Wearable Device for Blood Pressure Monitoring Based on a Novel Algorithm With a Few Morphological Feature Parameters of Pulse Waves

We proposed an algorithm for blood pressure estimation based on a few morphological feature parameters of pulse waves to solve the problem of over-reliance on high-quality pulse waves (involving numerous pulse wave morphological feature parameters) of blood pressure estimation based on single-channel photoplethysmography (PPG). We successfully implemented the algorithm in a self-developed miniaturized wearable device. Statistical analysis revealed that in the algorithm, a parameter calculated from pulse pressure difference (DP) was highly correlated (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R}^{{2}} = {0.942}$ </tex-math></inline-formula>) with systolic blood pressure (Ps). We defined this parameter as k. We used multiple linear regression to establish a mapping relationship between the morphological feature parameters of pulse waves and DP. This relationship had a correlation coefficient of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R}^{{2}} = {0.667}$ </tex-math></inline-formula>, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${P}~\lt 0.001$ </tex-math></inline-formula>. We built a miniaturized platform for blood pressure measurement and developed a reflective PPG sensor packaged with polydimethylsiloxane to enable comfortable wearable blood pressure measurement. In uncalibrated randomized experiments, we first evaluated the performance of the algorithm on separate datasets. A total of 116 objects showed a mean absolute error (MAE) of 11.15 and 8.29 mmHg for Ps and diastolic blood pressure (Pd), respectively. In a human experiment with 31 subjects, the errors for Ps and Pd were <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$5.29~\pm ~10.92$ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$- 2.77~\pm ~8.71$ </tex-math></inline-formula> mmHg, respectively. The results of multistate continuous blood pressure monitoring experiments improved after calibration was performed once. In a prolonged blood pressure monitoring experiment, the errors for the Ps and Pd of subjects were <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1.11~\pm ~7.05$ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.19~\pm ~2.69$ </tex-math></inline-formula> mmHg, respectively. The results of both experiments conducted in this work were close to the international standard (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$5~\pm ~8$ </tex-math></inline-formula> mmHg) issued by the American Association for the Advancement of Medical Instrumentation. Our algorithm also showed promising results in continuous blood pressure monitoring experiments after exercise and under breath-holding conditions.