Method for non-diagnostic fetal electrocardiogram acquisition

Abstract

The present application relates to the technical field of fetal electrocardiogram generation, and particularly relates to a fetal electrocardiogram acquisition method for non-diagnostic purposes. In clinical application, a non-invasive electrode method is used to acquire electrocardiogram signals of a pregnant woman's chest and electrocardiogram signals of the pregnant woman's abdomen, and then a de-noising and data processing method is used to indirectly acquire a fetal electrocardiogram. The method is passive and non-invasive, does not have any radiation problem, is suitable for long-term monitoring and recording, and does not cause any adverse effects on the health of the fetus. In addition, the support vector regression method disclosed in the present application has strong non-linear analysis capability and better data extraction accuracy, is finally beneficial to accurate acquisition of the fetal electrocardiogram, and has faster data convergence speed, which is beneficial to shortening the acquisition time of the fetal electrocardiogram.

Description

Technical Field

[0001] This invention relates to the field of fetal electrocardiogram (ECG) generation technology, and in particular to a method for obtaining fetal ECGs for non-diagnostic purposes. Background Technology

[0002] Perinatal fetal electrocardiogram (ECG) monitoring is crucial for the health of both the mother and fetus. Currently, it primarily diagnoses fetal development in the uterus by monitoring changes in heart sounds, heart rate, and electrocardiogram (ECG). Among these, the fetal ECG best reflects fetal cardiac activity. Not only can the fetal heart rate be obtained from the ECG, but changes in the ECG waveform can also directly reveal any abnormalities in fetal development. Analyzing the ECG morphology can extract important physiological indicators. However, fetal ECG signals are very weak, making it difficult to obtain a clear fetal ECG.

[0003] Previously, scalp electrodes were primarily used to obtain pure fetal electrocardiogram (ECG) signals. However, this method required ruptured membranes, which was not conducive to perinatal monitoring and could potentially harm both the mother and fetus. In recent years, with technological advancements, the use of electrodes placed on the maternal abdomen to obtain fetal ECG signals has become more widespread. However, signals obtained from maternal abdominal electrodes not only contain fetal ECG signals but also include a mixture of noise signals such as maternal ECG, electrode interference, and electromyography (EMG) signals, resulting in extremely poor purity and making it difficult to accurately assess the fetus's development in the uterus.

[0004] In recent years, fetal electrocardiogram (ECG) extraction technology has been continuously improved and developed, with extensive research conducted by experts and scholars both domestically and internationally. Examples include adaptive filtering, independent component analysis (ICA), and blind extraction based on second-order statistics. While these techniques can efficiently extract fetal heart signals, they also present certain challenges. Specifically, when using adaptive filtering, differences in electrode placement, electromyography (EMG) signals, and other noise can lead to amplitude and phase discrepancies between the maternal ECG obtained from the abdomen and the reference ECG from the chest. This prevents effective cancellation of the maternal ECG and noise effects. Furthermore, when the maternal and fetal ECG QRS waves overlap, adaptive filtering fails to suppress the maternal ECG, making it difficult to obtain a clear fetal ECG. Although much research has been conducted on acquiring maternal ECG templates, this involves complex calculations, resulting in less than ideal practical applications. When using ICA, the number of channels is critical; if the number of channels is less than the number of source signals, the source signals become difficult to distinguish, preventing the extraction of a pure fetal ECG. If the number of channels exceeds the number of source signals, the extracted signal will have more noise, requiring dimensionality reduction to decrease the noise. In cases where blind extraction based on second-order statistics is applicable, the delay period of the fetal ECG needs to be estimated based on the difference between the fetal and maternal ECG cycles. This is highly sensitive to time delay; a large estimation error may prevent the extraction of the fetal heart rate. Therefore, our research group urgently needs to solve these technical problems. Summary of the Invention

[0005] Therefore, in view of the above-mentioned existing problems and defects, the research group of this invention collected relevant data, conducted multiple evaluations and considerations, and through continuous discussion and design improvements by the research group members, the method for obtaining fetal electrocardiograms for non-diagnostic purposes was finally developed.

[0006] To address the aforementioned technical problems, the present invention relates to a method for obtaining fetal electrocardiograms for non-diagnostic purposes, comprising the following steps:

[0007] S1: Simultaneously import the pregnant woman's abdominal and chest electrocardiogram signals into the computer;

[0008] Both abdominal and chest electrocardiogram (ECG) signals were obtained using the electrode method. Abdominal ECG signals were represented as a(n), n = 1, 2, ..., l; chest ECG signals were represented as t(n), n = 1, 2, ..., l.

[0009] Where 1 represents the number of signal sample points;

[0010] S2: Establish an approximate model between abdominal signals and chest ECG signals;

[0011] a(n) = T[t(n)] + f(n) + w(n);

[0012] Where a(n) is the abdominal electrocardiogram signal, including the maternal electrocardiogram signal MECG and the fetal electrocardiogram signal FECG; t(n) is the chest electrocardiogram signal; the operator T[·] represents the nonlinear mapping from the chest to the abdomen; w(n) is the background noise;

[0013] S3: Construct the kernel function and cost function;

[0014] Construct the kernel function K(x, y) = (1 + xy / 2) 3 We set a positive number ε and a penalty parameter C, and thus obtain the cost function;

[0015]

[0016]

[0017] in, It is a Lagrange multiplier;

[0018] S4: Solve the optimization problem of the cost function to obtain the regression function;

[0019] Import the training samples, solve the optimization problem of the cost function, and obtain the regression function f(·);

[0020] The training samples are represented as follows:

[0021] S=[(t(n),a(n))],n=1,2,...,l;

[0022] S5: The maternal electrocardiogram (MECG) signal of the pregnant woman is estimated by the regression function f(·);

[0023] Substituting the chest electrocardiogram signal t(n) into the regression function f(·), the output sequence f(t(n)) is the pregnant woman's maternal electrocardiogram signal MECG;

[0024]

[0025] S6: Remove the maternal electrocardiogram (MECG) signal from the abdominal electrocardiogram (a(n)) to obtain the fetal electrocardiogram estimate;

[0026] fecg(n)=a(n)-f(t(n)).

[0027] As a further improvement to the technical solution disclosed in this invention, in step 1, the abdominal electrocardiogram signal and the chest electrocardiogram signal are preprocessed to eliminate interference signals.

[0028] As a further improvement to the technical solution disclosed in this invention, in step 1, a bandpass filter is used to eliminate interference signals in the abdominal electrocardiogram signal and the chest electrocardiogram signal.

[0029] As a further improvement to the technical solution disclosed in this invention, in step 1, the interference signal includes a baseline drift interference signal and a 50Hz power line interference signal; in the specific implementation of the bandpass filtering method, the high-pass cutoff frequency is selected as 0.8Hz and the low-pass cutoff frequency is selected as 75Hz.

[0030] In clinical applications, non-invasive electrode methods are used to obtain electrocardiogram (ECG) signals from the pregnant woman's chest and abdomen. Subsequent data processing is then used to indirectly obtain the fetal ECG. This method is passive and non-invasive, with no radiation concerns, suitable for long-term monitoring and recording, and will not cause any adverse effects on the fetus's health.

[0031] It should be noted that obtaining the mother's electrocardiogram (ECG) signal (a mixture of the mother's and fetal ECG signals) using the electrode method cannot directly extract a completely pure fetal ECG signal. This is because the fetal ECG signal is relatively weak and susceptible to interference from the mother's ECG signal and other noise factors, making extraction difficult. However, the technical solution disclosed in this invention utilizes a support vector regression algorithm to estimate the fetal ECG. While the ECG signal measured on the mother's body surface exhibits nonlinear distortion, compared to the linear independent component analysis model, the support vector regression method has stronger nonlinear analysis capabilities, better extraction accuracy, and faster data convergence speed, thus shortening the time required to obtain the fetal ECG.

[0032] Furthermore, with the addition of a small training set, the support vector regression algorithm can achieve better results than other traditional machine learning methods. This is based on minimizing structural risk by seeking support from a small number of training samples, which ultimately helps to obtain more accurate fetal electrocardiograms. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a flowchart of the method for obtaining fetal electrocardiograms for non-diagnostic purposes in this invention.

[0035] Figure 2Chest electrocardiogram of a pregnant woman obtained by applying the method of obtaining fetal electrocardiograms for non-diagnostic purposes according to the present invention.

[0036] Figure 3 Abdominal electrocardiogram of a pregnant woman obtained by applying the method of obtaining fetal electrocardiogram for non-diagnostic purposes according to the present invention.

[0037] Figure 4 The method for obtaining fetal electrocardiograms for non-diagnostic purposes according to the present invention is used to obtain maternal heart rate circulation trajectory maps of pregnant women.

[0038] Figure 5 This is a schematic diagram of a partial SVR analysis of a dataset obtained by applying the method of obtaining fetal electrocardiograms for non-diagnostic purposes according to the present invention.

[0039] Figure 6 This is a diagram showing the experimental results of the method for obtaining fetal electrocardiograms for non-diagnostic purposes according to the present invention. Detailed Implementation

[0040] To make the objectives and technical solutions of the embodiments of the present invention clearer, the technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0041] The following detailed description, in conjunction with specific embodiments, further illustrates the content disclosed in this invention. The method for obtaining fetal electrocardiograms for non-diagnostic purposes is implemented based on the Support Vector Regression (SVR) algorithm and nonlinear state-space projection theory. Figure 1 As shown in the figure, it mainly includes the following steps:

[0042] S1: Simultaneously import the pregnant woman's abdominal and chest electrocardiogram signals (e.g., ...) into the computer. Figure 2 , 3 (as shown in the image);

[0043] Both abdominal and chest electrocardiogram (ECG) signals were obtained using the electrode method. Abdominal ECG signals were represented as a(n), n = 1, 2, ..., l; chest ECG signals were represented as t(n), n = 1, 2, ..., l.

[0044] Where 1 represents the number of signal sample points;

[0045] S2: Establish an approximate model between abdominal signals and chest ECG signals (e.g.) Figure 4 (as shown in the image);

[0046] a(n) = T[t(n)] + f(n) + w(n);

[0047] Where a(n) is the abdominal electrocardiogram signal, including the maternal electrocardiogram signal MECG and the fetal electrocardiogram signal FECG; t(n) is the chest electrocardiogram signal; the operator T[·] represents the nonlinear mapping from the chest to the abdomen; w(n) is the background noise;

[0048] S3: Construct the kernel function and cost function;

[0049] Construct the kernel function K(x, y) = (1 + xy / 2) 3 We set a positive number ε and a penalty parameter C, and thus obtain the cost function;

[0050]

[0051]

[0052] in, It is a Lagrange multiplier;

[0053] S4: Solve the optimization problem of the cost function to obtain the regression function (e.g., Figure 5 (as shown in the image);

[0054] Import the training samples, solve the optimization problem of the cost function, and obtain the regression function f(·);

[0055] The training samples are represented as follows:

[0056] S=[(t(n),a(n))],n=1,2,...,l;

[0057] It is important to note that by increasing the number of training samples, the support vector regression algorithm can achieve better results than other traditional machine learning methods. This is based on minimizing the structural risk supported by a small number of training samples, which ultimately helps to obtain more accurate fetal electrocardiograms.

[0058] S5: Estimate the maternal electrocardiogram (MECG) signal of the pregnant woman using the regression function f(·). Figure 6 (as shown in the image);

[0059] Substituting the chest electrocardiogram signal t(n) into the regression function f(·), the output sequence f(t(n)) is the pregnant woman's maternal electrocardiogram signal MECG;

[0060]

[0061] S6: Remove the maternal electrocardiogram (MECG) signal from the abdominal electrocardiogram (a(n)) to obtain a fetal electrocardiogram estimate (e.g., Figure 6 (as shown in the image);

[0062] fecg(n)=a(n)-f(t(n)).

[0063] It should be noted that, in the appendix Figure 6 In the diagram, waveform line a represents the chest ECG signal, waveform line b represents the abdominal ECG signal, waveform line c represents the maternal ECG signal estimated by the SVR algorithm, and waveform line d represents the final fetal ECG signal.

[0064] The method for obtaining fetal electrocardiograms for non-diagnostic purposes is based on support vector regression (SVR) theory. In clinical applications, non-invasive electrode methods are used to acquire electrocardiogram signals from the pregnant woman's chest and abdomen. Subsequent data processing is then used to indirectly obtain the fetal electrocardiogram. This method is passive and non-invasive, with no radiation concerns, suitable for long-term monitoring and recording, and will not cause any adverse effects on the fetus's health.

[0065] It should also be noted that obtaining the mother's electrocardiogram (ECG) signal (a mixture of the mother's and fetal ECG signals) using the electrode method cannot directly extract a completely pure fetal ECG signal. This is because the fetal ECG signal is relatively weak and susceptible to interference from the mother's ECG signal and other noise factors, making extraction difficult. However, the technical solution disclosed in this invention utilizes a support vector regression algorithm to estimate the fetal ECG. While the ECG signal measured on the mother's body surface exhibits nonlinear distortion, compared to the linear independent component analysis model, the support vector regression method has stronger nonlinear analysis capabilities, better extraction accuracy, and faster data convergence speed, thus shortening the time required to obtain the fetal ECG.

[0066] It is known, based on theoretical knowledge, that the abdominal and chest ECG signals acquired from pregnant women in clinical applications contain a large number of interference signals, such as baseline drift interference and power line interference (frequency 50Hz). These interference signals inevitably negatively impact subsequent data processing and the accuracy of the acquired fetal ECGs. Therefore, as a further optimization of the above technical solution, in step 1, bandpass filtering is preferentially used to preprocess the abdominal and chest ECG signals to eliminate interference signals. Furthermore, in the specific implementation of the bandpass filtering, the high-pass cutoff frequency is selected as 0.8Hz, and the low-pass cutoff frequency is selected as 75Hz, to remove as many interference factors as possible, such as baseline drift interference and power line interference, from the pregnant woman's ECG signal.

[0067] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for obtaining fetal electrocardiograms for non-diagnostic purposes, characterized in that, Includes the following steps: S1: Simultaneously import the pregnant woman's abdominal and chest electrocardiogram (ECG) signals into the computer; both abdominal and chest ECG signals are obtained using electrode methods; the abdominal ECG signal is expressed as... Chest electrocardiogram signals are expressed as Where 1 represents the number of signal sample points; S2: Establish an approximate model between abdominal signals and chest ECG signals; ； in, This refers to abdominal electrocardiogram signals, including the pregnant woman's maternal electrocardiogram (MECG) and the fetal electrocardiogram (FECG). For chest electrocardiogram signals; operator This can be represented as a non-linear mapping from the chest to the abdomen. This is background noise; This refers to the raw mixed component of fetal electrocardiogram (FECG) signals contained in the abdominal electrocardiogram signal. S3: Construct the kernel function and cost function; Construct kernel function Set a positive number And the penalty parameter C, and from this, the cost function is obtained; ； ； in, It is a Lagrange multiplier; ≤ ≤ , ≤ ≤ , ≤ ≤ ; S4: Solve the optimization problem of the cost function to obtain the regression function; Import training samples, solve the optimization problem of the cost function, and obtain the regression function. The training samples are represented as follows: ； S5: Through regression function Estimate the maternal electrocardiogram (MECG) signal of the pregnant woman; extract the chest electrocardiogram signal. Substitute into the regression function The output sequence f(t(n)) is the pregnant woman's maternal electrocardiogram (MECG). ； Here, i and n have the same physical meaning, both being the sequence number of the signal sample point; S6: Transfer the maternal electrocardiogram (MECG) signal from the abdominal electrocardiogram signal. Remove from the middle to obtain Based on fetal electrocardiogram estimation; 。 2. The method for obtaining fetal electrocardiograms for non-diagnostic purposes according to claim 1, characterized in that, In step 1, the abdominal and chest ECG signals are preprocessed to eliminate interference signals.

3. The method for obtaining fetal electrocardiograms for non-diagnostic purposes according to claim 2, characterized in that, In step 1, bandpass filtering is used to eliminate interference signals from the abdominal and chest ECG signals.

4. The method for obtaining fetal electrocardiograms for non-diagnostic purposes according to claim 3, characterized in that, In step 1, the interference signals include baseline drift interference signals and 50Hz power line interference signals; in the specific implementation of the bandpass filtering method, the high-pass cutoff frequency is selected as 0.8Hz and the low-pass cutoff frequency is selected as 75Hz.

Images