Wearable multichannel pulse condition monitoring system based on flexible pressure sensor arrays

Traditional Chinese medical science (TCMS) has long relied on pulse diagnosis, but its clinical use is limited by the lack of standardized criteria and dependence on physicians’ subjective sensations. Integrating these qualitative assessments with modern data platforms remains challenging. The key to TCMS modernization is to detect pulse signals and translate concepts such as position, number, shape, and trend into specific, parametric data. Prior approaches have used optical, imaging, acoustic, and pressure-based sensors, with wearable pressure sensors particularly suitable for capturing arterial pulse conduction. However, most platforms use single-point, rigid sensors, making multidimensional, small-area array integration difficult and limiting their applicability to wearable pulse diagnosis. According to TCMS theory of three positions and nine indicators, investigating pulse length, width, and strength distribution can reveal pulse conditions. Existing systems with one sensor per position cannot capture temporal and spatial dimensions akin to fingertip palpation, and few studies address spatial distributions across the three positions due to the lack of methods for multidimensional detection and weak-signal analysis. This work proposes a flexible, wearable multichannel platform that simultaneously acquires 3D pulse signals at Cun, Guan, and Chi to address these gaps.

Multiple sensing mechanisms (optical, image processing, acoustics, pressure) have been explored for pulse detection, with pressure sensors favored for arterial pulse monitoring due to sensitivity to dynamic loading. Reported systems often employ single-point, rigid sensors, limiting spatial resolution and wearability. Chu et al. advanced three-position monitoring but used one sensor per position, lacking spatial detail and temporal-spatial mapping akin to physicians’ palpation. There is limited consideration of spatial distributions among Cun, Guan, and Chi in prior work, largely due to challenges in multidimensional weak-signal acquisition and analysis. The present study builds upon these efforts by implementing flexible sensor arrays and surface fitting to provide real-time 3D pulse maps.

System architecture: The platform comprises flexible 3×3 ionogel-based pressure sensor arrays placed at TCMS pulse positions (Cun, Guan, Chi), along with signal acquisition, A/D conversion, amplification, denoising, detection, and 3D surface fitting/display modules. This enables simultaneous, synchronized multichannel pulse capture across positions. Sensor design and fabrication: Silver (Ag) electrode lines were screen-printed on a flexible PET substrate. Patterned PDMS moulds were used to define arrayed ionogel films deposited over the Ag electrodes. A PDMS layer was bonded to enhance stability. The fabrication flow included electrode printing, PDMS mould preparation, plasma treatment, PET–PDMS bonding, ionic liquid/ionogel deposition, curing, cleaning (KOH), and packaging to yield the pulse sensor arrays. Arrays were 3×3 elements per position; sensors exhibited high flexibility and sensitivity. Uniformity and characterization: Uniformity was assessed by applying identical pressures to all nine elements, showing similar resistance change ranges, indicating consistent and stable pressure sensing suitable for pulse detection. Data acquisition and processing: The system simultaneously recorded voltage signals from nine channels at each position. Acquisition involved amplification and filtering/denoising. Time-domain waveform features were extracted, including percussion (P), tidal (T), and diastolic (D) wave components, early systolic peak (P1), inflection point (P2), dicrotic notch, and dicrotic peak (P3). Time-domain parameters (e.g., ΔTDVP = TP2 − TP1, augmentation index AI = P2/P1) and composite indices (e.g., K = (Pm − Pd)/PHI) were computed to reflect vascular stiffness, blood flow velocity, and TCMS pulse qualities. 3D mapping and visualization: For spatial analysis, per-sample pulse strengths from the 3×3 arrays were plotted as columns across the wrist plane (X: width, Y: length/position, Z: amplitude). Smooth 3D pulse surfaces were generated in real time using cubic spline interpolation (LabVIEW) and cubic fitting (MATLAB) to emulate fingertip palpation sensations and visualize temporal-spatial pulse evolution. Experiments: Continuous pulse-taking experiments were performed on healthy volunteers, including comparisons across individuals and within individuals before vs. after meals. Stability across arrays and positions was assessed by comparing simultaneous multichannel recordings, checking synchronization of peaks/valleys and baseline stability.

• The flexible ionogel-based 3×3 pressure sensor arrays demonstrated high uniformity and stability; elements showed similar resistance change ranges under the same load, enabling reliable multidimensional sensing.
• Clear two-dimensional pulse waveforms were recorded with repeatable cycles (~670 ms cycle time), exhibiting forward and reflected waves and three characteristic peaks (P1, P2 inflection, P3 dicrotic peak), consistent with TCMS and cardiovascular physiology.
• Sensor stability: Multiple arrays at the same position produced similar waveforms with minimal baseline drift, confirming consistency necessary for multichannel analysis.
• Inter-individual differences: Example 5 s segments showed pulse frequencies of approximately 85/min and 80/min for two healthy volunteers. Time-domain parameters (Table S4) indicated ΔTDVP ≈ 210 ms and 230 ms, and AI ≈ 0.74 and 0.80, respectively—values consistent with healthy, elastic arteries (longer ΔTDVP, smaller AI), while also distinguishing individual cardiovascular characteristics (e.g., clearer dicrotic peak in one subject).
• Condition-dependent changes (before vs. after meal): Waveforms were steadier before meals; after meals, the early systolic peak increased, indicating stronger postprandial blood flow. Representative segments showed frequency changes (e.g., ~72/min before vs. ~82/min after), ΔTDVP increasing from ~200 ms to ~230 ms, and ΔVp decreasing from ~0.74 to ~0.68. Reported AI was lower before meals, consistent with reduced vascular stiffness during fasting.
• Three-dimensional mapping: The system provided synchronized 9-channel signals at each of Cun, Guan, and Chi, enabling real-time 3D pulse surfaces. Pulse strength was strongest along the radial artery; Guan exhibited greater amplitude than Cun and Chi, consistent with literature. The fitted surfaces captured spatial distributions of pulse length, width, and strength, emulating physicians’ tactile assessment and revealing organ-related positional differences per TCMS theory.

The study addresses the central challenge of translating subjective TCMS pulse palpation into objective, quantitative, and spatially resolved data. By deploying flexible, high-uniformity pressure sensor arrays at Cun, Guan, and Chi and coupling them with synchronized acquisition and surface fitting, the platform captures both temporal waveforms and spatial distributions that mirror physicians’ fingertip sensations. The system distinguished inter-individual variability and physiological state changes (e.g., pre- vs. postprandial), with quantitative parameters (ΔTDVP, AI, ΔVp, frequency) aligning with known cardiovascular behaviors and TCMS descriptors (e.g., normal vs. slippery pulse). The 3D mapping validates that Guan often presents stronger signals and that the radial artery ridge dominates spatial strength, supporting TCMS positional interpretations. Collectively, these results demonstrate feasibility for modern, data-driven TCMS pulse diagnosis and open avenues for intelligent health monitoring that integrates traditional insights with contemporary sensing and analytics.

This work presents a wearable multichannel pulse monitoring platform that integrates flexible ionogel pressure sensor arrays with real-time 3D surface fitting to simultaneously acquire temporal and spatial pulse information at Cun, Guan, and Chi. The system overcomes limitations of single-point, rigid sensors by providing robust, synchronized, multidimensional data that reflect classical pulse features and TCMS positional nuances. It effectively differentiates individuals and physiological states (e.g., before vs. after meals) using quantitative metrics (ΔTDVP, AI, frequency, ΔVp), demonstrating potential for objective TCMS modernization and intelligent healthcare applications. Future work could expand clinical validation across broader populations and conditions, refine algorithms for automated pulse-type classification, and further optimize wearable integration for long-term, ambulatory monitoring.