基于PPG收缩期上升波形特征参数的无创血压检测模型

旨在提出一种基于光电容积脉搏波（PPG）收缩期上升波形（SUW）特征提取的血压计算模型。首先,采集130例成年志愿者的II导心电图、手指PPG信号以及袖带血压值。数据被随机分为训练集（共80例,其中46例血压正常）和测试集（共50例,其中35例血压正常）。然后从PPG信号的SUW中提取时间、面积、斜率最大点与波谷点连线与时间轴夹角的正切值、上升期占脉宽的比例等特征,采用偏最小二乘回归法分别建立收缩压和舒张压的检测模型。最后与基于脉冲到达时间（PAT）的模型以及PAT联合PPG的模型进行对比。结果显示,本模型计算的收缩压和舒张压与参考值的相关系数分别为0.80和0.62,判断血压异常的准确率为90%。这些指标均显著高于基于PAT和PAT联合PPG参数的血压模型（P<0.05）。研究表明,基于PPG收缩上升波形特征参数的无创血压检测模型具有较好的鲁棒性和较强的异常血压检测能力。

Noninvasive blood pressure measurement model based on characteristic parameters of PPG systolic upstroke waveform

Abstract: A model for noninvasive measurement of blood pressure based on the characteristics of the systolic upstroke waveform (SUW) of photoplethysmography (PPG) is proposed in this study. Firstly, lead II electrocardiogram (ECG), finger PPG signals and the blood pressure measured from a cuff-base electronic sphygmomanometer were recorded from 130 adult volunteers, and the data were randomly divided into a training set (n=80, 46 of which had normal blood pressure) and a test set (n=50, 35 of which had normal blood pressure). Some characteristics, such as upstroke time, area under SUW, tangent of the angle between the line of the maximum slope point to the trough point and the time axis, and the ratio of upstroke period to pulse width, were extracted from the SUW of PPG. The measurement models of systolic pressure and diastolic pressure were established by partial least square regression method. Finally, the performance of the proposed method was compared with the model based on pulse arrival time (PAT) and the model based on PAT combined with PPG. The correlation coefficients between the calculation results by the proposed model and the reference values were 0.80 and 0.62 for systolic pressure and diastolic pressure, respectively. The accuracy of the proposed model in finding abnormal blood pressure was 90%. These indicators were all significantly higher than those obtained by the model based on PAT combined with PPG (P<0.05). The study reveals that the SUW-based noninvasive blood pressure measurement model has improved robustness and better performance in finding abnormal blood pressure.