针对现有基于脉搏波传输时间的无创连续性血压测量算法精度不高的问题,该文综合考虑心电信号和血氧容积波与血压变化的相关性,提出一种基于BP神经网络的无创连续性血压测量方法.该文首先利用改进的心电信号算法提取出心电信号的 R 点,利用差分、阈值的方法提取出血氧容积波的特征参数,再经过特征解析,提取出与血压相关的15维特征向量,构建基于BP神经网络的血压计算模型,计算出逐拍的血压值.该方法在天坛医院等单位进行了医学临床比对测试,并通过因子分析法分析了15个特征参数的权重比.实验证明:在预测血压上,脉搏波传输时间的权重,大于相邻特征点之间的时间信息权重,大于脉搏波面积信息权重,大于脉搏波幅值信息权重;该方法精度优于其它相近方法,单次测量的舒张压和收缩压误差的平均值±标准差分别是-1.57±6.12 mmHg和-0.62±4.82 mmHg,重复测量误差的平均值±标准差分别是-2.12±5.10 mmHg和-2.52±4.41 mmHg.收缩压和舒张压的测量精度均达到了BHS血压标准的Grade A类和AAMI标准.%For the problem of noninvasive continuous blood pressure algorithm with un-accuracy, a novel multi-parameter fusion algorithm based on BP neural network is proposed, according to the formation from electrocardiogram and photoplethysmograph of arterial blood pressure. The improved Pan Tompkins algorithm is used to extract the R peak of electrocardiogram, and difference-threshold algorithm is used to extract the features points of photo-plethysmograph, and the fifteen feature parameters relative to blood pressure are extracted and used to establish the model of blood pressure to estimate the beat-to-beat systolic blood pressure and diastolic blood pressure. The factor analysis method is used to analyze the weight of each parameter. The results show that the weight order is pulse transit time, time information, photoplethysmography area information, amplitude information and area ratio. The algorithm is tested in the TianTan Hospital, and the means±standard difference of single measurement errors are respectively -1.57±6.12 mmHg and -0.62±4.82 mmHg, the means± standard difference, D. of repeated measurement errors are respectively -2.12±5.10 mmHg and -2.52±4.41 mmHg, for systolic blood pressure and diastolic blood pressure. And the measurement accuracy for systolic blood pressure and diastolic blood pressure reaches Grade A of BHS standard and AAMI standard.
展开▼
