CN104644151B - A kind of pressure pulse wave wave travel Forecasting Methodology based on photoelectricity volume pulse signal - Google Patents

Abstract

A device for predicting the propagation of pressure pulse wave waveform based on photoplethysmographic signals is characterized in that it includes a waveform input module, a waveform conditioning module, a waveform fitting module, a waveform conversion module, and a waveform output module. Among them: the waveform conditioning module includes a preprocessing circuit, a single-shot separator and a normalization circuit, the waveform fitting module includes a fitting function setter, a waveform fitter and a waveform quality discriminator, and the waveform conversion module includes a position setter , Target function setter, characteristic crowd setter, parameter converter and waveform synthesizer. The device can use the photoplethysmographic pulse signals of various parts of the human body to predict the pressure pulse wave waveform signals of various parts of the human body according to physiological statistical laws.

Description

A Prediction Method of Pressure Pulse Wave Waveform Propagation Based on Photoplethysmography Signal

technical field

The invention relates to the technical field of medical equipment, in particular to a method for predicting pressure pulse wave waveforms of various parts based on photoplethysmography waves.

Background technique:

Pulse waves contain rich hemodynamic information. It has always been used as the basis of clinical diagnosis and treatment, and a large number of clinical measurement results have confirmed that the characteristics of pulse wave are closely related to the cardiovascular physiological state. The comprehensive information of the form (shape of wave), intensity (amplitude of wave), velocity (speed of wave) and rhythm (period of wave) exhibited by pulse wave reflects the cardiovascular system of the human body to a considerable extent. Many physiological and pathological features.

In terms of acquisition principles, the current pulse wave acquisition methods mainly include pressure pulse wave acquisition using a pressure sensor or photoplethysmography pulse wave acquisition using a photoelectric sensor. The arterial pressure pulse wave has been relatively comprehensively analyzed and studied, the fluid dynamics model is relatively clear, and the corresponding hemodynamic characteristics and cardiovascular physiological significance are widely used. However, the collection of human pressure pulse waves is easily interfered by various aspects such as the collection location, requires high operational skills, and lacks repeatability, making continuous detection inconvenient. Photoplethysmography fingertip acquisition has good stability and repeatability, but its accuracy in judging cardiovascular function is not as good as that of pressure pulse wave, which has been well studied.

Invention content:

Existing technical solutions mainly use non-physiological parameter models such as transfer functions established by the power spectrum analysis of the photoelectric volume signal of the acquisition site and the pressure pulse wave signal of the corresponding target site, and mainly focus on the fingertip volume pulse signal and radial artery pressure pulse signal. Due to the large individual differences in pulse signals, the above-mentioned devices have poor stability and single functions when applied on a large scale, and the established model itself has no physiological significance, which is not conducive to the correction and improvement of the model.

In order to solve the above problems, the present invention respectively proposes a waveform containing parameter expression of volume pulse signal and pressure pulse signal based on pulse wave physiological characteristics, and uses prior statistical laws to establish the photoplethysmography expression parameters and corresponding A group of regression equations between the parameters of the pressure pulse waveform expression at the target site. In this way, the input photoplethysmogram signal is converted into the pressure pulse wave signal of the target site. The waveform fitting method has good stability, and the regression equation group has a relatively clear physiological meaning, which is convenient for fine-tuning and improvement for different physiological states. Therefore, the problems of poor stability of predicted waveform and inability to correct the model in the prior art can be solved.

In order to achieve the above object, the technical solution adopted by the present invention is: a pressure pulse wave propagation prediction device based on photoplethysmography signal, which is characterized in that it includes a signal input conditioning module, a waveform fitting module, a waveform conversion module, and a waveform synthesis modules and output modules.

The waveform input module receives a time-domain photoplethysmography signal measured from a certain part of the human body.

The signal conditioning module preprocesses the input time-domain photoplethysmogram signal, decomposes it into single-beat signals corresponding to a single cardiac cycle by using existing technology, and normalizes the amplitude and wavelength of each single-beat signal treatment.

The waveform fitting module receives the normalized single beat pulse signal, and utilizes the given waveform containing the parameter expression f _I to apply a curve fitting algorithm to fit it, and the waveform containing the parameter expression is represented by respectively pulse The parametric expressions of main wave, dicrotic wave, and reflected wave are determined by summing, and the parameter vectors of the analytical expressions obtained by fitting are I _I , which are used as waveform characteristic parameter vectors.

The parametric expression of the photoplethysmogram signal is:

Among them, H _in , bin _in , Win _in are parameters, and t is an independent variable, indicating the number of sampling points. Parts of n=1, 2, and 3 in the expression correspond to the main wave, reflected wave, and dicrotic wave waveform of the pulse wave waveform respectively. Among them, H _in represents the amplitude of the fluctuation, _{bin represents the center position of the fluctuation, and W in represents} the width of the fluctuation.

The curve fitting process algorithm adopts the least squares algorithm, and limits the range of each parameter according to the physiological meaning of each fluctuation, and then sets the initial fitting condition H _i1 >H _i2 >H _i3 , b _i1 <b _i2 <b _i3 , _Win >0. The characteristic parameter vector _II determined by fitting is:

I _I ＝[H _i1 ,H _i2 ,H _i3 , _bi1 , _bi2 , _bi3 ,W _i1 ,W _i2 ,W _i3 ]

Under the condition that the above-mentioned waveform contains parametric expressions, the fitting effect mainly depends ^on the degree of interference received by the pulse wave acquisition, so the fitting determination coefficient R2 is used as the quantitative standard for the judgment of the waveform quality. The fitting determination coefficient R2 is a commonly used calculation method for judging the similarity of ^two curves. In the present invention, R ² is also used as an acquisition quality discrimination parameter, and the single-shot waveform corresponding to R ² less than a certain value is considered to be of poor acquisition quality and discarded.

R2 calculation ^formula is as follows:

Among them, f _i , Respectively represent the measured photoplethysmography data points, the average value of the measured photoplethysmography data and the expected value of the data point fitting, and pl is the number of single-wave data points.

The waveform conversion module is grouped according to the gender, age, and average arterial pressure index of the subject, and according to the prior statistical law between the measured part of the photoplethysmography wave and the waveform characteristic parameter of the required predicted pressure pulse wave part under the corresponding grouping, The measured photoplethysmographic characteristic parameters are used to calculate and predict the characteristic parameters of the pressure pulse wave waveform at the corresponding part.

The method for establishing the prior statistical law between the measured photoplethysmography position and the required predicted pressure pulse wave position waveform characteristic parameters under the above corresponding classification is as follows:

(1) First, the people participating in the experiment were divided into groups according to gender, age, and mean arterial pressure (MAP), where the age started at 20 years old and the interval was 5 years old. The mean arterial pressure starts at 70mmHg, with intervals of 10mmHg. Group the people who participated in the experiment. The photoelectric plethysmography waves at the ear, fingertips, and toes were simultaneously detected for the above-mentioned experimental groups, and the pressure pulse wave signals at the radial artery, brachial artery, and carotid artery were detected by pressure sensors. Thereby, the measured pressure pulse wave and photoplethysmography signals of different population characteristics are obtained.

(2) Afterwards, the statistical relationship between the measured photoplethysmography wave and the characteristic parameters of the pressure pulse wave waveform at the target site is established. Similar to the feature extraction method of the photoplethysmogram above, in order to extract the characteristic parameters of the pressure pulse wave, the pressure pulse wave expression is used to fit the measured pressure pulse wave waveform. The pressure pulse wave contains parameter expressions:

The expression parameters and domain are the same as f _I. The fitting process algorithm adopts the least squares algorithm, and limits the range of each parameter according to the physiological meaning of each fluctuation, and then sets the initial fitting condition H _o1 >H _o2 >H _o3 , b _o1 <b _o2 <b _o3 ,W _on >0.

Its characteristic parameter vector is I _O ＝[H _o1 ,H _o2 ,H _o3 ,b _o1 ,b _o2 ,b _o3 ,W _o1 ,W _o2 ,W _o3 ]

(3) Use f _I and f _O for the measured waveforms of different groups and different parts, respectively, to fit the measured photoplethysmography waveform and pressure pulse wave waveform with the waveform fitting module, and obtain the corresponding I of each photoplethysmogram signal measured. _I and the I _O vector of the pressure pulse signal. For every parameter of each position pressure pulse signal I _O vector, set up the multivariate linear regression equation of the photoplethysmogram signal feature parameter vector that corresponds to different groupings and collect simultaneously, namely:

H _o1 =TM ₁₁ ×H _i1 +TM ₁₂ ×H _i2 +...+TM ₁₉ ×W _i3 +CM ₁

H _o2 ＝TM ₂₁ ×H _i1 +TM ₂₂ ×H _i2 +......+TM ₂₉ ×W _i3 +CM ₂

H _o3 ＝TM ₃₁ ×H _i1 +TM ₃₂ ×H _i2 +......+TM ₃₉ ×W _i3 +CM ₃

b _o1 =TM ₄₁ ×H _i1 +TM ₄₂ ×H _i2 +...+TM ₄₉ ×W _i3 +CM ₄

b _o2 ＝TM ₅₁ ×H _i1 +TM ₅₂ ×H _i2 +......+TM ₅₉ ×W _i3 +CM ₅

b _o3 =TM ₆₁ ×H _i1 +TM ₆₂ ×H _i2 +......+TM ₆₉ ×W _i3 +CM ₆

W _o1 =TM ₇₁ ×H _i1 +TM ₇₂ ×H _i2 +...+TM ₇₉ ×W _i3 +CM ₇

W _o2 =TM ₈₁ ×H _i1 +TM ₈₂ ×H _i2 +......+TM ₈₉ ×W _i3 +CM ₈

W _o3 =TM ₉₁ ×H _i1 +TM ₉₂ ×H _i2 +......+TM ₉₉ ×W _i3 +CM ₉

The coefficients of TM and CM were determined by multiple linear regression equations. Arranging various parameters of I _O to establish a system of equations, arranging the coefficient matrix and constant matrix, TM and CM can be obtained, where TM is a 9-order square matrix, and CM is a 9-element column vector.

According to the TM and CM matrices obtained by using the prior probability, in the application process, the characteristic parameter vector I _O of the target part is calculated by using the waveform characteristic parameter vector I _I of the acquisition part. According to the measured photoplethysmography wave, extract its characteristic I _I , then the characteristic vector I _O of the target part can be obtained as:

I _O =TM×I _I +CM

Then, the obtained characteristic parameter vector I _O of the target part is substituted into the parametric expression f _O of the pressure pulse wave at the corresponding part to obtain the corresponding analytical expression of the pressure pulse wave. Complete the waveform conversion.

The waveform output module outputs the result of the above waveform synthesis module and the target part characteristic parameter vector I _O in the required form.

The beneficial effects of the present invention are:

(1) The device of the present invention only needs to detect the photoplethysmographic pulse signal of a single part of the human body, and can obtain the pressure pulse wave signals of different parts with a certain accuracy. The operation is simple and convenient, and the waveform quality is stable. It overcomes the problems of complicated test process, inconvenient use, and difficulty in obtaining a stable pulse waveform in the conventional pressure pulse wave acquisition process. In the actual application process, it can effectively reduce the number of direct collection of pressure pulse waves, reduce the demand for operating skills, and improve the comfort level of the testee. After a lot of verification experiments, the effect is good.

(2) For the first time, a waveform fitting device based on the characteristics of pulse wave physiological fluctuations is applied to multiple pulse waveforms, so as to analyze the physiological changes during its propagation. It facilitates the adjustment and correction of the model, and lays a certain foundation for large-scale clinical application.

Description of drawings:

Fig. 1 is a structural block diagram of the present invention

Fig. 2 is the operation flowchart of the present invention

Figure 3 is a schematic diagram of waveform fitting and characteristic parameters

Figure 4 is a measured example, and gives a comparison with the measured pressure waveform

Fig. 5 is an experimental effect diagram of the present invention in predicting the pressure pulse wave of the radial artery by using the fingertip photoplethysmography wave. R ² data is the cross-validation effect between the measured pressure pulse single-beat waveform and the predicted single-beat waveform, a total of 426 cases were tested.

detailed description:

A more typical specific implementation manner of the present invention will be described in detail below with reference to the accompanying drawings.

A typical application scenario of the present invention is to predict the pressure pulse wave signal of the radial artery by using the photoplethysmographic pulse wave signal of the fingertip of the human body. Therefore, the mature technology and high-quality waveform of fingertip acquisition can be used to obtain radial artery pressure pulse wave data with clearer physiological meaning.

As shown in Figure 1, after the prediction process starts, first select the corresponding characteristic population according to the sex, age and blood pressure of the subject, and select the corresponding prediction part according to the need. Taking a 56-year-old female with a blood pressure of 90/130mmHg as an example, set gender, age and blood pressure in step S201, the collection site is the fingertip, and the target prediction site is the radial artery. In S202, according to the settings, the system automatically selects the TM and CM matrices corresponding to the groups of people and parts according to the grouping situation of the prior statistical rules.

In step S203, the system starts to receive the measured waveform, and in step S204, the input waveform is adjusted, the baseline and power frequency interference are filtered out, the continuous waveform is separated into discrete waveforms according to the cardiac cycle, and the waveform amplitude and wavelength are normalized in units of waveforms , where the normalization of the wavelength is achieved by interpolation.

In step S205, according to the acquisition waveform fitting expression given in S203, curve fitting is performed on each single-beat waveform using the least square method.

The fitting expression is:

Set initial conditions. Thus, I _I corresponding to each single-beat waveform is obtained and R ² is calculated.

In this example, the waveform amplitude and wavelength are both defined as 100 units, then

I _I =[H _i1 ,H _i2 ,H _i3 , _{bi1 ,bi2 ,bi3 ,W i1 , W i2 , W i3} ]=[43,69,49,15,27,48,14,25,52 ]

R ² >0.99

In step S206, the waveform quality is judged according to the value of R ² , the waveform with R ² <0.95 is discarded, and the discarding ratio is recorded. If the waveform is discarded, re-extract the next beat waveform for analysis. Generally, the proportion of waveforms with R ² <0.95 is less than 3%, and when the discarding proportion is greater than 5%, consideration should be given to adjusting the acquisition method.

Step S207 converts the I _I of the qualified waveform in S206, and calculates the eigenvector I _O of the waveform of the target part, according to the formula:

I _O =TM×I _I +CM

In this example, both TM and CM are fitted by the radial artery pressure pulse wave and the fingertip photoplethysmography pulse wave signal measured at the same time, according to f _i and f _o respectively, and the corresponding regression equations are respectively established from the obtained parameters get.

Calculated to get:

I _O ＝[H _o1 ,H _o2 ,H _o3 ,b _o1 ,b _o2 ,b _o3 ,W _o1 ,W _o2 ,W _o3 ]＝[64,71,35,14,25,49,13,22,48 ]

In step _S208 , each parameter in 10 is substituted into the given target site pressure pulse wave expression, namely

The predicted waveform expression of the target part and the corresponding main wave, reflected wave and dicrotic wave predicted waveforms are obtained.

Step S209 outputs the above-mentioned waveform and parameters according to the specified format.

Claims (1)

1. A pressure pulse wave waveform propagation prediction method based on photoplethysmography signals is characterized by comprising the following steps: the device comprises a waveform input module, a signal conditioning module, a waveform fitting module, a waveform conversion module and a waveform output module;

the waveform input module receives a time domain photoelectric volume pulse signal actually measured from a certain part of a human body;

the signal conditioning module is used for preprocessing the input time domain photoplethysmography pulse signals, decomposing the preprocessed time domain photoplethysmography pulse signals into single-beat pulse signals corresponding to a single cardiac cycle, and normalizing the amplitude and wavelength of each single-beat pulse signal;

the waveform fitting module receives the normalized single-beat pulse signal and utilizes a given waveform containing a parameter expression f_IFitting the pulse waveform by using a curve fitting algorithm, wherein the parameter-containing expressions of the waveform are determined by adding the parameter-containing expressions respectively representing the main wave, the dicrotic wave and the reflected wave of the pulse waveform, and each parameter vector of the analytical expression obtained by fitting is I_IAs a waveform characteristic parameter vector;

the waveform of the photoplethysmography signal contains a reference expression as follows:

wherein H_in，b_in，W_inIs a parameter, t is an independent variable and represents the number of sampling points; in the expression, the parts of n being 1, 2 and 3 respectively correspond to a main wave, a reflected wave and a dicrotic wave of the pulse wave waveform; wherein H_inRepresenting the amplitude of the wave, b_inIndicating the center position of the wave, W_inThe width of the undulations;

the curve fitting process algorithm adopts a least square algorithm, limits each parameter range according to each fluctuation physiological significance, and then sets fitting initial conditions H_i1>H_i2>H_i3，b_i1<b_i2<b_i3,W_in>0; fitting the determined characteristic parameter vector I_IComprises the following steps:

I_I＝[H_i1,H_i2,H_i3,b_i1,b_i2,b_i3,W_i1,W_i2,W_i3]

under the condition that the waveform contains the parameter expression, the fitting effect mainly depends on the interference degree of pulse wave acquisition, so that the coefficient R is determined by fitting²As the quantification standard for waveform quality discrimination; fitting to determine the coefficient R²The method is a commonly used calculation method for judging the similarity degree of two curves; r²As an acquisition quality discrimination parameter, corresponds to R²A single beat waveform less than a certain value is considered to be sampledCollecting the quality difference and abandoning;

R²the calculation formula is as follows:

wherein,respectively representing an actually measured photoplethysmogram data point, an actually measured photoplethysmogram data average value and a data point fitting expected value, wherein pl is the number of data points of a single-beat pulse signal;

the waveform conversion module is used for grouping according to sex, age and average arterial pressure indexes of a measured person, and calculating and predicting pressure pulse wave waveform characteristic parameters of a corresponding part by utilizing measured photoplethysmography pulse wave characteristic parameters according to a priori statistical rule between the measured part of the photoplethysmography pulse wave and the waveform characteristic parameters of the part of the pressure pulse wave to be predicted under the corresponding grouping;

the method for establishing the prior statistical rule between the waveform characteristic parameters of the actual measurement part of the photoplethysmogram and the part of the pressure pulse wave needing to be predicted under the corresponding grouping comprises the following steps:

(1) firstly, grouping the population participating in the experiment according to gender, age and mean arterial pressure, wherein the age is started at 20 years and is separated from the age at 5 years; mean arterial pressure started at 70mmHg, spaced at 10 mmHg; grouping the population participating in the experiment; respectively and simultaneously detecting photoelectric volume pulse waves at ear, finger and toe positions of each group of experimental population, and detecting pressure pulse wave signals at radial artery, brachial artery and carotid artery positions by using pressure sensors; thereby obtaining actually measured pressure pulse waves and photoplethysmography signals of different crowd characteristics;

(2) then, establishing a statistical relationship between the waveform characteristic parameters of the actually measured photoplethysmogram and the actually measured pressure pulse wave of the target part; similar with above-mentioned photoplethysmography pulse wave waveform characteristic parameter vector extraction mode, for extracting pressure pulse wave waveform characteristic parameter, utilize pressure pulse wave waveform to contain the parameter expression and carry out the fitting to actual measurement pressure pulse wave waveform, pressure pulse wave waveform contains the parameter expression and is:

the expression parameters and the definition domain are all equal to f_IThe same; the fitting process algorithm adopts a least square algorithm, limits each parameter range according to each fluctuation physiological significance, and then sets fitting initial conditions H_o1>H_o2>H_o3，b_o1<b_o2<b_o3,W_on>0；

The characteristic parameter vector is I_O＝[H_o1,H_o2,H_o3,b_o1,b_o2,b_o3,W_o1,W_o2,W_o3]

(3) Respectively using f for the measured waveforms of different groups and different parts_IAnd f_OFitting the actually measured photoplethysmogram waveform and pressure pulse waveform by a waveform fitting module to obtain I corresponding to each actually measured photoplethysmogram signal_IAnd I of pressure pulse signal_OVector quantity; for each part of the pressure pulse signal I_OEstablishing a multiple linear regression equation of the feature parameter vectors of the photoplethysmography signals which correspond to different groups and are acquired simultaneously for each parameter of the vector, namely:

H_o1＝TM₁₁×H_i1+TM₁₂×H_i2+......+TM₁₉×W_i3+CM₁

H_o2＝TM₂₁×H_i1+TM₂₂×H_i2+......+TM₂₉×W_i3+CM₂

H_o3＝TM₃₁×H_i1+TM₃₂×H_i2+......+TM₃₉×W_i3+CM₃

b_o1＝TM₄₁×H_i1+TM₄₂×H_i2+......+TM₄₉×W_i3+CM₄

b_o2＝TM₅₁×H_i1+TM₅₂×H_i2+......+TM₅₉×W_i3+CM₅

b_o3＝TM₆₁×H_i1+TM₆₂×H_i2+......+TM₆₉×W_i3+CM₆

W_o1＝TM₇₁×H_i1+TM₇₂×H_i2+......+TM₇₉×W_i3+CM₇

W_o2＝TM₈₁×H_i1+TM₈₂×H_i2+......+TM₈₉×W_i3+CM₈

W_o3＝TM₉₁×H_i1+TM₉₂×H_i2+......+TM₉₉×W_i3+CM₉

each coefficient of TM and CM is determined by multiple linear regression equation; finishing I_OEstablishing an equation set by each parameter, and sorting a coefficient matrix and a constant matrix to obtain TM and CM, wherein TM is a 9-order square matrix, and CM is a 9-element column vector;

according to the TM and CM matrixes obtained by using the prior probability, in the application process, the waveform characteristic parameter vector I of the collected part is used_ICalculating a feature parameter vector I of the target region_O(ii) a Extracting characteristic I from the measured photoplethysmography_IObtaining the characteristic parameter vector I of the target part_OComprises the following steps:

I_O＝TM×I_I+CM

then obtaining the target part characteristic parameter vector I_OSubstituting into the corresponding part of the pressure pulse waveform containing parameter expression f_OObtaining a corresponding pressure pulse wave analytic expression; completing the waveform conversion;

the waveform output module converts the result of the waveform conversion module and the characteristic parameter vector I of the target part_OAnd outputting according to a required form.