A multi-indicator pulse monitoring system based on an ultra-sensitive and stable self-powered wearable triboelectric sensor with assistance of personalized deep learning

Abstract

The non-invasive detection of pulse signals has significant applications in the prevention and diagnosis of cardiovascular diseases. The dicrotic pulse wave (P₃) is crucial in cardiovascular information analysis. However, conventional flexible pressure sensors struggle to sensitively detect this fluctuation and lack sufficient accuracy, limiting their application in physiological monitoring. To overcome these limitations, this study developed a triboelectric pulse sensor (TPS) utilizing ripple-shaped flexible metal electrodes and electro-spun polyvinyl alcohol (PVA) fiber membrane with a 3D interconnected structure. The TPS exhibited a sensitivity of 6.875 V/kPa and a response time of 7.9 ms, significantly surpassing existing flexible pressure sensors. After 3000 testing cycles, the TPS demonstrated excellent stability, enabling high-precision and rapid detection of the P₃ peak. Comparative tests with a commercial flexible pulse pressure sensor revealed a high degree of consistency in waveform detection at the same location (Pearson correlation coefficient P = 0.9735). By integrating a predictive regression-type deep learning neural network with a customized host computer, we developed a non-invasive multi-indicator pulse monitoring system (MIPMS) based on the TPS, facilitating real-time display of pulse waves and predictive analysis of various physiological information contained within the pulse waves. With its sensitive detection of the P₃ peak and excellent long-term stability, the MIPMS can identify cardiovascular issues that conventional flexible pressure sensors may overlook.