基于血压标定的中医脉图信号分析识别研究

Analysis and Recognition of TCM Pulse Signals Based on Blood Pressure Calibration

目的为了赋予脉图波形特征更明确的生理意义、提高脉图分类识别率。方法基于提取的脉图信号时域特征,分别运用随机森林算法等多种机器学习算法建立血压预测模型,并用预测的血压对脉图进行标定,使脉图波形特征点生理意义更明确;对标定后的脉图提取时域特征参数和血液动力学参数,并应用随机森林等多种机器学习算法对脉图进行分类识别。结果建立的血压预测模型对收缩压与舒张压的预测精度满足AAMI国际电子血压计标准要求;基于血压标定后的脉图特征建立的分类识别模型,其识别效果优于血压未标定的脉图特征。结论该研究为脉图分类识别提供了一种新的思路和方法,对脉诊客观化有一定的实际应用价值。

Objective To give a more clear physiological meaning to the waveform characteristics of the pulse waveform and improve the classification accuracy of the pulse pattern.Methods Firstly we extracted common time-domain characteristic parameters of pulse signals and established a model for predicting blood pressure continuously based on Random Forest and so on.Secondly we calibrated the pulse signal using predicted blood pressure and extracted time-domain characteristic parameters and hemodynamic parameters from calibrated pulse signal.Thirdly,classification model of pulse signal was built based on some machine learning algorithm such as Random Forest.Results The accuracy of the model established for predicting systolic and diastolic blood pressure met the American Medical Device AAMI standard,and classification effect of the pulse signal after calibration was better than the uncalibrated pulse signal.Conclusion This study provided a kind of new idea or method for analysis and identification of pulse signal,which is practically valuable for objectification of pulse diagnosis.