Fetal vector electrocardiogram reconstruction method, device and processing equipment

Abstract

The present application provides a fetal vector electrocardiogram reconstruction method, apparatus, and processing equipment, which are used to design a novel fetal vector electrocardiogram reconstruction mechanism and combine deep learning technology to create an electrocardiogram reconstruction model. There is no need to arrange orthogonal leads as in the prior art. Only a small number of abdominal wall lead measurements are required to reconstruct the corresponding fetal vector electrocardiogram, complete the vector electrocardiogram detection of the fetus, and obtain more comprehensive fetal heart electrical activity information. It can provide convenient and high-quality data support, which is of great value for comprehensive diagnosis of congenital heart disease and other diagnostic / treatment services in clinical work.

Description

Technical Field

[0001] The present application relates to the field of fetal electrocardiogram detection, and specifically to a fetal vector electrocardiogram reconstruction method, device, and processing equipment. Background Art

[0002] The vector electrocardiogram is a spatial vector ring that displays the heart's electrical activity by recording and analyzing the electrocardiogram (ECG) vector information during the heart's depolarization and repolarization processes. The standard twelve (12) lead electrocardiogram records the curve of the potential changes over time between different parts of the body surface during the heart's electrical activity. In essence, it is the projection of the ECG vector ring in twelve different directions. Compared with the conventional electrocardiogram, the vector electrocardiogram is a further expansion of the twelve-lead electrocardiogram ECG vector concept in the spatial dimension. It contains complete cardiac electrical activity information and is more sensitive to the diagnosis of ECG abnormalities.

[0003] Fetal electrocardiogram technology is a non-invasive examination method used to monitor fetal heart activity. It is of great significance in clinical applications and can be used as a routine prenatal examination item. It is especially important for high-risk pregnant women. It can dynamically monitor fetal heart development and functional changes at different stages of pregnancy, and detect heart abnormalities early so that further examination and treatment can be taken in time.

[0004] The fetal electrocardiogram (ECG) is measured based on the same principle as the adult ECG. It is collected by a fetal ECG machine, usually from the maternal abdominal wall, and contains ECG waveforms of both the mother and the fetus. However, the inventors of this application have discovered that the current fetal ECG collection technology has the following two main problems:

[0005] 1) Fetal vector electrocardiograms cannot be directly obtained. The spatial orientation of the adult heart is fixed, and it is easy to arrange three orthogonal leads to directly obtain the spatial coordinates of the electrocardiogram vector. However, for the fetus, the orientation of the fetus during pregnancy is uncertain.

[0006] 2) Fetal electrocardiogram (ECG) measurements are usually collected from the maternal abdominal wall. Common clinical methods are single-lead or four (4)-lead systems. The abdominal wall has limited space and cannot accommodate orthogonal leads, so fetal vector electrocardiogram (VECG) cannot be obtained. Summary of the Invention

[0007] The present application provides a fetal vector electrocardiogram reconstruction method, apparatus, and processing equipment, which are used to design a novel fetal vector electrocardiogram reconstruction mechanism and combine deep learning technology to create an electrocardiogram reconstruction model. There is no need to arrange orthogonal leads as in the prior art. Only a small number of abdominal wall lead measurements are required to reconstruct the corresponding vector electrocardiogram, complete the vector electrocardiogram detection of the fetus, and obtain more comprehensive fetal heart electrical activity information. It can provide convenient and high-quality data support, which is of great value for comprehensive diagnosis of congenital heart disease and other diagnostic / treatment services in clinical work.

[0008] In a first aspect, the present application provides a fetal vector electrocardiogram reconstruction method, the method comprising:

[0009] Obtaining an initial fetal electrocardiogram collected from the maternal abdominal wall, wherein the initial fetal electrocardiogram is collected under acquisition conditions of a preset lead configuration scheme, wherein the preset lead configuration scheme includes three measurement leads arranged in a triangle on the maternal abdominal wall and a corresponding ground electrode arranged at the mid-lower position;

[0010] Inputting the initial fetal electrocardiogram into a preconfigured electrocardiogram reconstruction model, wherein the electrocardiogram reconstruction model is used to reconstruct the fetal electrocardiogram input into a vector electrocardiogram;

[0011] A target fetal vector electrocardiogram (VECG) corresponding to the initial fetal electrocardiogram output by the electrocardiogram reconstruction model is extracted.

[0012] In a second aspect, the present application provides a fetal vector electrocardiogram reconstruction device, comprising:

[0013] an acquisition unit, configured to acquire an initial fetal electrocardiogram acquired from the maternal abdominal wall, wherein the initial fetal electrocardiogram is acquired under acquisition conditions of a preset lead configuration scheme, wherein the preset lead configuration scheme comprises three measurement leads arranged in a triangle on the maternal abdominal wall, and a corresponding ground electrode arranged at the mid-lower back;

[0014] a reconstruction unit, configured to input the initial fetal electrocardiogram into a preconfigured electrocardiogram reconstruction model, wherein the electrocardiogram reconstruction model is configured to reconstruct the fetal electrocardiogram input into a vector electrocardiogram;

[0015] The extraction unit is used to extract the target fetal vector electrocardiogram corresponding to the initial fetal electrocardiogram output by the electrocardiogram reconstruction model.

[0016] In a third aspect, the present application provides a processing device comprising a processor and a memory, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the method provided in the first aspect of the present application or any possible implementation of the first aspect of the present application is executed.

[0017] In a fourth aspect, the present application provides a computer-readable storage medium, which stores multiple instructions, and the instructions are suitable for a processor to load to execute the method provided in the first aspect of the present application or any possible implementation of the first aspect of the present application.

[0018] From the above content, it can be concluded that this application has the following beneficial effects:

[0019] Aiming at the goal of fetal vector ECG reconstruction, this application designs a novel fetal vector ECG reconstruction mechanism and combines deep learning technology to create an ECG reconstruction model. It does not require the arrangement of orthogonal leads as in the existing technology. Only a small number of abdominal wall lead measurements are required to reconstruct the corresponding fetal vector ECG, complete the vector ECG detection of the fetus, and obtain more comprehensive fetal heart electrical activity information. It can provide convenient and high-quality data support, which is of great value for comprehensive diagnosis of congenital heart disease and other diagnostic / treatment services in clinical work. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 A schematic diagram of a process of the fetal vector electrocardiogram reconstruction method of the present application;

[0022] Figure 2 A schematic diagram of a scenario for the preset lead configuration scheme of this application;

[0023] Figure 3 A schematic diagram of another scenario of the preset lead configuration scheme for this application;

[0024] Figure 4 A schematic diagram of a scenario for the model training architecture of this application;

[0025] Figure 5 A schematic diagram of the structure of the fetal vector electrocardiogram reconstruction device of the present application;

[0026] Figure 6 This is a structural diagram of the processing equipment for this application. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0028] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or that are inherent to these processes, methods, products or devices. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can be changed in the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0029] The division of modules in this application is a logical division. In actual application, there may be other division methods. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in this application. Moreover, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed into multiple circuit modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this application.

[0030] Before introducing the fetal vector electrocardiogram reconstruction method provided by this application, the background content involved in this application is first introduced.

[0031] The fetal vector electrocardiogram reconstruction method, device and computer-readable storage medium provided in this application can be applied to processing equipment to design a novel fetal vector electrocardiogram reconstruction mechanism and combine deep learning technology to create an electrocardiogram reconstruction model. There is no need to arrange orthogonal leads as in the existing technology. Only a small number of abdominal wall lead measurements are required to reconstruct the corresponding fetal vector electrocardiogram, complete the vector electrocardiogram detection of the fetus, and obtain more comprehensive fetal heart electrical activity information. It can provide convenient and high-quality data support, which is of great value for comprehensive diagnosis of congenital heart disease and other diagnostic / treatment services in clinical work.

[0032] The fetal vector electrocardiogram (VECG) reconstruction method mentioned in this application can be implemented by a fetal vector electrocardiogram (VECG) reconstruction device, or by various types of processing devices, such as an electrocardiogram (ECG) machine, server, physical host, or user equipment (UE) that integrates the fetal vector electrocardiogram (VECG) reconstruction device. The fetal vector electrocardiogram (VECG) reconstruction device can be implemented using hardware or software, and the UE can be a terminal device such as a smartphone, tablet computer, laptop computer, desktop computer, or personal digital assistant (PDA). The processing device can be configured as a device cluster.

[0033] Among them, it is necessary to understand that, in actual applications, the processing equipment that executes the fetal vector electrocardiogram reconstruction method of the present application or is equipped with the corresponding application service of the fetal vector electrocardiogram reconstruction method of the present application, considering that the present application is aimed at the reconstruction goal of the fetal vector electrocardiogram, can be directly for the fetal electrocardiograph, that is, it is mainly from the software level to optimize the existing fetal electrocardiograph. In addition, the present application scheme may also involve a fetal electrocardiogram optimization service based on the existing fetal electrocardiograph, which can be expanded to different electrocardiographs. Therefore, in this case, the specific device type and device deployment form of the processing equipment have significant flexibility, which can meet the fetal electrocardiogram acquisition / application needs in the clinical work of the hospital, and thus can promote the comprehensive diagnosis of congenital heart disease and other diagnostic / treatment services in clinical work to be promoted with higher quality.

[0034] Next, the fetal vector electrocardiogram reconstruction method provided by this application is introduced.

[0035] First, see Figure 1 , Figure 1 A schematic flow chart of the fetal vector electrocardiogram reconstruction method of the present application is shown. The fetal vector electrocardiogram reconstruction method provided by the present application may specifically include the following steps S101 to S103:

[0036] Step S101, obtaining an initial fetal electrocardiogram collected from the maternal abdominal wall, wherein the initial fetal electrocardiogram is collected under the collection conditions of a preset lead configuration scheme, wherein the preset lead configuration scheme has three measurement leads arranged in a triangle on the maternal abdominal wall, and the corresponding ground electrode is arranged at the middle of the lower back;

[0037] It can be seen that this application designs a novel fetal vector ECG reconstruction mechanism and combines deep learning technology to create an ECG reconstruction model, aiming to reconstruct the fetal ECG collected clinically into the corresponding fetal vector ECG through this ECG reconstruction model.

[0038] In this case, during the application of the model, the processing of the present application solution can be started by obtaining the initial fetal electrocardiogram collected from the maternal abdominal wall that needs to be reconstructed.

[0039] It should be noted that the acquisition and processing of the initial fetal electrocardiogram here can be either manual entry, local extraction or extraction from other equipment, or real-time collection. It is relatively flexible and can be flexibly configured according to actual conditions.

[0040] It can be seen that this application corresponds to the novel fetal vector electrocardiogram reconstruction mechanism designed, and has created a set of preset lead configuration schemes, which can also be said to be a novel electrocardiograph usage scheme. In the preset lead configuration scheme, three pre-selected measurement leads are arranged in a triangle on the mother's abdominal wall (surface), and the corresponding ground electrode is arranged in the middle position of the lower back (surface).

[0041] In specific applications, the fetal electrocardiograph that collects the initial fetal electrocardiograph here can be either an existing fetal electrocardiograph or a fetal electrocardiograph that is adaptively configured according to the setting of the three measurement leads of the present application. Both are possible, as long as the electrocardiograph collection requirements of the present application based on three predetermined measurement leads are met.

[0042] Step S102, inputting the initial fetal electrocardiogram into a pre-configured electrocardiogram reconstruction model, wherein the electrocardiogram reconstruction model is used to reconstruct the fetal electrocardiogram input into a vector electrocardiogram;

[0043] It can be understood that this application introduces a deep learning model, which is a neural network model with strong learning ability. It creates an electrocardiogram reconstruction model equipped with the fetal vector electrocardiogram reconstruction logic designed in this application. The model is initialized by training in advance through the corresponding training samples, that is, the annotated three-lead electrocardiogram.

[0044] Therefore, after obtaining the initial fetal electrocardiogram that needs to be reconstructed, it can be input / imported into a pre-configured electrocardiogram reconstruction model, allowing the electrocardiogram reconstruction model to carry out the corresponding reconstruction processing. After the reconstruction processing is completed, the electrocardiogram reconstruction model will output the corresponding processing result, that is, the target fetal vector electrocardiogram, which will help promote comprehensive diagnosis of congenital heart disease and other diagnostic / treatment services.

[0045] Step S103 : extracting the target fetal vector electrocardiogram (VECG) output by the electrocardiogram reconstruction model and corresponding to the initial fetal electrocardiogram.

[0046] It is easy to understand that when the ECG reconstruction model completes the reconstruction process and outputs the target fetal vector ECG corresponding to the initial fetal ECG, the target fetal vector ECG can be extracted, completing a model application and completing a reconstruction / restoration process from the fetal three-lead ECG to the fetal vector ECG.

[0047] The reconstructed target fetal vector electrocardiogram is specifically represented by xyz coordinates and may also involve a time dimension (t), which can be specifically represented by a function q=f(x, y, z, t).

[0048] At this time, corresponding to the data application needs, further data application links may be involved. For example, local storage, remote storage, result display, result printing, result push or result analysis (disease diagnosis, etc.) can be performed based on the target fetal vector electrocardiogram.

[0049] It is understandable that specific data application processing can be adaptively adjusted according to pre-configured and real-time configured data application strategies, which is relatively flexible in actual situations.

[0050] So, from Figure 1 As can be seen from the illustrated embodiments, in order to achieve the goal of fetal vector ECG reconstruction, the present application designs a novel fetal vector ECG reconstruction mechanism and combines deep learning technology to create an ECG reconstruction model. There is no need to arrange orthogonal leads as in the prior art. Only a small number of abdominal wall lead measurements are required to reconstruct the corresponding fetal vector ECG, complete the vector ECG detection of the fetus, and obtain more comprehensive fetal heart electrical activity information. It can provide convenient and high-quality data support, which is of great value for comprehensive diagnosis of congenital heart disease and other diagnostic / treatment services in clinical work.

[0051] Among them, it is worth mentioning that in existing clinical work and theoretical research, vector electrocardiograms only involve adults, and there is no fetal vector electrocardiogram. This is because in actual situations, the fetal electrocardiogram cannot be directly collected by the fetal electrocardiogram machine, and the fetal vector electrocardiogram cannot be processed based on the fetal electrocardiogram machine. Under the present application scheme, the corresponding vector electrocardiogram can be reconstructed based on the electrocardiogram measured by the three measurement leads. This can be said to be a breakthrough in providing a fetal vector electrocardiogram measurement scheme, which can provide fetal vector electrocardiograms conveniently and accurately for clinical work, creating a new application scenario, and can better carry out comprehensive diagnosis of congenital heart disease and other diagnostic / treatment services.

[0052] Continue to the above Figure 1 Each step of the illustrated embodiment and its possible implementation in practical applications are described in detail.

[0053] For the three measurement leads in the preset lead configuration scheme of this application, this application further provides specific preferred solutions.

[0054] Correspondingly, reference Figure 2 A schematic diagram of a scenario of a preset lead configuration scheme of the present application is shown. As an exemplary embodiment, in the preset lead configuration scheme, the three measurement leads are respectively recorded as the first measurement lead, the second measurement lead, and the third measurement lead. On the front of the mother, the arrangement point of the first measurement lead is specifically arranged along the waistline on the right side of the mother's abdominal wall (corresponding to the right hand side), the arrangement points of the second measurement lead and the arrangement points of the third measurement lead are specifically arranged on the left side of the mother's abdominal wall, and the midpoint between the arrangement points of the second measurement lead and the third measurement lead is located at the waistline;

[0055] The projections of the layout points of the first measurement lead, the second measurement lead, and the third measurement lead on the front face (which can be understood as a plane perpendicular to the horizontal plane and facing the mother body) form an isosceles triangle. The length of the line connecting the projection of the layout point of the first measurement lead and the projection of the layout point of the second measurement lead is equal to the length of the line connecting the projection of the layout point of the first measurement lead and the projection of the layout point of the third measurement lead. The straight line formed by the projection of the layout point of the second measurement lead and the projection of the layout point of the third measurement lead is perpendicular to the horizontal plane (i.e., viewed from the front, the three form an isosceles triangle with upper and lower sides of equal length and the base perpendicular to the horizontal plane).

[0056] In the positional relationship defined by the projection here, it can be understood that the maternal abdominal wall usually exists in an arc shape. Therefore, in actual situations, the arrangement points of the second measurement lead and the arrangement points of the third measurement lead will not / are difficult to be located on the same straight line perpendicular to the horizontal plane in most cases, that is, the projections of the two on the horizontal plane will not overlap. On the contrary, in a few cases, they can be located on the same straight line perpendicular to the horizontal plane. Figure 3 As shown in another scenario schematic diagram of the preset lead configuration scheme of the present application, the layout point of the second measurement lead and the layout point of the third measurement lead are located on the same straight line perpendicular to the horizontal plane, that is, the projections of the two on the horizontal plane overlap.

[0057] It can be understood that the above measurement lead configuration scheme is convenient to operate in actual operation, and can also bring about excellent electrocardiogram reconstruction effect.

[0058] Further, from Figure 3 It can also be seen that this application also provides a more specific implementation plan for the designed preset lead configuration scheme from a specific perspective. While providing a more specific implementation supporting plan, it can continue to enhance the convenient operation characteristics and electrocardiogram reconstruction effect.

[0059] Specifically, as another exemplary embodiment, the arrangement points of the first measurement lead and the arrangement points of the ground electrode constitute a first vector, the arrangement points of the second measurement lead and the arrangement points of the ground electrode constitute a second vector, and the angle between the first vector and the second vector is 50°. Similarly, the third measurement lead and the second measurement lead also have the corresponding 50° angle under the above position conditions.

[0060] The arrangement point of the third measurement lead and the arrangement point of the ground electrode form a third vector, and the angle between the second vector and the third vector is 20°.

[0061] As an example, the layout point of the first measurement lead is recorded as R, the layout point of the second measurement lead is recorded as L1, the layout point of the third measurement lead is recorded as L2, and the layout point of the ground electrode is recorded as N. Under this point position condition, the angle N in the triangle formed by points R, N and L1 is 50°; the angle N in the triangle formed by points R, N and L2 is 50°; the angle N in the triangle formed by points L1, N and L2 is 20°.

[0062] At the same time, corresponding to the application of the model, the present application may also involve a prior model training process. To this end, before step S102 inputs the initial fetal electrocardiogram into the pre-configured electrocardiogram reconstruction model, the present application method may further include:

[0063] The ECG reconstruction model is trained based on the labeled training samples.

[0064] It can be understood that the training sample is an electrocardiogram of the preset lead configuration scheme involved in this application, which can be an electrocardiogram configured for the preset lead configuration scheme or an electrocardiogram collected under the preset lead configuration scheme. Both are acceptable.

[0065] The annotation / labeling of training samples can also be understood in terms of true values ​​and theoretically desired model prediction results. Annotation can be used to guide the model's sensitivity and specificity for reconstructed vector electrocardiograms, thereby guiding model training.

[0066] The model training process usually includes the following training logic:

[0067] In each round of model training, a training sample is input into the model, allowing the model to perform corresponding ECG reconstruction processing to achieve forward propagation. Then, based on the reconstruction results output by the model and combined with the annotations, the loss function is calculated, and the model parameters are optimized according to the loss function calculation results to achieve reverse propagation. In this way, when the preset model training requirements such as training time, number of training times or prediction accuracy are met, the model training can be completed and an ECG reconstruction model that can be put into practical use can be obtained.

[0068] In addition, it is understandable that after the training of the ECG reconstruction model is completed in the background and the ECG reconstruction model is put into actual use in clinical work, the model can still be iteratively updated.

[0069] For example, medical staff can manually correct the ECG reconstruction results and feed back the involved ECG inputs, ECG reconstruction results, and correction results of the ECG reconstruction results to the backend / cloud as new training samples for iterative training of the model. Alternatively, they can directly perform iterative training of the model locally. The former facilitates unified updates and iterations at the overall level, while the latter facilitates personalized updates and iterations locally.

[0070] Furthermore, this application also provides further implementation solutions on how to carry out model training more efficiently and accurately to promote more efficient and accurate model processing performance.

[0071] Specifically, as another exemplary embodiment, the model training of the electrocardiogram reconstruction model includes two stages: the first stage and the second stage. The first stage can also be referred to as pre-training. In this case, the aforementioned training of the electrocardiogram reconstruction model based on the labeled training samples specifically includes:

[0072] 1. Phase 1

[0073] 1.1 In the first stage, a 12-lead adult sample electrocardiogram is obtained, and an adult three-lead sample electrocardiogram corresponding to the three measurement leads (of the fetus) is extracted from the 12-lead adult sample electrocardiogram. The three measurement leads are specifically lead I, lead V5, and lead V6.

[0074] It is easy to see that in the embodiment herein, the present application specifically selects a ready-made, complete adult twelve-lead sample electrocardiogram corresponding to the preset lead configuration scheme involved in the present application, and extracts an adult three-lead sample electrocardiogram corresponding to the three fetal measurement leads.

[0075] It is worth noting that the three measurement leads involved in the fetal vector electrocardiogram reconstruction logic designed in this application can specifically be lead I, lead V5, and lead V6 in the twelve leads.

[0076] It can be understood that, among the twelve selectable leads, this application specifically selects lead I among the limb leads, and leads V5 and V6 among the chest leads.

[0077] Of course, it is understandable that in actual applications, other three-lead combinations can be selected from the twelve-lead system, or three-lead combinations can be selected from other lead systems (such as eighteen-lead system), as long as the measurement and calculation requirements can be met in actual situations.

[0078] In this case, it is understandable that the selection of lead I, lead V5, and lead V6 among the twelve leads is the optimal solution, which can achieve the best application effect in both acquisition cost (the lead layout points are relatively concentrated, not too dispersed or even dispersed to the side of the waist, easy to operate, and with better feasibility) and reconstruction cost (the complexity and accuracy of the calculation required).

[0079] As for the twelve leads, it specifically involves three limb leads, namely lead I, lead II and lead III, three enhanced limb leads, namely lead aVR, lead aVL and lead aVF, and six chest leads, namely lead V1, lead V2, lead V3, lead V4, lead V5 and lead V6.

[0080] It is worth noting that in the first stage, which is the pre-training stage, this application first pre-trains the model based on adult data, so that the model has a certain degree of sensitivity and pertinence for the ECG reconstruction task.

[0081] As an example, the present application may choose to obtain the required adult twelve-lead sample electrocardiogram from the open-source PTBDiagnostic ECG dataset (an open-source electrocardiogram data product) on a certain complex physiological signal source website.

[0082] 1.2 For each measurement lead, assign a learnable embedding vector and fuse the learnable embedding vector with the corresponding ECG in the adult three-lead sample ECG to obtain the adult sample ECG;

[0083] It can be understood that, corresponding to the purpose of lead identity embedding, the present application then assigns a unique learnable embedding vector to each measurement lead, which can be recorded as , and then the ECG signal of each measurement lead , and fused with its corresponding learnable embedding vector.

[0084] Among them, the fusion method or combination method can be specifically a splicing method or an addition method. Copy As , and the signal is spliced ​​into For the additive method, Mapped to , and the signal After point-by-point addition, the fusion process forms an ECG signal with lead identity embedded.

[0085] 1.3 Configure the xyz lead vector ECG corresponding to the Frank lead system for the adult sample ECG and complete the annotation;

[0086] For the labeling process, this application also uses existing data, and specifically uses the xyz lead electrocardiogram of the Frank lead system.

[0087] The vector electrocardiogram obtained by the Frank lead system has multiple leads including the X, Y, and Z directions of the human body. It simultaneously measures the electrocardiogram in the frontal, transverse, and right frontal planes of the human body to obtain the XYZ coordinate information of the electrocardiogram vector ring.

[0088] Among them, it can be understood that the pairing between the adult sample ECG and the xyz lead vector ECG of the Frank lead system can be specifically performed in the 1.3 processing link here, or the previous pairing relationship between the adult twelve-lead sample ECG itself and the xyz lead vector ECG of the Frank lead system can be used.

[0089] At the same time, the xyz lead vector electrocardiogram of the Frank lead system can also be obtained from the PTB Diagnostic ECG dataset as an example.

[0090] 1.4 Train the ECG reconstruction model based on labeled adult sample ECGs.

[0091] It can be understood that after the adult sample electrocardiograms are labeled and the configuration of the training samples is completed, the specific model training process can be promoted.

[0092] As a training example, the training scheme adopted can divide the labeled adult sample electrocardiogram into a training set, a validation set, and a test set in an 8:1:1 ratio to perform model training.

[0093] At the same time, it is easy to understand that in the process of processing training samples, data preprocessing is usually involved to enhance data quality and improve model training effects.

[0094] In this regard, in a preprocessing operation specially designed in the present application, the electrocardiogram signal can also be sliced ​​using a sliding window, where the sliding window length is configured to be 0.5 seconds to 2 seconds and includes at least one complete heart beat.

[0095] 2. Second stage

[0096] 2.1 In the second stage, a fetal electrocardiogram sample is obtained under the acquisition conditions of the preset lead configuration scheme, and an additional fetal electrocardiogram sample is obtained at the same time using an additional measurement lead deployed on the maternal abdominal wall. The additional fetal electrocardiogram sample serves as an annotation of the fetal electrocardiogram sample, and the frontal projection of at least one additional measurement lead is located at the midpoint of a triangle formed by the frontal projections of the three measurement leads.

[0097] It is easy to see that in the first stage, which is the pre-training stage, the training is centered around the adult electrocardiogram, while in the specific model training in the second stage, the model is transferred to the fetal electrocardiogram, and specific targeted training is carried out corresponding to the model application goals of this application.

[0098] In this process, it is necessary to understand that in actual situations it is impossible to collect the corresponding fetal vector electrocardiogram as a training sample. As mentioned above, the existing technology does not involve the collection and application of fetal vector electrocardiograms, and there is no ready-made fetal vector electrocardiogram that can be used. Therefore, corresponding to the requirements of the present application, if the fetal vector electrocardiogram is directly configured as a label, or the additional fetal sample electrocardiogram collected by the additional measurement lead is directly used as a label, it may make the model training difficult to converge and the generalization ability is too poor. However, in the case of the early pre-training introduced in this application, the model can be prompted to obtain a certain degree of electrocardiogram reconstruction performance, effectively prompting the second stage of model training to converge and obtain an electrocardiogram reconstruction model with better generalization ability.

[0099] In this regard, here you can configure the fetal sample electrocardiogram collected under the collection conditions of the preset lead configuration scheme of this application and the additional fetal sample electrocardiogram collected at the same time through the additional measurement lead deployed on the maternal abdominal wall. The latter is used as a annotation. When it is not a vector electrocardiogram itself, it is compared with the reconstructed fetal vector electrocardiogram output by the model. Figure 1 It serves as the input of the corresponding loss function to quantify the loss, cleverly participates in the model training process, and effectively promotes model training.

[0100] With respect to the additional measurement leads, it can be noted that the present application has further defined their specific configuration positions, i.e., the layout points. That is, in the front projection, at least one of the leads is located within the triangle formed by the three measurement leads on the maternal abdominal wall (surface), and is located at the center of the triangle (i.e., the midpoint between the layout points of the second measurement lead and the third measurement lead, and the midpoint between the layout points of the first measurement lead). If there are other additional measurement leads, they can be configured in the triangle or in the vicinity outside the triangle.

[0101] by Figure 2For example, three additional measurement leads are configured. Viewed from the front (corresponding to the projection relationship on the front), one is located at the midpoint of the triangle, one is located outside the triangle and directly above the midpoint of the triangle, and one is located diagonally below the triangle.

[0102] Among them, the additional measurement leads can be any leads, such as the leads other than the three measurement leads in the twelve leads, or other leads in other leads (such as eighteen leads), or even leads in a non-existing lead system. This is also possible as long as it can meet the requirements of model training.

[0103] 2.2 Continue to train the ECG reconstruction model based on the annotated fetal sample ECG.

[0104] In this way, after obtaining the fetal sample ECG and the additional fetal sample ECG, that is, obtaining the fetal sample ECG with annotations, we can specifically advance the second stage of model training here, make more delicate adjustments to the model parameters, and perform fine-tuning.

[0105] As an example, the model training scheme for the second phase can also adopt the model training scheme for the first phase, that is, the model training work is carried out by dividing the set into training set, validation set and test set at an 8:1:1 ratio.

[0106] Among them, corresponding to the model training situation where additional fetal sample electrocardiograms collected by additional measurement leads deployed on the maternal abdominal wall are used as annotations, and the situation where the model normally outputs a vector electrocardiogram, the model can specifically add an additional fully connected layer and a new output layer in the output layer to transform the xyz coordinates (vector electrocardiogram) normally output by the model into an electrocardiogram in any lead direction. At this time, the additional fetal sample electrocardiogram can be combined to calculate the corresponding loss function for use in optimizing the model parameters.

[0107] It can be understood that the additionally configured model structure is specifically used to assist the second-stage model training. It can be configured after the model training is completed in the first stage, or it can be configured at the beginning. After the model training is completed in the second stage, it can be directly removed and then put into actual application of the model, or its output can be ignored during the actual application of the model, or it can be configured to be in an inactive state, i.e., an idle state, during the actual application of the model. All of these are possible.

[0108] In addition, it is understandable that the specific model architecture that can be adopted by the electrocardiogram reconstruction model of this application and the specific loss function that can be adopted during the model training process are similar to the model training schemes involved above. Existing schemes can be directly adopted. Of course, in actual situations, the possibility of adopting optimized schemes improved on the basis of existing technologies or even self-developed novel schemes is not ruled out. This is also possible.

[0109] For example, for the loss quantification work involving the electrocardiogram of any lead direction obtained by transforming the xyz coordinates (vector electrocardiogram) of the normal output of the model, as well as the electrocardiogram of additional fetal samples, the corresponding loss function needs to be adaptively configured, that is, a specially configured loss function may be involved.

[0110] Among them, as a preferred solution, the model architecture adopted in this application can specifically adopt the model architecture of the UNet model. For example, it can involve 4 layers of downsampling and 4 layers of upsampling. For example, the number of downsampling and upsampling layers can be increased or decreased. Considering that the specific model architecture itself is an existing technology, no further explanation will be given here.

[0111] For the above model training, you can also combine Figure 4 A scenario diagram of the model training architecture of this application is shown for a more vivid understanding.

[0112] The above is an introduction to the fetal vector electrocardiogram reconstruction method provided by this application. In order to facilitate better implementation of the fetal vector electrocardiogram reconstruction method provided by this application, this application also provides a fetal vector electrocardiogram reconstruction device from the perspective of functional modules.

[0113] See Figure 5 , Figure 5 This is a schematic diagram of the structure of the fetal vector electrocardiogram reconstruction device of the present application. In the present application, the fetal vector electrocardiogram reconstruction device 500 may specifically include the following structure:

[0114] An acquisition unit 501 is configured to acquire an initial fetal electrocardiogram (ECG) collected from the maternal abdominal wall, wherein the initial fetal ECG is collected under acquisition conditions of a preset lead configuration scheme, wherein the preset lead configuration scheme comprises three measurement leads arranged in a triangle on the maternal abdominal wall, and a corresponding ground electrode arranged at the mid-lower back.

[0115] The reconstruction unit 502 is configured to input the initial fetal electrocardiogram into a pre-configured electrocardiogram reconstruction model, wherein the electrocardiogram reconstruction model is configured to reconstruct the fetal electrocardiogram input into a vector electrocardiogram;

[0116] The extraction unit 503 is configured to extract the target fetal vector electrocardiogram (VECG) output by the electrocardiogram reconstruction model and corresponding to the initial fetal electrocardiogram.

[0117] In an exemplary embodiment, in a preset lead configuration scheme, the arrangement point of the first measurement lead is arranged along the waistline on the right side of the maternal abdominal wall, the arrangement points of the second measurement lead and the third measurement lead are arranged on the left side of the maternal abdominal wall, and the midpoint between the arrangement points of the second measurement lead and the third measurement lead is located at the waistline;

[0118] The projections of the arrangement points of the first measurement lead, the second measurement lead, and the third measurement lead on the front face form an isosceles triangle; the length of the line connecting the projection of the arrangement points of the first measurement lead and the second measurement lead is equal to the length of the line connecting the projection of the arrangement points of the first measurement lead and the projection of the third measurement lead; the straight line formed by the projection of the arrangement points of the second measurement lead and the projection of the third measurement lead is perpendicular to the horizontal plane.

[0119] In another exemplary embodiment, the arrangement points of the first measurement lead and the arrangement points of the ground electrode constitute a first vector, the arrangement points of the second measurement lead and the arrangement points of the ground electrode constitute a second vector, and the angle between the first vector and the second vector is 50°;

[0120] The arrangement point of the third measurement lead and the arrangement point of the ground electrode form a third vector, and the angle between the second vector and the third vector is 20°.

[0121] In another exemplary embodiment, the apparatus further includes a training unit 504 configured to:

[0122] The ECG reconstruction model is trained based on the labeled training samples.

[0123] In another exemplary embodiment, the training unit 504 is specifically configured to:

[0124] In the first stage, a 12-lead adult sample electrocardiogram is obtained, and an adult three-lead sample electrocardiogram corresponding to three measurement leads is extracted from the 12-lead adult sample electrocardiogram, where the three measurement leads are specifically lead I, lead V5, and lead V6;

[0125] For each measurement lead, a learnable embedding vector is assigned and fused with the corresponding ECG in the adult three-lead sample ECG to obtain the adult sample ECG;

[0126] Configure the xyz lead vector ECG corresponding to the Frank lead system for the adult sample ECG and complete the annotation;

[0127] Training an ECG reconstruction model based on labeled adult sample ECGs;

[0128] In the second stage, a fetal electrocardiogram sample is acquired under acquisition conditions of a preset lead configuration scheme, and an additional fetal electrocardiogram sample is acquired at the same time through an additional measurement lead additionally deployed on the maternal abdominal wall, wherein the additional fetal electrocardiogram sample serves as an annotation of the fetal electrocardiogram sample, and a frontal projection of at least one additional measurement lead is located at the midpoint of a triangle formed by the frontal projections of the three measurement leads;

[0129] Continue to train the ECG reconstruction model based on the annotated fetal ECG samples.

[0130] In another exemplary embodiment, the training unit 504 is further configured to:

[0131] Sliding window slicing is performed on the annotated adult sample electrocardiogram. The sliding window length is configured to be 0.5 seconds to 2 seconds and contains at least one complete heart beat.

[0132] In another exemplary embodiment, the device further includes a display unit 505, configured to:

[0133] Display target fetal vector electrocardiogram.

[0134] This application also provides a processing device from the perspective of hardware structure, see Figure 6 , Figure 6 The schematic diagram of the structure of the processing device of the present application is shown. Specifically, the processing device of the present application may include a processor 601, a memory 602 and an input / output device 603. The processor 601 is used to execute the computer program stored in the memory 602 to implement the following Figure 1 Each step of the fetal vector electrocardiogram reconstruction method in the corresponding embodiment; or, when the processor 601 is used to execute the computer program stored in the memory 602, the following is implemented Figure 5 The memory 602 is used to store the functions of each unit in the embodiment corresponding to the processor 601. Figure 1 The computer program required for the fetal vector electrocardiogram reconstruction method in the corresponding embodiment.

[0135] For example, the computer program may be divided into one or more modules / units, one or more of which are stored in the memory 602 and executed by the processor 601 to complete the present application. One or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in a computer device.

[0136] The processing device may include, but is not limited to, a processor 601, a memory 602, and an input / output device 603. Those skilled in the art will appreciate that the illustrations are merely examples of processing devices and do not limit the processing device. The processing device may include more or fewer components than shown, or a combination of certain components, or different components. For example, the processing device may also include a network access device, a bus, etc., and the processor 601, the memory 602, the input / output device 603, etc. are connected via a bus.

[0137] The processor 601 may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the processing device and connects various parts of the entire device using various interfaces and lines.

[0138] Memory 602 can be used to store computer programs and / or modules. Processor 601 implements various functions of the computer device by running or executing computer programs and / or modules stored in memory 602 and accessing data stored in memory 602. Memory 602 may primarily include a program storage area and a data storage area. The program storage area may store an operating system, at least one application required for a function, and the data storage area may store data generated based on the use of the processing device. Furthermore, memory may include high-speed random access memory (RAM) and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0139] When the processor 601 is used to execute the computer program stored in the memory 602, it can specifically implement the following functions:

[0140] Obtaining an initial fetal electrocardiogram collected from the maternal abdominal wall, wherein the initial fetal electrocardiogram is collected under acquisition conditions of a preset lead configuration scheme, wherein the preset lead configuration scheme includes three measurement leads arranged in a triangle on the maternal abdominal wall and a corresponding ground electrode arranged at the mid-lower position;

[0141] Inputting the initial fetal electrocardiogram into a preconfigured electrocardiogram reconstruction model, wherein the electrocardiogram reconstruction model is used to reconstruct the fetal electrocardiogram input into a vector electrocardiogram;

[0142] A target fetal vector electrocardiogram (VECG) corresponding to the initial fetal electrocardiogram output by the electrocardiogram reconstruction model is extracted.

[0143] Those skilled in the art will clearly understand that for the convenience and brevity of description, the specific working process of the fetal vector electrocardiogram reconstruction device, processing equipment and corresponding units described above can refer to the following. Figure 1 The description of the fetal vector electrocardiogram reconstruction method in the corresponding embodiment will not be repeated here.

[0144] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be accomplished by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0145] To this end, the present application provides a computer-readable storage medium, which stores a plurality of instructions, which can be loaded by a processor to execute the present application as follows: Figure 1 The steps of the fetal vector electrocardiogram reconstruction method in the corresponding embodiment, the specific operations can be referred to as follows Figure 1 The description of the fetal vector electrocardiogram reconstruction method in the corresponding embodiment will not be repeated here.

[0146] The computer-readable storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0147] Due to the instructions stored in the computer readable storage medium, the present application can be executed as follows: Figure 1 The steps of the fetal vector electrocardiogram reconstruction method in the corresponding embodiment can thus be implemented as follows: Figure 1 The beneficial effects that can be achieved by the fetal vector electrocardiogram reconstruction method in the corresponding embodiment are detailed in the previous description and will not be repeated here.

[0148] The above is a detailed introduction to the fetal vector electrocardiogram reconstruction method, device, processing equipment and computer-readable storage medium provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the core idea of ​​the present application; at the same time, for technical personnel in this field, based on the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A fetal vector electrocardiogram reconstruction method, characterized in that: The method comprises: Obtaining an initial fetal electrocardiogram collected from the maternal abdominal wall, wherein the initial fetal electrocardiogram is collected under collection conditions of a preset lead configuration scheme, wherein the preset lead configuration scheme comprises three measurement leads arranged in a triangle on the maternal abdominal wall, and a corresponding ground electrode arranged at the mid-lower back; Inputting the initial fetal electrocardiogram into a preconfigured electrocardiogram reconstruction model, wherein the electrocardiogram reconstruction model is used to reconstruct the fetal electrocardiogram input into a vector electrocardiogram; extracting a target fetal vector electrocardiogram output by the electrocardiogram reconstruction model and corresponding to the initial fetal electrocardiogram; Before inputting the initial fetal electrocardiogram into a preconfigured electrocardiogram reconstruction model, the method further includes: Training the electrocardiogram reconstruction model based on labeled training samples; The model training of the electrocardiogram reconstruction model includes a first stage and a second stage, wherein the training of the electrocardiogram reconstruction model based on the labeled training samples includes: In the first stage, a 12-lead adult sample electrocardiogram is obtained, and an adult three-lead sample electrocardiogram corresponding to the three measurement leads is extracted from the 12-lead adult sample electrocardiogram, where the three measurement leads are specifically lead I, lead V5, and lead V6; Assigning a learnable embedding vector to each of the measurement leads, and fusing the learnable embedding vector with a corresponding electrocardiogram in the adult three-lead sample electrocardiogram to obtain an adult sample electrocardiogram; Configure an xyz lead vector electrocardiogram corresponding to the Frank lead system for the adult sample electrocardiogram and complete the labeling; Training the electrocardiogram reconstruction model based on the adult sample electrocardiogram configured with annotations; In the second stage, a fetal electrocardiogram sample collected under the collection conditions of the preset lead configuration scheme and an additional fetal electrocardiogram sample collected at the same time by an additional measurement lead additionally deployed on the maternal abdominal wall are obtained, wherein the additional fetal electrocardiogram sample is used as a label of the fetal electrocardiogram sample, the additional measurement lead is at least one, and the front projection of at least one of the additional measurement leads is located at the midpoint of a triangle formed by the front projections of the three measurement leads; The electrocardiogram reconstruction model is continuously trained based on the fetal sample electrocardiogram with annotations.

2. The method according to claim 1, characterized in that In the preset lead configuration scheme, the arrangement point of the first measurement lead is arranged on the right side of the maternal abdominal wall along the waistline, the arrangement point of the second measurement lead and the arrangement point of the third measurement lead are arranged on the left side of the maternal abdominal wall, and the midpoint between the arrangement point of the second measurement lead and the arrangement point of the third measurement lead is located on the waistline; The projections of the arrangement points of the first measurement lead, the second measurement lead, and the third measurement lead on the front surface form an isosceles triangle, the length of the line connecting the projection of the arrangement points of the first measurement lead and the projection of the arrangement points of the second measurement lead is equal to the length of the line connecting the projection of the arrangement points of the first measurement lead and the projection of the arrangement points of the third measurement lead, and the straight line formed by the projection of the arrangement points of the second measurement lead and the projection of the third measurement lead is perpendicular to the horizontal plane.

3. The method according to claim 2, characterized in that The arrangement points of the first measurement lead and the arrangement points of the ground electrode constitute a first vector, the arrangement points of the second measurement lead and the arrangement points of the ground electrode constitute a second vector, and the angle between the first vector and the second vector is 50°; The arrangement point of the third measurement lead and the arrangement point of the ground electrode form a third vector, and the angle between the second vector and the third vector is 20°.

4. The method according to claim 1, wherein Before training the electrocardiogram reconstruction model based on the adult sample electrocardiogram with annotations, the method further includes: The adult sample electrocardiogram with annotations is sliced ​​using a sliding window, where the sliding window length is configured to be 0.5 seconds to 2 seconds and includes at least one complete heart beat.

5. The method according to claim 1, wherein After extracting the target fetal vector electrocardiogram output by the electrocardiogram reconstruction model and corresponding to the initial fetal electrocardiogram, the method further includes: The target fetus vector electrocardiogram is displayed.

6. A fetal vector electrocardiogram reconstruction device, characterized in that: The device comprises: an acquisition unit, configured to acquire an initial fetal electrocardiogram acquired from the maternal abdominal wall, wherein the initial fetal electrocardiogram is acquired under acquisition conditions of a preset lead configuration scheme, wherein the preset lead configuration scheme comprises three measurement leads arranged in a triangle on the maternal abdominal wall, and a corresponding ground electrode arranged at the middle of the lower back; a reconstruction unit, configured to input the initial fetal electrocardiogram into a preconfigured electrocardiogram reconstruction model, wherein the electrocardiogram reconstruction model is configured to reconstruct the fetal electrocardiogram input into a vector electrocardiogram; an extraction unit, configured to extract a target fetal vector electrocardiogram output by the electrocardiogram reconstruction model and corresponding to the initial fetal electrocardiogram; The apparatus further comprises a training unit, configured to: Training the electrocardiogram reconstruction model based on labeled training samples; The model training of the electrocardiogram reconstruction model includes a first stage and a second stage, and the training unit is specifically used to: In the first stage, a 12-lead adult sample electrocardiogram is obtained, and an adult three-lead sample electrocardiogram corresponding to the three measurement leads is extracted from the 12-lead adult sample electrocardiogram, where the three measurement leads are specifically lead I, lead V5, and lead V6; Assigning a learnable embedding vector to each of the measurement leads, and fusing the learnable embedding vector with a corresponding electrocardiogram in the adult three-lead sample electrocardiogram to obtain an adult sample electrocardiogram; Configure an xyz lead vector electrocardiogram corresponding to the Frank lead system for the adult sample electrocardiogram and complete the labeling; Training the electrocardiogram reconstruction model based on the adult sample electrocardiogram configured with annotations; In the second stage, a fetal electrocardiogram sample collected under the collection conditions of the preset lead configuration scheme and an additional fetal electrocardiogram sample collected at the same time by an additional measurement lead additionally deployed on the maternal abdominal wall are obtained, wherein the additional fetal electrocardiogram sample is used as a label of the fetal electrocardiogram sample, the additional measurement lead is at least one, and the front projection of at least one of the additional measurement leads is located at the midpoint of a triangle formed by the front projections of the three measurement leads; The electrocardiogram reconstruction model is continuously trained based on the fetal sample electrocardiogram with annotations.

7. A processing device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and the processor executes the method according to any one of claims 1 to 5 when calling the computer program in the memory.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the method according to any one of claims 1 to 5.

Images