Device fabrication and structural characterization
In TCM pulse diagnosis, it is believed that the health of human organs is related to the pressure pulse wave at corresponding mapping points (Cun, Guan, Chi) on the radial artery (Fig. 1a). In this study, we propose a wearable, flexible wristband that can be actively pressurized to mimic TCM pulse collection (Fig. 1b). The system comprises flexible pressure sensing units for collecting pulse waves at the Cun, Guan, Chi positions, an active pressure control unit providing different pressures, a wireless transmission unit for signal transmission and processing, a wireless charging unit for system power supply, and a power management unit.
a Method of TCM pulse diagnosis. b Optical image of the wireless wristband worn on the users wrist joint. c Block functional diagram of the sensing system, including the power supply, signal acquisition, processing, communication, and user interface. d Schematic illustration of the wireless wristband worn on the wrist, where the airbag provides backpressure to effectively collect pulse wave changes under different pressures. e Detailed diagram of the overall structural design of the sensor system. f Detailed diagram of the overall structural design of the pressure sensor. g, h, and i Digital optical image and FEA results of the wristband, flexible circuit and sensing array under mechanical deformation
The active pressure control unit, comprising silicone airbags, piezoelectric micropumps, a digital pressure sensor, electromagnetic valves, and one-way valves, works synergistically to provide precise pressure modulation. The micropump regulates airbag inflation, and pressure sensors and electromagnetic valves provide pressure feedback control (Fig. 1c, d). The hardware and software architecture of the system, including sensor integration, data processing modules, and user interface components, is comprehensively depicted in Fig. 1c. All components, such as the sensor array, micro-airbag array, micropump, and flexible printed circuit board (FPCB) and their interconnections, are encapsulated in soft silicone to create a fully flexible, wearable, multichannel active pressure pulse-sensing platform (WAPPP). This design allows the device to flex and stretch, ensuring tight and soft contact between the sensors and the arterial regions of the skin (Fig. 1d).
Figure 1e shows the hardware and software architecture of the system, including sensor integration, data processing modules, and user interface components. As shown in Fig. 1f, a 3-channel pulse sensor array was used to simulate three fingers for pulse wave acquisition. The overall structure of the pressure sensor includes three independent circular interdigital electrode resistance sensors, each with a diameter of 8mm, which is slightly larger than the fingertip area of the human finger (Fig. S1). The three sensor units are connected by serpentine wires, significantly improving the deformability of the device and preventing mechanical interference between adjacent units. The pressurization of micro airbags ensures close contact between the sensor unit and the skin, enabling the precise conversion of local skin deformations caused by arterial expansion/contraction into electrical signal output. Figure 1g and h show that the system and its built-in flexible circuit board have excellent bending performance and can maintain good flexibility and equipment integration despite deformation. A flexible sensor array is easy to bend and mechanically stable. Figure 1i shows a digital optical image of the pressure sensor array and corresponding finite element analysis (FEA), demonstrating its applicability for wrist pulse measurements.
As a key part of pulse sensing systems, flexible pressure sensing arrays have high requirements for sensor performance. Resistance-type pressure sensors based on interdigital electrodes have advantages such as high sensitivity, high accuracy, high stability, convenient data collection, and simple device structures. In this paper, we used an interdigital electrode with a polyimide film (PI) substrate manufactured by FPCB technology as the induction electrode and thermoplastic polyether polyurethane (TPU)-ionic liquid (ILD)-h-BN as the ionic membrane.
The sandwich structure is combined through bonding layers and a hot-pressing process to form an iontronic pressure sensor. The sensitive layer of the sensor was manufactured using screen printing, a process that is controllable in batches, as depicted in Figs. 2a and S2. After heat curing (Fig. 2a), the sensitive layer was endowed with microcolumnar microstructures via laser engraving (illustrated in Fig. 2a). The surface morphology of this layer is presented in microphotographs (Fig. 2b), scanning electron microscopy (SEM) maps (Fig. 2c), and laser scanning confocal microscopy (LSCM) images (Fig. 2d). These microcolumnar structures substantially enhance the deformation capability of the sensitive layer under compression, thereby significantly improving its sensitivity. Figure 2e shows the corresponding equivalent circuit, which indicates that the main variation in resistance within the circuit is due to the internal resistance (Rin).
a Fabrication process of the pressure sensor. b Optical image of the sensitive layer with microstructure. c Illustration and scanning electron microscopy (SEM) images of the sensitive layer. d Sense LSCM image. e Schematic illustration and sensing mechanism of the pulse pressure sensor. f Current variation in sensors prepared with different ionic liquid contents. g Current variation in sensors prepared by different laser etching times
The doping of h-BN in the sensitive layer increased the viscosity of the printing paste and significantly improved the conductivity variation of the sensitive layer during deformation through the ion pump effect27. To explore the optimal performance of the sensor, we investigated the effect of different laser irradiation times (0, 1, 2, and 3) and various ionic concentrations (1, 1.2, 1.5 and 2mL) on the sensor sensitivity. The results showed that the best performance for the sensitive layer was achieved with 1.5mL of ionic liquid and 2 laser engravings. This was selected as the final sensor fabrication parameter (Fig. 2f, g).
To provide further evidence of the performance of the sensor, we conducted a series of tests and measurements to characterize its electrical performance (Fig. S7). The pressure sensor exhibits high sensitivity and good linearity within the pressure range of 050kPa. As a crucial parameter for sensors, sensitivity is defined as S=(I/I0)/P. Here, I0 and I represent the initial current under a 1V voltage before loading and the change in the output current when pressure is applied, respectively. Figure 3a shows that the sensitivity of the pressure sensor is S=460.1kPa1, and the fitting coefficient is R2>0.999. It is noteworthy that the performance of this sensor surpasses that of most reported pressure sensors, enabling its suitability for testing scenarios such as human pulses and BP. We tested a series of continuous pressures to evaluate the sensors performance in this context. The sensor exhibits excellent consistency and mechanical robustness in the pressure sensing range of 050kPa, making it highly effective for real-world applications and enhancing its practical applicability (Fig. 3b). Figure 3c shows that the pressure sensor response time and recovery time are 25 and 30ms, respectively, which meet the requirements for pulse monitoring applications. To demonstrate the good resolution of the sensor, we characterized the limit of detection (LOD) of the sensor, which produces a response of ~0.035A at a pressure of 150Pa, further verifying that the LOD of the sensor is approximately 150Pa (Fig. 3d). Furthermore, the sensor demonstrated high stability and durability in long-term (12,000 cycles) pressure loadingunloading cycles at 40kPa (Fig. 3e).
a Current variation versus pressure change of the pressure sensor. b Current variation of the pressure sensor under various pressures. c Fast response of the pressure sensor. d LOD of the sensor. e Long-term cycling ability of the sensor at 40kPa for 12,000 cycles
The active pressurization device comprises a micropump (19mm21mm3.6mm, Fig. S4), a soft silicone (Ecoflex) airbag array and a one-way valve. Under pressure from the airbag array, the sensor array can detect mechanical pulses caused by the propagation of blood (Fig. 4a). Figure 4b shows the fabrication process of the micro airbag array. The piezoelectric micropump (Murata Machinery) controls the internal pressure of the silicone airbag and provides a controllable back pressure to the sensor array through conformal contact. FEA showed that the protruding displacement of the airbag surface was 2.223mm when the pressure inside the airbag was 40kPa, demonstrating the feasibility of using microairbags for the pressurized detection of pulse signals (Fig. 4c). The micropump supplies sufficient pressure (up to 50kPa) to the airbag array, enabling steady pressure support for the sensor array (Supplemental Movie 1). The one-way valve at the outlet of the micropump serves as a pressure regulator to maintain pressure within the airbag while acting as a damping valve to stabilize the active pressure adaptive system. Figure 4d shows that the pressure in the airbag is basically unchanged when the air pump is used to inflate it to 10, 20, 30, 40, and 50kPa at specific time intervals.
a Digital optical image of the sensor patch on skin. b Fabrication process of the airbag. c Optical image of the airbag and stressstrain simulation at 40kPa. d The pressure inside the airbag is maintained within a stable range of 050kPa. e and f With increasing pressure (525kPa), the pulse amplitude and corresponding FFT results change
During actual pulse acquisition, with increasing external pressure, the coupling degree between the sensor and blood vessel changes. The amplitude of the pulse wave gradually increases and then decreases, as confirmed by the FFT results, which also demonstrate corresponding changes in frequency components with variations in the amplitude of the pulse wave (Fig. 4e).
The device can wirelessly connect to a compatible smartphone app via Bluetooth, enabling the transmission of pressure sensor signals to the mobile device for data storage and analysis (Fig. 5a and Supplement Movie 2). The WAPPP is based on controllable active airbag pressurization, which allows for the control of the sensors press depth, enabling the collection of pulse waves at different static pressures. The test results indicate that as the pressing force and depth increase, the amplitude of the pulse wave gradually increases, followed by a decrease (Fig. 5b), which is consistent with the theory of pulse diagnosis in TCM.
a The display interface for mobile devices. b Pulse wave changes under 9 different static pressures. c BP prediction model. d BlandAltman plots to validate the accuracy of the pulse sensing system for SBP and DBP
To validate the systems applicability, we integrated a pulse wave test with a machine learning model and constructed a blood pressure prediction model based on a back-propagation neural network. This allows for accurate monitoring of blood pressure and cardiac status using the applied pressure and its corresponding pulse wave magnitude as input variables, inspired by the principle of blood pressure measurement. The back-propagation neural network was chosen for its flexible network structure and excellent nonlinear expression capabilities and is widely employed in BP prediction. In this study, we extracted pulse waveforms at nine pressure stages. After stabilizing the waveforms, we recorded the pulse amplitude values from the sensor and their corresponding airbag pressure values as inputs.
As illustrated in Fig. 5c, our approach utilizes a three-layer network structure comprising an input layer, a hidden layer, and an output layer. During model training, a single hidden layer is sufficient to fit high-precision functions. Using too many hidden layers can lead to overfitting and slow down the training process. The output layer consists of 2 nodes representing systolic and diastolic pressures. The pulse dataset is divided into three sets: training group, validation group, and testing group, with proportions of 70%, 15%, and 15%, respectively.
In the model training phase, as the back-propagation neural network receives data, it performs computations from the input layer through the hidden layer to the output layer, generating BP predictions. Through the adjustment of model parameters and correction with actual BP values, the corrected values are fed back into the input layer, enhancing the accuracy of the BP predictions. The results indicate a strong correlation (R-square value close to 0.99) between the output of the transfer function and that of commercial BP monitors (Fig. S5). Clinical validation of BP prediction was conducted using a test set of 21 BP data points. The average differences between our device and commercial BP monitors were 0.779.0mmHg for systolic blood pressure (SBP) and 3.229.72mmHg for diastolic blood pressure (DBP) (Fig. 5e, f). These BP prediction results met the American Association of Medical Instruments (AAMI) international criteria for BP testing.
By wearing the system on the users body, continuous and accurate monitoring of pulse variations can be achieved, allowing for the prediction of blood pressure. These results highlight the potential applications of the pulse acquisition system.
Go here to see the original:
Wearable multichannel-active pressurized pulse sensing platform | Microsystems & Nanoengineering - Nature.com
