A dynamic blood pressure conversion method based on synchronous heart vibration signal and ballistocardiogram

By calculating the AO-I peak time interval of synchronized cardiac oscillation and cardiac impact signals, and combining time-frequency characteristics and neural network algorithms, the nonlinear error problem in non-invasive blood pressure monitoring was solved, and accurate conversion of dynamic blood pressure was achieved.

CN120419925BActive Publication Date: 2026-11-10XI AN JIAOTONG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202510487199.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-04-18
Publication Date
2026-11-10
Estimated Expiration
2045-04-18

AI Technical Summary

Technical Problem

In existing non-invasive continuous blood pressure monitoring methods, the time difference between the R wave of the electrocardiogram signal and the pulse wave signal and the pulse conduction time have non-linear errors, which affect the accuracy of dynamic blood pressure monitoring.

Method used

Synchronous cardiac oscillation signal and cardiac impact signal are used. Noise and baseline drift are removed by bandpass filtering, and differential processing is used to find the maximum point, locate the AO peak and I peak, calculate the AO-I peak time interval, and calculate dynamic blood pressure by combining time-frequency characteristics and neural network algorithm.

Benefits of technology

It accurately represents the pulse conduction time, eliminates the nonlinear error in traditional methods, and improves the accuracy of dynamic blood pressure monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120419925B_ABST
    Figure CN120419925B_ABST
Patent Text Reader

Abstract

The application discloses a dynamic blood pressure conversion method based on synchronous heart vibration signals and ballistocardiogram signals, and realizes accurate conversion of dynamic blood pressure by using a noninvasive technical means; the method adopts an aortic valve opening time difference AO-I peak time interval of synchronous heart vibration signals and ballistocardiogram signals to represent the conduction time of a vibration signal in a same heart cycle from a proximal end to a distal end, and then accurately represents the pulse conduction time, effectively eliminates the nonlinear error between the time difference between a central electrical signal R wave and a pulse wave signal and the pulse conduction time in a traditional dynamic blood pressure conversion method, and makes the prediction result more accurate. In addition, the method also extracts time-frequency features of the synchronous heart vibration signals and the ballistocardiogram signals, and assists the pulse conduction time in dynamic blood pressure conversion, and further improves the accuracy of the method.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of medical device information processing technology, specifically relating to a dynamic blood pressure conversion method based on synchronous cardiac vibration signal and cardiac impact signal. Background Technology

[0002] With the aging population and increasing urbanization, the prevalence of hypertension in my country is rising year by year, making it a major disease seriously threatening the health of the Chinese people. According to the latest Chinese Hypertension Survey, the awareness, treatment, control, and treatment-controlled rates of hypertension in different age groups are all below 20%, indicating that the situation regarding hypertension control is far from optimistic. Continuous blood pressure monitoring, as an effective control method, can rapidly, directly, and sensitively reflect instantaneous changes in arterial blood pressure, and is of great significance for real-time monitoring and quality control of patients with severe hypertension.

[0003] Currently, continuous blood pressure measurement is mainly divided into two types: one is invasive measurement, which is the arterial puncture method commonly used in clinical practice; the other is non-invasive measurement, which mostly uses physiological signals to convert blood pressure.

[0004] Invasive blood pressure measurement is considered the gold standard for continuous blood pressure monitoring due to its accuracy; however, it poses certain risks to the body and is an invasive monitoring method. Improper operation can lead to complications such as thrombosis, embolism, bleeding, and infection. Non-invasive measurement, on the other hand, offers advantages such as being non-invasive, highly repeatable, time-saving, labor-saving, simple to operate, easy to master, and safe and convenient to use. Therefore, non-invasive continuous blood pressure measurement has greater clinical value for blood pressure control in critically ill patients.

[0005] The principle behind non-invasive continuous blood pressure monitoring (NVH) is primarily based on converting blood pressure into pulse conduction time. Pulse conduction time refers to the time it takes for the pulse wave to travel from the heart to the body's periphery with each heartbeat. This time is influenced by various factors, including vascular elasticity and hemodynamics. Higher blood pressure corresponds to poorer vascular compliance (elasticity), resulting in a shorter pulse wave conduction time. Conversely, lower blood pressure leads to a longer pulse wave conduction time. Current research typically uses the time difference between the R wave in an electrocardiogram (ECG) signal and the pulse wave signal to represent pulse conduction time. However, since the R wave in an ECG is merely a marker of cardiac electrical activity, this time difference introduces a difficult-to-correct nonlinear error compared to the true pulse conduction time, thus affecting the accuracy of ambulatory blood pressure monitoring. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, the present invention aims to provide a dynamic blood pressure conversion method based on synchronous cardiac vibration signal and cardiac impact signal. The method aims to clearly locate the aortic valve opening time through cardiac vibration signal and cardiac impact signal, that is, the AO peak of cardiac vibration signal and the I peak of cardiac impact signal, which represent the conduction time of the vibration signal of the same cardiac cycle from the proximal end to the distal end, thereby accurately representing the pulse conduction time and eliminating the error of electromechanical signal conversion.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A dynamic blood pressure conversion method based on synchronized cardiac vibration signals and cardiac impact signals, characterized in that:

[0009] S1, the original synchronous signal data of cardiac vibration and cardiac impact are collected and filtered out by bandpass filtering to remove high-frequency noise components and baseline drift components, so as to obtain the effective components of cardiac vibration and cardiac impact signals;

[0010] S2: Obtain the synchronous cardiac vibration and cardiac impact signals from S1 after noise removal, perform differential processing, find the maximum point and store it as a characteristic peak sequence;

[0011] S3, using the characteristic peak sequence of S2, obtain the location point of the J peak of the cardiac impact signal and store it as a signal cutting sequence;

[0012] S4, using the signal cutting sequence of S3, the noise-removed heart impact signal is cut into sub-peak clusters, and the position of peak I is located by waveform matching method;

[0013] S5, using the I peak position of S4 to determine the AO peak positioning window range of the synchronized cardiac vibration signal;

[0014] S6, using the AO peak positioning window range of S5 to extract the feature peak sequence and locate the AO peak position;

[0015] S7 extracts the time-frequency features of cardiac oscillation and cardiac impact signals and the AO-I peak time interval, which are used as the transmission time of the same cardiac cycle vibration signal from the proximal end to the distal end. The signal is then input into a neural network algorithm to calculate dynamic blood pressure.

[0016] Further, step S1 specifically involves: filtering out high-frequency noise and baseline drift components from the acquired original synchronous signal data of cardiac vibration and cardiac impact using bandpass filtering. The cardiac vibration signal uses a bandpass filter of 1Hz to 35Hz, and the cardiac impact signal uses a bandpass filter of 5Hz to 10Hz to obtain the effective components of the cardiac vibration and cardiac impact signals.

[0017] Further, step S2 specifically involves: obtaining the synchronous oscillation and cardiac impact signals after noise removal in S1, performing differential processing, finding the maximum value points, and storing the maximum values ​​of the oscillation and cardiac impact signals as characteristic peak sequences respectively.

[0018] Further, step S3 specifically involves: using the characteristic peak sequence of the cardiac impact signal from S2, starting from the second characteristic peak, obtaining the signal segment with smaller amplitudes of the two left and right characteristic peaks and the largest amplitude of the middle characteristic peak, and positioning the middle characteristic peak with the largest amplitude as the J peak of the cardiac impact signal and storing it as a signal segmentation sequence.

[0019] Further, step S4 specifically involves: using the signal cutting sequence from S3 to cut the noise-removed cardiac impact signal into sub-peak clusters, calculating the average value of the sub-peak clusters and using it as the template waveform of the signal, matching and filtering the J-peak location of the cardiac impact signal, and finding the first trough before the J-peak through the filtered J-peak location, which is the I-peak location of the cardiac impact signal.

[0020] Further, step S5 specifically involves: using the I-peak position of S4 to precede the position by 0.5 cardiac cycles, the corresponding signal segment of the synchronous cardiac vibration signal from 0.5 cardiac cycles before the I-peak position of the cardiac impact signal to the I-peak position of the cardiac impact signal is determined as the AO-peak positioning window range of the cardiac vibration signal.

[0021] Further, step S6 specifically involves: using the AO peak positioning window range of S5, extracting the maximum value of the cardiac vibration signal within this range as the characteristic peak sequence of the cardiac vibration signal; starting from the second characteristic peak, obtaining the signal segment with smaller amplitudes of the two left and right characteristic peaks and the largest amplitude of the middle characteristic peak, and positioning the characteristic peak with the largest amplitude in the middle as the AO peak of the cardiac vibration signal.

[0022] Further, step S7 specifically involves: extracting the time-frequency features of the cardiac oscillation and cardiac impact signals, calculating the AO-I peak time interval to represent the pulse conduction time, inputting the time-frequency features and pulse conduction time into a neural network algorithm after filtering based on the correlation coefficient and cross-power spectral density, extracting features step by step through four fully connected layers with the feature dimension decreasing layer by layer, and calculating dynamic blood pressure.

[0023] Compared with the prior art, the present invention has the following beneficial technical effects:

[0024] This invention discloses a dynamic blood pressure conversion method based on synchronized cardiac vibration and cardiac impact signals. This method achieves accurate dynamic blood pressure conversion using non-invasive techniques. It uses the aortic valve opening time difference (AO-I peak time interval) between synchronized cardiac vibration and cardiac impact signals to represent the conduction time of the vibration signal from the proximal to distal end of the same cardiac cycle, thus accurately representing the pulse conduction time. This effectively eliminates the nonlinear error in the time difference between the R-wave and pulse wave signals and the pulse conduction time in traditional dynamic blood pressure conversion methods, making the prediction results more accurate. Furthermore, this method extracts the time-frequency characteristics of synchronized cardiac vibration and cardiac impact signals to assist in dynamic blood pressure conversion using pulse conduction time, further improving the accuracy of the method. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method of the present invention.

[0026] Figure 2 This is a diagram showing the result of processing the synchronized cardiac vibration signal and cardiac impact signal through a bandpass filter in an embodiment of the present invention.

[0027] Figure 3 This is a diagram showing the location results of the AO and I peaks of the synchronized cardiac vibration signal and cardiac impact signal in an embodiment of the present invention. Detailed Implementation

[0028] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. These descriptions are intended to explain the invention and not to limit it.

[0029] like Figure 1 The present invention provides a dynamic blood pressure conversion method based on synchronous oscillation signals and cardiac impact signals, comprising:

[0030] S1, the original synchronous signal data of cardiac vibration and cardiac impact are collected and filtered out by bandpass filtering to remove high-frequency noise components and baseline drift components. The cardiac vibration signal is filtered by a bandpass filter of 1Hz to 35Hz and the cardiac impact signal is filtered by a bandpass filter of 5Hz to 10Hz to obtain the effective components of the cardiac vibration and cardiac impact signals.

[0031] S2, obtain the synchronous cardiac vibration and cardiac impact signals with noise removed in S1, perform differential processing, find the maximum point, and store the maximum values ​​of the cardiac vibration signal and cardiac impact signal as characteristic peak sequences respectively;

[0032] S3, taking the second characteristic peak of the cardiac impact signal characteristic peak sequence of S2 as the starting point, obtain the signal segment with smaller amplitudes of the two left and right characteristic peaks and the largest amplitude of the middle characteristic peak, and locate the middle characteristic peak with the largest amplitude as the cardiac impact signal J peak and store it as a signal segmentation sequence;

[0033] S4. Using the signal cutting sequence of S3, the noise-removed cardiac impact signal is cut into sub-peak clusters. The average value of the sub-peak clusters is calculated and used as the template waveform of the signal. The J-peak of the cardiac impact signal is matched and filtered. The first trough before the J-peak is found through the filtered J-peak location. This trough is the position of the I-peak of the cardiac impact signal.

[0034] S5, using the I peak position of S4 0.5 cardiac cycles before, the corresponding signal segment of the synchronous cardiac vibration signal from 0.5 cardiac cycles before the I peak position of the cardiac impact signal to the I peak position of the cardiac impact signal is determined as the AO peak positioning window range of the cardiac vibration signal.

[0035] S6. Using the AO peak positioning window range of S5, extract the maximum value of the cardiac vibration signal within this range as the characteristic peak sequence of the cardiac vibration signal. Starting from the second characteristic peak, obtain the signal segment with smaller amplitudes of the two left and right characteristic peaks and the largest amplitude of the middle characteristic peak, and locate the characteristic peak with the largest amplitude in the middle as the AO peak of the cardiac vibration signal.

[0036] S7 extracts the time-frequency features of cardiac tremor and cardiac impact signals, calculates the AO-I peak time interval to represent the pulse conduction time, and inputs the time-frequency features and pulse conduction time into the neural network algorithm after filtering according to the correlation coefficient and cross power spectral density to calculate dynamic blood pressure.

[0037] The time-frequency features and pulse conduction time are filtered based on correlation coefficient and cross-power spectral density and then input into a neural network algorithm. The specific method for calculating dynamic blood pressure is as follows: features are extracted step by step through four fully connected layers, with the feature dimension decreasing layer by layer, and finally the dynamic blood pressure value is calculated.

[0038] The present invention will be further explained and illustrated below with a specific embodiment:

[0039] like Figure 2 As shown, the original synchronous signal data of cardiac tremors and cardiac impacts (see...) Figure 2 (The two figures above) High-frequency noise and baseline drift components are removed using bandpass filtering. A bandpass filter of 1Hz–35Hz is used for the tremor signal, and a bandpass filter of 5Hz–10Hz is used for the impact signal, to obtain the effective components of the tremor and impact signals (see...). Figure 2 (The two images below).

[0040] like Figure 3As shown in (a), the noise-removed synchronous cardiac vibration and cardiac impact signals are differentially processed to find the maximum value point. The maximum values ​​of the cardiac vibration signal and the cardiac impact signal are stored as characteristic peak sequences respectively. Taking the second characteristic peak of the cardiac impact signal characteristic peak sequence as the initial point, the signal segment with smaller amplitudes of the two left and right characteristic peaks and the largest amplitude of the middle characteristic peak is obtained. The middle characteristic peak with the largest amplitude is located as the J peak of the cardiac impact signal and stored as a signal cutting sequence. The noise-removed cardiac impact signal is cut into sub-peak clusters using the signal cutting sequence. The average value of the sub-peak clusters is calculated and used as the template waveform of the signal. The J peak of the cardiac impact signal is matched and filtered. The first trough before the J peak is found through the filtered J peak location. This trough is the position of the I peak of the cardiac impact signal.

[0041] like Figure 3 As shown in (b), the corresponding signal segment of the cardiac vibration signal synchronized with the cardiac vibration signal from 0.5 cardiac cycles before the position of the cardiac impact signal I peak to the position of the cardiac impact signal I peak is determined as the positioning window range of the cardiac vibration signal AO peak. The maximum value of the cardiac vibration signal within this range is extracted as the characteristic peak sequence of the cardiac vibration signal. Starting from the second characteristic peak, the signal segment with smaller amplitudes of the two left and right characteristic peaks and the largest amplitude of the middle characteristic peak is obtained, and the characteristic peak with the largest amplitude in the middle is positioned as the cardiac vibration signal AO peak.

[0042] The time-frequency features of cardiac tremor and cardiac impact signals are extracted, and the AO-I peak time interval is calculated to represent the pulse conduction time. The above features are then filtered according to the correlation coefficient and cross power spectral density and input into a neural network algorithm to calculate dynamic blood pressure.

Claims

1. A dynamic blood pressure conversion method based on synchronous cardiac vibration signal and cardiac impact signal, characterized in that, include: S1, the original synchronous signal data of cardiac vibration and cardiac impact are collected and filtered out by bandpass filtering to remove high-frequency noise components and baseline drift components, so as to obtain the effective components of cardiac vibration and cardiac impact signals; S2: Obtain the synchronous cardiac vibration and cardiac impact signals from S1 after noise removal, perform differential processing, find the maximum point and store it as a characteristic peak sequence; S3, using the characteristic peak sequence of S2, obtain the location point of the J peak of the cardiac impact signal and store it as a signal cutting sequence; S4, using the signal cutting sequence of S3, the noise-removed heart impact signal is cut into sub-peak clusters, and the position of peak I is located by waveform matching method; S5, using the I peak position of S4 to determine the AO peak positioning window range of the synchronized cardiac vibration signal; S6, using the AO peak positioning window range of S5 to extract the feature peak sequence and locate the AO peak position; S7 extracts the time-frequency features of cardiac oscillation and cardiac impact signals and the AO-I peak time interval, which are used as the transmission time of the same cardiac cycle vibration signal from the proximal end to the distal end. The signal is then input into a neural network algorithm to calculate dynamic blood pressure.

2. The dynamic blood pressure conversion method based on synchronous oscillation signal and cardiac impact signal according to claim 1, characterized in that, Step S1 specifically involves: using bandpass filtering to remove high-frequency noise and baseline drift components from the acquired original synchronous signal data of cardiac vibration and cardiac impact. A bandpass filter of 1Hz to 35Hz is used for the cardiac vibration signal, and a bandpass filter of 5Hz to 10Hz is used for the cardiac impact signal to obtain the effective components of the cardiac vibration and cardiac impact signals.

3. The dynamic blood pressure conversion method based on synchronous oscillation signal and cardiac impact signal according to claim 1, characterized in that, Step S2 specifically involves: obtaining the synchronous oscillation and cardiac impact signals from S1 after noise removal, performing differential processing, finding the maximum value points, and storing the maximum values ​​of the oscillation and cardiac impact signals as characteristic peak sequences respectively.

4. The dynamic blood pressure conversion method based on synchronous oscillation signal and cardiac impact signal according to claim 1, characterized in that, Step S3 is as follows: Using the characteristic peak sequence of the cardiac impact signal in S2, starting from the second characteristic peak, obtain the signal segment with smaller amplitudes of the two left and right characteristic peaks and the largest amplitude of the middle characteristic peak, and locate the middle characteristic peak with the largest amplitude as the J peak of the cardiac impact signal and store it as a signal segmentation sequence.

5. The dynamic blood pressure conversion method based on synchronous oscillation signal and cardiac impact signal according to claim 1, characterized in that, Step S4 specifically involves: using the signal cutting sequence from S3 to cut the noise-removed cardiac impact signal into sub-peak clusters, calculating the average value of the sub-peak clusters and using it as the template waveform of the signal, matching and filtering the J-peak location of the cardiac impact signal, and finding the first trough before the J-peak through the filtered J-peak location, which is the position of the I-peak of the cardiac impact signal.

6. The dynamic blood pressure conversion method based on synchronous oscillation signal and cardiac impact signal according to claim 1, characterized in that, Step S5 specifically involves: using the I-peak position from S4, which is 0.5 cardiac cycles prior, to determine the corresponding signal segment of the synchronous cardiac vibration signal from 0.5 cardiac cycles prior to the I-peak position of the cardiac impact signal to the I-peak position of the cardiac impact signal as the AO-peak positioning window range of the cardiac vibration signal.

7. The dynamic blood pressure conversion method based on synchronous oscillation signal and cardiac impact signal according to claim 1, characterized in that, Step S6 is as follows: Using the AO peak positioning window range of S5, extract the maximum value of the cardiac vibration signal within this range as the characteristic peak sequence of the cardiac vibration signal. Starting from the second characteristic peak, obtain the signal segment with smaller amplitudes of the two left and right characteristic peaks and the largest amplitude of the middle characteristic peak, and locate the middle characteristic peak with the largest amplitude as the AO peak of the cardiac vibration signal.

8. The dynamic blood pressure conversion method based on synchronous oscillation signal and cardiac impact signal according to claim 1, characterized in that, Step S7 specifically involves: extracting the time-frequency features of the cardiac oscillation and cardiac impact signals, calculating the AO-I peak time interval to represent the pulse conduction time, filtering the time-frequency features and pulse conduction time according to the correlation coefficient and cross-power spectral density, inputting them into a neural network algorithm, extracting features step by step through 4 fully connected layers, with the feature dimension decreasing layer by layer, and calculating the dynamic blood pressure value.

Citation Information

Patent Citations

  • Blood pressure monitoring using multi-function wrist-worn device

    CN107106054A

  • Signal processing method and device

    CN108354612A