Research On The Key Issues Of Correlation Between Photoplethysmography Pulse Wave And Blood Pressure

Blood pressure is an important index which can reflect the physiological status of the human body.And it is an important basis for clinical observation of cardiac function and diagnosis of cardiovascular disease.Blood pressure is affected by many factors,such as cardiovascular status,emotion,nervous regulation and so on.The common auscultation is difficult to be portable and continuous to measure blood pressure.With the continuous attention of health problems in recent years,the concept of "body area network" has been formed,and it has become a concern for people to collect blood pressure and other physiological parameters in a portable and real-time manner.Photoplethysmography(PPG)is an effective way to implement portability in the body domain network.It is usually collected at the fingertip or ear end,and is the terminal manifestation of the human cardiovascular system.It contains not only the information of the heart and the arteries,but also the nervous system,the body fluid regulation and so on.Previous studies have suggested that PPG pulse waves are related to blood pressure and are usually expressed in terms of linear relationship between pulse transit time(PTT)and blood pressure.However,affected by the complex factors of human body,the linear equation can not track BP change dynamiclly.In order to explore the possibility of portable measurement of blood pressure by PPG,this paper mainly discusses the correlation between the PPG pulse wave and the blood pressure from the view of solving complex problems.And the signal acquisition problems involved were also studied.The main content of this dissertation includes five parts.(1)A phase demodulator based method is proposed.The pulse wave detected by smart phone camera is multiplied by a sine wave reference signal and then processed by a low pass filter to extract phase.The phase can be used to calculate the frequency beat to beat.Experimental results on an Android smart phone indicated that there was no obvious difference of instantaneous heart rate between the new method and the ECG method.Compared with derivative based and filter bank based algorithms,the new algorithm consumes less phone resources and obtains instantaneous heart rate more quickly.The instantaneous heart rate is used to analysis blood pressure in frequency domain.(2)A PTT extraction method based on fuzzy logic is proposed.The R peaks of ECG and the main peaks of PPG signal were extracted by an efficiency improved derivative-based algorithm which only search extreme points in the rising side of the signal.And then three parameters was calculated as the input of the fuzzy logic disriminator:the period ratio of ECG and PPG signal and the angle of the current peak and the previous peak.Fuzzy rules were designed according to the pulse wave shape and heart rate distribution to determine whether the peak detected is true.The experiment result showed that the proposed algorithm had a higher accuracy than the other two algorithms with large disturbance.(3)The correlation between PTT and blood pressure was analyzed by complex network method.Firstly,we found that there was a linear relationship between PTT and systolic blood pressure using the MIMIC database offered by Massachusetts Institute of Technology(MIT).However,the correlation was weak for the diastolic blood pressure.After that,the correlation coefficient is used as a time series.The results of the detrended correlation analysis(DCCA)for the different time length of SBP-PTT correlation series showed that the Hurst indexs were greater than 0.5,indicating the long term memory.Finally,to eliminate the inference of nosie in the judgement of long-term memory,we transformed the series into a textual sequence then divide it into fixed length words.Each word represented a node in the network and the corresponding nodes of two adjacent words in the text were linked.The degree distribution of the blood pressure associated complex network was analyzed.The results showed that the distribution of systolic blood pressure had power-law characteristics,which was consistent with the conclusion of long-term memory.In addition,the node growth characteristics and the average shortest path length in the blood pressure mapping network were analyzed.(4)The morphological difference of PPG pulse wave between hypertensive patients and healthy subjects was analyzed by the view of complex network.The PPG pulse wave time series was mapped into complex networks by visual graph method,and the following conclusions were obtained by analyzing network parameters.First,the pulse waves of hypertensive patients are prone to appear relatively obvious tidal waves,higher perfusion peak and lower dicrotic notch peak,and these morphological characteristics can be represented by the change of motif distribution of the network.Second,visceral vascular damage caused by hypertension can affect heart rate variability,thus affecting the degree of distribution of the associated network.The above two points can not distinguish between hypertension patients and healthy people linearly,so the support vector machines based on Gauss radial kernel function is designed with the two complex network characters as the input.The experimental results of 16 patients with hypertension showed that the accuracy of SVM is more than 90%with 100Hz sample rate.(5)A method to estimate beat-to-beat systolic blood pressure variability(SBPv)during exercise is proposed.First,the heart rate variability,systolic blood pressure,pulse transit time and a parameter extracted from pulse wave shape named AM/BL were analyzed in the frequency domain before and after exercise.We found that PTT can reflect the energy change of high frequency and AM/BL can well reflect the change of lower frequency.Then total variability was obtained from the combination of high frequency component calculated by PTT and low requency component calculated by AM/BL.Experiments on 19 individuals during exercise showed that the SBPv estimated accuracy was improved compared to the the method using PTT only.Various factors have influence on the state of human body.The research on the complexity and correlation of blood pressure and pulse wave is helpful to evaluate the human health condition more accurately in the evrionment of big data and provides valuable reference for the design of sensor in the body area network.