Stretchable triboelectric sensor for measurement of the forearm muscles movements and fingers motion for Parkinson's disease assessment and assisting technologies

2.1 Fabrication of Ecoflex™

The polymer triboelectric layer is fabricated with a 1:1 mixture of components A and B of commercial Ecoflex™ 00-30 from Smooth-On as described in Zou et al. (2019). After 5 h in ambient temperature, the material is retired from the mould and cut in the dimensions described in Section 3.1.

2.2 Fabrication of the printed circuit board

The board is designed in the licensed software Altium and fabricated with a commercial manufacturer. All the components except the Battery are Surface Mounted Devices (SMD) to achieve a small size. The BLE board is an MDBT50Q module from Raytac Corporation, based on the NRF52840 Bluetooth chip from Nordic Electronics. The small lithium-ion battery is attached with Isolating tape to the PCB. Finally, the Ecoflex™ and Al layers are attached to the sensor as graphically explained in Figure S2. In Figure S4, the signal is received by the Sparkfun Pro nRF52840 commercial board which sends the data to a Python code for posterior visualization and storage.

3.1 Triboelectric-based forearm motion sensor integrated into a PCB: Design and characterization

As illustrated in Figure 1a, the sensor system is assembled by attaching a thin, flexible layer of Ecoflex™ to a small printed circuit board (PCB), which is in charge of the signal processing and transmission, and possesses a small electrode made with aluminium tape. The Ecoflex™ layer is attached to the PCB through two small rectangles of Polyurethane (PU) Foam that allow a separation space between the Ecoflex™ and the Al electrode. The dimensions of the PCB, the Ecoflex™ layer and the Al electrode are 3.5 cm × 4 cm, 4 cm × 1.4 cm × 1.5 mm, and 5 mm × 3 mm, respectively (Figure S1 shows a clearer assembly process of the sensor).

The sensor is attached to the forearm on top of the flexor muscles/Tendons community (Figure 1b,c). As observed in Figure 1c, the sensor is located on top of the skin above the muscles or their inserted tendons such as the flexor carpi radialis (FCR), the flexor digitorium superficialis (FDS), flexor digitorium profundus (FDP), the palmaris longus (PL) and the flexor pollicis longus (FPL) which together are in charge of moving and bending the fingers (proximal, intermediate and distal Phalanges), and the wrist. When the hand is opened or closed, the muscles and tendons contract or relax accordingly. The sensor also could be placed on top of the community of tendons connected to such muscles. In both cases, the contraction or relaxation changes the volume of the muscles and the skin on top of them also changes. As observed in Figure 1di, when the hand opens, the volume on the skin seems to expand the tissue making the Ecoflex™ layer moves towards the AL electrode on the PCB (Clear view in Figure S2). When the polymer and the conductor make contact, the triboelectric effect allows positive charge transfer from the dielectric to the conductor as described in the triboelectric series (Lee, no date). When the hand starts closing, the volume of the forearm reduces and the skin and the muscles go back to their initial positions, moving the Ecoflex™ (negatively charged) away from the AL (Figure 1dii) until it reaches the maximum separation (Figure 1diii). Due to the distance, the positively charged Al attracts negative charges from the ground reference through the resistive load. The last, results in current and voltage generation, in this case, positive according to the sign convention on Figure 1d transfer negatively and positively according to the triboelectric series (Lee, no date). Posteriorly, when the hand opens again, the negative Ecoflex™ moves closer to the AL conductor, pushing the negative charges on the electrode back to the ground reference and resulting in an opposite voltage. When the motion dynamics of the fingers, hand or wrist are periodic, the voltage signal behaves the same. Furthermore, the characteristic of the electric output also depends on the movement amplitude and speed of the forearm's skin and muscles that control the hand and the phalanges.

Additionally, as observed from Figure S2, the maximum gap of the sensor varies approximately from 2 mm to 0 mm. Besides, the area of the Ecoflex™ is also small, and the area of the Al electrode is even smaller. For such reason, The PCB has a processing circuit equal to the one used in the past in (Vera Anaya et al., 2020) for the sensing of the small contraction–relaxation motion of the orbicularis oculi muscle surrounding the eye. The circuit serves as a high impedance load for the sensor, amplifies the voltage signal and fixes a DC offset of 1 volt with a maximum amplitude swing from 0 V to 3.3 V (Figure S3). The signal is later digitalized by the Analog-to-Digital Converter (ADC) of one of the input channels of the BLE module NRF5832 from Nordic Electronics which transmits the data wirelessly to a computer (Figure S4).

3.2 Hand, arm and fingers movement examination for preliminary assessment of localized Bradykinesia and Rigidity for Parkinson's disease

3.2.1 Hand and pronation–Supination Movement Examination

According to the MDS-UPDRS rating scale for diagnosing Parkinson's disease, physicians must test specific body motor characteristics to determine the grade of dementia of the patient. The output can be classified with 0 (Normal), 1 (Slight), 2 (Mild), 3 (Moderate) and 4 (Severe). Some physical tests to assess localized bradykinesia (slowness of movement) or rigidity due to PD closely evaluate the hand and arm motion characteristics. In Figure 2a, it shows the voltage output of three different severity levels of PD observed in the hand opening/closing motion, classified using the MDS-UPDRS scale. For this initial experiment, the hand motion was emulated by the user under test, simulating the specific amplitude and time characteristics of each classification level in exercise 3.6 of the MDS-UPDRS manuscript, related to ‘Hands Movement Examination’. In this assessment, the user is required to open and close the fist approximately 10 times (Figure 2ai). Figure 2aii presents the output when no PD problem is simulated (level 0). The signal has a DC offset in 1 V and is negative (lower than 1 V) when the hands close and positive (Higher than 1 V) when the hand opens. Having in mind that the amplifier in Figure S3 in the supplementary information has negative gain, the waveform in Figure 2aii shows coherence with the theory explained in Section 3.1. When the hand opens the sensor voltage is negative and changes sign when the signal is passed through the negative gain amplifier, being positive at the output and similarly happens when the hand closes. Furthermore, there are 9 clean and recognizable cycles with the maximum amplitude allowed by the ADC of the BLE (0 V–3.3 V). In contrast, in Figure 2aiii, a slight (level 1) slowing of the fist opening and closing is emulated. The waveform characteristic shows a negative peak followed by a positive peak for closing fist and a positive peak followed by a negative peak for opening fist. Evidently, the cycle duration now increases and the fluctuation of the signal may probably be due to the non-uniform muscles of the muscles and the skin more notorious in this slow movement. On the other hand, the amplitude of the signal did not change, since the opening and closing amplitude was not changed. Differently, Figure 2aiv presents a difference in the duration of the opening–closing cycle and a decaying amplitude which is the product of the mimic performed by the user emulating moderate symptoms (level 3). As expressed on the MDS-UPDRS manuscript, a PD patient in level 3 stops or reduces the amplitude of the movement in a short time after starting the exercise. The waveform clearly shows such behaviour, demonstrating the potential of the sensor to be used for the quantization of the symptoms in this test. The physical demonstration of the outputs in Figure 2 can be observed in Video S1.

In Figure 2b, it also presented the result for exercise 3.6 in the MDS-UPDRS scale related to the ‘Pronation–Supination Movement Examination’, which is an important indicator for PD diagnosing. For this test, the user is asked to perform up to 10 pronation–supination movements of the forearm (Figure 2bi, Video S2). Once again, the user emulates the action based on the different severity levels. Figure 2bii portraits the output when the person performs the action naturally without any problem (Level 0). The supinating action generates a negative voltage (<1 v in this case) followed by a positive voltage. The pronation action also generates a similar voltage, with a high amplitude, probably higher than 3.3 V, explaining why the signal looks clipped or saturated at the peaks. Besides, the signal looks clean and consistent for different cycles. In contrast, in Figure 2biii is the waveform that results when the user emulates actions with mild slowing without decreasing the amplitude of the motion. The time for each cycle is longer now, and there is a difference between both actions, with a longer time for the supination as well as a more distorted signal shape than the pronation. The amplitude is practically unchanged because the user was asked to keep the same as in the test in Figure 2bii. Finally, in Figure 2 is evident the high power of the signal compared with the inherent electric noise, which is due to the electronic filtering in the PCB.

3.2.2 Finger Tapping Examination

Another experiment is performed to assess exercise 3.4 on the scale related to the ‘Finger Tapping Examination’. The user is required no to tap the index finger with the thumb as in Figure 3a up to 10 times. In Figure 3 and Video S3, the output from finger tapping is also portrayed for different levels of rigidity/slowness. Figure 3b presents the signal when the user taps the finger without emulating any bradykinesia problem. When the finger extends, the signal is above 1 V (considered negative for explanation purposes), and when the finger taps, the signal is negative or below 1 V, coherent with the results for opening and closing hand. Nonetheless, the voltage amplitude is lower than the one obtained in Figure 2. This is expected since now just one finger is moving instead of the whole fist. The thumb, in this case, is in the same position so the action of the FPL muscle is less compared when the whole fist closes. A similar rational inference can be done with the rest of the forearm muscles and tendons that may have less influence in the motion of one individual finger. A mild slowing with lower motion amplitude of finger tapping (Level 2) results in the waveform in Figure 3c. The cycle duration increases due to the user slowing the motion on purpose, but there are particular characteristics in the amplitude. When the finger closes, the amplitude is less negative compared with Figure 3b. In some cycles, the amplitude when the finger taps is also lower than the one obtained when no rigidity problems, but it may be due to involuntary inconsistent tapping amplitude from the user under test. Finally, the user was also asked to emulate moderate symptoms (level 3) whose result is presented in Figure 3c. As observed, the amplitude decreases soon after starting the tapping action, which is contemplated in the description of a moderate level. Hence, the results also demonstrate the positive potential of the sensor in the assessment of the localized symptoms due to finger kinetics.

3.3 Tremor vibration assessment

Tremor is another important symptom in patients with Parkinson's disease. Its assessment is contemplated on exercises 3.15–3.17 of the MDS-UPDRS scale which evaluates postural, kinetic and rest tremor. The physician must observe these symptoms during all the evaluation process of the patient and annotate the estimated amplitude of the tremor. In Figure 4a, the user under test is asked to vibrate with different amplitudes the entire forearm separated from the body while keeping the wrist and hand static, to emulate a similar movement to postural tremor. The individual moves the arm with a low amplitude tremor until reaching a high amplitude tremor of about 5 cm. In Figure 4aii, it is observed how the peak voltage amplitude increases and then continuously decreases when the forearm vibration is stopped. According to this result, the amplitude is directly dependent on the tremor amplitude and if further characterization is performed, a more accurate model can be used to associate the amplitude of the signal with the physical amplitude of the tremor.

Posteriorly, the person under test is also asked to just vibrate the wrist keeping the forearm straight, starting from a low amplitude vibration to a greater amplitude vibration as in the first experiment (Figure 4bi). The outcome of this experiment in Figure 4bii shows also a signal that increases related to the vibration amplitude and then stops. Although the behaviour of the two vibration waveform is similar, the vibration from the wrist generates lower voltage than the vibration of the forearm. Thus, the preliminary experiment in Figure 4 and Video S4 exhibits a direct association between the voltage and the amplitude, and more research and more testing should be done on actual patients to understand more characteristics of the tremor and rate the vibration according to the main PD severity scale.

3.4 Output of the forearm TENG-based sensor due to different moving fingers

As noticed in Figure 3, the assembled sensor is sensible to finger motion. In the specific case of Parkinson's disease assessment, the finger under testing must be the index. However, such an experiment can also be done for the rest of the fingers (Video S5). In Figure 5a, the waveform corresponding to a fast and slow taping motion of the middle finger is presented. In this case, the output is even greater than the result for the index finger in Figure 3. Besides, Figure 5b shows the output when the middle finger is moved with a decaying amplitude, showing a direct relationship between the motion and the voltage. When the same test is repeated in Figure 5c using the third finger (ring finger), the signal amplitude is smaller (but not neglectable), in comparison with the result obtained with the index and middle finger. Something similar happens in Figure 5d, where the output due to the taping of the small finger is weak compared with the rest of the fingers. Interestingly, the results here indicate the dependency of the output due to the motion of a specific finger, proof that the placement of the sensor in the forearm influences the detection of the phalanges. Each of them is connected to each muscle through a tendon, and each specific tendon under contraction changes the shape of the moving skin. Thus, the skin which does not deform equally in all parts affects the movement of the flexible polymeric layer towards the AL electrode. Therefore, in Section 3.5, we study the signal resulting due to different placements of the sensor on the forearm, where specific motion is more sensitive to certain finger movements.

3.5 Output due to different finger motions for specific sensor positions on top of the forearm flexor muscles and tendons

The results in Section 3.4 indicated that when measuring the forearm muscle dynamic for PD assessments the signal output due to each finger is influenced depending on the placement of the sensor on top of the forearm. As observed in Figure 6a, the muscles that control the finger flexion like the FDS or FDP are connected to four tendons to a specific phalange. In literature, for example the measurement and recognition of finger motions are done by using either EMG sensors or Inertial sensors (Li et al., 2015; Yu et al., 2010). Accelerometers as in (Yu et al., 2010) use the skin stretching due to the motion of muscles and tendons to achieve recognition of the fingers. However, in this report, we can achieve recognition of the fingers implementing the designed sensor. As in Figure 6b, the sensor can be placed closer to each of the tendons to aim single finger monitoring. When the muscle and tendon contracts to allow flexion of the finger, the skin close to such tendon also changes in volume, and the triboelectric sensor on top generates voltage due to such small motion (See Video S6). For instance, when the sensor is placed in position A, closer to the tendon that controls the index finger, the tapping motion of such a digit generates a greater voltage than the motion from the tapping of other fingers (Figure 6ci). We can observe in the same figure (Figure 6cii), how different is the signal value when the finger is taped against the thumb, and when the finger is just bent back and forth without tapping, with lower amplitude. In this position, the other fingers also generate a small signal, but not comparable to the values obtained from the index. When the sensor is placed in position B in Figure 6b, it is evident that the motion of the middle finger generates a higher peak value than the motion of the other fingers as illustrated in Figure 6di. The tapping motion also generates a higher signal than the no tapping motion (Figure 6dii). Differently, when the sensor is placed in position C in Figure 6b the signal coming from the motion of the ring and the small finger is quite similar, for taping and no taping motion (Figure 6e). This may be because, as portrayed in Figure 6a, both tendons are very close to each other and the triboelectric sensor is not small enough to detect the individual motion. Finally, we can observe that the result when the location is position D (Figure 6f), only the small finger contributes to a significant voltage output when moved, compared with the outcome of other fingers. In this case, the thumb alone is not evaluated because is mostly controlled by the FPL muscle which is below the FDS and the FDP, which are the main muscles and tendon communities under evaluation in this case.

3.6 Other applications for finger motion monitoring: Hand activity monitoring when using keyboard and finger bending sensing

Finally, some applications related to finger kinetics recognition were tested and graphically summarized in Figure 7 to demonstrate more features of the developed triboelectric sensor for forearm muscles/tendons monitoring. With the sensor, it is possible to track the activity of the hand. For instance, in Figure 7a and Video S7, it is presented the output signal when the person is typing on the keyboard. It differentiated the moments when the person uses the right hand and the left hand, which has the sensor. This can be helpful in activity and productivity monitoring to indicate when the person should stop typing and start resting. This will be important in the future of ergonomic and health fields to avoid the symptoms of certain upper-limb problems due to repetitive movements or bad hand posture, such as Carpal Tunnel Syndrome (Thurston, 2013).

Finally, the other experiment is related to finger bending monitoring (Figure 7b, Video S8). Multiple finger bending measurement systems based on TENG have been achieved before as the one proposed in (He et al., 2019). However, those devices require to be attached to each finger to sense the bending angle. In contrast, in this project, we used the designed forearm TENG to track finger bending. Figure 7b presents the result when the middle finger is bent 15, 45 and 75 degrees. The bending angle is directly related to the voltage peak value, being more evident in the positive peak (above 1 V for this manuscript) which corresponds to the finger closing action. Similar to Figure 3 finger extension generates a lower voltage absolute value than the finger contraction, which is the resulting peak after the triboelectric contact. Finally, the last designed sensor topology improves the portability and comfortability for long-term use if continuous finger bending monitoring is needed in the future, leaving the hands and finger free from any wearable transducer. The result also indicates that the proposed topology can be used for the development of future human–machine interfaces controlled by human muscles.