SCG feature point identification method based on ECG reference and waveform curvature features

Abstract

The invention relates to an SCG feature point recognition method based on ECG reference and waveform curvature features, belongs to the technical field of SCG signal feature extraction, and solves the problems that in the prior art, the robustness is reduced, the feature point recognition application range is limited, and the recognition precision is low due to the fact that the form of an SCG signal changes along with different parameters such as motion artifacts, respiratory cycles, gender and age. The method comprises the following steps: acquiring synchronous ECG and SCG signals, and preprocessing the ECG and SCG signals; performing single-period waveform extraction on the SCG preprocessed signal based on the ECG preprocessed signal to obtain a single-period waveform of the SCG signal; performing curvature method processing on the single-period waveform of the SCG signal to obtain curvature characteristics of the single-period waveform; dividing the curvature characteristics into a systolic period area SCG1 and a diastolic period area SCG2 according to the period length proportion of the predicted systolic period to the equal-length SCG signal segments; and feature point marking is carried out through an extreme value taking method after peak searching. Accurate mapping of heart mechanical activities is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of SCG signal feature extraction, and in particular to a SCG feature point recognition method based on ECG reference and waveform curvature features. Background Art

[0002] Seismic cardiography (SCG) is an effective way to capture cardiac events in the form of mechanical vibrations. In healthy subjects, each SCG cardiac cycle can display 9 separate mechanical events in the form of signal peaks. The SCG peaks and corresponding mechanical events correspond to: AS = atrial contraction peak; MC = mitral valve closure; IM = isovolumetric movement; AO = aortic valve opening; IC = isobaric contraction; RE = systolic rapid ejection peak; AC = aortic valve closure; MO = mitral valve opening; RF = diastolic rapid filling peak. Accurately extracting the characteristic points of SCG can help better diagnose and monitor cardiac mechanical activity and provide an effective method for distinguishing between healthy and diseased states.

[0003] In the related technologies, the feature point recognition of SCG signals mainly includes high-frequency envelope, wavelet transform, statistical learning, etc. However, the morphology of SCG signals will change with different parameters such as motion artifacts, respiratory cycle, different genders, ages, chest sensor positions, health conditions, cardiac contractility, heart rate and body positions. Different morphological changes will greatly reduce the robustness of the above methods, resulting in limited application scope of feature point recognition and reduced accuracy of feature point recognition. Summary of the invention

[0004] In view of the above analysis, an embodiment of the present invention aims to provide an SCG feature point recognition method based on ECG reference and waveform curvature features, so as to solve the problems in the prior art that the robustness is reduced, the application scope of feature point recognition is limited, and the recognition accuracy is low due to the change of SCG signal morphology with different parameters such as motion artifacts, respiratory cycle, gender and age.

[0005] An embodiment of the present invention provides a method for identifying SCG feature points based on ECG reference and waveform curvature features, the method comprising:

[0006] Acquire synchronous ECG and SCG signals, and preprocess the ECG and SCG signals to obtain ECG preprocessing signals and SCG preprocessing signals;

[0007] Extracting a single-cycle waveform of the SCG preprocessing signal based on the ECG preprocessing signal to obtain a single-cycle waveform of the SCG signal;

[0008] Processing the single-cycle waveform of the SCG signal by a curvature method to obtain a curvature feature of the single-cycle waveform;

[0009] The curvature characteristics are divided into the systolic region SCG1 and the diastolic region SCG2 according to the predicted systolic period occupied equal length SCG signal segment cycle length ratio;

[0010] Feature points are marked on the systolic region SCG1 and the diastolic region SCG2 by using the peak-finding and extremum-taking method.

[0011] Further, obtaining synchronous ECG and SCG signals and preprocessing the ECG and SCG signals include:

[0012] The ECG signal is extracted through the ECG electrodes and preprocessed through a 0.05-40Hz second-order Butterworth filter;

[0013] The SCG signal is collected by an accelerometer or a piezoelectric film. The SCG signal collected by the accelerometer is preprocessed by a 0.08-40Hz second-order Butterworth filter. The SCG signal collected by the piezoelectric film is preprocessed by a 5-40Hz second-order Butterworth bandpass filter.

[0014] Furthermore, the step of extracting a single-cycle waveform from the SCG preprocessed signal to obtain a single-cycle waveform of the SCG signal includes:

[0015] R wave is extracted from the ECG preprocessed signal, and the position data of the R wave is used as a reference benchmark. The synchronized SCG preprocessed signal is truncated to obtain multiple single-cycle SCG segments of unequal lengths. The multiple single-cycle SCG segments of unequal lengths are processed with equal length to obtain multiple SCG segments of equal length. The multiple SCG segments of equal length are averaged in the time domain to obtain a single-cycle waveform of the SCG signal with a high signal-to-noise ratio.

[0016] Furthermore, the equal-length processing of a plurality of single-cycle SCG segments of unequal lengths comprises:

[0017] The average value of the periods of n single-period SCG segments of unequal length is selected as the period length of the equal-length SCG segments, that is, the period length of the equal-length SCG signal segments:

[0018]

[0019] Wherein, T is the period length of the equal-length SCG signal segments, n is the number of single-period SCG segments of unequal length, and Ti is the period of each single-period SCG segment of unequal length;

[0020] Cubic spline interpolation is performed on multiple SCG signal segments of unequal length to obtain multiple single-cycle SCG segments of equal length.

[0021] Furthermore, multiple equal-length SCG signal segments are averaged in the time domain to obtain a single-cycle waveform of the SCG signal with a high signal-to-noise ratio, including:

[0022] The multiple equal-length SCG signal segments are averaged in the time domain to obtain a single-cycle waveform of the SCG signal:

[0023]

[0024] Among them, x SCG_ave is the single-cycle waveform of the SCG signal, x SCG_i It is a single-cycle waveform of each SCG signal segment of equal length.

[0025] Furthermore, the curvature formula of the curvature method is:

[0026]

[0027] Where x is the single-cycle waveform of the SCG signal, is the first-order derivative of a single-cycle waveform, is the second-order derivative of the single-cycle waveform, and K is the curvature characteristic of the single-cycle waveform.

[0028] Furthermore, the curvature feature is divided into a systolic region SCG1 and a diastolic region SCG2 according to the predicted systolic period occupied by the equal length SCG signal segments, including:

[0029] The time period from the beginning of one cycle of the equal-length SCG signal segment to Ps·T is taken as the systolic region SCG1, and the time period from the beginning of one cycle of the equal-length SCG signal segment to the end of the cycle is taken as the diastolic region SCG2;

[0030] That is, the systolic period region SCG1 is [0, Ps·T), and the diastolic period region SCG2 is [Ps·T, T].

[0031] Furthermore, the feature point marking on the systolic region SCG1 and the diastolic region SCG2 by peak-finding and extreme value-taking method includes:

[0032] In the systolic region SCG1, the two values ​​with the largest negative curvature are selected, and the one with the earlier time sequence position is marked as isovolumetric movement IM, and the one with the later time sequence position is marked as isobaric contraction IC;

[0033] In the diastolic region SCG2, the two values ​​with the largest negative curvature are selected, and the one in the front time sequence position is marked as the aortic valve closure AC, and the one in the back time sequence position is marked as the mitral valve opening MO.

[0034] Furthermore, after obtaining the characteristic points of isovolumetric movement IM, isobaric contraction IC, aortic valve closure AC, and mitral valve opening MO, the characteristic point marking on the systolic region SCG1 and the diastolic region SCG2 by peak finding and extreme value taking method also includes:

[0035] The first peak after isovolumetric exercise IM is marked as aortic valve opening AO, the first trough after aortic valve opening AO is marked as isobaric contraction IC, the circular positive wave peak after isobaric contraction IC is marked as systolic rapid ejection peak RE, and the second circular peak after mitral valve opening MO is marked as diastolic rapid filling peak RF; the first peak before isovolumetric exercise IM is mitral valve closure MC, and the first peak before mitral valve closure MC is atrial contraction peak AS.

[0036] Furthermore, the position of the R wave is extracted by processing the ECG preprocessing signal using the Pan-Tompkins detection algorithm.

[0037] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0038] The present application obtains synchronous ECG signals and SCG signals, performs signal preprocessing, and extracts the single-cycle waveform of the SCG signal. The single-cycle waveform after time-domain averaging has a higher signal-to-noise ratio, which makes the subsequent feature point extraction more accurate; the single-cycle waveform of the SCG signal is processed by the curvature method to obtain the curvature characteristics of the single-cycle waveform, and the feature point recognition through the curvature characteristics can improve the robustness of feature point recognition; the curvature characteristics are divided into the systolic region SCG1 and the diastolic region SCG2 according to the ratio of the cycle lengths of the equal-length SCG signal segments occupied by the predicted systolic period; the feature points are marked on the systolic region SCG1 and the diastolic region SCG2 by the peak-finding and then extreme value method; it can effectively target the different morphological changes of SCG and improve the application scope of feature point recognition.

[0039] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0041] Figure 1 This is a flow chart of a feature point recognition method in an embodiment of the present invention;

[0042] Figure 2 : is a waveform diagram of SCG and ECG synchronization signals in an embodiment of the present invention;

[0043] Figure 3 Schematic diagram of a feature point recognition method in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0045] A specific embodiment of the present invention discloses a SCG feature point recognition method based on ECG reference and waveform curvature characteristics, such as Figure 1 The method comprises:

[0046] S11, acquiring synchronous ECG and SCG signals, and preprocessing the ECG and SCG signals to obtain ECG preprocessing signals and SCG preprocessing signals;

[0047] S12, extracting a single-cycle waveform of the SCG preprocessing signal based on the ECG preprocessing signal to obtain a single-cycle waveform of the SCG signal;

[0048] S13, performing curvature processing on the single-cycle waveform of the SCG signal to obtain the curvature characteristics of the single-cycle waveform;

[0049] S14, dividing the curvature feature into a systolic region SCG1 and a diastolic region SCG2 according to the predicted systolic period occupied equal length SCG signal segment cycle length ratio;

[0050] S15, marking characteristic points on the systolic region SCG1 and the diastolic region SCG2 by using the peak-finding and extremum-taking method.

[0051] Furthermore, step S11 acquires synchronous ECG and SCG signals, and preprocesses the ECG and SCG signals, including:

[0052] The ECG signal is extracted through the ECG electrodes and preprocessed through a 0.05-40Hz second-order Butterworth filter;

[0053] The SCG signal is collected by an accelerometer or a piezoelectric film. The SCG signal collected by the accelerometer is preprocessed by a 0.08-40Hz second-order Butterworth filter. The SCG signal collected by the piezoelectric film is preprocessed by a 5-40Hz second-order Butterworth bandpass filter.

[0054] Specifically, clear and clean synchronized ECG preprocessing signals and SCG preprocessing signals are obtained through signal preprocessing.

[0055] Different digital filters are used for different acquisition methods because the waveforms collected by different acquisition methods reflect different information.

[0056] Furthermore, the step S12 extracts the single-cycle waveform of the SCG preprocessed signal to obtain the single-cycle waveform of the SCG signal, including:

[0057] R wave is extracted from the ECG preprocessed signal, and the position data of the R wave is used as a reference benchmark. The synchronized SCG preprocessed signal is truncated to obtain multiple single-cycle SCG segments of unequal lengths. The multiple single-cycle SCG segments of unequal lengths are processed with equal length to obtain multiple SCG segments of equal length. The multiple SCG segments of equal length are averaged in the time domain to obtain a single-cycle waveform of the SCG signal with a high signal-to-noise ratio.

[0058] Specifically, Figure 2 As shown, according to two adjacent peaks of the ECG signal wave, namely the R wave, the truncation position of the SCG signal is determined according to the R wave, that is, the SCG signal is truncated at each R wave position to obtain multiple single-cycle SCG segments of unequal length.

[0059] Furthermore, the position of the R wave is extracted by processing the ECG preprocessing signal using the Pan-Tompkins detection algorithm.

[0060] Specifically, in addition to this, those skilled in the art may also select other R wave extraction methods to accurately extract the R wave according to actual needs, such as wavelet transform, EMD decomposition, etc.

[0061] Furthermore, the equal-length processing of a plurality of single-cycle SCG segments of unequal lengths comprises:

[0062] The average value of the periods of n single-period SCG segments of unequal length is selected as the period length of the equal-length SCG segments, that is, the period length of the equal-length SCG signal segments:

[0063]

[0064] Wherein, T is the period length of the equal-length SCG signal segments, n is the number of single-period SCG segments of unequal length, and Ti is the period of each single-period SCG segment of unequal length;

[0065] Cubic spline interpolation is performed on multiple SCG signal segments of unequal length to obtain multiple single-cycle SCG segments of equal length.

[0066] Specifically, the cubic spline interpolation method obtains the single-cycle waveform of each SCG signal segment of equal length:

[0067] x SCG_i =f spline (SCG i ,T)

[0068] Among them, f spline is the cubic spline interpolation function, SCG i It is a single-cycle waveform of SCG signals of unequal length.

[0069] Furthermore, multiple equal-length SCG signal segments are averaged in the time domain to obtain a single-cycle waveform of the SCG signal with a high signal-to-noise ratio, including:

[0070] The multiple equal-length SCG signal segments are averaged in the time domain to obtain a single-cycle waveform of the SCG signal:

[0071]

[0072] Among them, x SCG_ave is the single-cycle waveform of the SCG signal, x SCG_i It is a single-cycle waveform of each SCG signal segment of equal length.

[0073] Specifically, the single-cycle waveform after time-domain averaging has a higher signal-to-noise ratio, making the subsequent feature point extraction more accurate.

[0074] Furthermore, the curvature formula of the curvature method in step S13 is:

[0075]

[0076] Where x is the single-cycle waveform of the SCG signal, is the first-order derivative of a single-cycle waveform, is the second-order derivative of the single-cycle waveform, and K is the curvature characteristic of the single-cycle waveform.

[0077] Specifically, the curvature method is used to extract features under different SCG morphological changes. The curvature at the peak position is defined as positive curvature, and the curvature at the trough position is defined as negative curvature.

[0078] Furthermore, the step S14 divides the curvature feature into a systolic region SCG1 and a diastolic region SCG2 according to the predicted systolic period occupied period length ratio Ps of the SCG signal segments of equal length, including:

[0079] The time period from the beginning of one cycle of the equal-length SCG signal segment to Ps·T is taken as the systolic region SCG1, and the time period from the beginning of one cycle of the equal-length SCG signal segment to the end of the cycle is taken as the diastolic region SCG2;

[0080] That is, the contraction area SCG1 is T SCG1 =[0, Ps·T), the diastolic region SCG2 is T SCG2 =[Ps·T,T].

[0081] The relationship between the predicted systolic duration and the cycle length of the equal-length SCG signal segments can be obtained by the least squares method, and the predicted systolic duration can be obtained, and then the ratio of the predicted systolic duration to the cycle length of the equal-length SCG signal segments can be obtained:

[0082]

[0083] Where Ps is the ratio of the cycle length of the equal-length SCG signal segments occupied by the predicted contraction period, To predict the duration of contraction;

[0084] Specifically, the size of the systolic region SCG1 and the diastolic region SCG2 changes with the heart rate. Based on a large number of samples, the relationship between the predicted systolic duration S and the cycle length T of the equal-length SCG signal segment can be obtained by the least squares method:

[0085]

[0086] in, The proportion of contraction period, is the ratio of the cycle length of the systolic period to that of the equal-length SCG signal segment, both of which are constants.

[0087] Furthermore, in step S15, the characteristic point marking on the systolic region SCG1 and the diastolic region SCG2 by peak-finding and extremum-taking method includes:

[0088] In the systolic region SCG1, the two values ​​with the largest negative curvature are selected, and the one with the earlier time sequence position is marked as isovolumetric movement IM, and the one with the later time sequence position is marked as isobaric contraction IC;

[0089] In the diastolic region SCG2, the two values ​​with the largest negative curvature are selected, and the one in the front time sequence position is marked as the aortic valve closure AC, and the one in the back time sequence position is marked as the mitral valve opening MO.

[0090] Furthermore, after obtaining the characteristic points of isovolumetric movement IM, isobaric contraction IC, aortic valve closure AC, and mitral valve opening MO, the characteristic point marking on the systolic region SCG1 and the diastolic region SCG2 by peak finding and extreme value taking method also includes:

[0091] The first peak after isovolumetric exercise IM is marked as aortic valve opening AO, the first trough after aortic valve opening AO is marked as isobaric contraction IC, the circular positive wave peak after isobaric contraction IC is marked as systolic rapid ejection peak RE, and the second circular peak after mitral valve opening MO is marked as diastolic rapid filling peak RF; the first peak before isovolumetric exercise IM is mitral valve closure MC, and the first peak before mitral valve closure MC is atrial contraction peak AS.

[0092] Compared with the prior art, the present embodiment provides an SCG feature point recognition method based on ECG reference and waveform curvature features, which obtains synchronous ECG signals and SCG signals, performs signal preprocessing, and extracts the single-cycle waveform of the SCG signal. The single-cycle waveform after time domain averaging has a higher signal-to-noise ratio, which makes the subsequent feature point extraction more accurate; the single-cycle waveform of the SCG signal is processed by the curvature method to obtain the curvature characteristics of the single-cycle waveform, and the feature point recognition through the curvature characteristics can improve the robustness of feature point recognition; the curvature characteristics are divided into a systolic region SCG1 and a diastolic region SCG2 according to the predicted ratio of the period length of the SCG signal segments of equal length occupied by the systolic period; the feature points are marked on the systolic region SCG1 and the diastolic region SCG2 by the peak finding and then extreme value method; it can effectively target the different morphological changes of SCG and improve the scope of application of feature point recognition.

[0093] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0094] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for identifying SCG feature points based on ECG reference and waveform curvature features, wherein the method comprises: Acquire synchronous ECG and SCG signals, and preprocess the ECG and SCG signals to obtain ECG preprocessing signals and SCG preprocessing signals; Extracting a single-cycle waveform of the SCG preprocessing signal based on the ECG preprocessing signal to obtain a single-cycle waveform of the SCG signal; Processing the single-cycle waveform of the SCG signal by a curvature method to obtain a curvature feature of the single-cycle waveform; The curvature characteristics are divided into the systolic region SCG1 and the diastolic region SCG2 according to the predicted systolic period occupied equal length SCG signal segment cycle length ratio; Feature points are marked on the systolic region SCG1 and the diastolic region SCG2 by using the peak-finding and extremum-taking method.

2. According to claim 1, a SCG feature point recognition method based on ECG reference and waveform curvature features, wherein the characteristic point is that obtaining synchronous ECG and SCG signals and preprocessing the ECG and SCG signals comprises: The ECG signal is extracted through the ECG electrodes and preprocessed through a 0.05-40Hz second-order Butterworth filter; The SCG signal is collected by an accelerometer or a piezoelectric film. The SCG signal collected by the accelerometer is preprocessed by a 0.08-40Hz second-order Butterworth filter. The SCG signal collected by the piezoelectric film is preprocessed by a 5-40Hz second-order Butterworth bandpass filter.

3. The SCG feature point recognition method based on ECG reference and waveform curvature features according to claim 2, wherein the single-cycle waveform of the SCG signal obtained by extracting the single-cycle waveform of the SCG preprocessed signal comprises: R wave is extracted from the ECG preprocessed signal, and the position data of the R wave is used as a reference benchmark. The synchronized SCG preprocessed signal is truncated to obtain multiple single-cycle SCG segments of unequal lengths. The multiple single-cycle SCG segments of unequal lengths are processed with equal length to obtain multiple SCG segments of equal length. The multiple SCG segments of equal length are averaged in the time domain to obtain a single-cycle waveform of the SCG signal with a high signal-to-noise ratio.

4. The SCG feature point recognition method based on ECG reference and waveform curvature features according to claim 3, wherein the equal-length processing of a plurality of single-cycle SCG segments of unequal lengths comprises: The average value of the periods of n single-period SCG segments of unequal length is selected as the period length of the equal-length SCG segments, that is, the period length of the equal-length SCG signal segments: Wherein, T is the period length of the equal-length SCG signal segments, n is the number of single-period SCG segments of unequal length, and Ti is the period of each single-period SCG segment of unequal length; Cubic spline interpolation is performed on multiple SCG signal segments of unequal length to obtain multiple single-cycle SCG segments of equal length.

5. According to claim 4, a SCG feature point identification method based on ECG reference and waveform curvature features, wherein the characteristic point is that a single-cycle waveform of an SCG signal with a high signal-to-noise ratio is obtained by performing time domain averaging on multiple equal-length SCG signal segments, and comprising: The multiple equal-length SCG signal segments are averaged in the time domain to obtain a single-cycle waveform of the SCG signal: Among them, x SCG_ave is the single-cycle waveform of the SCG signal, x SCG_i It is a single-cycle waveform of each SCG signal segment of equal length.

6. The SCG feature point recognition method based on ECG reference and waveform curvature features according to claim 5, wherein the curvature formula of the curvature method is: in, x is the single-cycle waveform of the SCG signal, is the first-order derivative of a single-cycle waveform, is the second-order derivative of the single-cycle waveform, and K is the curvature characteristic of the single-cycle waveform.

7. The SCG feature point recognition method based on ECG reference and waveform curvature features according to claim 6, wherein the curvature features are divided into a systolic region SCG1 and a diastolic region SCG2 according to the ratio of the cycle lengths of the SCG signal segments of equal length occupied by the predicted systolic period, comprising: The time period from the beginning of one cycle of the equal-length SCG signal segment to Ps·T is taken as the systolic region SCG1, and the time period from the beginning of one cycle of the equal-length SCG signal segment to the end of the cycle is taken as the diastolic region SCG2; That is, the systolic period region SCG1 is [0, Ps·T), and the diastolic period region SCG2 is [Ps·T, T].

8. The SCG feature point recognition method based on ECG reference and waveform curvature features according to claim 7, wherein the feature point marking on the systolic region SCG1 and the diastolic region SCG2 by peak-finding and extremum-taking method comprises: In the systolic region SCG1, the two values ​​with the largest negative curvature are selected, and the one with the earlier time sequence position is marked as isovolumetric movement IM, and the one with the later time sequence position is marked as isobaric contraction IC; In the diastolic region SCG2, the two values ​​with the largest negative curvature are selected, and the one in the front time sequence position is marked as the aortic valve closure AC, and the one in the back time sequence position is marked as the mitral valve opening MO.

9. According to claim 8, a SCG feature point recognition method based on ECG reference and waveform curvature features, wherein after acquiring the feature points of isovolumetric movement IM, isobaric contraction IC, aortic valve closure AC, and mitral valve opening MO, the feature point marking on the systolic region SCG1 and the diastolic region SCG2 by peak-finding and then extreme value method further comprises: The first peak after isovolumetric exercise IM is marked as aortic valve opening AO, the first trough after aortic valve opening AO is marked as isobaric contraction IC, the circular positive wave peak after isobaric contraction IC is marked as systolic rapid ejection peak RE, and the second circular peak after mitral valve opening MO is marked as diastolic rapid filling peak RF; the first peak before isovolumetric exercise IM is mitral valve closure MC, and the first peak before mitral valve closure MC is atrial contraction peak AS.

10. The SCG feature point recognition method based on ECG reference and waveform curvature features according to claim 9, wherein the feature point is that the position of the R wave is extracted by processing the ECG preprocessing signal through the Pan-Tompkins detection algorithm.

Images