A method and device for locating myocardial infarction using electrocardiography based on geometric iterative networks

Abstract

This application provides a method and apparatus for locating myocardial infarction using electrocardiogram (ECG) based on a geometric iterative network. The method establishes a correspondence between ECG data and key nodes of myocardial infarction using an artificial intelligence model. This correspondence includes a first sub-correspondence, a second sub-correspondence, and a third sub-correspondence. Specifically, the first artificial intelligence sub-model establishes a first sub-correspondence between ECG data and ventricular surface potential distribution; the second artificial intelligence sub-model establishes a second sub-correspondence between ventricular surface potential distribution and key regions of myocardial infarction; and the third artificial intelligence sub-model establishes a third sub-correspondence between key regions of myocardial infarction and key nodes of myocardial infarction. Target ECG data of a target patient is acquired, and the target key nodes of myocardial infarction corresponding to the target ECG data are determined using the correspondence. The geometric iterative network can accurately identify the most critical nodes in the myocardial infarction region.

Description

Technical Field

[0001] This application relates to the field of medical testing, and in particular to a method and device for locating myocardial infarction using electrocardiography based on geometric iterative networks. Background Technology

[0002] Myocardial infarction, also known as acute myocardial infarction or coronary artery obstruction, is a potentially life-threatening acute condition caused by acute blockage of the coronary arteries, leading to necrosis of the heart muscle due to lack of blood supply and impaired heart function.

[0003] Localizing myocardial infarction is essential for informed clinical management and treatment intervention planning. It is crucial in guiding patients in selecting revascularization procedures, helping to identify coronary artery blockages and provide an assessment of their severity. Furthermore, localizing myocardial infarction helps estimate the risk of related complications, such as arrhythmias. Infarction creates a predisposition to arrhythmias by altering the electrophysiological state of the myocardium. Therefore, localization of myocardial infarction contributes to the development of arrhythmia ablation strategies.

[0004] Determining the location of myocardial infarction is crucial for clinical management and treatment strategies. However, existing diagnostic tools for myocardial infarction either sacrifice practicality or are limited by spatial resolution. Summary of the Invention

[0005] In view of the aforementioned problems, this application is proposed to provide a method and apparatus for electrocardiographic localization of myocardial infarction based on geometric iterative networks to overcome or at least partially solve the aforementioned problems, comprising:

[0006] A method for locating myocardial infarction using electrocardiography based on a geometric iterative network, the method involving an artificial intelligence model, the artificial intelligence model including a first artificial intelligence sub-model, a second artificial intelligence sub-model, and a third artificial intelligence sub-model; the method includes:

[0007] An artificial intelligence model is used to establish a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction. This correspondence includes a first sub-correspondence, a second sub-correspondence, and a third sub-correspondence. Specifically, the first AI sub-model establishes a first correspondence between ECG data and ventricular surface potential distribution; the second AI sub-model establishes a second correspondence between ventricular surface potential distribution and key regions of myocardial infarction; and the third AI sub-model establishes a third correspondence between key regions of myocardial infarction and key nodes of myocardial infarction.

[0008] The target electrocardiogram (ECG) data of the target patient is obtained, and the target key nodes of myocardial infarction corresponding to the target ECG data are determined through the pre-established correspondence relationship of the artificial intelligence model.

[0009] Further, the step of establishing the first correspondence between the electrocardiogram data and the ventricular surface potential distribution through the first artificial intelligence sub-model includes:

[0010] Discrete multi-wavelet operation was used to determine the potential features of the electrocardiogram data in the time and frequency domains;

[0011] Based on the preset sandwich structure operator, feature compensation is performed on the latent features to determine the graphic information whose data dimension is consistent with the ventricular surface potential distribution.

[0012] Geometric constraint embedding is performed based on the graphical information of nodes and edges, and the first correspondence between the electrocardiogram data and the ventricular surface potential distribution is established by utilizing the self-learning capability of the first artificial intelligence sub-model.

[0013] Furthermore, the step of determining the potential features of the electrocardiogram data in the time and frequency domains using discrete multi-wavelet operations includes:

[0014] The electrocardiogram data was decomposed using orthogonal multiwavelets to obtain decomposed electrocardiogram data.

[0015] Simplified ECG information corresponding to the decomposed ECG data is determined by using downsampling and zero-filling operations.

[0016] Discrete multi-wavelet operation is used to determine the potential features of the simplified electrocardiogram information in the time and frequency domains.

[0017] Further, the step of embedding geometric constraints based on the graph information of nodes and edges, and establishing the first correspondence between the electrocardiogram data and the ventricular surface potential distribution using the self-learning capability of the first artificial intelligence sub-model, includes:

[0018]

[0019] u k =P l (y k -γf(y k (2)

[0020] y k+1 =u k +τ k+1 (u k -u k-1 (3)

[0021]

[0022] y k+1 =u k +τk (u k -u k-1 (5)

[0023] In the formula, T k+1 The iteration step size is used to control the magnitude of variable updates during the iteration process; u k y is an intermediate variable in the iteration process. k γ is the target variable in the iteration process; γ is the regularization parameter used to control the size of the gradient descent step size. The transpose of the adaptive gradient operator during iteration is used to adjust u. k Update direction; Hy k Initialize application terms H to y k The result; where in step k, H is the initialization term; P represents the proximal operator supported by the network with embedded constraints; G is the adaptive gradient operator during iteration; τ k >0 can be learned through training as defined in equation (1); ρ is the contraction operator ρ(u k )=(|u k |-ρ)+sgn(u k ), ()+ is |u k |-ρ is the maximum value between 0 and 0, where sgn represents the sign function, limited to ±1.

[0024] Further, the step of establishing the second correspondence between the ventricular surface potential distribution and the key region of myocardial infarction through the second artificial intelligence sub-model includes:

[0025] Obtain a medical imaging dataset; wherein the medical imaging dataset includes a normal myocardial image dataset, a myocardial infarction area image dataset, and a pathological change image dataset related to myocardial infarction;

[0026] The second artificial intelligence sub-model is trained based on the medical image dataset to determine the second correspondence between the ventricular surface potential distribution and the key region of myocardial infarction.

[0027] Further, the second artificial intelligence sub-model includes a physical sub-model and a neural network. The step of training the second artificial intelligence sub-model based on the medical image dataset to determine the second correspondence between the ventricular surface potential distribution and the key region of myocardial infarction includes:

[0028] The expression for the electric potential is approximated based on the neural network; the formula is as follows:

[0029] u(x,y,z,t)≈N(x,y,z,t;θ);

[0030] In the formula, θ is the parameter of the neural network; x, y, z are the spatial coordinates; and t represents time.

[0031] The expression for the electric potential is substituted into the physical sub-model used to describe the electrophysiological process of the heart, and the second artificial intelligence sub-model is trained. Based on the loss function, the neural network is made to satisfy the physical laws; the formula is as follows:

[0032]

[0033] In the formula, α is the weighting parameter that balances the two loss terms; p i At position (x) i ,y i ,t i ) The observed probability of myocardial infarction; N data It is the total number of observed data points; It is the actual potential value at position j; It is the potential value predicted by the neural network;

[0034] The weights W and bias b are updated using the backpropagation algorithm; the formula is as follows:

[0035]

[0036] The update rules are as follows:

[0037]

[0038] In the formula, L is the loss function, used to evaluate the difference between the model's prediction and the actual observation; y is the output of the neural network, representing the predicted potential value; and η is the learning rate.

[0039] After each training iteration, if the difference between the predicted value and the true value is still greater than the preset minimum tolerance difference, the neural network is returned to continue training in a loop until the difference between the predicted value and the true value is no greater than the preset minimum tolerance difference.

[0040] Further, the step of establishing the third correspondence between the key regions and key nodes of myocardial infarction through the third artificial intelligence sub-model includes:

[0041] The node features in the critical region of myocardial infarction are input into a preset cross-entropy loss function, and the output parameters of the cross-entropy loss function are determined as the corresponding critical nodes of myocardial infarction; wherein, the cross-entropy loss function is as follows:

[0042]

[0043] In the formula, N is the number of nodes; y i It is the actual label of node i; It is the probability of infarction predicted by the model.

[0044] A device for locating myocardial infarction using electrocardiography based on a geometric iterative network, the device involving an artificial intelligence model, the artificial intelligence model including a first artificial intelligence sub-model, a second artificial intelligence sub-model, and a third artificial intelligence sub-model; comprising:

[0045] A correspondence establishment module is used to establish a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction using an artificial intelligence model. The correspondence includes a first sub-correspondence, a second sub-correspondence, and a third sub-correspondence. Specifically, the first AI sub-model establishes a first correspondence between ECG data and ventricular surface potential distribution; the second AI sub-model establishes a second correspondence between ventricular surface potential distribution and key regions of myocardial infarction; and the third AI sub-model establishes a third correspondence between key regions of myocardial infarction and key nodes of myocardial infarction.

[0046] The data processing module is used to acquire the target electrocardiogram data of the target patient and determine the target myocardial infarction key node corresponding to the target electrocardiogram data through the correspondence pre-established by the artificial intelligence model.

[0047] An apparatus includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the electrocardiogram-based method for locating myocardial infarction based on a geometric iterative network as described above.

[0048] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the electrocardiogram-based method for locating myocardial infarction based on a geometric iterative network as described above.

[0049] This application has the following advantages:

[0050] In the embodiments of this application, in response to the technical problem in the prior art that "existing myocardial infarction diagnostic tools either sacrifice practicality or are limited by spatial resolution," this application provides a solution for accurately locating myocardial infarction by reconstructing ventricular surface potentials from electrocardiogram (ECG) data using an artificial intelligence (AI) model. Specifically, this involves establishing a correspondence between ECG data and key nodes of myocardial infarction using an AI model; wherein the correspondence includes a first sub-correspondence, a second sub-correspondence, and a third sub-correspondence; specifically, a first sub-correspondence between ECG data and ventricular surface potential distribution is established using the first AI sub-model; a second sub-correspondence between ventricular surface potential distribution and key regions of myocardial infarction is established using the second AI sub-model; a third sub-correspondence between key regions of myocardial infarction and key nodes of myocardial infarction is established using the third AI sub-model; target ECG data of the target patient is obtained, and the target key nodes of myocardial infarction corresponding to the target ECG data are determined using the correspondence. The first AI sub-model reconstructs the ventricular surface potential distribution of electrocardiogram data, demonstrating excellent accuracy on both synthetic and clinical datasets. The second AI sub-model enables the network to not only learn patterns in the data but also follow physical laws to identify key areas of myocardial infarction. The third AI sub-model processes graph-structured data and learns the relationships between nodes in the graph, accurately identifying the most critical nodes in the myocardial infarction area. Attached Figure Description

[0051] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description of this application will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart illustrating the steps of an electrocardiogram-based method for locating myocardial infarction using a geometric iterative network, as provided in one embodiment of this application.

[0053] Figure 2 This is a schematic diagram of the structure of a geometric iterative network provided in an embodiment of this application;

[0054] Figure 3 This is a structural block diagram of an electrocardiogram-based myocardial infarction localization device based on a geometric iterative network according to an embodiment of this application;

[0055] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0056] To make the objectives, features, and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0057] The inventors, through analysis of existing technologies, discovered that late-stage gadolinium-enhanced cardiac MRI is a standard for locating myocardial infarction in clinical practice. However, this technique has some limitations. Late-stage gadolinium-enhanced cardiac MRI requires patients to have sufficient breath-holding ability. It also requires gadolinium-based contrast agents, which carry risks such as renal systemic sclerosis and long-term brain retention. Furthermore, large quantities of gadolinium are quite expensive for both patients and the healthcare system due to the high operating costs associated with the equipment.

[0058] Body surface potential mapping has become an advanced tool for localizing myocardial infarction. As an extension of 12-lead electrocardiography (ECG), body surface potential mapping utilizes additional electrodes to capture a comprehensive ECG profile. It can be combined with cardiac anatomy models to visualize and analyze cardiac electrical activity through inverse kinematics. With this inverse kinematics, cardiac electrical activity can be precisely visualized and analyzed, thus achieving a fusion of the heart and data. Recently, various deep learning and machine learning techniques have been used to solve the inverse problem in ECG, aiming to obtain robust inverse kinematics. However, due to its additional operational complexity and cost, body surface potential mapping has not yet been widely adopted in clinical practice. This hinders the rapid clinical promotion and adoption of inverse techniques. Therefore, it is important to find a method for localizing myocardial infarction using 12-lead ECG.

[0059] By reconstructing the transmembrane potential distribution using a 12-lead electrocardiogram and identifying regions of abnormal potential distribution, myocardial infarction can be localized. This reconstruction of ventricular surface potentials will significantly enhance its clinical value. This method is expected to be more widely adopted and disseminated in clinical settings.

[0060] However, 12-lead electrocardiograms (ECGs) have limitations in localization resolution, only able to distinguish relatively wide regions. These regions include the inferior, lateral, and posterior parts, or combinations thereof, such as inferior lateral and inferior posterior infarctions. This limitation relates to the indirect representation of electrical activity. A 12-lead ECG provides only a narrowed view of electrical activity unfolding in space and time. Therefore, it obscures crucial spatial information essential for localization. Reconstructing the ventricular surface potential distribution from a 12-lead ECG exhibits substantial ill-defined characteristics, meaning that the same ECG data may correspond to different potential distributions. This can lead to physiologically illogical or even erroneous distribution results. This problem stems from the complex propagation of cardiac potentials in tissues such as the lungs and skin. The varying conductivities in these tissues cause attenuation and distortion. When these potentials are captured by surface electrodes, the ECG signal contains mixed noise. This noise is amplified during reconstruction. This affects feature mapping and complicates the accurate capture of electrical activity. Furthermore, a 12-lead ECG provides only 12 surface electrical recordings, limiting its data dimensionality.

[0061] To overcome the aforementioned problems, the inventors introduced a frequency-enhanced geometrically constrained iterative network. This network first mines latent features in the time and frequency domains from ECG data. Subsequently, the method increases the dimensionality of the ECG data and utilizes convolutional layers to capture complex features. Finally, the geometrically constrained iterative network uses the ventricular geometry as a constraint on the surface potential distribution. It assigns variable weights to different edges.

[0062] Reference Figure 1 This paper illustrates an embodiment of an electrocardiogram method for locating myocardial infarction based on a geometric iterative network.

[0063] The method includes:

[0064] S1. Establish a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction using an artificial intelligence (AI) model; wherein, the AI ​​model includes a first AI sub-model, a second AI sub-model, and a third AI sub-model; specifically, the first AI sub-model establishes a first correspondence between ECG data and ventricular surface potential distribution; the second AI sub-model establishes a second correspondence between ventricular surface potential distribution and key regions of myocardial infarction; and the third AI sub-model establishes a third correspondence between key regions of myocardial infarction and key nodes of myocardial infarction.

[0065] S2. Obtain the target electrocardiogram data of the target patient, and determine the key nodes of myocardial infarction corresponding to the target electrocardiogram data through the correspondence pre-established by the artificial intelligence model.

[0066] In the embodiments of this application, in response to the technical problem in the prior art that "existing myocardial infarction diagnostic tools either sacrifice practicality or are limited by spatial resolution," this application provides a solution for accurately locating myocardial infarction by reconstructing ventricular surface potentials from electrocardiogram (ECG) data using an artificial intelligence (AI) model. Specifically, this involves establishing a correspondence between ECG data and key nodes of myocardial infarction using an AI model; wherein the correspondence includes a first sub-correspondence, a second sub-correspondence, and a third sub-correspondence; specifically, a first sub-correspondence between ECG data and ventricular surface potential distribution is established using the first AI sub-model; a second sub-correspondence between ventricular surface potential distribution and key regions of myocardial infarction is established using the second AI sub-model; a third sub-correspondence between key regions of myocardial infarction and key nodes of myocardial infarction is established using the third AI sub-model; target ECG data of the target patient is obtained, and the target key nodes of myocardial infarction corresponding to the target ECG data are determined using the correspondence. The first AI sub-model reconstructs the ventricular surface potential distribution of electrocardiogram data, demonstrating excellent accuracy on both synthetic and clinical datasets. The second AI sub-model enables the network to not only learn patterns in the data but also follow physical laws to identify key areas of myocardial infarction. The third AI sub-model processes graph-structured data and learns the relationships between nodes in the graph, accurately identifying the most critical nodes in the myocardial infarction area.

[0067] The following will further describe an electrocardiogram method for locating myocardial infarction based on a geometric iterative network in this exemplary embodiment.

[0068] As described in step S1, an artificial intelligence model is used to establish a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction. Specifically, a first correspondence is established between ECG data and ventricular surface potential distribution using the first artificial intelligence sub-model; a second correspondence is established between ventricular surface potential distribution and key regions of myocardial infarction using the second artificial intelligence sub-model; and a third correspondence is established between key regions of myocardial infarction and key nodes of myocardial infarction using the third artificial intelligence sub-model.

[0069] It should be noted that, referring to Figure 2 The artificial intelligence model is a geometrically constrained iterative network, and the artificial intelligence model includes a first artificial intelligence sub-model, a second artificial intelligence sub-model, and a third artificial intelligence sub-model.

[0070] The first AI sub-model is a frequency-enhanced geometrically constrained iterative network, comprising a basic framework, a frequency-enhancing feature module, and a geometrically constrained embedding module. By introducing a frequency-enhanced geometrically constrained iterative network to reconstruct the ventricular surface potential distribution of a standard 12-lead electrocardiogram, it demonstrates excellent accuracy on both synthetic and clinical datasets.

[0071] The second artificial intelligence sub-model is a physical information neural network, which combines deep learning with physical laws. It utilizes the powerful nonlinear mapping capabilities of neural networks to learn complex functional relationships through massive amounts of data. The physical information neural network combines physical laws with the powerful function approximation capabilities of neural networks in a deep learning model. By embedding physical equations into the neural network, it enables the network not only to learn patterns in the data but also to follow physical laws to identify key areas of myocardial infarction.

[0072] The third AI sub-model is a graph attention mechanism, used to enhance inter-node interaction and information transmission, iteratively aggregating information from neighboring nodes to learn nodes. The graph attention mechanism processes graph-structured data, learns the relationships between nodes in the graph, and assigns different attention weights to each node based on these relationships to identify the most critical nodes in the myocardial infarction region.

[0073] In one embodiment of the present invention, the specific process of "establishing the first correspondence between the electrocardiogram data and the ventricular surface potential distribution through the first artificial intelligence sub-model" in step S1 can be further described in conjunction with the following description.

[0074] S111. Discrete multi-wavelet operation is used to determine the potential features of the electrocardiogram data in the time domain and frequency domain;

[0075] It should be noted that the electrocardiogram data refers to the electrocardiogram sequence obtained from the patient's 12-lead electrocardiogram, which contains information on the electrical activity of the heart at different time points.

[0076] It should be noted that the time domain describes the change of a signal over time, captures the instantaneous characteristics, and focuses on the instantaneous behavior and timing relationship of the signal; the frequency domain reveals the frequency distribution and spectral characteristics of the signal, and understands the contributions of different frequencies. Through time-domain analysis, the dynamic behavior of the electrocardiogram (ECG) signal on the time axis can be intuitively observed, including key information such as the instantaneous characteristics, waveform shape, amplitude, and phase of the signal. In the analysis of electrocardiogram (ECG) signals, time-domain analysis can help doctors identify abnormal rhythms and waveform changes of the heart by analyzing the frequency components of the electrocardiogram signal and understanding the activity characteristics of the heart in different states. Through frequency-domain analysis, the frequency components of the signal are revealed. By representing the signal as a superposition of frequency components, frequency-domain analysis can describe the frequency characteristics of the signal, including the different frequency components contained in the signal and their relative contributions.

[0077] It should be noted that the geometric constraint iterative network with frequency enhancement (the first artificial intelligence sub-model) consists of a geometric constraint iterative network framework, a frequency enhancement feature attention mechanism, and corresponding geometric constraint embeddings. Among them, the frequency enhancement feature attention mechanism includes a frequency operation module, discrete multi-wavelet operation, a long-term memory self-attention mechanism, and feature compensation. In this step, the potential features of the electrocardiogram data in the time domain and the frequency domain are calculated based on the frequency operation module, discrete multi-wavelet operation, and long-term memory self-attention mechanism.

[0078] In an embodiment of the present invention, the specific process of "determining the potential features of the electrocardiogram data in the time domain and the frequency domain by using discrete multi-wavelet operation" can be further described in combination with the following description.

[0079] As described in the following steps, the electrocardiogram data is decomposed by using orthogonal multi-wavelets to obtain decomposed electrocardiogram data; downsampling and zero-padding operations are used to determine the simplified electrocardiogram information corresponding to the decomposed electrocardiogram data; discrete multi-wavelet operation is used to determine the potential features of the simplified electrocardiogram information in the time domain and the frequency domain.

[0080] [[ID=​​​​​​​​​​​​​​​​​​Represents the original frequency domain sequence; C represents the complex domain; L represents the number of original frequency components; T represents time; M represents the number of frequency components after downsampling; t i ∈[1,T] represents the input channel and lead number; t o ∈[1,T] represents the output channel; m represents the index of the frequency component; to restore the data dimension reduced by the S operation, a zero-padding operation is then applied to obtain...

[0083]

[0084] In the formula, This represents the frequency domain sequence after the downsampling operation; Padding represents the zero-padding operation.

[0085] It should be noted that discrete multiwavelet operations include two key operations: decomposition and representation. Given an input sequence, orthogonal multiwavelets are used for decomposition due to their robust performance in denoising and signal recovery applications. The multiwavelet transform coefficients at scale j and position l are defined as follows:

[0086]

[0087] In the formula, represents the approximation coefficients at scale j, used to represent the low-frequency components of the signal; f represents the original input signal; k represents the number of approximation coefficients at scale j; i represents the index of the approximation coefficients; and μ is used as the measure. j and Characterization, Indicates the inner product; Describe a wavelet orthogonal basis for piecewise polynomials; This represents the detail function at scale j, used to capture detailed information about the signal. Cross-scale decomposition is defined as:

[0088]

[0089] In the formula, the linear coefficients (H(0), H(1), G(0), G(1)) are used as a multi-wavelet decomposition filter using Legendre polynomials, determined by Gaussian orthogonality and Gram-Schmidt orthogonality. Based on empirical evidence and the length of the ECG data, the decomposition level is set to 4. The filter length is set to 480. Zero-padding is chosen as the boundary treatment method. Each ECG signal is clipped to 480 time steps (removing some redundant parts after the T wave). The representation in the multi-wavelet frequency domain is defined as... in Representing multi-scale, multi-wavelet coefficients, with H and G as low-pass and high-pass filters respectively, (A j B j Cj ) represents the independent frequency operation module that processes different sequences during decomposition, and r is the coarsest scale during recursive decomposition. This is a fully connected layer for processing the decomposed sequence. The settings for multiple wavelet bases, reconstruction levels, and filter coefficients are consistent with those used during the decomposition process.

[0090] By analyzing the distribution of different frequency components in an electrocardiogram (ECG) signal, it is possible to determine whether there are abnormal rhythms in the heart, such as arrhythmias. Decomposing the ECG signal into frequency components of different scales helps to filter out noise and interference, enhance frequency domain characteristics, and thus improve signal quality, providing more accurate data support for subsequent ECG signal analysis and processing.

[0091] It should be noted that the long memory self-attention mechanism: in analyzing electrocardiogram data y∈R L×T At that time, we used an independent fully connected layer with an adaptive weight matrix w Q ,w K ,w V ∈R L×L Modify y to get y Q =w Q ·y,y K =w K ·y,y V =w V ·y. Then, for y Q ,y K ,y V Discrete multi-wavelet operations are performed to enhance frequency domain features, ultimately yielding... Finally, the self-attention mechanism is applied to the frequency domain, with the activation function being... Where d K The dimension is L. This self-attention mechanism ultimately produces in the time domain. Preserve basic features ( It is also y during iteration step k in equation (2) k The benefits of applying self-attention mechanisms are as follows: First, self-attention enhances ECG analysis by adaptively focusing on diagnostically critical regions associated with the infarct site. Therefore, it can improve the efficiency of providing useful features for subsequent data. Second, it can also capture long-term dependencies throughout the ECG data. This is highly beneficial because information about the location of myocardial infarction in the ECG data center typically spans multiple time steps.

[0092] S112. Based on the preset sandwich structure operator, feature compensation is performed on the latent features to determine the graphic information whose data dimension is consistent with the ventricular surface potential distribution.

[0093] It should be noted that feature compensation in the frequency-enhanced geometrically constrained iterative network is used to output data dimensionality consistent with the ventricular surface potential distribution. Feature compensation involves creating a sandwich-structured operator Γ, consisting of a two-layer convolution (layer C and layer D) containing an active ReLU layer: Γ(·) = D(ReLU(C(·))). To encapsulate the complex forward relations, an additional convolutional layer E is added before the double Γ operation. To maintain a closed process, we introduce the left inverse of Φ, denoted as Φinverse. -1 , satisfying Φ -1 ·Φ=I (where I represents the unit operator), resulting in

[0094] This block implements a proximal mapping from features extracted from a 12-lead electrocardiogram (input from a long-memory self-attention block) to the output, with the output data dimension consistent with the ventricular surface potential distribution (input geometric constraint embedding mechanism).

[0095] S113. Based on the graphical information of nodes and edges, perform geometric constraint embedding, and utilize the self-learning capability of the first artificial intelligence sub-model to establish the first correspondence between the electrocardiogram data and the ventricular surface potential distribution.

[0096] It should be noted that a first artificial intelligence model is established, and geometric constraints are set for the first artificial intelligence sub-model. The geometric constraint embedding module uses the ventricular geometry as a constraint on the surface potential distribution. By assigning variable weights to different edges, it simulates the propagation pattern of cardiac electrical activity, thereby outputting a reconstructed ventricular surface potential distribution with high accuracy and robustness, overcoming the ill-posedness of reconstructing the ventricular surface potential distribution from a 12-lead electrocardiogram.

[0097] It should be noted that the geometrically constrained iterative network framework integrates the neural network into equations (2) and (3), and... The estimation is transformed into the following equations (4) and (5):

[0098]

[0099] u k =P l (y k -γf(y k (2)

[0100] y k+1 =u k +τ k+1 (u k -u k-1 (3)

[0101]

[0102] y k+1 =u k +τ k (u k -u k-1 (5)

[0103] In the formula, T k+1 The iteration step size is used to control the magnitude of variable updates during the iteration process; u k y is an intermediate variable in the iteration process. k γ is the target variable in the iteration process; γ is the regularization parameter used to control the size of the gradient descent step size. The transpose of the adaptive gradient operator during iteration is used to adjust u. k Update direction; Hy k Initialize application terms H to y k The result; where in step k, H is the initialization term; P represents the proximal operator supported by the network with embedded constraints; G is the adaptive gradient operator during iteration; τ k >0 can be learned through training as defined in equation (1); ρ is the contraction operator ρ(u k )=(|u k |-ρ)+sgn(u k ), ()+ is |u k |-ρ is the maximum value between 0 and 0, where sgn represents the sign function, limited to ±1.

[0104] Implementation details in P: The network implementation in P is twofold: a proximal part and a constraint part. The proximal part mainly involves frequency-enhanced feature attention mechanisms. The constraint part involves geometric constraint embedding.

[0105] It is important to note the geometric constraint embedding: Analyzing the spatial relationships between nodes is crucial for correctly implementing the regularizer in non-Euclidean space. These relationships are intrinsically linked to the ventricular anatomy. Given that the focus is solely on the ventricular surface, we restrict our geometric representation to use triangular surface elements. This representation comprises node indices υ = [N, 3] and an adjacency matrix A = [N, N], defined by a ventricular triangular mesh. Therefore, the undirected ventricular graph is described as G = (υ, A). The edge index ε = [2, num_edges] is formed by merging node indices. Edge weights are obtained from the adjacency matrix using the corresponding indices.

[0106] In summary, this study addresses the challenges of complex ECG signal propagation and noise interference by mining latent features in the time and frequency domains, increasing data dimensionality, utilizing convolutional layers to capture complex features, and employing geometrically constrained embedding. Operations such as downsampling, feature extraction kernels, and zero-padding enhance the frequency domain features of ECG signals, providing high-quality data for subsequent processing. Decomposition and representation operations break down ECG signals into frequency components of different scales, filtering out noise and interference and enhancing frequency domain features. A self-attention mechanism improves the efficiency of ECG analysis and increases the accuracy of myocardial infarction localization. Operations such as convolutional layers and active ReLU layers encapsulate complex forward relationships, maintaining a closed processing flow. Steps such as initialization terms, proximal operators, and adaptive gradient operators progressively approximate the true ventricular surface potential distribution. Geometrically constrained embedding simulates the propagation pattern of cardiac electrical activity, ensuring that the potential around the infarct boundary is consistent with the physiological local structural features.

[0107] In one embodiment of the present invention, the specific process of step S1, "training the second artificial intelligence sub-model based on the medical image dataset to determine the second correspondence between the ventricular surface potential distribution and the key region of myocardial infarction," can be further explained in conjunction with the following description.

[0108] It should be noted that the Physical Information Neural Network (the second artificial intelligence sub-model) is a machine learning method that combines a physical model and a deep learning network. In locating myocardial infarction, this embodiment constructs a neural network that not only learns patterns in the data but also must satisfy the physical laws of the cardiac electrophysiological model. The Physical Information Neural Network consists of five parts: data collection, physical model, neural network, loss function, and iterative training. The five parts of the Physical Information Neural Network will be described in detail below.

[0109] S121. Obtain a medical image dataset; wherein, the medical image dataset includes a normal myocardial image dataset, a myocardial infarction area image dataset, and a pathological change image dataset related to myocardial infarction.

[0110] It's important to note the data collection process: First, a medical imaging dataset containing myocardial infarction images was collected, including normal myocardium, infarcted areas, and pathological changes associated with myocardial infarction. This data can borrow some of the data used in frequency-enhanced geometrically constrained iterative networks, provided they meet certain criteria. To ensure effective training, a sufficient number of CT, MRI, or echocardiographic images of normal myocardium were obtained from hospital medical imaging databases. The dataset was ensured to cover individuals of different ages, sexes, body types, and health conditions to increase the model's generalization ability. Next, the normal myocardial areas were accurately labeled for the supervised learning process during model training.

[0111] S122. The second artificial intelligence sub-model is trained based on the medical image dataset to determine the second correspondence between the ventricular surface potential distribution and the key region of myocardial infarction.

[0112] It should be noted that physical models are used to describe cardiac electrophysiological processes. A common cardiac electrophysiological model is the two-domain model, in which cardiomyocytes are divided into intracellular and extracellular regions. Physical models include:

[0113] ① Intracellular potential equation

[0114]

[0115] In the formula, It is electric potential u i Rate of change over time; · represents the divergence operator; σ i It is the intracellular conductivity; ρ i It is the resistivity within the cell; I i It is the ionic current within the cell.

[0116] ② Extracellular potential equation

[0117]

[0118] In the formula, It is electric potential u e Rate of change over time; · represents the divergence operator; σ e It is the extracellular conductivity; ρ e It is the extracellular resistivity; I e It is an extracellular ion current.

[0119] ③ Ion current

[0120] Ion current I i and I e It is generated by the activity of ion channels, and can usually be represented as:

[0121]

[0122] In the formula, g oj and g ej It is the current density of the ion channel; and It is the equilibrium potential of the ion channel.

[0123] It's important to note that in the physical model, the key function we use is the electric potential *u*. Using the previously collected and labeled data, we can use a neural network to approximate the expression for *u*. Let's define a neural network *N*, whose input is (x, y, z, t) and whose output is the electric potential *u*.

[0124] u(x,y,z,t)≈N(x,y,z,t;θ)

[0125] In the formula, θ is the parameter of the neural network; x, y, z are the spatial coordinates; and t represents time.

[0126] In this way, the known approximate expression of u can be substituted back into the formula of the physical model to train the neural network.

[0127] It should be noted that the loss function ensures that the neural network satisfies the physical laws.

[0128]

[0129] In the formula, α is the weighting parameter that balances the two loss terms; p i At position (x) i ,y i ,t i ) The observed probability of myocardial infarction; N data It is the total number of observed data points; It is the actual potential value at position j; It is the potential value predicted by the neural network.

[0130] It should be noted that during iterative training, we use the backpropagation algorithm to update the weights W and biases b.

[0131]

[0132] The update rules are as follows:

[0133]

[0134] In the formula, L is the loss function, used to evaluate the difference between the model's prediction and the actual observation; y is the output of the neural network, representing the predicted potential value; and η is the learning rate.

[0135] A minimum tolerance difference between the predicted value and the true value is preset. After each training session, if the difference is still greater than the preset minimum tolerance difference, the neural network is returned to continue training in a loop. After a certain number of loops, when the difference is no greater than the preset minimum tolerance difference, the training of the fitting function is complete.

[0136] In summary, by leveraging the powerful function approximation capabilities of neural networks, learning patterns in data, and following physical laws, key areas of myocardial infarction can be identified.

[0137] In one embodiment of the present invention, the specific process of "establishing the third correspondence between the key region of myocardial infarction and the key node of myocardial infarction through the third artificial intelligence sub-model" in step S1 can be further described in conjunction with the following description.

[0138] S131. Input the node features in the key region of myocardial infarction into a preset cross-entropy loss function, and determine the output parameters of the cross-entropy loss function as the corresponding key nodes of myocardial infarction; wherein, the cross-entropy loss function is as follows:

[0139]

[0140] In the formula, N is the number of nodes; y i It is the actual label of node i; It is the probability of infarction predicted by the model.

[0141] It should be noted that the graph attention mechanism (the third AI sub-model) applies graph attention to myocardial infarction localization. It utilizes the vascular network structure of the heart, representing its various parts graphically, and then uses the graph attention mechanism to identify and locate the myocardial infarction area. The graph attention mechanism includes:

[0142] (1) Input layer

[0143] Input layer receives node feature matrix X∈R N×D , where N is the number of nodes and D is the feature dimension.

[0144] (2) Attention weight calculation

[0145] For any edge e between nodes i and j ij Calculate the attention weights a between them. ij (h i ,h j This can be achieved using the following formula:

[0146]

[0147] In the formula, h i h is the eigenvector of node i; j W is the feature vector of node j; a It is a weight matrix used for linear transformations; b a It is a bias term; This represents the dot product of eigenvectors; LeakyReLU is a non-linear activation function used to introduce non-linear relationships.

[0148] (3) Model Training

[0149] We introduce a cross-entropy loss function to optimize the model parameters:

[0150]

[0151] In the formula, y i It is the actual label of node i (e.g., whether it is a blockage area); It is the probability of infarction predicted by the model.

[0152] In summary, by using graph attention mechanisms, the most critical nodes in the myocardial infarction region can be identified, thereby improving the accuracy of myocardial infarction localization.

[0153] As described in step S2, target electrocardiogram data of the target patient is obtained, and the key nodes of myocardial infarction corresponding to the target electrocardiogram data are determined through the correspondence pre-established by the artificial intelligence model.

[0154] The target 12-lead electrocardiogram (ECG) data of the target patient is collected using an ECG acquisition device, consisting of six limb leads and six precordial leads. This target 12-lead ECG data is then input into the trained artificial intelligence model, which analyzes the target patient's ECG data and outputs key nodes of myocardial infarction.

[0155] The beneficial effects achieved by this invention include:

[0156] (1) Compared with existing surface potential mapping, this invention addresses the pathological nature of surface potential reconstruction by resolving two causes: complex propagation and noise. For complex propagation, geometrically constrained embedding is performed using graphical information of nodes and edges to learn the relationship between a 12-lead electrocardiogram and the distribution of ventricular surface potentials. This embedding controls the direction of characteristic changes in potential levels by introducing weight assignments at the edges between geometric nodes on the ventricular surface, thereby simulating the propagation pattern of cardiac electrical activity. This effectively limits the output of the reconstruction to conform to the ventricular transmembrane potential distribution pattern, particularly ensuring that the potential around the infarct boundary is consistent with the physiological local structural features.

[0157] (2) For noise, the frequency-enhanced geometrically constrained iterative network applies a frequency-enhanced feature attention mechanism to reduce noise components while highlighting target features during reconstruction. This mechanism focuses not only on time-domain features but also on frequency-domain features through discrete multi-wavelet decomposition. This decomposition uses a set of wavelet bases to decompose the ECG signal into approximate components (low-frequency, long-time scale) and detail components (high-frequency, short-time scale) at different scales. Each wavelet base captures specific frequency information at a specific time point without interfering with each other. This allows for analysis of the high-frequency components of the signal at finer time scales while maintaining a global view of the low-frequency components. It provides frequency information at different scales while preserving the temporal distribution of these frequency components.

[0158] (3) By leveraging a self-attention mechanism, the frequency-enhanced geometrically constrained iterative network can focus on target features related to infarct localization from these decomposed components while reducing noise components. After signal denoising and feature extraction, the signal can be represented using the corresponding inverse transform. Therefore, noise that may affect the reconstruction process in a 12-lead electrocardiogram can be significantly reduced.

[0159] As the apparatus embodiment is basically similar to the method embodiment, it is described in a relatively simple manner. For relevant details, please refer to the description of the method embodiment.

[0160] Reference Figure 3 This application illustrates an embodiment of an electrocardiogram-based myocardial infarction localization device based on a geometric iterative network. The device relates to an artificial intelligence model, which includes a first artificial intelligence sub-model, a second artificial intelligence sub-model, and a third artificial intelligence sub-model.

[0161] Specifically, it includes:

[0162] The correspondence establishment module 310 is used to establish a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction through an artificial intelligence model; wherein, the correspondence includes a first correspondence, a second correspondence, and a third correspondence; specifically, the first AI sub-model establishes a first correspondence between ECG data and ventricular surface potential distribution; the second AI sub-model establishes a second correspondence between ventricular surface potential distribution and key regions of myocardial infarction; and the third AI sub-model establishes a third correspondence between key regions of myocardial infarction and key nodes of myocardial infarction.

[0163] The data processing module 320 is used to acquire the target electrocardiogram data of the target patient and determine the target myocardial infarction key node corresponding to the target electrocardiogram data through the correspondence pre-established by the artificial intelligence model.

[0164] In one embodiment of the present invention, the correspondence establishment module 310 includes:

[0165] The discrete multi-wavelet operation submodule is used to determine the potential features of the electrocardiogram data in the time and frequency domains using discrete multi-wavelet operations.

[0166] The feature compensation submodule is used to perform feature compensation on the potential features based on a preset sandwich structure operator to determine graphic information whose data dimension is consistent with the ventricular surface potential distribution.

[0167] The geometric constraint embedding submodule is used to perform geometric constraint embedding based on the graphical information of nodes and edges, and to establish the first correspondence between the electrocardiogram data and the ventricular surface potential distribution by utilizing the self-learning capability of the first artificial intelligence sub-model.

[0168] In one embodiment of the present invention, the discrete multi-wavelet operation submodule includes:

[0169] The data decomposition unit is used to decompose the electrocardiogram data using orthogonal multiwavelets to obtain decomposed electrocardiogram data;

[0170] An information simplification unit is used to determine the simplified ECG information corresponding to the decomposed ECG data by employing downsampling and zero-filling operations.

[0171] A feature capture unit is used to determine the potential features of the simplified electrocardiogram information in the time and frequency domains using discrete multiwavelet operations.

[0172] In one embodiment of the present invention, the geometric constraint embedding submodule includes:

[0173] The first sub-model building unit is used to establish the first correspondence between the electrocardiogram data and the ventricular surface potential distribution, and the formula for setting geometric constraints on the first artificial intelligence sub-model is as follows:

[0174]

[0175] u k =P l (y k -γf(y k (2)

[0176] y k+1 =u k +τ k+1 (u k -u k-1 (3)

[0177]

[0178] y k+1 =u k +τ k (u k -u k-1 (5)

[0179] In the formula, T k+1 The iteration step size is used to control the magnitude of variable updates during the iteration process; u k y is an intermediate variable in the iteration process. kγ is the target variable in the iteration process; γ is the regularization parameter used to control the size of the gradient descent step size. The transpose of the adaptive gradient operator during iteration is used to adjust u. k Update direction; Hy k Initialize application terms H to y k The result; where in step k, H is the initialization term; P represents the proximal operator supported by the network with embedded constraints; G is the adaptive gradient operator during iteration; τ k >0 can be learned through training as defined in equation (1); ρ is the contraction operator ρ(u k )=(|u k |-ρ)+sgn(u k ), ()+ is |u k |-ρ is the maximum value between 0 and 0, where sgn represents the sign function, limited to ±1.

[0180] In one embodiment of the present invention, the correspondence establishment module 310 further includes:

[0181] The data acquisition submodule is used to acquire medical image datasets; wherein, the medical image datasets include normal myocardial image datasets, myocardial infarction area image datasets, and pathological change image datasets related to myocardial infarction.

[0182] The second sub-model building sub-module is used to train the second artificial intelligence sub-model based on the medical image dataset to determine the second correspondence between the ventricular surface potential distribution and the key region of myocardial infarction.

[0183] In one embodiment of the present invention, the second artificial intelligence sub-model includes a physical sub-model and a neural network, and the second sub-model building sub-module includes:

[0184] The potential expression unit is used to approximate the expression for the potential based on the neural network; the formula is as follows:

[0185] u(x,y,z,t)≈N(x,y,z,t;θ);

[0186] In the formula, θ is the parameter of the neural network; x, y, z are the spatial coordinates; and t represents time.

[0187] The training unit is used to substitute the expression for the electric potential into the physical sub-model used to describe the electrophysiological process of the heart, and to train the second artificial intelligence sub-model, making the neural network satisfy the physical laws based on the loss function; the formula is as follows:

[0188]

[0189] In the formula, α is the weighting parameter that balances the two loss terms; p i At position (x) i ,y i ,t i ) The observed probability of myocardial infarction; N data It is the total number of observed data points; It is the actual potential value at position j; It is the potential value predicted by the neural network;

[0190] The update unit is used to update the weights W and bias b using the backpropagation algorithm; the formula is as follows:

[0191]

[0192] The update rules are as follows:

[0193]

[0194]

[0195] In the formula, L is the loss function, used to evaluate the difference between the model's prediction and the actual observation; y is the output of the neural network, representing the predicted potential value; and η is the learning rate.

[0196] The loop unit is used to return to the neural network for continuous training after each training session if the difference between the predicted value and the true value is still greater than the preset minimum tolerance difference, until the difference between the predicted value and the true value is no greater than the preset minimum tolerance difference.

[0197] In one embodiment of the present invention, the correspondence establishment module 310 further includes:

[0198] The third sub-model building module is used to input the node features in the key region of myocardial infarction into a preset cross-entropy loss function, and determine the output parameters of the cross-entropy loss function as the corresponding key nodes of myocardial infarction; wherein, the cross-entropy loss function is as follows:

[0199]

[0200] In the formula, N is the number of nodes; y i It is the actual label of node i; It is the probability of infarction predicted by the model.

[0201] Reference Figure 4 The present invention illustrates a computer device for a method of locating myocardial infarction using electrocardiography based on a geometric iterative network, which may specifically include the following:

[0202] The computer device 12 described above is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0203] Bus 18 refers to one or more of several types of bus 18 architectures, including memory bus 18 or memory controller, peripheral bus 18, graphics acceleration port, processor, or local bus 18 using any of the various bus 18 architectures. For example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus 18, Micro Channel Architecture (MAC) bus 18, Enhanced ISA bus 18, Audio / Video Electronics Standards Association (VESA) local bus 18, and Peripheral Component Interconnect (PCI) bus 18.

[0204] Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0205] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard disk drive"). Figure 4 Not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 42 configured to perform the functions of the embodiments of the present invention.

[0206] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory. Such program modules 42 include—but are not limited to—an operating system, one or more application programs, other program modules 42, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0207] Computer device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, camera, etc.), and with one or more devices that enable a user to interact with the computer device 12, and / or with any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through input / output (I / O) interface 22. Furthermore, computer device 12 can also communicate with one or more networks (e.g., local area network (LAN)), wide area network (WAN), and / or public networks (e.g., the Internet) via network adapter 20. Figure 4 As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18. It should be understood that, although... Figure 4 Not shown, it can be combined with computer device 12 to use other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing unit 16, external disk drive array, RAID system, tape drive and data backup storage system 34, etc.

[0208] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the electrocardiogram localization method for myocardial infarction based on geometric iterative networks provided in the embodiments of the present invention.

[0209] That is, when the processing unit 16 executes the above program, it achieves the following: establishing a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction through an artificial intelligence model; wherein, the correspondence includes a first correspondence, a second correspondence, and a third correspondence; specifically, establishing a first correspondence between ECG data and ventricular surface potential distribution through the first artificial intelligence sub-model; establishing a second correspondence between ventricular surface potential distribution and key areas of myocardial infarction through the second artificial intelligence sub-model; establishing a third correspondence between key areas of myocardial infarction and key nodes of myocardial infarction through the third artificial intelligence sub-model; acquiring target ECG data of the target patient, and determining the target key nodes of myocardial infarction corresponding to the target ECG data through the correspondence.

[0210] In this embodiment of the invention, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements an electrocardiogram method for locating myocardial infarction based on a geometric iterative network as provided in all embodiments of this application.

[0211] That is, when the program is executed by the processor, it implements the following: establishing a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction through an artificial intelligence model; wherein, the correspondence includes a first correspondence, a second correspondence, and a third correspondence; specifically, establishing a first correspondence between ECG data and ventricular surface potential distribution through the first artificial intelligence sub-model; establishing a second correspondence between ventricular surface potential distribution and key regions of myocardial infarction through the second artificial intelligence sub-model; establishing a third correspondence between key regions of myocardial infarction and key nodes of myocardial infarction through the third artificial intelligence sub-model; acquiring target ECG data of the target patient, and determining the target key nodes of myocardial infarction corresponding to the target ECG data through the correspondence.

[0212] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0213] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0214] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider). The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments; similar or identical parts between embodiments can be referred to interchangeably.

[0215] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0216] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0217] The above provides a detailed description of the electrocardiogram method and apparatus for locating myocardial infarction based on geometric iterative networks provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for locating myocardial infarction using electrocardiography based on geometric iterative networks, characterized in that, The method involves an artificial intelligence model, which includes a first artificial intelligence sub-model, a second artificial intelligence sub-model, and a third artificial intelligence sub-model; the method includes: An artificial intelligence model is used to establish a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction. This correspondence includes a first sub-correspondence, a second sub-correspondence, and a third sub-correspondence. Specifically, the first AI sub-model establishes a first correspondence between ECG data and ventricular surface potential distribution; the second AI sub-model establishes a second correspondence between ventricular surface potential distribution and key regions of myocardial infarction; and the third AI sub-model establishes a third correspondence between key regions of myocardial infarction and key nodes of myocardial infarction. Obtain the target electrocardiogram data of the target patient, and determine the target myocardial infarction key nodes corresponding to the target electrocardiogram data through the correspondence relationship; The step of establishing the first correspondence between the electrocardiogram (ECG) data and the ventricular surface potential distribution through the first artificial intelligence sub-model includes: determining the potential features of the ECG data in the time and frequency domains using discrete multi-wavelet operations; performing feature compensation on the potential features based on a preset sandwich structure operator to determine graphical information with a data dimension consistent with the ventricular surface potential distribution; performing geometric constraint embedding based on the graphical information of nodes and edges, and utilizing the self-learning capability of the first artificial intelligence sub-model to establish the first correspondence between the ECG data and the ventricular surface potential distribution; The step of determining the potential features of the electrocardiogram (ECG) data in the time and frequency domains using discrete multiwavelet operations includes: decomposing the ECG data using orthogonal multiwavelets to obtain decomposed ECG data; determining simplified ECG information corresponding to the decomposed ECG data using downsampling and zero-filling operations; and determining the potential features of the simplified ECG information in the time and frequency domains using discrete multiwavelet operations.

2. The method according to claim 1, characterized in that, The step of establishing the second correspondence between the ventricular surface potential distribution and the key region of myocardial infarction through the second artificial intelligence sub-model includes: Obtain a medical imaging dataset; wherein the medical imaging dataset includes a normal myocardial image dataset, a myocardial infarction area image dataset, and a pathological change image dataset related to myocardial infarction; The second artificial intelligence sub-model is trained based on the medical image dataset to determine the second correspondence between the ventricular surface potential distribution and the key region of myocardial infarction.

3. A device for locating myocardial infarction using electrocardiogram based on a geometric iterative network, characterized in that, The device relates to an artificial intelligence model, which includes a first artificial intelligence sub-model, a second artificial intelligence sub-model, and a third artificial intelligence sub-model; including: A correspondence establishment module is used to establish a correspondence between electrocardiogram (ECG) data and key nodes of myocardial infarction using an artificial intelligence model. The correspondence includes a first sub-correspondence, a second sub-correspondence, and a third sub-correspondence. Specifically, the first AI sub-model establishes a first correspondence between ECG data and ventricular surface potential distribution; the second AI sub-model establishes a second correspondence between ventricular surface potential distribution and key regions of myocardial infarction; and the third AI sub-model establishes a third correspondence between key regions of myocardial infarction and key nodes of myocardial infarction. The data processing module is used to acquire the target electrocardiogram data of the target patient and determine the target myocardial infarction key node corresponding to the target electrocardiogram data through the correspondence pre-established by the artificial intelligence model. The correspondence establishment module includes: a discrete multi-wavelet operation submodule, used to determine the potential features of the electrocardiogram data in the time and frequency domains using discrete multi-wavelet operation; a feature compensation submodule, used to perform feature compensation on the potential features based on a preset sandwich structure operator to determine graphical information whose data dimension is consistent with the ventricular surface potential distribution; and a geometric constraint embedding submodule, used to perform geometric constraint embedding based on the graphical information of nodes and edges, and to establish the first correspondence between the electrocardiogram data and the ventricular surface potential distribution using the self-learning capability of the first artificial intelligence sub-model. The discrete multi-wavelet operation submodule includes: a data decomposition unit, used to decompose the electrocardiogram data using orthogonal multi-wavelets to obtain decomposed electrocardiogram data; an information simplification unit, used to determine the simplified electrocardiogram information corresponding to the decomposed electrocardiogram data using downsampling and zero-filling operations; and a feature capture unit, used to determine the potential features of the simplified electrocardiogram information in the time domain and frequency domain using discrete multi-wavelet operation.

4. A device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the method as described in any one of claims 1 to 2.

5. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the method as described in any one of claims 1 to 2.

Images