Design and Implementation of a Wearable System Based on a Flexible Capacitive Sensor, Monitoring Knee Laxity

The anterior cruciate ligament (ACL) tear is one of the most common knee injuries causing instability to the knee joint. Current methods of diagnosis fail to meet the ergonomic, reliability, and reproducibility requirements. Thus, the wearable sensors are gaining momentum to overcome current challenges. This paper aims at proposing a wearable capacitive based sensor system that shows a good potential in substituting the currently used methods for the diagnosis of ACL rupture. The developed sensor system measures the internal tibial rotation of the knee. It is compact and lightweight. Being cable free, it can be worn as a patch, without impeding the freedom of movement of the physician. Moreover, it can be powered with a battery or wireless. Both methods make it compact, ergonomic, easy for the patient to wear and for the doctor to use. To analyze the suitability of the developed sensing system, data from a knee simulator setup and three healthy volunteers (2 Males and 1 Female) are compared and analyzed. In all the patients, above 15° for every 5° angle variation, a relative change of capacitance with respect to its initial value of 0.01 is observed. These results are comparable with the knee simulator's data with a max RMSE of 0.002. Below 15° the system is additionally able to measure a gender‐based difference of rotation due to the higher flexibility of ligaments in females. For them the sensitivity below and above 15° is comparable, for male the sensitivity below 15° is lower. The results show that the developed system has good potential in substituting the currently used method for the diagnosis of ACL rupture and paves the way toward the continuous observation in free movement of knee laxity.


Introduction
Over the last few decades, general interest in sports activities has increased and this has resulted in a corresponding increase in the occurrence of knee injuries. [1]One of the most common knee injuries is the anterior cruciate ligament (ACL) tear, which accounts for 20% of all knee injuries. [2]In Germany, around 45 cases occur for every 100 000 inhabitants per year, which is a total of 36 900 out of 82 million people. [2]he United States report 37 cases for every 100 000 inhabitants annually, which results in a total number of 121 000 injuries out of 328 million individuals. [3]herefore, the importance of devices and methods to measure knee laxity caused by ACL tear has been raised in the last decades.
Currently, there are different examination methods used for the diagnosis of ACL ruptures through internal tibial rotation: Physical examination and instrument-based measuring methods.The physical examination includes the Lachman test, the pivot-shift test, and the anterior drawer test.In these methods, the range of motion between the stationary (upper) and the movable (lower) knee is measured to analyze the degree of laxity. [4,5]These physical examination methods are, in principle, standardized.However, outcomes often vary when the same patient is seen by different examiners.This can be traced back to the fact that, despite the standardized procedure, there are still variable parameters such as the mechanical pressure applied or the maneuver techniques of the examiner. [6][8] This collides with the requirement for a precise monitoring of the healing process, as 20-25% of patients reinjure their knee joint or experience secondary damage. [9]everal technical instruments have been developed since the early 1980s to examine knee injuries by internally rotating the tibia (lower leg) versus the upper leg.A short overview of the existing instruments, their development, and some approaches which have emerged in recent years follows.A well-known technical instrument for measuring rotational knee laxity is the KT-1000 arthrometer which was developed in the 1980s by MEDmetric Corp. (San Diego, CA, USA).The instrument is strapped to the leg and, using a handle, the examiner pulls the tibia anteriorly with forces of 67, 89, and 134 N. At each applied force, the movement of the tibia is measured in millimeters (mm).These measurements are conducted on both the injured and healthy leg.The differences between the movement of each leg are defined by the KT-1000 score.As one of the first commercially available instruments, it has been considered a gold standard.However, many studies have reported a lack of reproducibility. [10]Many scientists compared the results of the KT-1000 with clinical findings.[15][16][17] Alternative devices are the GNRB arthrometer (Genourob, Laval, France), [18,19] the Rolimeter, [19,20] and the Telos (Telos GmbH, Laubscher, Hölstein, Switzerland). [21]In, [21,22] a comparative study between Telos and GNRB proves that the GNRB produced more reliable and reproducible results than the Telos, irrespective of age and gender.However, when the GNRB was applied for ACL laxity measurements at 60 healthy subjects, the data was not reliable and reproducible under the optimum conditions of comparability set for the experiment. [23]The Rolimeter uses a knee laxity measurement principle comparable to the KT-1000.However, the measurements had to be repeated multiple times by the same examiner to produce reproducible results. [24]In addition, these instruments are large and expensive.
A compact, economical, and wearable knee laxity-measuring device would be the ideal alternative to both the physical and technical instrumental examination methods discussed above. [25][32] Particularly, the inertial sensors or Inertial Measurement Units (IMUs) are comparatively cheaper and more compact among the lot.In, [25] the IMUs are easily integrated into a knee brace, however the positioning of the brace on the knee is not accurate for different people due to the elastic nature of the brace.This results in reproducibility issues in a clinical setting.Moreover, the device is quoted as just satisfactory, implying a need for improvement in both design and electronics to make it more user-friendly. [25]n alternative promising new approach consists of using capacitive strain gauges, [33] which have the advantage of exhibiting a small hysteresis, high repeatability, and stretchability. [33]To the knowledge of the authors, the suitability of this type of sensor for the intended application has not yet been tested.Therefore in this work, a fully wearable, compact, and lightweight sensor system, based on the capacitive strain gauge developed in, [33] is presented to measure the internal tibial rotation for a diagnosis of the ACL ligament tear.In, [33] a detailed, innovative, reliable and reproducible fabrication process for sensor production is designed, here the sensor is embedded on a wearable system and it is tested on a healthy human leg and the measurements are compared to a knee simulator test bench setup.
Additional features, which increase the competitiveness of the proposed solution, are a wireless data transmission capability and the possibility to power it with a battery and, alternatively, with an environmentally friendly wireless powering technology.These features allow for a comfortable wearability during clinical diagnosis.

System Design
The overall integrated system design is shown in Figure .It consists of four main parts: The capacitive strain gauge, [33] the sensor front-end, its power supply circuit, and a stationary computer that receives the transferred data.The flexible strain gauge contact pads are connected to the flex PCB by screws as shown in Figure 1.The screws are pierced from the bottom to the PCB, top metal plate, sensor, and bottom metal plate tightened with the bolt.The metal plates are used to ensure a robust connection between the sensor and the PCB board.Therefore, this sandwich ensures that the sensor when stretched does not tear from the point of connection thus establishing a reliable contact between the board and the sensor.
The sensor front-end acquires, records, processes, and transmits the sensor reading.For that, it consists of a capacitive-todigital converter, a microcontroller (c), and a wireless communication unit.
A flex PCB is chosen over a rigid PCB for integrating the sensor front-end, capacitive sensor, and power supply circuit so that Sensor Structure with 279 fingers and contact parts.Adapted with permission. [33] can adapt to the shape of the human knees of different individuals and can be easily attached.The board is produced using standard FR4 material.For the surface (metal trace), chemical gold traces are usually recommended but the nickel layer is built over the gold layer.The presence of nickel may cause allergic reactions in human skin. [34]This is a concern because, in the current design, the bottom surface of the board has metal via prints.Therefore, chemical silver is used as a top layer on all metal traces to avoid allergies to human skin.The board is fabricated by Multi-Circuit Boards Ltd. (Pole, GB).The following sections describe the elements of the electronic interface in detail.

Capacitive Strain Gauge Sensor
The polymer-based capacitive strain gauge developed in [33] is used.This consists of an interdigitated capacitor with 279 fingers and two contact areas, as shown in Figure 2. The innovative and reproducible fabrication process is explained in detail in the paper. [33]The sensor is fabricated using 0.5 mm thick carbon black polydimethylsiloxane (C-PDMS) for the sensing part and pure PDMS as the substrate.The whole sensor structure is lasered using an Nd: YAG laser with a wavelength of 1064 nm (DPL Smart Marker II, ACI Laser GmbH, Nohra, Germany). [33]he total length and width of one strain gauge sensor are 65 and 11.5 mm, respectively.The strain-sensitive area covered by the capacitive sensor is 55.8 mm long and 7.5 mm wide.Also, the connecting sidelines for each set of fingers are enlarged to a width of 2 mm to provide for lower electrical resistance.
This choice was made because the limited material resistivity of C-PDMS in combination with the small track dimensions increases the internal resistance of the sensor structure.The strain gauge can be modeled as a series of electrical resistances for the connected traces and of capacitances for the individual interdigital capacitor (IDC) elements.As a result, the whole sensor is a series of low-pass filters.It follows that the parasitic resistance affects the sensor bandwidth: above the cut-off frequency, the capacitor does not get fully charged.This would result in an unwanted decrease of the capacitance measured, which can be interpreted as a not-present sensor elongation.An alternative to limit the parasitic resistance is to increase the number of carbon particles in the C-PDMS.However, this causes an increase of material stiffness, which is a drawback for this application in which high elasticity is required: the sensor must be able to follow the movement of the knee without limiting it.
The sensor relies on the fundamental principle that the distance between the fingers increases when stretched, which in turn decreases the capacitance.The capacitance C is given by [35] Where N is the number of fingers,  0 is the permittivity of vacuum,  r is the relative permittivity of the dielectric, l o is the length of the finger, t is the thickness of the finger, and d is the distance between the fingers. [35]

Microcontroller
The Launchpad CC2652R1 consists of a microcontroller using an Arm Cortex-M4F processor (System CPU), a radio frequency core (RF core) supporting the Bluetooth low energy protocol using an Arm Cortex-M0 processor, and a sensor controller unit that can control a time-to-digital converter (TDC) for capacitive sensing.It also implements the Bluetooth low energy protocol up to generation 5.0.This microcontroller was chosen because its architecture allows the sampling and transmission of the data at 1 kHz.This bandwidth was a requirement set by the clinicians, with a future perspective to test the functioning potential of this system under dynamic conditions (walking).The microcontroller includes two CPUs: the main CPU and the sensor controller.The sensor controller can be selectively switched between a standby and active power mode.It is meant to read and monitor the capacitive sensor and run in parallel with the main CPU.We use this feature of the microcontroller to achieve the required sampling and transmission frequency of 1 kHz.The sensor controller samples data at a rate of 1 kHz and stores them in a memory segment accessible also to the main CPU.The main CPU wakes up from sleep with a frequency of 200 Hz, then updates the characteristic and sends out a block of 10 subsequent sensor readings via Bluetooth.Hence the c samples data every 1 ms and transmits 10 data every 5 ms.This method also has the advantage of reducing the c power consumption, as the main CPU wakes up only for a short time when required.
The power consumption of the microcontroller is calculated using the built-in software EnergyTrace. [36]The results show that, while advertising data via Bluetooth, the current consumption at 3.3 V is ≈0.72 mA with peaks of up to 8.63 mA when broadcasting.Therefore, the average current lies at 0.81 mA.After connecting to the receiver, the supply current is at an average of 2.10 mA with maximal peaks of 11.40 mA while measuring.
Based on the current consumption of the electronics, the energy consumption per day of the device is estimated.After enquiring about the consultation process to the doctors, it is understood that a physician can see a maximum of 7 patients per day for sessions of 90 min each.During these sessions, the physical examinations take only about 20 min while the remaining 70 min are used for consultation.In the worst-case scenario, the medical device will be switched on for the entirety of each session while sensor measurements are only taken for 20 min.
Given the examination time in Table 1, the capacity drains from the system's battery, calculated for consultation and physical examinations, are 0.95 and 0.70 mAh, respectively.Therefore, the total electric charge used is 1.65 mAh and, as a result, a total capacity drain of 11.55 mAh is calculated for one day and 7 patients.With an average voltage of 3.7 V for the rechargeable battery in use, the energy consumption amounts to 42.74 mWh per day.The next essential requirement in designing is powering the system.

Power Supply Circuit
Two alternatives were considered for the power supply method: (1) a rechargeable battery, and (2) a battery-free wireless power transfer circuit.

Battery-Powered System (BPS)
Based on the power consumption calculations, a battery is chosen that at least provides twice the required power (7 patients/day) of the system, to have a buffer in case of a scenario requiring more energy.Therefore, the battery should have a capacity above 23.10 mAh, be small, and be lightweight from an ergonomic point of view.It should be also rechargeable, and the voltage should range between 1.8 and 3.8 V which is determined by the microcontroller unit.The CP1254 (Varta Microbattery GmbH) coin battery which is 12.1 mm in diameter and 1.8 g in weight satisfies all these requirements. [37]A BQ29700 battery management IC (Texas Instruments, Dallas, TX, USA) is used to protect the battery from over-charge, over-discharge, over-current, and short-circuit.A step-down converter, the TPS82740B (Texas Instruments, Dallas, TX, USA) [38] is used to convert the battery output to a suitable voltage for the electronics.The schematic of the BPS is shown in Figure 3 below.

Wireless Power Transfer System (WPTS)
As an alternative to battery power, a more environmentally friendly solution is implemented with WPTS.The schematic of the WPTS is shown in Figure 4 below, in which the power supply circuit of the BPS on the flex board is replaced with the transmitter (Tx) and receiver (Rx) coils.
Based on previous works on inductive wireless power transfer systems (WPTS) for biomedical implants [39][40][41] an optimized WPTS has been developed.It has size constraints on the Rx coil (as it must comfortably fit on the patient's leg) and distance constraints between the Tx coil and the patient, as the Tx coil must not be on the same table as the patient so as not to interfere with the knee tests.Using the design process outlined in, [42] a WPTS has been constructed that successfully powers the knee  sensor while adhering to the imposed constraints.The WPTS comprises a ZHL-100W-GAN+ Class-A power amplifier (Mini-Circuits, Brooklyn NY, USA), a Tx coil, and an Rx coil.The Tx and Rx coil specifications are described in Table 2.
The transmit coil is built into a series LC circuit, with the tuning capacitance added in series to avoid saturating the voltage limits of the power amplifier.The Rx coil is built into a parallel LC circuit to achieve the necessary input voltage level for the sensor's MCU.
The power conditioning circuit is composed of a half-wave rectifier, a 100 nF smoothing capacitor, and a low-dropout (LDO) regulator that ensures that the voltage delivered will not exceed the 3.3 V supply limit of the sensor's MCU.The WPTS can deliver 50.4 mW to a 56 Ω resistor at 5 cm and 6.0 mW to a 56 Ω resistor at a 10 cm distance.The 56 Ω is not the optimal load for the system and therefore does not reflect the maximum achievable efficiency since no impedance matching is done in the WPTS.

Measurement Methods and Setup
To analyze the functionality of the sensor system the rotational knee laxity is first tested on a knee simulator and afterwards on voluntarily healthy test subjects.These two studies are conducted Figure 5. Movement parts of knee simulator measuring stand.Adapted with permission. [44]th both the BPS and WPTS powering systems in a quasi-static condition.

Knee Simulator Setup
The knee simulator shown in Figure 5 consists of controllable kinematics, which can simulate all relevant knee movements necessary to quantify the sensor system.The moving parts of the knee simulator are shown in Figure 6.The motion base of the measuring stand is an X-Y rotary table (IntelLiDrives Inc., Philadelphia, PA, USA), which is equipped with three NEMA17 motors.A c-shaped arc coupled with a NEMA23 spindle motor permits the simulation of the flexion, or extension, of the joint. [43]igure 6 shows the CAD drawing of the artificial knee joint mechanism using the two aluminum cylinders. [44]or the measurement analysis, the BPS and the WPTS are installed on the knee simulator as shown in Figures 7 and 8.The sensor is positioned along the line of the varus/valgus tendon.The BPS, consisting of a sensor and PCB, can be applied as a patch on the knee (Figure 7).
The WPTS has an additional Rx coil which currently is placed separately from the PCB in any part of the leg and that in the next version can be integrated into the PCB.The WPTS is used to power the capacitive knee sensor on both the robotic knee and with human patients.The WPTS could deliver adequate power to the sensor with a 10 cm distance between the transmit (Tx) and receive (Rx) coils (Figure 8).For the measurement, one of the sensors from the reproducible batch sensors in [33] is used for both Figure 6.CAD drawing -Aluminum cylinders joined to articulate the artificial knee joint.Adapted with permission. [44]S and WPTS.When the tibia rotates between 10°and 45°internally with steps of 5°the corresponding capacitive measurement is sampled and stored.The relative change of capacitance is used for the analysis, which is obtained by the ratio of the raw data and the capacitance measured when the knee is in the starting position.By doing this, the influence of the operator on the measurements is limited.When the sensor is placed on a knee simulator, the sensor is stretched slightly and differently, depending on the person that applies it.This operator-dependent variation is mainly due to the highly elastic material characteristics which in turn results in a deviation of the measurement starting values.The effect of the prestretching, which is not relevant for the measurement, does not influence the relative variation.Therefore, it is chosen for the analysis.

Participant Selection
Statement from the ethics committee "The procedure described in the project (Application-Nr.EK-Freiburg: 487/16) and the data collection are based on the self and your colleagues in the project

Healthy Human Knee Setup
For the measurements with the human knee, three subjects are examined in the trial.Also, one of the sensors from the reproducible batch sensors in [33] is used for both BPS and WPTS.Measurements are acquired from both the right and left knee of these three subjects.The experimental protocol for the sensor system is described in Table 3 below.Although only 3 different people are presented in this paper, both the legs (right and left) of the individual are subjected to an internal tibial rotation.As a result, a total of 6 different data points is collected to further characterize the sensor system.
For the measurement setup, the sensor is placed, like before, along the line of the anterolateral ligament (ALL).The knee skeletal system is shown in Figure 9 below.The upper thigh bone is the femur, the lower bone is the tibia and the patella are the knee cap which is a flat, rounded triangular bone.
A test setup is developed featuring the fixation of the sensor covering the femur and tibia where the varus/valgus stress posts can be measured. [45]The sensor-to-body alignment requires medical expertise and is a crucial factor for securing a reliable measurement.Therefore, it is done with the help of an orthopedic doctor who can precisely identify this ligament stretch through physical examination.Then, the sensor is mounted with the help of a Skin Tite (Smooth-On, Inc.Macungie, Pennsylvania, USA) adhesive that holds the sensor firmly during stretching which is ideal for our application.The electronic flex board is secured to the knee with the help of medical-grade adhesive tape.
To calibrate the measurements of the newly developed sensor system, a Laxitester (ORTEMA Sports Protection, Markgroeningen, Germany) is used to measure the internal rotation of the knee joint.As shown in Figures 10 and 11, the laxitester setup was placed on the table and the femoral condyles are fixed by a sliding post to provide resistance during the internal rotation of the knee.The subjects were placed in a supine position.First,   the left and then the right leg were clinically examined one after the other and documented through an examination form.Exclusion criteria were the presence of pathological ligament findings in the lateral/cruciate ligaments, previous knee operations, or restrictions in the range of motion.The foot was then placed on a measurement plate of the laxitester where the adjustable cheeks can do the lateral fixation of the foot.The tibial internal rotation was subjected to different angles until a 2 Nm force is seen on the scale of the laxitester. [45]Due to the viscoelastic properties of the soft tissue, the internal rotation of the tibia is carried out three times with a torque of 2 Nm before the definitive measurement.This procedure is necessary for the preconditioning of the soft tissue. [46]One of the subjects is tested with the BPS and the other two with the WPTS.Figures 10 and 11 show the whole setup to measure the tibial rotation with both the BPS and the WPTS.This choice was taken to show that the power method does not affect the sensor measurements.For the analysis of the recorded data, 10 measurements per angle are averaged and plotted.

Statistical Analysis
The pre-processing of data included Mean of the measured capacitance and normalization to obtain the relative change of capacitance.The sample size of each measurement at each angle is a mean of 10 data points per angle rotation.The statistical method to assess the significant differences is done by a second order polynomial fitting and root mean square error (RMSE).The software used for the statistical analysis is OriginLab.

System Validation Using Knee Simulator Setup
The cyclic elongation of the sensor subjected to the knee simulator's internal tibial rotation at different angles is shown in Figure 12.The measurements are taken using both the BPS and WPTS.Irrespective of the powering supply, data from both the BPS and WPTS are corresponding.Therefore, one of the data is plotted as a reference for the rest of the analysis.The characterization of the sensor for 3 different cycles of rotation has a good correspondence and is fitted with a second-order polynomial equation as shown in Figure 12.
In Table 4, the R-square value is very close to 1 which determines a good fitting of the data with the chosen model.The high R-square shows how the scatter is limited also between the data belonging to different cycles of rotation of the knee simulator.This proves the reliability of the test setup and the reproducibility of the test measurements.

System Validation In Vivo Using the Battery-Powered System
Figure 13 shows the left knee data of subject 1 (Male).The degree of tibial rotation of subject 1 is from 10°to 25°when applying a maximum of 2 Nm force as given by the clinical standards. [45]igure 14 shows the measurements executed on the right leg of subject 1.In both plots, the data of the knee simulator and the different cycles of rotation are compared.
From the plot, we can observe a limited rotation of the knee in comparison to the range considered for the knee simulator measurements.The extent to which a knee can rotate internally differs from person to person and gender.In the current study, female subjects had a relatively larger range of tibial rotation (10°-35°) than male subjects who fell within a much more restricted range (10°-20°).This is due to the differences in the gender biomechanical properties of tendons and ligaments.Also, the hormonal profile between women and men is significant as testosterone can indirectly influence tendon and ligament laxity.Therefore, the female has a higher ligament laxity than the male. [47]From the clinical analysis, differences can be observed between the right and the left knee of the same individual.This analysis is also observable in our measurements.
Both in Figures 13 and 14, a difference in the slope can be observed below and above 15°in comparison to the characteristic obtained with the knee simulator in Figure 12.This is due to the resistance of the ACL ligament at smaller tibial rotation for men.This characteristic is not modeled in the knee simulator: The setup has two cylinders corresponding to the upper and lower leg moving freely in the concave socket as shown in Figure 6.Therefore, this contributes to the difference between the two characteristics below 15°.
The deviation in the data between 3 different cycles of rotation in the subject's knees is also observable.As such disturbance is not observable in the knee simulator data, it can be attributed to the fact that the patient tends to move when force is applied to rotate the knee.This opposition to the rotation causes the movement of the whole body, which can disturb the measurement of the actual knee, but does not correspond to a dysfunctionality of the sensor.

System Validation In Vivo Using the Wireless Power Transmission System
The newly developed sensor system was further tested on two subjects (subject 2 male and subject 3 female).For this measurement, the sensor was powered with the WPTS.
Figures 15 and 16 show, respectively, the measurements on the right and left leg of patient 2. Figures 17 and 18 show the measurements on the right and left of For patients, 3 cycles of measurements are recorded and the knee simulator characteristic is plotted for comparison.
As already discussed in paragraph IV B, there is a slope difference between males and females for the measurement below and above 15°, due to the resistance of the ACL ligament at  smaller tibial rotation for men.Figures 17 and 18 do not show this discrepancy of the slope above and below 15°.According to the claimed biomedical characteristic in females, [47] the resistance of the ACL ligament for small rotation angles is limited.
The data of subject 3 (female) exhibits a better correspondence to the mean of knee simulator data from 10°to 35°.To statistically prove the correspondence of the measurements, the data above and below 15°of all three subjects are fitted with a second-order polynomial equation (see equation in Table 5).The root-meansquare error (RMSE) and the adjusted R-square values are calculated based on the predicted polynomial model measurements.Above 15°, the RMSE is comparable for all the patients, which shows good correspondence between measurements and fitting curves.The higher RMSE values below 15°are due to the biomedical characteristic difference between males and females.This behavior is confirmed by the R-squared values in Table 6.The R-square values are close to 1 for all the subjects above 15°.

Conclusion
The development of a wireless, lightweight, and compact sensor system for the potential diagnosis of an anterior cruciate ligament tear has been achieved.The developed system satisfies the main requirements like wearability, and suitability for everyday use in a clinical setting to assist in the diagnosis of an ACL rupture.The system uses a capacitive strain gauge sensor connected to an electronic system enabling wireless data transmission.Additionally, two types of power supply systems were introduced in the work (1) BPS, and (2) WPTS.To verify the functioning of the sensor system irrespective of the two different power supply systems, the whole sensor unit was mounted on a knee simulator setup and at the knees of three healthy test subjects (6 data points from both knees).The results obtained from the knee simulator exhibited a relative capacitance change of ≈0.01 for every 5°angle change from 10°through 45°.The data from the male test subjects 1 and 2 showed a comparable behavior to the knee simulator's data above 15°and a lower sensitivity below.For females, the data was comparable even at lower tibial rotation angles.This difference in lower angle measurement data between the test subjects' knees is due to the varied biomechanical characteristics based on gender.This implies that there is more resistance at lower angles for tibial rotation on an actual knee for males than for females.This behavior, which is visible in the in-vivo measurements is not currently modeled from the knee simulator.Its implementation will be the objective of future work.
Above 15°the data from the in vivo measurements and the knee simulator are comparable, proving the reliability of the sensor system and the reproducibility of the The values different subjects above 15°indicate that the data has good correspondence to the predicted model for different cycles of rotation of the same sensor.This analysis proves that the developed sensor system, which is wireless, compact, and lightweight is potentially suitable for the diagnosis of ACL rupture.Its further validation will be the object of future research work.
In this work, a wearable sensor system has been developed able to sample and transmit the data at 1 kHz.The software the data reception has not been optimized for this frequency Future work will aim to achieve this goal by reducing data losses.Also, the system must be tested for people with injury in a clinical setting and more data samples must be collected.Additionally, there is a possibility to damage the sensor due to repeated handling.A reliable protective housing is necessary to increase the life span of the sensor unit.

Figure 1 .
Figure 1.Flex board with integrated sensor front-end and power supply circuit, connected to a capacitive strain gauge.

Figure 3 .
Figure 3. Schematic of the battery-powered system, consisting of the capacitive sensor, the sensor front-end, and the power supply circuit.Data are transmitted to a stationary computer.

Figure 4 .
Figure 4. Schematic of the wireless power transfer system consisting of the capacitive sensor, the sensor front-end, and a WPT section with transmitter (Tx) and receiver (Rx) coils.Data are transmitted to a stationary computer.

Figure 7 .
Figure 7. Sensor-mounted knee simulator setup with the BPS, front view (left) and side view (right).
does not require an expert opinion in accordance with the professional code of conduct of the State Medical Association of Baden-Württemberg or positive vote by an ethics committee.Therefore, the ethics committee has no objections to the use of the results."Ethics approval was gained as Low Risk Research for this study under the local Human Research and Ethics Committee.Due to the Covid19 pandemic, considering the rules of associated isolation and protective measures, the decision of the ethics committee was to allow only the members of the respective research group as it was necessary to reduce the number of subjects than initially planned to avoid exposure risks.All experiments on human skin were performed with the written consent of the participants, in accordance with all local laws, and without relevant ethical issues.

Figure 10 .
Figure 10.Battery-powered sensor and electronics mounted on the healthy knee of a test subject.

Figure 11 .
Figure 11.Wireless power transmission sensor and electronics mounted the healthy knee of a test subject.

Figure 12 .
Figure 12.Knee simulator data analysis: angle vs relative capacitance change.

Figure 13 .
Figure13.Healthy subject's data analysis (subject 1 (male): left knee): angle vs relative capacitance change with the error interval defining data variability.

Figure 14 .
Figure14.Healthy subject's data analysis (subject 1 (male): right knee): angle vs relative capacitance with the error interval defining data variability.

Figure 15 .
Figure 15.Healthy subject's data analysis (subject 2 (male): right knee): angle vs relative capacitance change with the error interval defining data variability.

Figure 16 .
Figure 16.Healthy subject's data analysis 2 (male): left knee): angle vs relative capacitance change with the error interval defining data variability.

Figure 17 .
Figure 17.Healthy subject's data analysis (subject 3 (female): right knee): angle vs relative capacitance change with the error confidence interval defining data variability.

Figure 18 .
Figure 18.Healthy subject's data analysis (subject 3 (female): left knee): angle vs relative capacitance change with the error confidence interval defining data variability.

Table 1 .
Allocation of the examination phase to the device mode, and related drain of battery capacitance.

Table 2 .
Wireless power transfer system.

Table 4 .
Parameter values of second-order polynomial fit.
Table 5 below shows the comparison of the RMSE values of all three healthy subjects' knees (right and left) below and above 15°.

Table 5 .
Comparison of rmse below and above 15°.

Table 6 .
Comparison of r-square below and above 15°.