Fetal electrocardiosignal extraction method, device, equipment and readable storage medium

Abstract

This application provides a method, apparatus, device, and readable storage medium for extracting fetal electrocardiogram (ECG) signals, relating to the field of artificial intelligence. The method includes: acquiring a pregnant woman's abdominal ECG signal; determining a pregnant woman's chest ECG signal related to the pregnant woman's ECG component within the abdominal ECG signal; eliminating the pregnant woman's ECG component from the abdominal ECG signal based on the chest ECG signal to obtain a target ECG signal; and enhancing the target ECG signal to extract the fetal ECG signal. This application addresses the problem that existing fetal ECG signal extraction methods and devices cannot accurately acquire relatively weak fetal ECG signals.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a method, apparatus, device, and readable storage medium for extracting fetal electrocardiogram signals. Background Technology

[0002] With the development of artificial intelligence, its application in the medical field is gradually increasing, such as in the extraction of fetal electrocardiogram signals.

[0003] Currently, electrocardiogram (ECG) signals are collected by placing one or more electrodes in the pregnant woman's abdomen. However, the ECG signal from the pregnant woman's abdomen is mainly composed of fetal ECG signals, maternal ECG signals, and noise. The amplitude of the maternal ECG signal is usually much larger than that of the fetus, and their frequencies are similar, which seriously affects the separation of fetal and maternal ECG signals in the time and frequency domains. Therefore, existing methods and equipment for extracting fetal ECG signals cannot accurately acquire the relatively weak fetal ECG signals, greatly affecting the reliability of fetal ECG monitoring. Summary of the Invention

[0004] This application provides a method, apparatus, device, and readable storage medium for extracting fetal electrocardiogram (ECG) signals, which solves the problem that existing fetal ECG signal extraction methods and devices cannot accurately obtain relatively weak fetal ECG signals.

[0005] According to one aspect of the embodiments of this application, a method for extracting fetal electrocardiogram signals is provided, comprising:

[0006] Obtain abdominal electrocardiogram signals from pregnant women;

[0007] Determine the pregnant woman's chest electrocardiogram signal that is related to the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal;

[0008] Based on the pregnant woman's chest electrocardiogram signal, the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal is eliminated to obtain the target electrocardiogram signal;

[0009] The target electrocardiogram (ECG) signal is enhanced to extract the fetal ECG signal.

[0010] In one possible implementation, the step of eliminating the pregnant woman's electrocardiogram component from the pregnant woman's abdominal electrocardiogram signal based on the pregnant woman's chest electrocardiogram signal to obtain the target electrocardiogram signal includes:

[0011] The electrocardiogram signal from the pregnant woman's chest was transformed to generate a fitted signal;

[0012] Minimize the difference between the pregnant woman's abdominal electrocardiogram signal and the fitted signal to indicate an update to the fitted signal;

[0013] When the difference is less than a preset threshold, the updated fitting signal is determined to be an abdominal electrocardiogram fitting signal used to characterize the electrocardiogram components of the pregnant woman.

[0014] The abdominal ECG fitting signal is removed from the pregnant woman's abdominal ECG signal to obtain the target ECG signal.

[0015] In one possible implementation, the abdominal electrocardiogram fitting signal is obtained through a temporal convolutional neural network; the temporal convolutional neural network includes an encoding unit and a decoding unit; the encoding unit includes multiple convolutional layers arranged in a cascaded order; the decoding unit includes multiple deconvolutional layers arranged in a cascaded order; at least one layer between the convolutional layers and the deconvolutional layers has a skip connection structure;

[0016] The following steps are performed using the temporal convolutional neural network to obtain the abdominal electrocardiogram fitting signal:

[0017] For each convolutional layer in the coding unit, the following steps are performed: extracting the pregnant woman's chest ECG signal and abdominal ECG signal input to the top-level convolutional layer, or the fitted signal output by the convolutional layer above it, to obtain the fitted signal output by the convolutional layer, and transmitting the fitted signal to the deconvolutional layer connected to it.

[0018] For each deconvolutional layer in the decoding unit, the following steps are performed: the fitted signal transmitted by the convolutional layer connected to it, and / or the fitted signal output by the previous deconvolutional layer, are recovered to obtain the fitted signal output by the deconvolutional layer, and the fitted signal output by the last deconvolutional layer is used as the abdominal electrocardiogram fitted signal.

[0019] In one possible implementation, each of the skip connection structures is provided with a dilated convolutional structure consisting of dilation factors of its respective level; the output of the convolutional layer serves as the input of the dilated convolutional structure, and the output of the dilated convolutional structure serves as the input of the deconvolutional layer.

[0020] In one possible implementation, the step of signal enhancement of the target electrocardiogram signal to extract the fetal electrocardiogram signal includes:

[0021] Using a preset feature matrix combined with an approximate diagonalization algorithm, residual pregnant woman ECG components and / or noise in the target ECG signal are separated and removed, thereby extracting the fetal ECG signal.

[0022] According to another aspect of the embodiments of this application, a fetal electrocardiogram signal extraction device is provided, comprising:

[0023] The first signal acquisition module is used to acquire the abdominal electrocardiogram signal of the pregnant woman;

[0024] The second signal acquisition module is used to determine the pregnant woman's chest electrocardiogram signal related to the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal.

[0025] The signal component elimination module is used to eliminate the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal based on the pregnant woman's chest electrocardiogram signal in order to obtain the target electrocardiogram signal.

[0026] The target signal enhancement module is used to enhance the target electrocardiogram signal in order to extract the fetal electrocardiogram signal.

[0027] In one possible implementation, the signal component cancellation module includes:

[0028] A signal transformation unit is used to transform the electrocardiogram signal from the pregnant woman's chest to generate a fitted signal;

[0029] A signal update unit is used to minimize the difference between the pregnant woman's abdominal electrocardiogram signal and the fitted signal, so as to indicate the update of the fitted signal;

[0030] An abdominal electrocardiogram fitting signal extraction unit is used to determine, when the difference is less than a preset threshold, the updated fitting signal as an abdominal electrocardiogram fitting signal used to characterize the electrocardiogram components of the pregnant woman.

[0031] The target electrocardiogram (ECG) signal extraction unit is used to remove the abdominal ECG fitting signal from the pregnant woman's abdominal ECG signal to obtain the target ECG signal.

[0032] In one possible implementation, the abdominal ECG fitting signal extraction unit is further configured to acquire the abdominal ECG fitting signal through a temporal convolutional neural network; the temporal convolutional neural network includes an encoding unit and a decoding unit; the encoding unit includes multiple convolutional layers arranged in a cascaded order; the decoding unit includes multiple deconvolutional layers arranged in a cascaded order; at least one layer between the convolutional layers and the deconvolutional layers has a skip connection structure.

[0033] The following steps are performed using the temporal convolutional neural network to obtain the abdominal electrocardiogram fitting signal:

[0034] For each convolutional layer in the coding unit, the following steps are performed: extracting the pregnant woman's chest ECG signal and abdominal ECG signal input to the top-level convolutional layer, or the fitted signal output by the convolutional layer above it, to obtain the fitted signal output by the convolutional layer, and transmitting the fitted signal to the deconvolutional layer connected to it.

[0035] For each deconvolutional layer in the decoding unit, the following steps are performed: the fitted signal transmitted by the convolutional layer connected to it, and / or the fitted signal output by the previous deconvolutional layer, are recovered to obtain the fitted signal output by the deconvolutional layer, and the fitted signal output by the last deconvolutional layer is used as the abdominal electrocardiogram fitted signal.

[0036] In one possible implementation, each of the skip connection structures is provided with a dilated convolutional structure consisting of dilation factors of its respective level; the output of the convolutional layer serves as the input of the dilated convolutional structure, and the output of the dilated convolutional structure serves as the input of the deconvolutional layer.

[0037] In one possible implementation, the target signal enhancement module includes:

[0038] The noise reduction unit is used to separate and remove residual pregnant woman ECG components and / or noise in the target ECG signal by using a preset feature matrix combined with an approximate diagonalization algorithm, so as to extract the fetal ECG signal.

[0039] According to another aspect of the embodiments of this application, a fetal electrocardiogram signal extraction device is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the above embodiments.

[0040] According to another aspect of the embodiments of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method described in the above embodiments.

[0041] The beneficial effects of the technical solutions provided in this application are:

[0042] This application provides a method, apparatus, device, and readable storage medium for extracting fetal electrocardiogram (ECG) signals. Using a pregnant woman's chest ECG signal, which is related to the pregnant woman's ECG component in the abdominal ECG signal, as a reference, the nonlinear relationship between the pregnant woman's chest ECG signal and the pregnant woman's ECG component is fitted to eliminate the pregnant woman's ECG component in the abdominal ECG signal, extracting the target ECG signal. The target ECG signal is then enhanced to obtain the fetal ECG signal, effectively separating the components in the pregnant woman's abdominal ECG signal. This solves the problem that existing fetal ECG signal extraction methods and devices cannot accurately acquire relatively weak fetal ECG signals, improving the accuracy of fetal ECG signal extraction and thus enhancing the reliability of fetal ECG monitoring. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0044] Figure 1 A schematic diagram of the computer system architecture provided for an embodiment of this application;

[0045] Figure 2 A flowchart illustrating a method for extracting fetal electrocardiogram signals provided in this application embodiment;

[0046] Figure 3 A schematic diagram of a process for extracting target electrocardiogram signals provided as an exemplary embodiment of this application;

[0047] Figure 4 This is a schematic diagram of the structure of a temporal convolutional neural network provided in an embodiment of this application;

[0048] Figure 5 This is a schematic diagram of the structure of the temporal convolution module provided in an embodiment of this application;

[0049] Figure 6 A schematic diagram of the process for extracting fetal electrocardiogram signals based on a temporal convolutional neural network, provided for an embodiment of this application;

[0050] Figure 7 This is a schematic diagram of the structure of a fetal electrocardiogram signal extraction device provided in an embodiment of this application;

[0051] Figure 8 This is a schematic diagram of the structure of a fetal electrocardiogram signal extraction device provided in an embodiment of this application. Detailed Implementation

[0052] The embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the embodiments described below with reference to the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions of the embodiments of this application.

[0053] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the terms “comprising” and “including” as used in embodiments of this application mean that the corresponding feature can be implemented as the presented feature, information, data, step, operation, element, and / or component, but do not exclude implementation as other features, information, data, step, operation, element, component, and / or combinations thereof supported by the art. It should be understood that when we say that an element is “connected” or “coupled” to another element, the one element can be directly connected or coupled to the other element, or it can mean that the one element and the other element establish a connection relationship through an intermediate element. Furthermore, “connected” or “coupled” as used herein can include wireless connection or wireless coupling. The term “and / or” as used herein indicates at least one of the items defined by the term; for example, “A and / or B” can be implemented as “A,” or as “B,” or as “A and B.”

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0055] Figure 1 This is a schematic diagram of a computer system architecture provided in an embodiment of this application. The computer system 100 includes multiple terminal devices 101 and a server 102. Each terminal device 101 and the server 102 are connected through a communication network, and the terminal devices 101 and the server 102 can be directly or indirectly connected through wired or wireless communication methods. This application does not impose any restrictions on this.

[0056] Terminal device 101 can be any terminal device with an application installed or capable of running a program, such as a smart camera device, smartphone, tablet computer, laptop computer, desktop computer, smart wearable device, in-vehicle device, etc., and this application embodiment does not limit it to this. Regarding the hardware structure, the terminal device 101 mentioned above includes a camera, display screen, memory, processor, and input device, but is not limited thereto. For example, the application mentioned above is a terminal-side application of a multimedia platform.

[0057] Server 102 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and big data and artificial intelligence platforms. Furthermore, in this application, server 102 can also be other processing devices with data processing capabilities. The server hosts a multimedia platform, providing background services for applications developed and running on multiple terminal devices. For example, a terminal device acquires the pregnant woman's abdominal electrocardiogram (ECG) signal, determines the pregnant woman's chest ECG signal related to the pregnant woman's ECG component in the abdominal ECG signal, and sends both the abdominal and chest ECG signals to the server. The server then eliminates the pregnant woman's ECG component from the abdominal ECG signal based on the chest ECG signal to obtain a target ECG signal, and subsequently enhances the target ECG signal to extract the fetal ECG signal. It should be noted that this exemplary embodiment achieves accurate extraction of the fetal ECG signal based on the pregnant woman's abdominal and chest ECG signals through collaborative operation between the terminal device and the server.

[0058] In an optional embodiment, the various steps of the fetal electrocardiogram (ECG) signal extraction method can be configured in the same device, such as a fetal ECG detection device. This device has the ability to collect and extract ECG signals, so that the fetal ECG signal extraction method can be run through this device to accurately extract the fetal ECG signal, thereby realizing an integrated operation for extracting the fetal ECG signal and improving operational efficiency.

[0059] Figure 2 This is a flowchart illustrating a method for extracting fetal electrocardiogram signals according to an embodiment of this application. The method includes steps S201 to S204.

[0060] S201. Obtain the abdominal electrocardiogram signal of the pregnant woman.

[0061] It should be noted that the abdominal electrocardiogram (AECG) signal of a pregnant woman is a mixed signal, containing maternal electrocardiogram (MECG) components, fetal electrocardiogram (FECG) components, and various noises such as maternal electromyogram (EMG), electric field interference, baseline drift, and random noise. More specifically, the energy of the maternal electrocardiogram component in the abdominal electrocardiogram signal is much greater than that of the fetal electrocardiogram component, and the maternal and fetal electrocardiogram components overlap in both the time and frequency domains. This makes the extraction of the fetal electrocardiogram signal highly susceptible to interference from the maternal electrocardiogram component and noise.

[0062] S202. Determine the pregnant woman's chest electrocardiogram signal related to the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal.

[0063] In this application, the electrocardiogram signal of the pregnant woman's chest is not related to the fetal electrocardiogram component or noise in the electrocardiogram signal of the pregnant woman's abdomen, but there is a nonlinear relationship with the electrocardiogram component of the pregnant woman in the electrocardiogram signal of the pregnant woman's abdomen.

[0064] S203. Based on the pregnant woman's chest electrocardiogram signal, the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal is eliminated to obtain the target electrocardiogram signal.

[0065] In this application, the pregnant woman's chest electrocardiogram (ECG) signal is used as a reference quantity. The nonlinear relationship between the pregnant woman's chest ECG signal and the pregnant woman's ECG component is fitted to realize the mapping of the pregnant woman's chest ECG signal to the pregnant woman's ECG component in the pregnant woman's abdominal ECG signal. At this time, the pregnant woman's chest ECG signal is transformed and fitted into a fitted signal that is similar to the pregnant woman's ECG component. Based on the fitted signal, the pregnant woman's ECG component in the pregnant woman's abdominal ECG signal is eliminated, thereby extracting the target ECG signal.

[0066] In one embodiment, the step of eliminating the pregnant woman's electrocardiogram component from the pregnant woman's abdominal electrocardiogram signal based on the pregnant woman's chest electrocardiogram signal to obtain the target electrocardiogram signal includes:

[0067] The electrocardiogram signal from the pregnant woman's chest was transformed to generate a fitted signal;

[0068] Minimize the difference between the pregnant woman's abdominal electrocardiogram signal and the fitted signal to indicate an update to the fitted signal;

[0069] When the difference is less than a preset threshold, the updated fitting signal is determined to be an abdominal electrocardiogram fitting signal used to characterize the electrocardiogram components of the pregnant woman.

[0070] The abdominal ECG fitting signal is removed from the pregnant woman's abdominal ECG signal to obtain the target ECG signal.

[0071] In this embodiment, the pregnant woman's chest ECG signal is transformed to generate a fitted signal, which is then updated based on the difference between the fitted signal and the pregnant woman's abdominal ECG signal. Specifically, this embodiment minimizes the difference between the fitted signal and the pregnant woman's abdominal ECG signal, making the fitted signal as close as possible to the pregnant woman's ECG component in the abdominal ECG signal that is non-linearly correlated with the pregnant woman's chest ECG signal. When the difference is less than a preset minimum error value, the output fitted signal can be used to characterize the pregnant woman's ECG component. This allows subtracting the fitted signal from the abdominal ECG signal to obtain a target ECG signal that includes the fetal ECG component but not the pregnant woman's ECG component, thus achieving a mapping from the pregnant woman's chest ECG signal to the pregnant woman's abdominal ECG signal and improving the accuracy of fetal ECG signal extraction.

[0072] For example, see Figure 3 The diagram below illustrates a flowchart for extracting a target electrocardiogram (ECG) signal, as provided in an exemplary embodiment of this application. On one hand, the pregnant woman's abdominal ECG signal is used as the main input signal s+n0 of an adaptive filter, where s represents the target ECG signal (e.g., fetal ECG signal) and n0 represents the pregnant woman's ECG component. On the other hand, the pregnant woman's chest ECG signal, which is unrelated to the fetal ECG signal but nonlinearly correlated with the pregnant woman's ECG component, is used as the reference input signal n1 of the adaptive filter. Based on the difference between the two input signals, the adaptive filter continuously transforms the reference input signal n1 to generate an output signal y. When the difference between the main input signal and the output signal y is less than a preset threshold, the output signal y approximates the pregnant woman's ECG component n0. Subtracting the output signal y from the main input signal s+n0 yields the final output signal z = s+n0 - y. At this point, signal z is the pregnant woman's abdominal ECG signal after removing the pregnant woman's ECG component, which is the target ECG signal in this application.

[0073] S204. Enhance the target electrocardiogram signal to extract the fetal electrocardiogram signal.

[0074] In this application, since the pregnant woman's abdominal electrocardiogram (ECG) signal contains fetal ECG components, pregnant woman ECG components, and various noises, after performing the above steps to eliminate the pregnant woman ECG component, the target ECG signal still contains the fetal ECG component, but also various noises, affecting the accuracy of the extracted fetal ECG signal. Therefore, this application enhances the target ECG signal to remove noise or previously residual pregnant woman ECG components, making the final extracted fetal ECG signal more accurate.

[0075] In one embodiment, the step of enhancing the target electrocardiogram signal to extract the fetal electrocardiogram signal includes:

[0076] Using a preset feature matrix combined with an approximate diagonalization algorithm, residual pregnant woman ECG components and / or noise in the target ECG signal are separated and removed, thereby extracting the fetal ECG signal.

[0077] In this embodiment, to ensure the stability of the algorithm, the Joint Approximate Diagonalization of Eigen Matrices (JADE) algorithm is used to separate the fetal ECG component, noise, and residual maternal ECG component from the target ECG signal. This removes noise and residual maternal ECG component from the target ECG signal, allowing the obtained signal to be used as the fetal ECG signal.

[0078] It should be noted that the signal enhancement algorithm used in this application is not limited to the joint approximate diagonalization algorithm of the feature matrix. For example, blind source signal separation algorithm, related algorithms for machine learning signal separation, etc., are not limited in this respect.

[0079] This application provides a method for extracting fetal electrocardiogram (ECG) signals. Using a pregnant woman's chest ECG signal, which is related to the pregnant woman's ECG component in the abdominal ECG signal, as a reference, a nonlinear relationship is fitted between the pregnant woman's chest ECG signal and the pregnant woman's ECG component to eliminate the pregnant woman's ECG component in the abdominal ECG signal, extracting the target ECG signal. The target ECG signal is then enhanced to obtain the fetal ECG signal, effectively separating the components in the pregnant woman's abdominal ECG signal. This method solves the problem that existing fetal ECG signal extraction methods and equipment cannot accurately acquire relatively weak fetal ECG signals, improving the accuracy of fetal ECG signal extraction and thus enhancing the reliability of fetal ECG monitoring.

[0080] In some embodiments, the abdominal electrocardiogram fitting signal is obtained by a temporal convolutional neural network; the temporal convolutional neural network includes an encoding unit and a decoding unit; the encoding unit includes multiple convolutional layers arranged in a cascaded order; the decoding unit includes multiple deconvolutional layers arranged in a cascaded order; at least one layer between the convolutional layers and the deconvolutional layers has a skip connection structure;

[0081] The following steps are performed using the temporal convolutional neural network to obtain the abdominal electrocardiogram fitting signal:

[0082] For each convolutional layer in the coding unit, the following steps are performed: extracting the pregnant woman's chest ECG signal and abdominal ECG signal input to the top-level convolutional layer, or the fitted signal output by the convolutional layer above it, to obtain the fitted signal output by the convolutional layer, and transmitting the fitted signal to the deconvolutional layer connected to it.

[0083] For each deconvolutional layer in the decoding unit, the following steps are performed: the fitted signal transmitted by the convolutional layer connected to it, and / or the fitted signal output by the previous deconvolutional layer, are recovered to obtain the fitted signal output by the deconvolutional layer, and the fitted signal output by the last deconvolutional layer is used as the abdominal electrocardiogram fitted signal.

[0084] In this embodiment, addressing the nonlinear correlation between the pregnant woman's abdominal ECG signal and her chest ECG signal, a temporal convolutional neural network is used to fit this nonlinear relationship, achieving a mapping from the pregnant woman's chest ECG signal to the pregnant woman's abdominal ECG component. The temporal convolutional neural network includes an encoding unit and a decoding unit. The encoding unit extracts features from the input signal at different scales to generate a fitted signal. The decoding unit upsamples and recovers the encoded fitted information to reconstruct the signal; specifically, it fits the pregnant woman's chest ECG signal into an abdominal ECG fitted signal that approximates the pregnant woman's abdominal ECG component.

[0085] In one exemplary embodiment, see Figure 4 This is a schematic diagram of the structure of the temporal convolutional neural network provided in the embodiments of this application, where the encoding unit (e.g.) Figure 4 The temporal convolutional module (located on the left side of the temporal convolutional neural network, as shown) includes multiple convolutional layers arranged in a cascaded order, and decoding units (such as...). Figure 4 The temporal convolutional module (located on the right side of the temporal convolutional neural network) includes multiple deconvolutional layers arranged in a cascaded order, with each level's convolutional and deconvolutional layers symmetrically arranged. At least one layer between each convolutional and deconvolutional layer in each level has a skip connection structure to facilitate signal transmission between the encoding and decoding units. Furthermore, residual structures are evenly distributed within each convolutional and deconvolutional layer. More specifically, the encoding unit uses a channel size of [16, 32, 64, 128, 256, 256] and a corresponding dilation factor D = [1, 2, 4, 8, 16, 16]. In this embodiment, for the execution operation of the convolutional layer, in the first level, the pregnant woman's chest ECG signal and abdominal ECG signal are used as inputs to the uppermost convolutional layer of the encoding unit for feature extraction, such as... Figure 4The average pooling operation shown generates a fitted signal, which is then passed to the next convolutional layer and, via a skip connection, to the deconvolutional layer in the first level. Next, in the second level, the fitted signal output from the previous convolutional layer (i.e., the topmost convolutional layer) is extracted to obtain the fitted signal output from that level's convolutional layer. This fitted signal is then passed to the next convolutional layer and, via a skip connection, to the deconvolutional layer in the second level. This process is repeated until the fifth convolutional layer transmits its fitted signal to the next convolutional module, which extracts the received fitted signal and uses the fitted signal output from that level's convolutional module as input to the decoding unit.

[0086] Therefore, the decoding unit upsamples and recovers the fitted signal output by the encoding unit to reconstruct the signal. For example... Figure 4 As shown, in the fifth layer, the deconvolution layer at this layer upsamples and restores the fitted signal transmitted by the convolutional layer at the fifth layer connected to it, and / or the fitted signal output by the convolutional module (e.g., Figure 4 The upsampling + temporal block shown in the diagram obtains the fitted signal output by the deconvolution layer at this level, and transmits it to the fourth-level deconvolution layer. Then, in the fourth level, the fitted signal transmitted from the convolution layer connected to it at the fourth level and / or the fitted signal output by the deconvolution layer at the fifth level are recovered to obtain the fitted signal output by that deconvolution layer, and this is transmitted to the third-level deconvolution layer. This process is repeated until the first-level deconvolution layer recovers the fitted signal, which is then used as the output of the decoding unit, i.e., the abdominal ECG fitted signal.

[0087] In one specific embodiment, both the convolutional layer and the deconvolutional layer include at least one temporal convolution module, such as... Figure 4 As shown. For more details, see [link to relevant documentation]. Figure 5 This is a schematic diagram of the structure of the temporal convolution module provided in an embodiment of this application. The temporal convolution module includes two connected processing units and a 1*1 convolutional layer (e.g., Figure 5The diagram shows a 1*1 CONV layer and a ReLU layer. The processing unit includes a Dilated Conv layer, a Batch Norm layer, a SELU layer, and a Dropout layer connected in sequence. The input of the temporal convolution module is connected to the Dilated Conv layer and the 1*1 convolution layer of one processing module, respectively. The output of the Dropout layer of this processing module is used as the input of the Dilated Conv layer in another processing module. Then, the output of the Dropout layer in this processing module is merged with the output of the 1*1 convolution layer. The merged result is used as the input of the ReLU layer, and the output of the ReLU layer is connected to the output of the temporal convolution module.

[0088] Therefore, this embodiment utilizes the nonlinear characteristics of a temporal convolutional neural network to fit the nonlinear relationship between the pregnant woman's abdominal ECG signal and the pregnant woman's chest ECG signal, realizing the mapping of the pregnant woman's chest ECG signal to the pregnant woman's abdominal ECG signal. Furthermore, this structure improves the fitting degree of the fitted signal, making the pregnant woman's chest ECG signal as close as possible to the pregnant woman's abdominal ECG signal, thus improving the accuracy of the subsequently extracted fetal ECG signal. Secondly, in this application, the pregnant woman's chest ECG signal is used as a reference signal. The model parameters of the temporal convolutional neural network are adaptively updated only when the currently input pregnant woman's abdominal ECG signal and pregnant woman's chest ECG signal are in an end-to-end learning state, without separation of training and testing phases. Considering that training and testing this neural network is a challenging machine learning task, this approach minimizes the risk of overfitting.

[0089] Based on the above embodiments, in some embodiments, each of the skip connection structures is provided with a dilated convolution structure composed of dilation factors of its respective level; the output of the convolution layer serves as the input of the dilated convolution structure, and the output of the dilated convolution structure serves as the input of the deconvolution layer.

[0090] In this embodiment, by introducing a dilated convolution structure into the temporal convolutional neural network and controlling the dilation factor in the structure, a larger receptive field is obtained and an excessive number of parameters is avoided, which greatly reduces the computational cost and depth of the network.

[0091] In an optional embodiment, to optimize the target ECG signal, reduce the risk of overfitting, and ensure the accuracy of the output results, the objective function shown in the following formula (1) is defined as the mean square error between the output of the temporal convolutional neural network and the pregnant woman's abdominal ECG signal:

[0092]

[0093] Where N represents the length of the sequence, and x(n) represents the abdominal electrocardiogram signal of the pregnant woman. This represents the output of the target ECG signal, i.e., the pregnant woman's chest ECG signal, fitted by a time convolutional neural network. The loss value for terminating the fitting is set to 0.01.

[0094] In one exemplary embodiment, see Figure 6 This is a flowchart illustrating the process of extracting fetal electrocardiogram (ECG) signals based on a temporal convolutional neural network (TCNN) according to an embodiment of this application. An adaptive filter based on the TCNN is used to eliminate the pregnant woman's ECG component from the abdominal ECG signal. The JADE algorithm is used to enhance the target ECG signal obtained after eliminating this component. Specifically, the pregnant woman's abdominal ECG signal A(t) and the pregnant woman's chest ECG signal T(t) are used as inputs to the adaptive filter based on the TCNN. The TCNN acts as a nonlinear fitting function, mapping the pregnant woman's chest ECG signal to the pregnant woman's ECG component in the abdominal ECG signal based on the minimum error criterion. M(t) is the abdominal ECG fitting signal generated by the network, and O1(t) is the target ECG signal obtained after eliminating M(t) from the pregnant woman's abdominal ECG signal. Further, the JADE algorithm is used to enhance the target ECG signal, removing residual pregnant woman ECG components and noise, which would otherwise affect subsequent fetal QRS complex detection. Finally, the fetal ECG signal is output, and O2(t) is the fetal ECG signal enhanced by the JADE algorithm.

[0095] See Figure 7 This is a schematic diagram of a fetal electrocardiogram (ECG) signal extraction device provided in an embodiment of this application. The fetal ECG signal extraction device 300 includes:

[0096] The first signal acquisition module 301 is used to acquire the abdominal electrocardiogram signal of the pregnant woman;

[0097] The second signal acquisition module 302 is used to determine the pregnant woman's chest electrocardiogram signal related to the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal.

[0098] The signal component elimination module 303 is used to eliminate the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal based on the pregnant woman's chest electrocardiogram signal in order to obtain the target electrocardiogram signal.

[0099] The target signal enhancement module 304 is used to enhance the target electrocardiogram signal in order to extract the fetal electrocardiogram signal.

[0100] In some embodiments, the signal component cancellation module 303 includes:

[0101] A signal transformation unit is used to transform the electrocardiogram signal from the pregnant woman's chest to generate a fitted signal;

[0102] A signal update unit is used to minimize the difference between the pregnant woman's abdominal electrocardiogram signal and the fitted signal, so as to indicate the update of the fitted signal;

[0103] An abdominal electrocardiogram fitting signal extraction unit is used to determine, when the difference is less than a preset threshold, the updated fitting signal as an abdominal electrocardiogram fitting signal used to characterize the electrocardiogram components of the pregnant woman.

[0104] The target electrocardiogram (ECG) signal extraction unit is used to remove the abdominal ECG fitting signal from the pregnant woman's abdominal ECG signal to obtain the target ECG signal.

[0105] In some embodiments, the abdominal electrocardiogram fitting signal extraction unit is further configured to acquire the abdominal electrocardiogram fitting signal through a temporal convolutional neural network; the temporal convolutional neural network includes an encoding unit and a decoding unit; the encoding unit includes multiple convolutional layers arranged in a cascaded order; the decoding unit includes multiple deconvolutional layers arranged in a cascaded order; at least one layer between the convolutional layers and the deconvolutional layers has a skip connection structure;

[0106] The following steps are performed using the temporal convolutional neural network to obtain the abdominal electrocardiogram fitting signal:

[0107] For each convolutional layer in the coding unit, the following steps are performed: extracting the pregnant woman's chest ECG signal and abdominal ECG signal input to the top-level convolutional layer, or the fitted signal output by the convolutional layer above it, to obtain the fitted signal output by the convolutional layer, and transmitting the fitted signal to the deconvolutional layer connected to it.

[0108] For each deconvolutional layer in the decoding unit, the following steps are performed: the fitted signal transmitted by the convolutional layer connected to it, and / or the fitted signal output by the previous deconvolutional layer, are recovered to obtain the fitted signal output by the deconvolutional layer, and the fitted signal output by the last deconvolutional layer is used as the abdominal electrocardiogram fitted signal.

[0109] In some embodiments, each of the skip connection structures is provided with a dilated convolution structure consisting of dilation factors of its respective level; the output of the convolution layer serves as the input of the dilated convolution structure, and the output of the dilated convolution structure serves as the input of the deconvolution layer.

[0110] In some embodiments, the target signal enhancement module 304 includes:

[0111] The noise reduction unit is used to separate and remove residual pregnant woman ECG components and / or noise in the target ECG signal by using a preset feature matrix combined with an approximate diagonalization algorithm, so as to extract the fetal ECG signal.

[0112] This application provides a fetal electrocardiogram (ECG) signal extraction device. This device uses a pregnant woman's chest ECG signal, which is related to the pregnant woman's ECG component in the pregnant woman's abdominal ECG signal, as a reference. It fits the nonlinear relationship between the pregnant woman's chest ECG signal and the pregnant woman's ECG component to eliminate the pregnant woman's ECG component in the pregnant woman's abdominal ECG signal, extracting the target ECG signal. Then, it enhances the target ECG signal to obtain the fetal ECG signal, effectively separating the components in the pregnant woman's abdominal ECG signal. This solves the problem that existing fetal ECG signal extraction methods and devices cannot accurately acquire relatively weak fetal ECG signals, improving the accuracy of fetal ECG signal extraction and thus enhancing the reliability of fetal ECG monitoring.

[0113] The apparatus in this application embodiment can execute the method provided in this application embodiment, and the implementation principle is similar. The actions performed by each module in the apparatus of each embodiment of this application correspond to the steps in the method of each embodiment of this application. For detailed functional descriptions of each module of the apparatus, please refer to the descriptions in the corresponding methods shown above, which will not be repeated here.

[0114] This application provides a fetal electrocardiogram (ECG) signal extraction device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps of the fetal ECG signal extraction method. Compared with related technologies, this device can achieve the following: by using a pregnant woman's chest ECG signal related to the pregnant woman's ECG component in the pregnant woman's abdominal ECG signal as a reference, it fits the nonlinear relationship between the pregnant woman's chest ECG signal and the pregnant woman's ECG component to eliminate the pregnant woman's ECG component in the pregnant woman's abdominal ECG signal, extract the target ECG signal, and then enhance the target ECG signal to obtain the fetal ECG signal. This effectively separates the components in the pregnant woman's abdominal ECG signal, solves the problem that existing fetal ECG signal extraction methods and devices cannot accurately obtain relatively weak fetal ECG signals, improves the accuracy of fetal ECG signal extraction, and thus improves the reliability of fetal ECG monitoring.

[0115] In one alternative embodiment, a fetal electrocardiogram (ECG) signal extraction device is provided, such as Figure 8 As shown, Figure 8 The fetal electrocardiogram (ECG) signal extraction device 4000 shown includes a processor 4001 and a memory 4003. The processor 4001 and memory 4003 are connected, for example, via a bus 4002. Optionally, the device 4000 may further include a transceiver 4004, which can be used for data interaction between the device and other devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the device 4000 does not constitute a limitation on the embodiments of this application.

[0116] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0117] Bus 4002 may include a pathway for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0118] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing computer programs and capable of being read by a computer, without limitation herein.

[0119] The memory 4003 stores computer programs that execute embodiments of this application, and its execution is controlled by the processor 4001. The processor 4001 executes the computer programs stored in the memory 4003 to implement the steps shown in the foregoing method embodiments.

[0120] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement the steps and corresponding content of the aforementioned method embodiments.

[0121] The terms "first," "second," "third," "fourth," "1," "2," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in a sequence other than that shown in the figures or text.

[0122] It should be understood that although arrows indicate various operation steps in the flowcharts of this application's embodiments, the order in which these steps are implemented is not limited to the order indicated by the arrows. Unless explicitly stated herein, in some implementation scenarios of this application's embodiments, the implementation steps in each flowchart can be executed in other orders as required. Furthermore, some or all steps in each flowchart, based on the actual implementation scenario, may include multiple sub-steps or multiple stages. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage can also be executed at different times. In scenarios where execution times differ, the execution order of these sub-steps or stages can be flexibly configured according to requirements, and this application's embodiments do not limit this.

[0123] The above description is only an optional implementation method for some implementation scenarios of this application. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this application without departing from the technical concept of this application also fall within the protection scope of the embodiments of this application.

Claims

1. A method for extracting fetal electrocardiogram signals, characterized in that, include: Obtain abdominal electrocardiogram signals from pregnant women; Determine the pregnant woman's chest electrocardiogram signal that is related to the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal; Based on the pregnant woman's chest ECG signal, the pregnant woman's ECG component in the pregnant woman's abdominal ECG signal is eliminated to obtain the target ECG signal, including: The electrocardiogram signal from the pregnant woman's chest was transformed to generate a fitted signal; Minimize the difference between the pregnant woman's abdominal electrocardiogram signal and the fitted signal to indicate an update to the fitted signal; When the difference is less than a preset threshold, the updated fitting signal is determined to be an abdominal electrocardiogram fitting signal used to characterize the electrocardiogram components of the pregnant woman. The abdominal electrocardiogram fitting signal is removed from the abdominal electrocardiogram signal of the pregnant woman to obtain the target electrocardiogram signal; The target electrocardiogram (ECG) signal is enhanced to extract the fetal ECG signal; The abdominal electrocardiogram fitting signal is obtained by a temporal convolutional neural network; the temporal convolutional neural network includes an encoding unit and a decoding unit; the encoding unit includes multiple convolutional layers arranged in a cascaded order; the decoding unit includes multiple deconvolutional layers arranged in a cascaded order; at least one layer between the convolutional layers and the deconvolutional layers has a skip connection structure; The following steps are performed using the temporal convolutional neural network to obtain the abdominal electrocardiogram fitting signal: For each convolutional layer in the coding unit, the following steps are performed: extracting the pregnant woman's chest ECG signal and abdominal ECG signal input to the top-level convolutional layer, or the fitted signal output by the convolutional layer above it, to obtain the fitted signal output by the convolutional layer, and transmitting the fitted signal to the deconvolutional layer connected to it. For each deconvolutional layer in the decoding unit, the following steps are performed: the fitted signal transmitted by the convolutional layer connected to it, and / or the fitted signal output by the previous deconvolutional layer, are recovered to obtain the fitted signal output by the deconvolutional layer, and the fitted signal output by the last deconvolutional layer is used as the abdominal electrocardiogram fitted signal.

2. The method according to claim 1, characterized in that, Each of the skip connection structures is provided with a dilated convolution structure consisting of dilation factors of its own level; the output of the convolution layer serves as the input of the dilated convolution structure, and the output of the dilated convolution structure serves as the input of the deconvolution layer.

3. The method according to claim 1, characterized in that, The step of enhancing the target electrocardiogram signal to extract the fetal electrocardiogram signal includes: Using a preset feature matrix combined with an approximate diagonalization algorithm, residual pregnant woman ECG components and / or noise in the target ECG signal are separated and removed, thereby extracting the fetal ECG signal.

4. A fetal electrocardiogram signal extraction device, characterized in that, include: The first signal acquisition module is used to acquire the abdominal electrocardiogram signal of the pregnant woman; The second signal acquisition module is used to determine the pregnant woman's chest electrocardiogram signal related to the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal. The signal component elimination module is used to eliminate the pregnant woman's electrocardiogram component in the pregnant woman's abdominal electrocardiogram signal based on the pregnant woman's chest electrocardiogram signal in order to obtain the target electrocardiogram signal. The target signal enhancement module is used to enhance the target electrocardiogram signal in order to extract the fetal electrocardiogram signal; The signal component cancellation module includes: A signal transformation unit is used to transform the electrocardiogram signal from the pregnant woman's chest to generate a fitted signal; A signal update unit is used to minimize the difference between the pregnant woman's abdominal electrocardiogram signal and the fitted signal, so as to indicate the update of the fitted signal; An abdominal electrocardiogram fitting signal extraction unit is used to determine, when the difference is less than a preset threshold, the updated fitting signal as an abdominal electrocardiogram fitting signal used to characterize the electrocardiogram components of the pregnant woman. A target electrocardiogram (ECG) signal extraction unit is used to remove the abdominal ECG fitting signal from the pregnant woman's abdominal ECG signal in order to obtain the target ECG signal. The abdominal electrocardiogram fitting signal extraction unit is further configured to acquire the abdominal electrocardiogram fitting signal through a temporal convolutional neural network; the temporal convolutional neural network includes an encoding unit and a decoding unit; the encoding unit includes multiple convolutional layers arranged in a cascaded order; the decoding unit includes multiple deconvolutional layers arranged in a cascaded order; at least one layer between the convolutional layers and the deconvolutional layers has a skip connection structure; The following steps are performed using the temporal convolutional neural network to obtain the abdominal electrocardiogram fitting signal: For each convolutional layer in the coding unit, the following steps are performed: extracting the pregnant woman's chest ECG signal and abdominal ECG signal input to the top-level convolutional layer, or the fitted signal output by the convolutional layer above it, to obtain the fitted signal output by the convolutional layer, and transmitting the fitted signal to the deconvolutional layer connected to it. For each deconvolutional layer in the decoding unit, the following steps are performed: the fitted signal transmitted by the convolutional layer connected to it, and / or the fitted signal output by the previous deconvolutional layer, are recovered to obtain the fitted signal output by the deconvolutional layer, and the fitted signal output by the last deconvolutional layer is used as the abdominal electrocardiogram fitted signal.

5. A fetal electrocardiogram signal extraction device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-3.

Images