are extracted from the raw voltage signals to improve the accuracy of the ANNs.

IV. CONCLUSION

Continuous and optimal monitoring of the load is a promising technique in improving the design of knee implants and detect unsafe loads. This paper presents a frontend sensor circuitry for both small and wide range of loads. A prototype PCB and measurement results with the harvester are demonstrated for small range of loads. These test results validate the proposed approach as PCB test results and circuit simulation results sufficiently match with high accuracy. Simulation results for the wide range of loads are also presented. The front-end electronic circuitry is powered entirely by the harvested power without requiring any external supply or bias voltage. Triboelectric energy harvesting is therefore a promising technique for self-monitoring the load inside knee implants.

Thus, in one aspect each system and method described above, combinations, and subcombinations thereof. All such permutations and combinations are intended to fall within the scope of the present disclosure. While the invention has been disclosed in connection with the preferred embodiments shown and described in detail, various modifications and improvements thereon will become readily apparent to those skilled in the art. Accordingly, the spirit and scope of the present invention is not to be limited by the foregoing examples, but is to be understood in the broadest sense allowable by law.

REFERENCES

Each reference cited herein is expressly incorporated herein by reference in its entirety for all purposes. Almouahed, S., M. Gouriou, C. Hamitouche, E. Stindel, and C. Roux, ‘The use of piezoceramics as electrical energy harvesters within instrumented knee implant during walking,” IEEE/ASME Trans. Mechatronics, vol. 16, no. 5, pp. 799-807, Oct. 2011.

Almouahed, S., M. Gouriou, C. Hamitouche, E. Stindel, and C. Roux, “Self-powered instrumented knee implant for early detection of postoperative complications,” in Proc. Annu. Int. Conf. IEEE Eng. Med. Biol., Aug. 2010, pp. 5121-5124.

Almouahed, Shaban, Chafiaa Hamitouche, Eric Stindel, and Christian Roux. New trends in instrumented knee prostheses. 2008 3rd International Conference on Information and Communication Technologies: From Theory to Applications, ICTTA, 2008.

Almouahed, Shaban, Chafiaa Hamitouche, Eric Stindel, and Christian Roux. Optimization of an instrumented knee implant prototype according to invivo use requirements. In Point-of-Care Healthcare Technologies (PHT), 2013 IEEE, pages 5-8. IEEE, 2013.

American Joint Replacement Registry. Annual report 2014, 2014.

Arami, Arash, JeanRomain Delaloye, Hossein Rouhani, Brigitte M Jolies, and Kamiar Aminian. Knee implant loosening detection: a vibration analysis investigation. Annals of biomedical engineering, 46(1 ):97- 107, 2018.

Arami, Arash, Matteo Simoncini, Oguz Atasoy, Shafqat Ali, Willyan Hasenkamp, Arnaud Bertsch, Eric Meurville, Steve Tanner, Philippe Renaud, Catherine Dehollain, et al. Instrumented knee prosthesis for force and kinematics measurements. IEEE transactions on automation science and engineering, 10(3):615-624, 2013.

Askari, Hassan, Ehsan Asadi, Zia Saadatnia, Amir Khajepour, Mir Behrad Khamesee, and Jean Zu. A flexible tubebased triboelectric-electromagnetic sensor for knee rehabilitation assessment. Sensors and Actuators A: Physical, 279:694-704, 2018.

Atmeh, Mohammad. "Efficient Energy Harvesting With Applications In Vibration And Implantable Biosensors." (2022).

Australian Orthopaedic Association National Joint Replacement Registry. Annual report, 2015.

Baronio, Gabriele, Vincenzo Luciano, Emilio Sardini, and Mauro Serpelloni. "Design considerations of an electromechanical generator implanted in human total knee prosthesis." International Journal of Mechatronics and Manufacturing Systems 1 6, no. 3 (2013): 270-284.

Barrios, Joaquin, and John Willson. Minimum detectable change in medial tibiofemoral contact force parameters: derivation and application to a load-altering intervention. Journal of applied biomechanics, 33(2): 171- 175, 2017.

Bergmann, G., A Bender, F Graichen, J Dymke, A Rohlmann, A Trepczynski, MO Heller, and I Kutzner. Standardized loads acting in knee implants. PLoS One, 9(1 ):e86035, 2014.

Bourns, RLB0913 Series Radial Lead Inductor, Datasheet, Riverside, CA, USA, 2011. Broberg, Jordan S., Silvio Ndoja, Steven J MacDonald, Brent A banting, and Matthew G Teeter. Compari son of contact kinematics in posterior-stabilized and cruciate-retaining total knee arthroplasty at long-term followup. The Journal of arthroplasty, 2019.

Brock, J Logan, and Atul F Kamath. Obesity and racial characteristics drive utilization of total joint arthroplasty at a younger age. Journal of clinical orthopaedics and trauma, 10(2): 334-339, 2019.

Callahan, Christopher M., Brenda G Drake, David A Heck, and Robert S Dittus. Patient outcomes following tricompartmental total knee replacement: a meta-analysis. Jama, 271 (17): 1349-1357, 1994.

Canadian Institute for Health Information. Hip and knee replacements in Canada, 2017-2018: Canadian joint replacement registry annual report., 2019.

Crottet, Denis, Thomas Maeder, Daniel Fritschy, Hannes Bleuler, LP Nolte, and Ion P Pappas. Development of a force amplitude and location sensing device designed to improve the ligament balancing procedure in tka. IEEE Transactions on Biomedical Engineering, 52(9): 1609-1611, 2005.

D’Lima DD, and CW Colwell. Intraoperative measurements and tools to assess stability. Journal of the American Academy of Orthopaedic Surgeons, Suppl 1 :S29-S32, 2017.

D’Lima, D., C Townsend, S Arms, B Morris, and CJ Colwell. An implantable telemetry device to measure intraarticular tibial forces. Journal of Biomechanics, 38(2):299-304, 2005.

D’Lima, D., S Patil, N Steklov, J Slamin, and CJ Colwell. In vivo knee moments and shear after total knee arthroplasty. Journal of Biomechanics, 40(1 Suppl 1 ):S11-S17, 2007.

D’Lima, D., S Patil, N Steklov, J Slamin, and CJ Colwell. Tibial forces measured in vivo after total knee arthroplasty. Journal of Arthroplasty, 21 (2):255-262, 2006.

Dayan, Ortal, Irina Spulber, Amir Eftekhar, Pantelis Georgiou, Jeroen Bergmann, and Alison McGregor. Applying emg spike and peak counting for a real-time muscle fatigue monitoring system. In 2012 IEEE Biomedical Circuits and Systems Conference (BioCAS), pages 41-44. IEEE, 2012.

Dorr, Lawrence D., and Richard A Boiardo. Technical considerations in total knee arthroplasty. Clinical orthopaedics and related research, (205):5-11, 1986.

Dusad, Anand, Sofia Pedro, Ted R. Mikuls, Curtis W. Hartman, Kevin L. Garvin, James R. O’Dell, and Kaleb Michaud. Impact of total knee arthroplasty as assessed using patient-reported pain and health-related quality of life indices: Rheumatoid arthritis versus osteoarthritis. Arthritis and Rheumatology, 67(9): 2503- 2511 , 2015.

Elwell, Josie A, George S Athwal, and Ryan Willing. Characterizing the tradeoff between range of motion and stability of reverse total shoulder arthroplasty. Journal of Shoulder and Elbow Surgery, 2021 .

Fregly, BJ, TF Besier, DG Lloyd, SL Delp, SA Banks, MG Pandy, and DD D’Lima. Grand challenge competition to predict in vivo knee loads. Journal of Orthopaedic Research, 30(4):503-513, 2012.

Galley, Alexandre, Samira Vakili, and Ryan Willing. Developmentof a muscle actuator system. In The Canadian Society for Mechanical Engineering, Online Meeting, Charlottetown, PEI, Canada. Canadian Society for Mechanical Engineering, 2021. Gardinier, Emily S, Kurt Manal, Thomas S Buchanan, and Lynn SnyderMackler. Minimum detectable change for knee joint contact force estimates using an emg-driven model. Gait & posture, 38(4): 1051- 1053, 2013.

Gouma, Perena, Krithika Kalyanasundaram, Xiao Yun, Milutin Stanacevic, and Lisheng Wang. Nanosensor and breath analyzer for ammonia detection in exhaled human breath. IEEE Sensors Journal, 10(1 ):49-53, 2009.

Guo, H. et al., “A highly sensitive, self-powered triboelectric auditory sensor for social robotics and hearing aids,” Sci. Robot, vol. 3, no. 20, pp. 1-9, 2018, Art. no. eaat2516.

Harpe, Pieter, Hao Gao, Rainier van Dommele, Eugenio Cantatore, and Arthur van Roermund. 21 .2 a 3nw signal acquisition ic integrating an amplifier with 2.1 nef and a 1.5 fj/conv step adc. In 2015 IEEE International SolidState Circuits Conference (ISSCC) Digest of Technical Papers, pages 1-3. IEEE, 2015.

Heinlein, B., F Graichen, A Bender, A Rohlmann, and G Bergmann. Design, calibration and preclinical testing of an instrumented tibial tray. Journal of Biomechanics, 40(Suppl 1 ):S4-S10, 2007.

Heinlein, B., I Kutzner, F Graichen, A Bender, A Rohlmann, A Halder, A Beier, and G Bergmann. Esb clinical biomechanics award 2008: Complete data of total knee replacement loading for level walking and stair climbing measured in vivo with a follow-up of 610 months. Clinical Biomechanics, 24:315-326, 2009.

High Operating Temperature Radial Leaded Multilayer Ceramic Capacitors for Automotive Applications, 50 VDC, 100 VDC, 200 VDC, Datasheet, Vishay IntertechnoL, Malvern, PA, USA, 2016.

Holmberg, J., L. Alexander, Rajesh Rajamani, and Joan E. Bechtold. "Battery-less wireless instrumented knee implant." Journal of Medical Devices 7, no. 1 (2013).

Hossain, Nabid Aunjum, Geofrey George Yamomo, Ryan Willing, and Shahrzad Towfighian. Characterization of a packaged triboelectric harvester under simulated gait loading for total knee replacement. IEEE/ASME Transactions on Mechatronics, 2021. IEEE/ASME Transactions on Mechatronics 26, no. 6 (2021): 2967-2976. doi: 10.1109/TMECH.2021.3049327.

Hossain, Nabid Aunjum, Geofrey George Yamomo, Ryan Willing, and Shahrzad Towfighian. "Effect of dielectric material and package stiffness on the power generation in a packaged triboelectric energy harvesting system for total knee replacement" Journal of Biomechanical Engineering 143, no. 10 pp. 101009 (2021).

Hossain, Nabid Aunjum, Mir Jalil Razavi, and Shahrzad Towfighian. Analysis of mechanical deformation effect on the voltage generation of a vertical contact mode triboelectric generator. Journal of Micromechanics and Microengineering, 30(4), 2020.

Ibrahim, A., G. Yamomo, R. Willing, and S. Towfighian, “Parametric study of a triboelectric transducer in total knee replacement application,” Journal of Intelligent Material Systems and Structures, 32(1): 16-28, 2021.

Ibrahim, A., M. Jain, E. Salman, R. Willing, and S. Towfighian, “A smart knee implant using triboelectric energy harvesters,” Smart Mater. Struct, vol. 28, no. 2, Feb. 2019, Art. no. 025040.

Ibrahim, Alwathiqbellah, Abdallah Ramini, and Shahrzad Towfighian. Experimental and theoretical investigation of an impact vibration harvester with triboelectric transduction. Journal of Sound and Vibration, 416: 111 -124, 2018. Ibrahim, Alwathiqbellah, Geofrey Yamomo, Ryan Willing, and Shahrzad Towfighian. Analysis of a triboelectric energy harvester for total knee replacements under gait loading. In Active and Passive Smart Structures and Integrated Systems XII, volume 10967, page 109671 D. International Society for Optics and Photonics, 2019.

Ibrahim, Alwathiqbellah, Manav Jain, Emre Salman, Ryan Willing, and Shahrzad Towfighian. Feasibility of triboelectric energy harvesting and load sensing in total knee replacement. In ASME 2018 Conference on Smart Materials, Adaptive Structures and Intelligent Systems, vol. 51951 , pages V002T07A016-V002T07A016. American Society of Mechanical Engineers, 2018.

Ibrahim, Alwathiqbellah, Shahrzad Towfighian, and Nael Barakat "Design Optimization of a Triboelectric Energy Harvesting Mechanism for Knee Implants." Engineering Letters 29, no. 3 (2021).

Institut fur Qualitatssicherung und Transparenz im Gesundheitswesen. Qualita tsreport 2015, 2015.

Jain, M., A. Ibrahim, E. Salman, M. Stanacevic, R. Willing, and S. Towfighian, “Frontend electronic system for triboelectric harvester in a smart knee implant,” in Proc. IEEE 62nd Int Midwest Symp. Circuits Syst (MWSCAS), Aug. 2019, pp. 386-389.

Jain, Manav, Nabid A Hossain, Shahrzad Towfighian, Milutin Stanacevic, and Emre Salman. System prototype for a triboelectric harvester in a smart knee implant. In Active and Passive Smart Structures and Integrated Systems IX, Proc. SPIE, vol. 11376, Apr. 2020, Art. no. 113761 F.

Jain, Manav, Nabid Aunjum Hossain, Shahrzad Towfighian, Ryan Willing, Milutin Stanacevic, and Emre Salman. "Self-Powered Load Sensing Circuitry for Total Knee Replacement." IEEE Sensors Journal 21, no. 20 (2021): 22967-22975.

Jian, J., and M. Stanacevic. Adaptive transmitting coil array for optimal power transfer in deeply implanted medical devices. In IEEE International Symposium on Circuits and Systems (ISCAS’16), pages 2030- 2033, 2016.

Karimi, Yasha, Adam Khalifa, Webert Montlouis, Milutin Stanacevic, and Ralph Etienne-Cummings. Coil array design for maximizing wireless power transfer to submm sized implantable devices. In IEEE Biomedical Circuits and Systems Conference (BioCAS’17), pages 1-4, 2017.

Karimi, Yasha, Yingkan Lin, Gagan Jodhani, Milutin Stanacevic, and Pelagia-Irene Gouma. Single exhale biomarker breathalyzer. Sensors, 19(2):270, 2019.

Kaufman, Kenton R., Nebojsa Kovacevic, Steven E Irby, and Clifford W Colwell. Instrumented implant for measuring tibiofemoral forces. Journal of biomechanics, 29(5):667-671, 1996.

Khalifa, A., Y. Liu, Y. Karimi, Q. Wang, A. Eisape, M. Stanac evic', N. Thakor, Z. Bao, and R. Etienne Cummings. The microbead: A 0.009 mm ³ implantable wireless neural stimulator. IEEE Transactions on Biomedical Circuits and Systems, 13(5):971 -985, 2019.

Khalifa, Adam, Yasha Karimi, Qihong Wang, Sahithi Garikapati, Webert Montlouis, Milutin Stanacevic, Nitish Thakor, and Ralph Etienne-Cummings. The microbead: A highly miniaturized wirelessly powered implantable neural stimulating system. IEEE transactions on biomedical circuits and systems, 12(3):521 -531 , 2018. Khan, Muhammad Ahmed, Mauro Serpelloni, and Emilio Sardini. "Optimized power harvesting module for an autonomous sensor system implanted in a total knee prosthesis." In 2017 IEEE International Instrumentation and Measurement Technology Conference (I2MTC), pp. 1-6. IEEE, 2017.

Kutzner, I., A. Trepczynski, M. Heller, and G. Bergmann, “Knee adduction moment and medial contact forcefacts about the correlation during gait,” J. Biomech., vol. 8, Dec. 2012, Art. no. e81036.

Kutzner, I., B Heinlein, F Graichen, A Bender, A Rohlmann, A Halder, A Beier, and G Bergmann. Loading of the knee joint during activities of daily living measured in vivo in five subjects. Journal of Biomechanics, 43:2164-2173, 2010.

Lajnef, Nizar, Niell G Elvin, and Shantanu Chakrabartty. A piezo-powered floating-gate sensor array for longterm fatigue monitoring in biomechanical implants. IEEE Transactions on Biomedical Circuits and Systems, 2(3):164- 172, 2008.

Lange, Hans-E., Dennis Hohlfeld, Rainer Bader, and Daniel Kluess. "A piezoelectric energy harvesting concept for an energy-autonomous instrumented total hip replacement." Smart Materials and Structures 29, no. 11 (2020): 115051.

Le, D. H., S. B. Goodman, W. J. Maloney, and J. I. Huddleston, “Current modes of failure in TKA: Infection, instability, and stiffness predominate,” Clin. Orthopaedics Rel. Res., vol. 472, no. 7, pp. 2197-2200, 2014.

Lee, B., and M. Ghovanloo. An overview of data telemetry in inductively powered implantable biomedical devices. IEEE Communications Magazine, 57(2)74-80, 2019.

Li, X., and Y. Sun, “WearETE: A scalable wearable E-textile triboelectric energy harvesting system for human motion scavenging,” Sensors, vol. 17, no. 11, pp. 2649-2663, 2017.

Liu, H., H. Rui, Y. Lu, Y. Wang, E. Salman, and J. Liang. Boosting the Efficiency of a Footstep Piezoelectric Stack Energy Harvester by Using the Synchronized Switch Technology. Journal of Intelligent Material Systems and Structures, 30(6), 2019.

Lombardi, AV Jr., KR Berend, and JB Adams. Why knee replacements fail in 2013: patient, surgeon, or implant? The Bone and Joint Journal, 96B (11 Supple A):101 -4, 2014.

Luciano, V., E. Sardini, M. Serpelloni, and G. Baronio. "An energy harvesting converter to power sensorized total human knee prosthesis." Measurement Science and Technology 25, no. 2 (2014): 025702.

Luciano, Vincenzo, Emilio Sardini, Mauro Serpelloni, and Gabriele Baronio. Analysis of an electromechanical generator implanted in a human total knee prosthesis. In Sensors Applications Symposium (SAS), 2012 IEEE, pages 1-5. IEEE, 2012.

Marioli, D., E. Sardini, and M. Serpelloni, “Power harvesting integrated in a knee implant for autonomous sensors implanted in human body,” in Proc. IEEE Int. Symp. Med. Meas. Appl. (MeMeA), May 2015, pp. 462-466.

Mathur, Gaurav, Peter Desnoyers, Paul Chukiu, Deepak Ganesan, and Prashant Shenoy. Ultralow power data storage for sensor networks. ACM Transactions on Sensor Networks (TOSN), 5(4):33, 2009. Mohsen Safaei, R Michael Meneghini, and Steven R Anton. Parametric analysis of electromechanical and fatigue performance of total knee replacement bearing with embedded piezoelectric transducers. Smart Materials and Structures, 26(9):094002, 2017. 2

Moslemian, Alireza, Michelle E Arakgi, Philip P Roessler, Rajeshwar Singh Sidhu, Ryan M Degen, Ryan Willing, and Alan MJ Getgood. The medial structures of the knee have a significant contribution to postero medial rotational laxity control in the pci-deficient knee. Knee Surgery, Sports T raumatology, Arthroscopy, pages 1 -10, 2021.

Moslemian, Alireza, Roger Sidhu, Philip Roessler, Ryan Wood, Ryan Degen, Alan Getgood, and Ryan Willing. Influence of the posterior cruciate ligament on kinematics of the knee during experimentally simulated clinical tests and activities of daily living. Journal of Biomechanics, 115:110133, 2021 .

Mousavi, Mohammad, Mohammad Alzgool, and Shahrzad Towfighian. Autonomous shock sensing using bi stable triboelectric generators and mems electrostatic levitation actuators. Smart Materials and Structures, 30(6):065019, 2021.

National Joint Registry. 12th annual report 2015 national joint registry for England, Wales, Northern Ireland and the Isle of Man, 2015.

Nelson, Daniel, Alwathiqbellah Ibrahim, and Shahrzad Towfighian. A tunable triboelectric wideband energy harvester. Journal of Intelligent Material Systems and Structures, 2019.

Nelson, Daniel, Alwathiqbellah Ibrahim, and Shahrzad Towfighian. Dynamics of a threshold shock sensor: Combining bistability and triboelectricity. Sensors and Actuators A: Physical, 285:666-675, 2019.

Nichicon, Aluminum Electrolytic Capacitors LLS, Datasheet, Kyoto, Japan, 2011.

Paidimarri, Arun, Danielle Griffith, Alice Wang, Gangadhar Burra, and Anantha P Chandrakasan. An rc oscillator with comparator offset cancellation. IEEE Journal of Solid State Circuits, 51 (8): 1866-1877, 2016.

Pallay, Mark, Alwathiqbellah I Ibrahim, Ronald N Miles, and Shahrzad Towfighian. Pairing electrostatic levitation with triboelectric transduction for high-performance Self-powered mems sensors and actuators. Applied Physics Letters, 115(13): 133503, 2019.

Paraskevopoulou, Sivylla E., and Timothy G Constandinou. A sub1 w neural spike-peak detection and spikecount rate encoding circuit. In 2011 1EEE Biomedical Circuits and Systems Conference (BioCAS), pages 29-32. IEEE, 2011.

Platt, Stephen R., Shane Farritor, Kevin Garvin, and Hani Haider. The use of piezoelectric ceramics for electric power generation within orthopedic implants. IEEE/ASME transactions on mechatronics, 10(4):455- 461 , 2005.

Ponder, Robert I., Mohsen Safaei, and Steven R Anton. Fabrication and selection of surrogate knee implant bearings for experimental evaluation of embedded in vivo sensors. Journal of the mechanical behavior of biomedical materials, 91 :237-246, 2019.

Postler, A., C Liitzner, F Beyer, T Tille, and J Liitzner. Analysis of total knee arthroplasty revision causes. BMC Musculoskeletal Disorders, 19(55):1 -6, 2018.

Radial Leaded Multilayer Ceramic Capacitors, Datasheet, KEMET, Fort Lauderdale, FL, USA, 2020. Rasel, M. Salauddin, Pukar Maharjan, Md Salauddin, M Toyabur Rahman, Hyun Ok Cho, Jae Woo Kim, and Jae Yeong Park. An impedance tunable and highly efficient triboelectric nanogenerator for largescale, ultrasensitive pressure sensing applications. Nano Energy, 49:603-613, 2018.

Roth, Joshua D., Stephen M Howell, and Maury L Hull. An improved tibial force sensor to compute contact forces and contact locations in vitro after total knee arthroplasty. Journal of biomechanical engineering, 139(4), 2017.

Safaei, Mohsen, and Steven R. Anton. "Sensing and energy harvesting performance, and fatigue life of embedded piezoelectric transducer in total knee arthroplasty." In Smart Materials, Adaptive Structures and Intelligent Systems, vol. 50497, p. V002T07A010. American Society of Mechanical Engineers, 2016.

Safaei, Mohsen, and Steven R. Anton. "The effects of dimensional parameters on sensing and energy harvesting of an embedded PZT in a total knee replacement." In Active and Passive Smart Structures and Integrated Systems 2016, vol. 9799, pp. 766-778. SPIE, 2016.

Safaei, Mohsen, Henry A Sodano, and Steven R Anton. A review of energy harvesting using piezoelectric materials: state-of-the-art a decade later (2008-2018). Smart Materials and Structures, 28(11 ): 113001 , 2019.

Safaei, Mohsen, R Michael Meneghini, and Steven R Anton. Energy harvesting and sensing with embedded piezoelectric ceramics in knee implants. IEEE/ASME Transactions on Mechatronics, vol. 23, no. 2, pp. 864-874, Apr. 2018.

Safaei, Mohsen, R. Michael Meneghini, and Steven R. Anton. "Force detection, center of pressure tracking, and energy harvesting from a piezoelectric knee implant." Smart materials and Structures 27, no. 11 (2018): 114007.

Safaei, Mohsen. "A piezoelectric instrumented total knee replacement for sensing and energy harvesting." PhD diss., Tennessee Technological University, 2019.

Salman, E. Onchip regulators for low voltage and portable systems-on-chip. In S. Kurinec and K. Iniewski, editors, Energy Efficient Technologies: Devices, Circuits, and Systems, pages 57-79. Springer, 2019.

Salman, Emreand Eby G Friedman. High-Performance Integrated Circuit Design. McGraw-Hill New York, 2012.

Sauer, Christian, Milutin Stanacevic, Gert Cauwenberghs, and Nitish Thakor. Power harvesting and telemetry in emos for implanted devices. IEEE Transactions on Circuits and Systems I: Regular Papers, 52(12):2605-2613, 2005.

Sekeitto, Allen, Jance McGale, Liam Montgomery, Edward Vasarhelyi, Ryan Willing, and Brent Lanting. Biomechanical analysis of polyethylene insert thickness in posterior stabilized total knee arthroplasty (tka) using tka implants linked to a virtual ligament model. In 2021 COA CORS Annual Meeting, Online Meeting, Canada. Canadian Orthopaedic Association, 2021.

Seo, Jai-Gon, Young-Wan Moon, Byung-Chul Cho, Su Cheol Kim, Young Hoo Ko, Seung Pil Jang, and Byung Hoon Lee. Is Total Knee Arthroplasty a Viable Treatment Option in Octogenarians with Advanced Osteoarthritis? Knee Surgery & Related Research, 27(4):221 -227, 2015.

Sharkey, PF, PM Lichstein, C Shen, AT Tokarski, and J Parvizi. Why are total knee arthroplasties failing today- has anything changed after 10 years? Journal of Arthroplasty, 29(9): 1774-8, 2014. Sidhu, Sahil Prabhnoor, Alireza Moslemian, Geofrey Yamomo, Samira Vakili, Paul Kelly, Ryan T Willing, and Brent A anting. Lateral subvastus lateralis versus medial parapatellar approach for total knee arthroplasty: A cadaveric biomechanical study. The Knee, 27(6): 1735-1745, 2020.

Stanacevic, M., Y. Lin, and E. Salman. Analysis and design of 3d potentiostat for deep brain implantable devices. In Z. Yang, editor, Neural Computation, Neural Devices, and Neural Prosthesis, pages 261-287. Springer, 2014.

Stanacevic, Milutin, Kartikeya Murari, Abhishek Rege, Gert Cauwenberghs, and Nitish VThakor. Vlsi potentiostat array with oversampling gain modulation for wide-range neurotransmitter sensing. IEEE T ransactions on Biomedical Circuits and Systems, 1 (1 ):63-72, 2007.

Stone, O. D., A. D. Duckworth, D. P. Curran, J. A. Ballantyne, and I. J. Brenkel. Severe arthritis predicts greater improvements in function following total knee arthroplasty. Knee Surgery, Sports Traumatology, Arthroscopy, 25(8):2573-2579, 2017.

Tanaka, Miho J., Luke S Oh, Scott D Martin, and Eric M Berkson. Telemedicine in the era of covidl 9: the virtual orthopaedic examination. The Journal of bone and joint surgery. American volume, 2020.

Teeter, Matthew G., Kevin I Perry, Xunhua Yuan, James L Howard, and Brent A Lanting. Contact kinematic differences between gap balanced vs measured resection techniques for single radius posterior-stabilized total knee arthroplasty. The Journal of arthroplasty, 32(6): 1834-1838, 2017.

Telehealth services. www.cms.gov/OutreachandEducation/ MedicareLearningNetworkMLN/MLNProducts/Downloads/TelehealthS rvcsfctshtpdf.

Texas Instruments, LMx58-NLow-Power, Dual-Operational Amplifiers, Datasheet, Dallas, TX, USA, 2014.

Thiele, K., C Perka, G Matziolis, HO Mayr, M Sostheim, and R Hube. Current failure mechanisms after knee arthroplasty have changed: polyethylene wear is less common in revision surgery. The Journal of Bone and Joint Surgery, American Edition, 97(9)715-20, 2015.

Towfighian, Shahrzad, Emre Salman, and Ryan Willing. An implantable Self-powered load sensor for total knee replacement health monitoring, September 01 2017. NIH1 R21 AR06857201 A1 .

Towfighian. Shahrzad, Mems high voltage triboelectric levitation: A generactuator, June 01 2019. NSF 1919608.

Tsou, ChiHui, BoJyue Kao, MingChien Yang, MawChemg Suen, YiHsuan Lee, JuiChin Chen, Wei Hua Yao, ShangMing Lin, ChihYuan Tsou, ShuHsien Huang, et al. Biocompatibility and characterization of polylactic acid/styrene-ethylene-butylene-styrene composites. Biomedical materials and engineering, 26(s1 ):S147-S154, 2015.

Wan, T., and E. Salman. Ultra low power simon core for lightweight encryption. In IEEE International Symposium on Circuits and Systems (ISCAS), pages 1-5, May 2018.

Wan, T., Y. Karimi, M. Stanacevic, and E. Salman. Perspective paper: Can AC computing be an alternative for wirelessly powered loT devices? IEEE Embedded Systems Letters, 9(1): 13-16, 2017. Wan, Tutu, Yasha Karimi, Milutin Stanacevic, and Emre Salman. Ac computing methodology for rf-powered lot devices. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 27(15), May 2019.

Wang, S., L. Lin, and Z. L. Wang, “Nanoscale triboelectric-effect-enabled energy conversion for sustainably powering portable electronics,” Nano Lett., vol. 12, no. 12, pp. 6339-6346, 2012.

Wang, Sihong, Simiao Niu, Jin Yang, Long Lin, and Zhong Lin Wang. Quantitative measurements of vibration amplitude using a contact-mode freestanding triboelectric nanogenerator. ACS nano, 8(12): 12004- 12013, 2014.

Wang, Z. L., “On Maxwell’s displacement current for energy and sensors: The origin of nanogenerators,” Mater. Today, vol. 20, no. 2, pp. 74-82, Mar. 2017.

Wang, Zhong Lin. Triboelectric nanogenerators as new energy technology for Self-powered systems and as active mechanical and chemical sensors. ACS nano, 7(11 ):9533-9557, 2013.

Wang, Z. L., J. Chen, and L. Lin, “Progress in triboelectric nanogenerators as a new energy technology and self- powered sensors,” Energy Environ. Sci., vol. 8, pp. 2250-2282, May 2015.

Weddell, Alex S, Geoff V Merrett, Tom J Kazmierski, and Bashir M Al-Hashimi. Accurate supercapacitor modeling for energy harvesting wireless sensor nodes. IEEE Transactions on Circuits and Systems II: Express Briefs, 58(12):911 -915, 2011.

Willing, R., and PS Walker. Measuring the sensitivity of total knee replacement kinematics and laxity to soft tissue imbalances. Journal of Biomechanics, 77:62-68, 2018.

Willing, Ryan, Alireza Moslemian, Geofrey Yamomo, Thomas Wood, James Howard, and Brent Lanting. Condylar stabilized tkr may not fully compensate for pci-deficiency: An in vitro cadaver study. Journal of Orthopaedic Research, 2019.

Wilson, Brooke E., Michael Meneghini, and Steven R. Anton. "Embedded piezoelectrics for sensing and energy harvesting in total knee replacement units." In Active and Passive Smart Structures and Integrated Systems 2015, vol. 9431, pp. 424-433. SPIE, 2015.

Wu, Mengjun, Ting Zheng, Haiwu Zheng, Jifang Li, Weichao Wang, Mingsai Zhu, Fengzhu Li, Gentian Yue, Yuzong Gu, and Jiagang Wu. High-performance piezo-electric energy harvester and Self-powered mechanosensing using leadfree potassium-sodium niobate flexible piezoelectric composites. Journal of Materials Chemistry A, 6(34): 16439-16449, 2018.

Yamomo, G., A Ibrahim, S Towfighian, and R Willing. Tuning of 3d printed energy harvester package design. In ISTA 2019 Toronto Scientific Program. International Society for Technology in Arthroplasty, 2019. Podium Presentation.

Yamomo, G., N. A. Hossain, S. Towfighian, and R. Willing, “Design and analysis of a compliant 3D printed energy harvester housing for knee implants,” Med. Eng. Phys., vol. 88, pp. 59-68, Feb. 2021.

Yamomo, Geofrey, Nabid Hossain, Shahrzad Towfighian, and Ryan Willing. Design and analysis of a compliant 3d printed energy harvester housing for knee implants. Medical Engineering & Physics, 88:59-68, 2021. Yang, Wei, and Shahrzad Towfighian. A hybrid nonlinear vibration energy harvester. Mechanical Systems and Signal Processing, 90:317-333, 2017.

Yang, Wei, and Shahrzad Towfighian. A parametric resonator with low threshold excitation for vibration energy harvesting. Journal of Sound and Vibration, 446:129-143, 2019.

Yang, Wei, and Shahrzad Towfighian. Low frequency energy harvesting with a variable potential function under random vibration. Smart Materials and Structures, 27(11 ): 114004, 2018.

Yaul, Frank M., and Anantha P Chandrakasan. 11.3 a 10b 0.6 nw sar adc with data dependent energy savings using Isb-first successive approximation. In 2014 IEEE International Solid State Circuits Conference Digest of Technical Papers (ISSCC), pages 198-199. IEEE, 2014.

Zhang, Hanlu, Shan Feng, Delong He, YiguoXu, Minhao Yang, and Jinbo Bai. An electret film-based triboelectric nanogenerator with largely improved performance via a tape-peeling charging method. Nano Energy, 48:256-265, 2018.

Zhang, Lei, Xiaosong Hu, Zhenpo Wang, Fengchun Sun, and David G Dorrell. A review of supercapacitor modeling, estimation, and applications: A control/management perspective. Renewable and Sustainable Energy Reviews, 81:1868-1878, 2018.

Zhang, X.-S. et al., “Frequency-multiplication high-output triboelectric nanogenerator for sustainably powering biomedical microsystems,” Nano Lett., vol. 13, no. 3, pp. 1168-1172, Mar. 2013.

Zheng, Qiang, Bojing Shi, Zhou Li, and Zhong Lin Wang. Recent progress on piezoelectric and triboelectric energy harvesters in biomedical systems. Advanced Science, 4(7):1700029, 2017.

Zhou, Chaochao, and Ryan Willing. Multi-objective design optimization of a biconcave mobile-bearing lumbar total artificial disk considering spinal kinematics, facet joint loading, and metal-on-polyethylene contact mechanics. Journal of Biomechanical Engineering, 142(4), 2020.

See also: CA2964127; CA3116979; CN103532430; CN105186922; CN105871249; CN107547003;

CN110121792; CN111564946; CN112165146; CN112350443; CN112564541; CN112688591; CN112803826;

CN112881908; CN113091888; CN113156230; CN113241967; CN113452278; CN113840955; CN114301255; CN214045473U; DE102017124556; EP0053005; EP0366591; EP0984707; EP0992211; EP1050955; EP1078331;

EP1095710; EP1120026; EP1154579; EP1173287; EP1254559; EP1278603; EP1454674; EP1960114;

EP2465417; EP2671702; EP2711412; EP2740645; EP2890891; EP2912214; EP2928063; EP2975758;

EP2975759; EP2978123; EP3001557; EP3010138; EP3014755; EP3016264; EP3021476; EP3031350;

EP3086460; EP3101801; EP3131195; EP3133375; EP3178299; EP3213302; EP3213361; EP3235116;

EP3247032; EP3247033; EP3264584; EP3277621; EP3278443; EP3288172; EP3330986; EP3345295;

EP3345296; EP3345297; EP3367537; EP3384590; EP3389486; EP3394976; EP3430697; EP3451971;

EP3547905; EP3566112; EP3652721; EP3654513; EP3659249; EP3698461; EP3762148; EP3857700; EP3867321; EP3881419; EP3888239; EP3908754; EP3910784; EP3946982; EP3949105; EP4011638;

EP4011639; IN201831024973; I N202041006974; IN397451; KR101532889; KR101549445; KR101611126; KR101648634; KR101691552; KR101705974; KR101850959; KR101867055; KR101870278; KR101984533;

KR102132098; KR102163962; KR102233550; KR102292573; KR102391307; KR20160039705; KR20160052171;

KR20160098843; KR20160118600; KR20160125710; KR20180002267; KR20180031451; KR20180045417;

KR20180077648; KR20190013069; KR20200085994; KR20200106264; KR20200113952; KR20200121510;

KR20200129064; KR20200135122; KR20210059933; KR20210080786; KR20210137306; KR20220034412;

KR20220046250; KR20220048531 ; KR20220086206; US10014799; US10027187; US10050562; US10075100;

US10103649; US10170686; US10199958; US10305395; US10333430; US10367431; US10425018; US10439517;

US10498259; US10565849; US10587207; US10587208; US10622918; US10622919; US10630207; US10635213;

US10716339; US10734920; US10734921; US10770989; US10770990; US10774618; US10804818; US10855205;

US10873274; US10873275; US10873277; US10874325; US10924036; US10931209; US10938323; US10978965;

US11027140; US11090501; US11101747; US11114955; US11161421; US11196357; US11228259; US11305295;

US11336207; US11431265; US20010052551; US20020058002; US20020085977; US20030067427;

US20040184974; US20040232801; US20060064826; US20070145167; US20070192972; US20070252047;

US20100133006; US20120299559; US20130049531; US20130075308; US20130323583; US20140084748;

US20140202630; US20140225475; US20140246950; US20140246951; US20140292138; US20140300248;

US20140313141; US20140338458; US20150001966; US20150001993; US20150061460; US20150061464;

US20150284874; US20150295101; US20150311823; US20160016364; US20160028327; US20160036351;

US20160040648; US20160043661; US20160065091; US20160123805; US20160149517; US20160149518;

US20160156282; US20160164434; US20160218640; US20160233794; US20160248345; US20160261210;

US20160294305; US20160315561; US20160344308; US20160359429; US20160365808; US20170005258;

US20170025857; US20170077838; US20170187307; US20170219391; US20170331396; US20170331397;

US20170346416; US20170367455; US20180006582; US20180013358; US20180041139; US20180062543;

US20180091064; US20180091065; US20180115258; US20180142382; US20180145613; US20180153450;

US20180191274; US20180205327; US20180263328; US20180294744; US20180294745; US20180316281;

US20180353101; US20180355840; US20180375443; US20190006961; US20190019632; US20190056603;

US20190058419; US20190061533; US20190087771; US20190125523; US20190131889; US20190157992;

US20190284907; US20190307368; US20190312529; US20190346952; US20190354222; US20190372482;

US20190379300; US20200005620; US20200028450; US20200067426; US20200070664; US20200106371;

US20200126557; US20200127585; US20200147398; US20200168788; US20200244188; US20200244189;

US20200254923; US20200266045; US20200374604; US20200395870; US20210008932; US20210061648;

US20210077251; US20210078016; US20210083600; US20210091683; US20210109480; US20210119554;

US20210170169; US20210172789; US20210211070; US20210211072; US20210249972; US20210355339;

US20210384847; US20220052628; US20220109384; US20220115962; US20220134135; US20220155160;

US20220216809; US20220239236; US20220255464; US20220257898; US2853009; US3086131; US3124457;

US3201620; US3218968; US3246239; US3257223; US3633544; US3690646; US3733544; US3893413; US3968637; US4120015; US4126822; US4133610; US4265998; US4359192; US4990813; US5506491;

US5614994; US5714007; US6007630; US6199244; US6395145; US6407483; US6473072; US6474563;

US6484328; US6756930; US7230751; US7416646; US7621471 ; US7637462; US7732974; US8179278;

US8298304; US8536760; US8810049; US8853583; US9102104; US9178446; US9245561; US9331603;

US9381703; US9394875; US9444031; US9484842; US9543860; US9571009; US9590532; US9595894;

US9604534; US9790928; US9812993; US9825557; US9837933; US9837934; US9843275; US9876443;

US9887644; US9906170; US9985554; WO1999023749; WO1999059101 ; W02000019791; W02000064592

W02001058135; W02001083116; W02005117679; W02007078438; W02008040145; W02009028991; WO2010065431 ; WO2013151590; WO2013170651 ; WO2013181372; WO2014000484; WO2014012403; WO2014032684; WO2014044077; WO2014064260; WO2014101395; WO2014146466; WO2014176937; WO2014193182; WO2014206098; WO2014209522; WO2015107878; WO2015154693; WO2015183822; WO2016022207; WO2016023490; WO2016048026; WO2016069441 ; WO2016070011 ; WO2016072606; WO2016072621 ; WO2016089046; WO2016092080; WO2016096492; WO2016156180; WO2016156471 ; WO2016199082; WO2017007359; WO2017019977; WO2017036938; WO2017037104; WO2017037291 ; WO2017045910; WO2017093424; WO2017102867; WO2017108546; WO2017119532; WO2017124772; WO2017158513; WO2017187838; WO2017191542; WO2018018790; WO2018019292; WO2018036237; WO2018049370; WO2018090003; WO2018095431 ; WO2018102011 ; WO2018127839; WO2018176893; WO2018182521 ; WO2018190668; WO2018205994; WO2018223474; WO2018228373; WO2019020172; WO2019020204; WO2019043446; WO2019069311 ; WO2019073317; WO2019075445; WO2019077077; WO2019082120; WO2019103374; WO2019160285; WO2019171011 ; WO2019177693; WO2019203389; WO2019212271 ; WO2019218164; W02020013620; W02020022453; W02020054916; W02020065533; W02020079473; W02020084456; W02020091161 ; W02020098280; W02020099030; W02020099812; W02020109783; W02020120218; W02020144519; W02020156321 ; W02020198451 ; W02020204545; WO2020235937; WO2020236895; WO2020241927; WO2020258614; WO2021016524; WO2021044329; WO2021086924; WO2021095947; WO2021137410; WO20211374103; WO2021167532; WO2021169758; WO2021172654; WO2021177489; WO2021181132; WO2021203533; WO2021206258; WO2021212125; WO2021221255; W02022008794; WO2022034497; WO2022041465; W02022050893; WO2022055427; W02022057073; WO2022062133; W02022080811 ; W02022093074; WO2022151542; WO2022167016; ZA202202194

What is claimed is: