An Intelligent Wristband for Simultaneous Multiparameter Measurement during Fumigation and Washing Therapy of Traditional Chinese Medicine

Traditional Chinese medicine (TCM) is one of the oldest healing systems, which plays an essential role in daily health management. However, TCM is frequently controversial due to its subjectivity and lack of scientific quantification. Modern sensing technologies, enabling the detection of various physiological signals, have the remarkable performance of multifunctionality, superintegration, and ultraminiaturization. Combining with modern sensing technology is a good opportunity for the modernization, objectification, and scientization of TCM. This article proposes an intelligent wristband for simultaneous multiparameter measurement of pulse, skin temperature, and perspiration, essential physiological signals in TCM. As a significant functional material, a carbon nanotube/poly(dimethylsiloxane) nanocomposite with 3D interpenetrating network structures exhibits great electrical conductivity, stress effect, and thermoresistive effect, ensuring intelligent wristband's high accuracy, high sensitivity, and good stability for pulse and temperature measurements. Furthermore, the perspiration sensing unit is also integrated. The as‐prepared intelligent wristband is applied during fumigation and washing therapy, and artificial intelligence is introduced to judge whether the user is during fumigation and washing therapy or in the resting state by classifying the user's physiological state, with an accuracy of up to 100%. This work offers a unique strategy for TCM and wearable sensing technology with an important practical significance.


Introduction
Traditional Chinese medicine (TCM), originating from oriental philosophy and culture, has played an essential role in the Chinese and global healthcare systems for thousands of years. [1]M, for example, has been widely utilized in COVID-19 therapy for the last 3 years. [2]owever, TCM is frequently controversial compared with Conventional Western Medicine due to its subjectivity and lack of scientific quantification. [3]Specifically, TCM employs four traditional examination methods: looking, listening and smelling, asking, and feeling the pulse. [4]TCM examination and diagnosis findings mainly depend on the ability and experience of TCM doctors themselves, which are sadly subjective and difficult to quantify scientifically.TCM's subjectivity and lack of scientific quantification impede the acceptance of TCM by broader groups and hamper TCM's inheritance and development. [5]owever, it is still difficult to scientifically explain the mechanism of traditional Chinese medicine at the present stage.It is critically necessary to conduct quantitative scientific research and objectification of TCM. [6]Wearable sensing technology has the remarkable performance of multifunctional, superintegration, and ultraminiaturization, thus significantly developing in medicine and health. [7]In recent years, wearable sensing technology has attracted wide attention in prosthetics, [8] health monitoring, [9] smart robots, and human-machine interfaces.The organic combination of wearable sensing technology with TCM will innovatively handle the subjective and lack of scientifically quantified problems, Traditional Chinese medicine (TCM) is one of the oldest healing systems, which plays an essential role in daily health management.However, TCM is frequently controversial due to its subjectivity and lack of scientific quantification.Modern sensing technologies, enabling the detection of various physiological signals, have the remarkable performance of multifunctionality, superintegration, and ultraminiaturization. Combining with modern sensing technology is a good opportunity for the modernization, objectification, and scientization of TCM.This article proposes an intelligent wristband for simultaneous multiparameter measurement of pulse, skin temperature, and perspiration, essential physiological signals in TCM.As a significant functional material, a carbon nanotube/ poly(dimethylsiloxane) nanocomposite with 3D interpenetrating network structures exhibits great electrical conductivity, stress effect, and thermoresistive effect, ensuring intelligent wristband's high accuracy, high sensitivity, and good stability for pulse and temperature measurements.Furthermore, the perspiration sensing unit is also integrated.The as-prepared intelligent wristband is applied during fumigation and washing therapy, and artificial intelligence is introduced to judge whether the user is during fumigation and washing therapy or in the resting state by classifying the user's physiological state, with an accuracy of up to 100%.This work offers a unique strategy for TCM and wearable sensing technology with an important practical significance.furthering TCM's scientific, objective, and contemporary development.Likewise, wearable sensing technology enters a new field with enormous developmental potential.
Although many researchers have done much work to this end, and several TCM-based sensors, such as the diaphragm-based optical fiber pulse sensor, [10] the robotic tonometry pulse sensor system, [11] Bi-Sensing Pulse Diagnosis Instrument (BSPDI), [12] and the smartphone-based sensor system for TCM pulse diagnosis [13] have recently been reported.The present TCM-based sensors still need further improvement.One of the common critical issues is that their functions primarily focus on pulse measurement.In fact, during the TCM diagnosis, doctors employ not only pulse but also many other physiological parameters such as body temperature, perspiration, and so on to judge the physiological condition of patients.The above examples of TCM-based pulse sensors are incomprehensive in parameter measurement, thus failing to meet users' needs during TCM diagnosis and treatment. [14]Then, most TCM-based sensors have large volumes and cannot be worn by users, reducing the comfort of use and limiting their application in primary hospitals, especially in family healthcare.Furthermore, many of these reported TCM-based sensors are data recorders rather than analysts, [11b] which means a lot of physiological data obtained by TCM-based sensors still need to be analyzed by doctors or other professionals, and the convenience will be significantly reduced.So, it is in desperate need to develop TCM-based wearable sensors that can realize simultaneous multiparameter measurement and combine artificial intelligence to analyze physiological data.
We herein propose and design a TCM-based intelligent wristband for simultaneous multiparameter measurement (Scheme 1).Based on the asking and pulse feeling in four traditional examination methods in TCM, we select three physiological parameters of great significance to the human body: pulse, body temperature, and perspiration, as the target parameters of the intelligent wristband.For simultaneous pulse, body temperature, and perspiration measurement, the intelligent wristband was designed into three independent sensing units: pulse, temperature, and perspiration.We also designed and fabricated the carbon nanotube (CNT)/poly(dimethylsiloxane) (PDMS) nanocomposites with a 3D interpenetrating network structure.The CNT/PDMS nanocomposites have great electrical conductivity, flexibility, and elasticity and are critical functional materials for a simultaneous pulse, body temperature, and perspiration measurement.Subsequently, the cloth-based CNT/PDMS nanocomposite was also fabricated as the main component of the perspiration sensing unit of the intelligent wristbands.We exhibited the application of the intelligent wristband during TCM's fumigation and washing therapy.The pulse, body temperature, and perspiration data of users in the resting state and during fumigation and washing therapy were collected by intelligent wristbands.Artificial intelligence was introduced to analyze the data acquired by intelligent wristbands.With the help of the tight combination of intelligent wristbands and artificial intelligence, whether the users were in the state of fumigation and washing therapy could be accurately distinguished by the intelligent wristband based on their biological data.To sum up, it is expected to promote TCM's further scientific, objective, and modern development through the organic integration of wearable sensor technology, artificial intelligence, and TCM.

Design of the Intelligent Wristband
The intelligent wristband with three independent sensing units, namely, the pulse sensing unit, the temperature sensing unit, and the perspiration sensing unit, was designed based on flexible and 3D interpenetrated CNT/PDMS nanocomposites (Figure 1A).To prepare such CNT/PDMS nanocomposites, a one-step solution-phase approach by polymerized predispersed CNT/PDMS homogeneous ink was adopted.CNTs were fixed via the cross-linking of PDMS to form a stable and 3D interpenetrating network structure (Figure S1, Supporting Information).After removing the solvent, a highly flexible and elastic black CNT/PDMS nanocomposite could be obtained (Figure 1B and S1, Supporting Information).Scanning electron microscope (SEM) images show both the CNT/PDMS films' surface and cross section with an isotropic and 3D "nano-felt" structure (Figure 1C,D).To investigate the possible reactions during the preparation, Raman spectra, Fourier transform infrared (FT-IR) spectra, and thermogravimetry (TG) were employed.Strong Raman peaks at 1350.0 cm À1 (D band), 1585.0 cm À1 (G band), and 2700.0 cm À1 (G 0 band), indicating the hexagonal lattice of CNT, could be observed in the spectra of both CNTs and the CNT/PDMS nanocomposite (Figure S2A, Supporting Information). [15]Raman spectra of PDMS that display adsorption peaks in 2940.6,2888.0,689.1, and 490.8 cm À1 represent -CH 3 asymmetric stretch vibration, -CH 3 symmetric stretch vibration, Si-CH 3 symmetric rocking vibration, and Si-O-Si stretch vibration, respectively, can also be observed in the CNT/PDMS nanocomposites' spectra. [16]The FT-IR spectra of PDMS that display adsorption peaks in 2960.0,1260.0,1080.0, and 860.0 cm À1 represent -CH 3 stretching vibration, -CH 3 bending vibration, Si-O-Si stretching vibration, and -CH 3 rocking vibration, Scheme 1. Wearable sensing technology and artificial intelligence are innovatively introduced into TCM and promote its further development.We designed and manufactured an intelligent wristband for simultaneous pulse, temperature, and perspiration measurement based on the CNT/PDMS nanocomposites.Assisted with artificial intelligence, the intelligent wristband can accurately identify whether users are during fumigation and washing therapy or in the resting state.
respectively, can also be observed in the CNT/PDMS nanocomposites' spectra (Figure S2B, Supporting Information). [16]urthermore, the weight loss data from the thermal degradation of the PDMS and the CNT/PDMS nanocomposites were well consistent (Figure S2C, Supporting Information).All results validate that no chemical reaction occurred between CNT and PDMS during nanocomposites preparation.
As the vital functional materials of proposed intelligent wristbands, the CNT/PDMS nanocomposites are urgently required to own relatively superior conductivity, mechanical properties, stability, and biosafety.The electric conductivity of the CNT/PDMS nanocomposites was investigated by linear scanning voltammetry (LSVs) (Figure 1E).It could be found that the conductivity of the CNT/PDMS nanocomposites increased with the increasing CNT content and reached 849.2 S m À1 at a CNT content of 13% (Figure S3A, Supporting Information).The results indicated that the electric conductivity of CNT/PDMS nanocomposites mainly came from the incorporated CNTs.The mechanical properties of the CNT/PDMS nanocomposites were represented by strain-stress data obtained from the fracture mechanics test (Figure 1E).Contrary to the conductivity, the flexibility and elasticity of the CNT/PDMS nanocomposites decreased with the increasing CNT content (Figure S3A, Supporting Information).It is worth mentioning that there were no apparent changes in conductance after 500 times continued electrical tests and only a slight decrease in elasticity modulus after 160 times multiple cyclic tensile tests, even at high CNT content (13% in this case), which implied both electrical and mechanical stability of the CNT/PDMS nanocomposites (Figure S3B,C, Supporting Information).The great flexibility, elasticity, and adhesive property of CNT/PDMS nanocomposites are intuitively displayed in Figure 1F.Considering the compromise between mechanical properties and electrical conductivity, the CNT/PDMS nanocomposites with a CNT content of 13% exhibit suitable mechanical properties (Young's modulus of 12.04 kPa) and electrical conductivity (849.2 S m À1 ) to meet the requirements of the intelligent wristband for simultaneous multiparameter measurement.To show the stability of CNT/PDMS nanocomposite applied to human beings, a prepared CNT/PDMS film was immersed in 36.0 °C artificial sweat to simulate the physical environment on the skin surface, and its conductivity was measured every day (Figure S3D, Supporting Information).Even after 8 days, the almost identical conductivity indicated that the CNT/PDMS nanocomposites showed good stability in their usage scenarios.The cytotoxicity and biocompatibility of the CNT/PDMS nanocomposites were investigated using human immortalized keratinocytes (HaCAT).After cells were cultured in the medium containing 0, 2.5, 5.0, and 10 μg mL À1 CNT/PDMS extracts for 24, 48, and 72 h, results showed that CNT/PDMS nanocomposites did not affect cell viability, implying the ideal biocompatibility of the CNT/PDMS nanocomposites (Figure S4, Supporting Information).Furthermore, the CNT/PDMS nanocomposites can easily be processed into various patterns, essential for designing and fabricating intelligent wristbands (Figure 1G).

Pulse Sensing Unit in the Intelligent Wristband
In TCM, the pulse at the radial artery is commonly divided into three sections: Chi, Guan, and Cun (Figure S5, Supporting Information), [17] and the whole pulse signals can only be obtained by simultaneously measuring the pulse at abovementioned three positions.Therefore, in the pulse sensing unit, three pieces of CNT/PDMS films were ranged and corresponded to Chi, Guan, and Cun, respectively (Figure 2A 1 ).When there was a pulse beat, the CNT/PDMS films were compressed vertically and stretched horizontally.Both deformations may lead to the resistance changes of CNT/PDMS films according to the law of resistance.Figure 2B shows the relationships between the resistance changes and the applied stresses on the CNT/PDMS nanocomposites.It could be found that the resistance increased linearly with force from 0 to 0.7 N (the red area in Figure 2B), which, however, increased exponentially when applied force over 0.7 N (the blue area in Figure 2B) and could not be explained by the law of resistance.Therefore, to explain the stress effect principle of the CNT/PDMS films, their stress effect was divided into two phases (Figure 2A 2 ).The relationship between stress and resistance in phase 1 (Equation 1) and phase 2 (Equation 2) was both fitted where R and σ represent resistance and stress.In phase 1, the resistance change was caused by the change in length and cross-sectional area.According to the resistance formula and conductivity formula where ΔR, ρ, ΔL, ΔS, and σ represent resistance changes, resistivity, length changes (equal to L p À L 0 ), cross-sectional area changes (equal to S p À S 0 ), and conductivity.The change in the length and the cross-sectional area of the CNT/PDMS film led to the linear change of its resistance.When the stress increases to 0.7 N, the stress effect of the CNT/PDMS film steps into phase 2, in which not only the length and the cross-sectional area of the CNT/PDMS film change but also the structure of the CNT/PDMS film at the molecular level changes as well.The relative displacement of CNTs caused by the stress increases the distance between some CNTs, resulting in the disappearance of the quantum tunnel effect and the overall conductivity decrease. [18]Hence, in phase 2, with the stress increase, CNT/PDMS film resistance increases exponentially.
The CNT/PDMS film's improved stress sensitivity in phase 2 is beneficial to measure the weak pulse beat.In this viewpoint, a pre-applied constant stress (CS) on CNT/PDMS nanocomposite may dramatically increase the sensitivity of the pulse sensors.Figure 2C compares the electric response to stress with and without a pre-applied CS.It has been found that the 0.2 N stress caused 3.014 Ω changes, which was 10.30 Ω when 0.7 N CS was pre-applied.The current change of the CNT/PDMS nanocomposites at different stresses was tested with and without CS (Figure 2D).With increasing stress, the current change of the CNT/PDMS nanocomposite increased for both situations.However, the CNT/PDMS nanocomposites under CS overall had larger current changes than that without CS.Therefore, with the help of CS, the sensitivity of pulse measurement by the pulse sensing unit was greatly improved from 15.07 Ω N À1 (without CS) to 41.87 Ω N À1 (with 0.7 N CS).In our pulse sensing unit, a water sac was utilized to provide CS via pushing against the wrist, enhancing the sensitivity of CNT/PDMS nanocomposites to pulse beat.The stability of the stress effect of the CNT/PDMS nanocomposites under CS was also evaluated (Figure 2E).After cyclic stress 2000 times under CS, the electrical conductivity of the CNT/PDMS nanocomposites remained stable, demonstrating the great stability of the stress effect of the CNT/PDMS nanocomposites.
We then tested the performance of the pulse sensing unit on the body.During the measurement, the intelligent wristband was tight worn at the wrist, and the pulse sensing unit was aimed at the radial artery.The stable pulse measurement comes from the sensitivity improvement of the pulse sensing unit with CS and the tight and stable attachment between the CNT/ PDMS films and the wrist.Three pieces of pulse data, recorded by the pulse sensing unit simultaneously, come from position Chi, Guan, and Cun, respectively (Figure 2F).The details about pulse rate and pulse waveform are presented in the figure.Each piece of pulse data was measured independently by corresponding CNT/PDMS films using individual measurement channels to avoid conflicts between pieces of CNT/PDMS films.One pulse beat was picked from the measured data, and its 3D map was drawn based on the pulse data at three positions simultaneously measured by the pulse sensing unit (Figure 2G).For validating the accuracy of the pulse sensing unit, a commercial pulse oximeter was applied to measure photoplethysmography (PPG) signals while the pulse data were simultaneously detected by the pulse sensing unit (Figure 2H).A bivariate correlation analysis was subsequently introduced to evaluate the correlativity between the pulse data obtained by the pulse sensing unit and the PPG data from the commercial pulse oximeter.The correlation coefficients revealed close correlations between the pulse data and PPG data (R = 1, p < 0.01), validating the accuracy of our pulse sensing unit.

Temperature Sensing Unit and Perspiration Sensing in the Intelligent Wristband
Skin temperature, especially in TCM, is a significant physiological parameter that has already been divided and described into many different situations, such as alternating chills and fever and menstrual fever, to indicate health conditions. [19]To measure skin temperature, we utilized the thermoresistive effect of the CNT/PDMS nanocomposites: with the temperature of the CNT/PDMS nanocomposites rising, their resistance decreases on the contrary.The 1D variable range hopping (VRH) model was introduced to explain the thermoresistive effect of the CNT/PDMS nanocomposites where Δσ, σ 1D , T 1D , and ΔT represent conductivity change, proportionality constants, activation energy in the 1D VRH model, and temperature change. [20]In the temperature measurement process, the temperature sensing unit was tightly attached to the skin, resulting in heat conduction from the skin to the temperature sensing unit (Figure 3A).With temperature increasing, electrons were thermally activated, and their tunneling probability would be significantly enhanced, enabling the conductivity of the CNT/PDMS nanocomposites to increase. [18]The thermoresistive effect of the CNT/PDMS nanocomposites was visually presented via a multistep heating test (Figure 3B).When temperature increases, the conductivity of the CNT/PDMS nanocomposites increases, increasing When the temperature was constant, the conductivity of the CNT/PDMS nanocomposites and the current were kept constant.Moreover, the electrical conductivity of the CNT/PDMS nanocomposites was also measured under 250 times heating-cooling cycles (Figure 3C).During the test, the speed and degrees of change the CNT/PDMS nanocomposites' conductivity remained relatively stable.All test results indicated that the thermoresistive effect of the CNT/PDMS nanocomposites has high sensitivity and good repeatability, which is of great significance for temperature measurement by the temperature sensing unit.
To test the actual use performance of the temperature sensing unit, the relationship of the temperature sensing unit between temperature and current needs to be confirmed first.The current curves of the temperature sensing unit at different temperatures were tested (Figure S6A, Supporting Information), and the relationship between temperature and current was fitted (Figure S6B,C, Supporting Information) where I and T represent current (mA) and temperature (°C).The current changes in the temperature sensing unit indicated the temperature changes of users.According to the calculation result of Equation ( 5), the sensitivity of the temperature sensing unit is 1.389 Â 10 À3 mA K À1 .Next, the temperature sensing unit measured room temperature at two positions and skin temperature at the wrist (Figure S6D, Supporting Information).Moreover, the temperature at three positions was simultaneously photoed by an infrared camera.The partial correlation analysis analyzed the correlation between the temperature obtained by the infrared camera and calculated by Equation (5) based on the temperature sensing unit data.The correlation coefficients revealed close correlations of the temperature obtained in two ways (R = 1, p < 0.01), validating the accuracy of the temperature sensing unit in the actual use.
In TCM, sweat, regarded as one kind of jin ye (body fluid), regulates body temperature, moisturizes skin, and discharges waste.The perspiration measurement by the intelligent wristband largely relies on the design of the perspiration sensing unit.First, the cloth-based CNT/PDMS nanocomposites were prepared and patterned (Figure S7A, Supporting Information).The CNT/PDMS nanocomposites coated the cloth fiber to make it conductive.An area of cloth between patterned cloth-based CNT/PDMS electrodes is considered the sensing area.The resistance changes in the sensing area will lead to the resistance change of the whole circuit, suggesting perspiration.Except for the sensing area and wire junctions, all the rest areas were coated with wax to avoid sweat absorption at the non-sensing area affecting the results of measurements (Figure 3E).In the process of perspiration sensing, the perspiration sensing unit of the intelligent wristband was attached to the skin.When the skin starts to sweat, the sweat is absorbed by the sensing area of the perspiration sensing unit.Ions in the sweat decrease the resistance of the sensing area and connect cloth-based CNT/PDMS electrodes to form conductive passages so that perspiration can be measured via the change of resistance.
Artificial sweat was added to the sensing area of the perspiration sensing unit to imitate skin perspiration for studying the property of the perspiration sensing unit.Figure 3F shows that the cyclic voltammetry (CV) curve of the perspiration sensing unit significantly differed with and without the artificial sweat addition.Through many measurements, the current change of 0.03 mA was determined as the threshold of the perspiration sensing unit.When the current changes of the perspiration sensing unit were over 0.03 mA, the user, at this time, can be classified as perspiration.Then, the perspiration sensing unit was subjected to repeated artificial sweat adding and cleaning 35 times, and its outstanding sensitivity and repeatability of perspiration sensing were maintained (Figure 3G).To validate the perspiration sensing unit's reliability, wrist perspiration situations at six different time points were measured by the perspiration sensing unit (Figure S7B, Supporting Information).The wrist skin was also recorded by taking photos simultaneously (Figure 3H).A bivariate correlation analysis was introduced to evaluate the correlativity between the current of the perspiration sensing unit and whether skin photos can observe perspiration.The correlation coefficients revealed close correlations (R = 0.865, p < 0.01), indicating that the perspiration sensing unit can accurately determine whether perspiration is and has high sensitivity and repeatability.

Simultaneous Multiparameter Measurement during Fumigation and Washing Therapy of TCM
The intelligent wristband was utilized to judge if the physical state of volunteers is in fumigation and washing therapy of TCM or not by its simultaneous multiparameter measurements results of pulse, temperature, and perspiration, for example.Fumigation and washing therapy are the external treatment in TCM that fumigates and washes the skin or affected area by warming herb liquids. [21]In the process of fumigation and washing therapy, on the one hand, warming liquids allow the expansion of the skin's blood capillaries and speed up blood circulation; on the other hand, the active ingredient in the herb liquids can enter the blood circulation through the skin. [22]herefore, fumigation and washing therapy has been widely used in the treatment of knee osteoarthritis, [23] development dysplasia of the hip in children, [24] psoriasis vulgaris, [25] and so on.Volunteers were asked to soak their feet in warm water at around 40.0 °C to simulate the fumigation and washing therapy.The intelligent wristband measured their pulse, skin temperature, and perspiration simultaneously during the fumigation and washing (Figure 4A).Furthermore, volunteers' pulse, skin temperature, and perspiration in their resting state (after sitting in the chair tranquility for 15 min) were also measured.Pulse data measured in two states were performed in Figure 4B 1 , showing that during fumigation and washing therapy, the volunteer's pulse rate was faster than the resting state.Similarly, temperature and perspiration data measured in the resting state and during fumigation and washing showed the temperature increase and perspiration appearance in the process of fumigation and washing therapy (Figure 4C 1 ,D 1 ).Then, after multiple measurements, the comparison results presented a significant difference in volunteers' pulse, temperature, and perspiration during fumigation and washing therapy and in the resting state (Figure 4B 2 ,C 2 ,D 2 ).
Assisted by the feedforward neural network (FNN) algorithm, the intelligent wristband can distinguish the user's physical state, whether during the fumigation and washing therapy or in the resting state.First, the pulse, temperature, and perspiration data The intelligent wristband was used during the fumigation and washing therapy of TCM to realize the simultaneous measurement of users' pulse, temperature, and perspiration.The FNN algorithm was introduced to identify users' physical states.A) Schematic diagram of fumigation and washing therapy of TCM.The intelligent wristband was used in this process for simultaneous multiparameter measurement of pulse, temperature, and perspiration.B 1 ) Two pieces of pulse curves obtained by the intelligent wristband respectively during fumigation and washing therapy and in the resting state.B 2 ) All pulse data of volunteers measured during fumigation and washing therapy and in the resting state were calculated for pulse rates.C 1 ) Two pieces of temperature curves of a volunteer obtained by the intelligent wristband respectively during fumigation and washing therapy and in the resting state.C 2 ) All temperature data of volunteers were measured during fumigation and washing therapy and in the resting state.D 1 ) Two pieces of perspiration curves of a volunteer obtained by the intelligent wristband respectively during fumigation and washing therapy and in the resting state.D 2 ) All perspiration data of volunteers were measured during fumigation and washing therapy and in the resting state.E) The structure of the FNN consisted of three fully connected layers to the physical situation of users when during fumigation and washing therapy or in the resting state.F) The evolution process of training loss during 100 epochs.G) The evolution condition of classifying accuracy during 100 epochs.H) The confusion matrix of the prediction result made using the FNN algorithm.
obtained during fumigation and washing therapy and in the resting state were sliced into fragments.Then, four eigenvalues were from three pieces of pulse, temperature, and perspiration fragments that were measured simultaneously and can be regarded as a group.The database contained 604 groups and was randomly divided into a training set (545 groups) and a test set (59 groups).The FNN model was proposed with three fully connected layers, and the size of the three fully connected layers was 4 Â 10, 10 Â 5, and 5 Â 2, respectively (Figure 4E).The other details of FNN are included in the "Details of Database and FNN" part.During the training process, the training loss decreased fleetly first and was saturated gradually (Figure 4F).Correspondingly, after 100 training epochs, the classification accuracy of the training set was up to 88.99% (Figure 4G).More importantly, the accuracy of the FNN to classify the test set can be up to 100%, and the confusion matrix of classifying the results is shown in Figure 4H.The results indicate that the intelligent wristband assisted with the FNN algorithm can continuously monitor the user's physical state and accurately distinguish whether the user is during fumigation and washing therapy or in the resting state.

Conclusion
We have demonstrated an intelligent wristband consisting of the pulse sensing unit, temperature sensing unit, and perspiration sensing unit for simultaneous multiparameter measurement of pulse, temperature, and perspiration.The CNT/PDMS nanocomposites were designed and fabricated as an essential functional material with a 3D interpenetrating network structure to realize the measurement of pulse, temperature, and perspiration via their great electrical conductivity, mechanical properties, stress effect, and thermoresistive effect.The performance of the intelligent wristband indicates that all these sensing units have high accuracy, sensitivity, repeatability, and reliability.During fumigation and washing therapy of TCM, as an example, artificial intelligence was introduced to distinguish users' physiological states based on pulse, temperature, and perspiration data simultaneously measured by the intelligent wristband, and the accuracy of classification can be up to 100%.It is indicated that wearable sensing technology and artificial intelligence are new ways to advance the modernization, objectification, and scientization of TCM to some extent.In the coming days, TCM-based wearable sensors assisted by artificial intelligence will offer fresh insights into TCM and wearable sensing technology.

Experimental Section
Materials: Multiwalled CNT (NC7000) was purchased from Nanocyl S.A., and its average diameter, average length, specific surface area, and purity are 9.5 nm, 1.5 μm, 250-300 m 2 g À1 , and 90%, respectively.PDMS, comprising prepolymer (component A) and cross-linker (component B), was purchased from Dow Corning (Sylgard 184, USA).Xylene (C 8 H 10 ) was obtained from Tianjin ZhiYuan Reagent Co., Ltd (China).Artificial sweat was purchased from Dongguan Yunfei Automation Equipment Technology Co., Ltd.Wax was supplied from Shanghai New Century Co., Ltd.(China).White plain weave cloth (polyester cotton) was purchased from Dongguan Fengxing Labor Protection Products Co., Ltd.Unless otherwise specified, all reagents are of analytical grade and were used as received and without further purification.
Fabrication of CNT/PDMS Nanocomposite: CNTs (0.329 g) and component A of PDMS (2.000 g) were both dissolved in xylene (200 mL) separately and then ultrasonic dispersed for 30 min.The well-dispersed CNTs in xylene and PDMS dissolved in xylene were mixed, and ultrasonic dispersion energy was applied for 30 min, followed by magnetic stirring for 5 min.After the cycle of ultrasonic dispersion and magnetic stirring 5 times, a uniformly dispersed CNTs and PDMS mixture in xylene was obtained.Next, component B of PDMS (0.200 g) was added while magnetic stirring.The mixture of CNTs and PDMS was then poured into a glass Petri dish and exposed in the fume hood for 8 h to volatilize the xylene.Finally, the mixture of CNTs and PDMS was heated in an oven at 60.0 °C for 30 min to acquire the CNT/PDMS nanocomposite eventually.
Characterization and Measurement of the CNT/PDMS Nanocomposite: Morphology of the well-dispersed mixture of CNTs and PDMS in xylene was characterized by a transmission electron microscope (TEM, JEM-1400).The structure of the surface and the cross section of the CNT/PDMS nanocomposites were examined by a SEM (ZEISS-AURIGA).Moreover, the prepared CNT/PDMS nanocomposites were characterized by a laser micro-Raman spectrometer, FT-IR spectroscopy, and TG.
All conductivity measurements were carried out by an electrochemical analyzer CHI627D (Shanghai Chenhua, China).The LSV of the CNT/PDMS nanocomposites was performed from À1.0 V to 1.0 V.And the conductance (σ) of the CNT/PDMS nanocomposites was calculated as follows where L, S, R, U, and I represent the length, cross-sectional area, resistance values, voltage, and current.Then, the CV curves of the CNT/PDMS nanocomposites were conducted between À1.0 and 1.0 V 500 times.The mechanical properties of the CNT/PDMS nanocomposites were evaluated by carrying out their strain-stress tests using a universal material testing machine (LR10K Plus).The CNT/PDMS films (40.0 mm in length, 9.0 mm in width) were clamped and prestressed, and then stretched at a tensile speed of 20.0 mm min À1 for the fracture mechanic test.The cyclic loading-unloading tests were also conducted with a strain of 50% for 160 cycles.
To study the biocompatibility and cytotoxicity of the CNT/PDMS nanocomposites, human immortalized keratinocytes (HaCAT) were chosen to investigate the biocompatibility of the CNT/PDMS nanocomposites.The CNT/PDMS films were sterilized by autoclaving and immersed in a culture medium for 24 h.Then the culture medium was filtered and diluted to obtain the medium containing different amounts of the CNT/PDMS nanocomposite extracts.The HaCAT cells were cultured in the medium (0, 2.5, 5.0, and 10 μg mL À1 ) for 24, 48, and 72 h.The cell viability was analyzed by the Cell Counting Kit-8 method using a microplate reader (450 nm).
Pulse Sensing Unit of the Intelligent Wristband: Three pieces of CNT/PDMS films (40.0 mm in length, 8.0 mm in width) were vertically ranged in arrays and fixed to the surface of a water sac to prepare the pulse sensing unit.Then, the CNT/PDMS arrays and the water sac were fixed to a wristband, and the CNT/PDMS arrays were toward the inside of the wristband.
For the stress effect of the CNT/PDMS nanocomposite, the current curve of the CNT/PDMS nanocomposites was tested with the stress exerted by the universal material testing machine with a moving speed of 25.0 mm min À1 .And the relationship between the current changes of the CNT/PDMS nanocomposites and stress was calculated.Afterward, the current-time curve of the CNT/PDMS nanocomposites was tested and analyzed with and without CS at multistep stresses of 0.01, 0.05, 0.10, 0.15, 0.20, 0.30, 0.40, and 0.50 N. The current-time curve of the CNT/PDMS nanocomposites also performed at 0.3 N cyclic stress 2000 times (XLD-100E, Guangzhou Precision Control Testing Instrument Co., Ltd.).
The intelligent wristband was tightly worn at the left wrist, and the pulse sensing unit was aimed at the left wrist's radial artery.Three current-time curves of the CNT/PDMS arrays were simultaneously performed.Moreover, the PPG signal of the middle finger of the left hand was also measured by a commercial pulse oximeter (CMS60D, CONTEC).
Temperature Sensing Unit of the Intelligent Wristband: To prepare the temperature sensing unit, a CNT/PDMS film (40.0 mm in length, 8.0 mm in width) was fixed to the internal surface of the wristband.
For the thermoresistive effect of the CNT/PDMS nanocomposite, the current-time curve of the CNT/PDMS nanocomposites was performed at a multistep temperature rise of 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, and 80.0 °C.Then, the cyclic heating-cooling tests were conducted between 29.0 and 31.0 °C 250 times.
The current-time curve of the temperature sensing unit was tested by temperature sensing unit at certain temperatures (30.0, 32.0, 34.0, 35.0, 36.0,37.0, 38.0, 39.0, and 40.0 °C).Then, the intelligent wristband was tight worn at the left wrist, and the temperature sensing unit was attached to the skin to measure the current-time curve of the temperature sensing unit.The temperature sensing unit was also used to sense room temperature at two different positions, and the corresponding current-time curves were also performed.Meanwhile, the skin temperature on the left wrist and the room temperature of the two positions mentioned above were also measured by an infrared camera (FLIR E30, FLIR systems, OÜ, Estonia) for validation.
Perspiration Sensing Unit of the Intelligent Wristband: The Teflon mask (100.0 mm Â 100.0 mm Â 0.2 mm) was prepared using a laser cutter (CMA1080, HAN'S YUEMING LASER, China) to fabricate the cloth-based CNT/PDMS nanocomposite.The cloth was placed under the mask, and the mixture of CNT and PDMS after the xylene evaporated was used as paint to be applied above the mask.The excess mixture of CNT and PDMS was scraped off with a glass rod, and then the cloth coated with the mixture of CNT and PDMS was heated in an oven at 60.0 °C for 30 min to obtain the cloth-based CNT/PDMS nanocomposite.
The wax was heated to melt and used to cover all the non-sensing parts of the cloth-based CNT/PDMS nanocomposite to prepare the perspiration sensing unit.Then, the wax-coated cloth-based CNT/PDMS nanocomposite was fixed to the internal surface of the wristband.
The CV curves of the perspiration sensing unit were performed with and without artificial sweat, adding a range from À1.0 V to 1.0 V.Then, the CV curves of the perspiration sensing unit after 35 times cyclic artificial sweat adding and cleaning were measured and analyzed.
For the on-body validation of the perspiration sensing unit, the intelligent wristband was tightly worn at the left wrist, and the perspiration sensing unit was attached to the skin for validation.The perspiration sensing unit's CV curves at different time points were performed and analyzed.In the meantime, the perspiration conditions of the skin were recorded simultaneously by taking photos.
Simultaneous Multiparameter Measurement during Fumigation and Washing Therapy of TCM: Volunteers were instructed to soak their foot in hot water at about 40.0 °C to simulate fumigation and washing therapy of TCM, and the intelligent wristband was tightly worn at their left wrist.During this process, the current-time curves of the pulse and temperature sensing unit and the CV curves of the perspiration sensing unit were performed simultaneously.In addition, the current-time curves of the pulse sensing unit, temperature sensing unit, and the CV curves of the perspiration sensing unit were also measured when volunteers were in the resting state after sitting for 15 min.
Details of Database and FNN: The continuous pulse, temperature, and perspiration signals, obtained at the same time by the intelligent wristband, were sliced into fragments, and each fragment has 300 data points (100 data points from pulse signals, 100 data points from temperature signals, and 100 data points from perspiration signals).Then, four eigenvalues were extracted from each fragment (two eigenvalues from pulse signals, one eigenvalue from temperature signals, and one eigenvalue from perspiration signals) and regarded as a group.There are 604 groups of eigenvalues in total, and the number of groups when volunteers were in the resting state and during fumigation and washing therapy was 278 and 326, respectively.Next, the database was randomly divided into a training set (545) and a test set (59).The batch size is set to 16, and the data are transferred to three fully connected layers.Cross-entropy was chosen as the loss function.Adam was chosen as the optimizer.The learning rate was 0.0005 from 1 to 50 epochs and 0.0001 from 51 to 100 epochs.Details of each neural network layer are listed as follows: Fully connected layer 1: size = 4 Â 10, dropout = 0.5.Fully connected layer 2: size = 10 Â 5, dropout = 0.5.Fully connected layer 3: size = 5 Â 2, dropout = 0.5.

Figure 1 .
Figure 1.Design of the intelligent wristband, including pulse sensing unit, temperature sensing unit, and perspiration sensing unit, and characterization of the CNT/PDMS nanocomposites.A) Schematic diagram of the intelligent wristband: pulse sensing unit, temperature sensing unit, and perspiration sensing unit were integrated into the intelligent wristband.The simultaneous multiparameter measurement of the wristband was based on the CNT/PDMS nanocomposite with a 3D interpenetrating network structure.B) Optical photo of the CNT/PDMS nanocomposites shown as a thin black film.C) SEM image of the surface of the CNT/PDMS film.D) SEM image of the cross section of the CNT/PDMS film.E) The tensile stress-strain curve (red) and the LSVs curve (black) of the CNT/PDMS nanocomposites with 13% CNT content.F) Optical photo showing the flexibility, stretchability, and adhesion of the CNT/PDMS films.G) Patterning of the CNT/PDMS films realized by a laser cutter.

Figure 2 .
Figure 2. Sensing principles, performances, and on-body validation of the pulse sensing unit.A 1 ) Schematic diagram shows the pulse sensing unit and it during the pulse measurement.A 2 ) Schematic diagram of the stress effect principle of the CNT/PDMS nanocomposites.In phase 1, the change in cross-sectional area (decreased from S 0 to S p ) and length (increased from L 0 to L p ) produced by stress increased the resistance of the CNT/PDMS film.Apart from the change in cross-sectional area and length, in phase 2, stress caused a relative displacement between CNTs, resulting in decreased electric channels between CNTs and higher resistance.B) The relationship between the stress exerted on the CNT/PDMS nanocomposites by the vertical direction and the corresponding current change can be separated into these two phases.The current change of the CNT/PDMS nanocomposites is linearly relative to stress in phase 1 and exponentially relative to stress in phase 2. C) The CS-induced sensitivity improvement principle of pulse measurements.When under CS (the gray area), the current change of the CNT/PDMS nanocomposites was bigger than without CS via the same stress (from the red area to the blue area).D) Current changes in the CNT/PDMS nanocomposites with multistep stresses when under CS or not.E) The current time curve of the CNT/PDMS nanocomposites with 2000 times cyclic stress when CS loading.F) Three pieces of pulse curve obtained by pulse sensing unit at Chi, Guan, and Cun positions.G) The 3D map of a single pulse beat from (F).H) The pulse curves obtained by the pulse sensing unit and the simultaneous PPG data provided by the commercial pulse oximeter.Via the bivariate correlation analysis, the correlation coefficients of the pulse and PPG are 1, and p < 0.01.

Figure 3 .
Figure 3. Sensing principles, performances, and on-body validations of temperature sensing and perspiration sensing unit.A) Schematic diagram of temperature sensing unit and the temperature measurement principle of temperature sensing unit based on the thermoresistive effect of the CNT/PDMS nanocomposites.B) The current time curve of the CNT/PDMS nanocomposites with multistep heating.C) The current time curve of the CNT/PDMS nanocomposites with 250 times heating-cooling cycles.D) The relationship of the temperature sensing unit between current and temperature was shown, and three temperatures (one skin temperature and two room temperature) were calculated based on the measured data.Also, the infrared camera photographed three positions' temperatures simultaneously, and the infrared photos were shown as the inset.Via the partial correlation analysis, the correlation coefficients of the temperature measured by the temperature sensing unit and the infrared camera are 1, and p < 0.01.E) Schematic diagram of perspiration sensing unit and the perspiration measurement principle of perspiration sensing unit.F) The CV curve of the perspiration sensing unit when adding artificial sweat or not.G) The current change of the perspiration sensing unit was tested with 35 times repeated artificial sweat addition and cleaning.H) The current change of the perspiration sensing unit at six different time points and the corresponding perspiration situation of the skin were recorded by photos simultaneously.Via the bivariate correlation analysis, the correlation coefficients of the current change of perspiration and whether perspiration observed from photos are 0.865, and p < 0.01.

Figure 4 .
Figure 4.The intelligent wristband was used during the fumigation and washing therapy of TCM to realize the simultaneous measurement of users' pulse, temperature, and perspiration.The FNN algorithm was introduced to identify users' physical states.A) Schematic diagram of fumigation and washing therapy of TCM.The intelligent wristband was used in this process for simultaneous multiparameter measurement of pulse, temperature, and perspiration.B 1 ) Two pieces of pulse curves obtained by the intelligent wristband respectively during fumigation and washing therapy and in the resting state.B 2 ) All pulse data of volunteers measured during fumigation and washing therapy and in the resting state were calculated for pulse rates.C 1 ) Two pieces of temperature curves of a volunteer obtained by the intelligent wristband respectively during fumigation and washing therapy and in the resting state.C 2 ) All temperature data of volunteers were measured during fumigation and washing therapy and in the resting state.D 1 ) Two pieces of perspiration curves of a volunteer obtained by the intelligent wristband respectively during fumigation and washing therapy and in the resting state.D 2 ) All perspiration data of volunteers were measured during fumigation and washing therapy and in the resting state.E) The structure of the FNN consisted of three fully connected layers to the physical situation of users when during fumigation and washing therapy or in the resting state.F) The evolution process of training loss during 100 epochs.G) The evolution condition of classifying accuracy during 100 epochs.H) The confusion matrix of the prediction result made using the FNN algorithm.