Arrhythmia identification method and system based on lead specificity and correlation

Abstract

The invention discloses an arrhythmia identification method and system based on lead specificity and correlation, and belongs to the technical field of physiological signal analysis. The method comprises the following steps: acquiring a 12-lead electrocardiosignal, and processing the 12-lead electrocardiosignal through a trained deep learning network to generate a first recognition result; carrying out correlation analysis on the 12-lead electrocardiosignals, and constructing a correlation graph by utilizing the correlation among leads; performing graph structure feature learning on the correlation graph through a trained graph neural network to obtain a second recognition result; the first recognition result and the second recognition result are integrated, an arrhythmia recognition result is obtained, the deep learning network comprises a plurality of recognition branches arranged in parallel, and each recognition branch learns specific information of the corresponding electrocardiosignal based on the lead channel. Specific information of each lead and correlation information between the leads can be fully learned, and the problem that the comprehensiveness and accuracy of existing arrhythmia recognition need to be improved is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of physiological signal analysis, and particularly to an arrhythmia recognition method and system based on lead specificity and correlation. Background Technique

[0002] The statements in this part only mention the background techniques related to the present invention and do not necessarily constitute prior art.

[0003] In recent years, the incidence and mortality of arrhythmia have been rising continuously, bringing serious burdens to residents and society. Early detection and intervention of arrhythmia have great clinical significance. At present, arrhythmia is mainly diagnosed by clinicians through observing electrocardiogram (ECG). Electrocardiogram is a non-invasive diagnostic tool that can record the bioelectricity generated by the heart during the excitation process and display the change process of the heart rhythm and the potential waveform of heart activity in a graphical way. It contains the basic information for heart function analysis, so it is widely used in the field of cardiology to evaluate the health status of the human body.

[0004] With the increase in the number of arrhythmia patients, the amount of electrocardiogram data is also increasing day by day. In addition, due to the sporadic characteristics of arrhythmia, Holter devices are usually required to record long-term electrocardiogram data, which further increases the amount of electrocardiogram data. Clinicians will spend a lot of time and energy in viewing and analyzing these electrocardiograms, increasing their work burden. At the same time, most patients often seek medical treatment when they feel obvious discomfort in the heart, which may lead to the aggravation of the condition and the delay of treatment. The lack of early detection and diagnosis means makes many heart problems fail to be discovered and treated in time, thus increasing the complexity and risk of treatment.

[0005] In view of the high efficiency and robustness of computers, using them to assist in the diagnosis of arrhythmia has gradually become the focus of attention and research in the medical field. Using a computer to assist in classifying and identifying arrhythmia can, on the one hand, reduce the burden on doctors, and at the same time add an extra guarantee for disease diagnosis to avoid misdiagnosis and wrong diagnosis. On the other hand, this method can be used for daily monitoring and popularized in families to achieve early intervention of arrhythmia.

[0006] Since the leads showing changes in electrocardiograms for different types of arrhythmias are not the same, and the same type of arrhythmia also changes not only in one lead of the electrocardiogram, in actual clinical practice, the 12-lead electrocardiogram remains the most reliable and comprehensive tool for diagnosing arrhythmias. At present, a large number of studies based on deep neural networks have used 12-lead electrocardiograms to conduct arrhythmia recognition research. However, these studies usually concatenate the 12 leads of the electrocardiogram signals into a matrix and then input the matrix into the deep neural network to extract effective information. Although this framework has the ability to learn the holistic information of 12-lead electrocardiogram signals, it will cause the information contained in each lead to affect each other during the training process, ignoring the specific information contained in different leads, resulting in insufficient learning of the specific information contained in each lead of the 12-lead electrocardiogram signals. In addition, this framework also lacks the learning of the correlation information between each lead. Summary of the Invention

[0007] To solve the deficiencies of the prior art, the present invention provides an arrhythmia recognition method, system, electronic device, computer-readable storage medium, and computer program product based on lead specificity and correlation, which fully considers the specific information of each lead and the correlation information between leads, further excavates the effective information contained in the electrocardiogram signals, and improves the comprehensiveness and accuracy of the evaluation.

[0008] In a first aspect, the present invention provides an arrhythmia recognition method based on lead specificity and correlation; An arrhythmia recognition method based on lead specificity and correlation includes: Obtain 12-lead electrocardiogram signals, process the 12-lead electrocardiogram signals through a trained deep learning network, and generate a first recognition result; Conduct a correlation analysis on the 12-lead electrocardiogram signals, and construct a correlation graph using the correlation between each lead; Learn the graph structure features of the correlation graph through a trained graph neural network to obtain a second recognition result; Integrate the first recognition result and the second recognition result to obtain an arrhythmia recognition result; Wherein, the deep learning network includes a plurality of recognition branches arranged in parallel, and each recognition branch learns the specific information of the corresponding electrocardiogram signal based on the lead channel.

[0009] In some embodiments, the process of processing the 12-lead electrocardiogram signals through a trained deep learning network to generate a first recognition result includes: Input the 12-lead electrocardiogram signals into the corresponding recognition branches for processing according to the lead channels to obtain the corresponding lead feature representations; Cascade all lead feature representations to obtain a high-dimensional feature vector; process the high-dimensional feature vector through a fully connected layer to obtain a first recognition result.

[0010] In some embodiments, the recognition branch includes a convolutional neural network, a bidirectional long short-term memory network, and an attention mechanism connected in sequence; The convolutional neural network is used to extract local features from the electrocardiogram signal to generate local feature representations; the bidirectional long short-term memory network is used to extract temporal features from the local feature representations to generate a temporal feature vector; the attention mechanism is used to weight the temporal feature vector according to the importance of different parts of the electrocardiogram signal.

[0011] In some embodiments, the correlation analysis of the 12-lead electrocardiogram signal and constructing a correlation graph using the correlation between leads includes: Calculate the Pearson correlation coefficient between the electrocardiogram signals of each lead and construct a correlation matrix; Based on the correlation matrix, using the comparison result between the Pearson correlation coefficient between the electrocardiogram signals of each lead and a preset threshold, construct a correlation graph with lead channels as nodes.

[0012] In some embodiments, the integrating the first recognition result and the second recognition result to obtain an arrhythmia recognition result is specifically: determining the corresponding voting weights according to the classification performance of the deep neural network and the graph neural network, and using the voting weights to perform weighted voting on the first recognition result and the second recognition result to obtain an arrhythmia recognition result.

[0013] In some embodiments, when processing the 12-lead electrocardiogram signal through a trained deep learning network, it further includes: Perform filtering processing on the 12-lead electrocardiogram signal in sequence through an IIR notch filter, a Savitzky-Golay filter, and a Butterworth band-pass filter.

[0014] In a second aspect, the present invention provides an arrhythmia recognition system based on lead specificity and correlation; An arrhythmia recognition system based on lead specificity and correlation includes: A specificity information acquisition module, configured to: acquire a 12-lead electrocardiogram signal, process the 12-lead electrocardiogram signal through a trained deep learning network, and generate a first recognition result; A correlation information acquisition module, configured to: perform correlation analysis on the 12-lead electrocardiogram signal, construct a correlation graph using the correlation between leads; perform graph structure feature learning on the correlation graph through a trained graph neural network to obtain a second recognition result; An integrated module, configured to: integrate the first recognition result and the second recognition result to obtain an arrhythmia recognition result; Wherein, the deep learning network includes a plurality of recognition branches arranged in parallel, and each recognition branch learns the specific information of the corresponding electrocardiogram signal based on the lead channel.

[0015] In a third aspect, the present invention provides an electronic device; An electronic device includes a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the above-mentioned arrhythmia recognition method based on lead specificity and correlation.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium; A computer-readable storage medium, on which a computer program / instructions are stored, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned arrhythmia recognition method based on lead specificity and correlation are implemented.

[0017] In a fifth aspect, the present invention provides a computer program product; A computer program product includes a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned arrhythmia recognition method based on lead specificity and correlation are implemented.

[0018] Compared with the prior art, the beneficial effects of the present invention are: 1. The technical solution provided by the present invention constructs a deep learning network including multiple recognition branches, and each branch independently learns the specific information contained in each lead, avoiding mutual interference and loss of information among the information; by jointly using the correlation analysis and graphical analysis methods, the relevant information between leads is fully mined, strengthening the learning of the potential associations between leads, and providing more hierarchical feature supports for the recognition of arrhythmia.

[0019] 2. The technical solution provided by the present invention, through ensemble learning, fuses the output results of the two networks to obtain the final recognition result, effectively reducing the bias and overfitting problems existing in a single model, and making the model have stronger generalization ability and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention, and the schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0021] Figure 1 It is a schematic flowchart of the arrhythmia recognition method based on lead specificity and correlation provided by an embodiment of the present invention; Figure 2 Schematic diagram of the network architecture of the arrhythmia recognition method based on lead specificity and correlation provided by the embodiments of the present invention; Figure 3 Schematic diagram of the network architecture of the deep learning network provided by the embodiments of the present invention; Figure 4 Schematic diagram of the construction process of the correlation graph provided by the embodiments of the present invention. Detailed implementation manners

[0022] It should be noted that the following detailed descriptions are all exemplary and are intended to provide further explanations of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0023] It should be noted that the terms used herein are only for describing specific implementation manners and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these process, method, product or device.

[0024] In the case of no conflict, the embodiments in the present invention and the features in the embodiments may be combined with each other.

[0025] Embodiment 1 Existing arrhythmia recognition treats the 12-lead electrocardiogram (ECG) signals as a whole for processing, ignoring the mutual influence between leads, the specific information of different leads, and the correlation information between two leads; therefore, this embodiment provides an arrhythmia recognition method based on lead specificity and correlation, which separately processes each lead signal in the 12-lead ECG signals in parallel through a multi-branch deep neural network, and mines the correlation information through a graph neural network.

[0026] Next, in combination with Figures 1 - 4 , a detailed description will be given of an arrhythmia recognition method based on lead specificity and correlation disclosed in this embodiment. The arrhythmia recognition method based on lead specificity and correlation includes: S1. Obtain 12-lead ECG signals and perform preprocessing.

[0027] Specifically, acquire the 12-lead electrocardiogram (ECG) signal of the subject, use an IIR notch filter to remove the power frequency interference in the 12-lead ECG signal, use a polynomial 3rd-order Savitzky-Golay filter to remove the baseline drift in the 12-lead ECG signal, and use a second-order Butterworth band-pass filter with a frequency range of 0.05 - 75 Hz to remove high-frequency noises such as electromyogram interference in the 12-lead ECG signal.

[0028] S2. Process the preprocessed 12-lead ECG signal through a trained deep learning network to generate a first recognition result.

[0029] Furthermore, the deep learning network includes 12 parallel recognition branches. The outputs of the 12 recognition branches are cascaded and then input into a fully connected layer, and the first recognition result is output through the processing of the fully connected layer.

[0030] Specifically, the recognition branch includes a convolutional neural network, a bidirectional long short-term memory network, and an attention mechanism connected in sequence. The convolutional neural network is used to extract local features from the ECG signal to generate a local feature representation. The bidirectional long short-term memory network is used to extract temporal features from the local feature representation to generate a temporal feature vector. The attention mechanism is used to weight the temporal feature vector according to the importance of different parts of the ECG signal.

[0031] The bidirectional long short-term memory network can capture the temporal dependence relationship of the front and back of the ECG signal. Using the attention mechanism, it can automatically adjust the weights of the features of each part according to the importance of different parts of the ECG signal, enabling the network to focus on the key regions related to arrhythmia in the ECG signal while ignoring irrelevant information.

[0032] In this embodiment, no improvement is made to the network architectures of the convolutional neural network, the bidirectional long short-term memory network, and the attention mechanism, and their specific network structure compositions, etc., will not be elaborated here.

[0033] As an implementation manner, S2 includes: S201. Input each lead ECG signal in the 12-lead ECG signal into the corresponding recognition branch for processing according to the lead channels to obtain the corresponding lead feature representation.

[0034] Furthermore, the processing of the input corresponding lead ECG signal by the recognition branch (T represents the duration of the signal) is specifically as follows: First, extract the local features in the ECG signal through the convolutional neural network. After a series of convolutional operations and pooling operations of the convolutional neural network, generate the corresponding local feature representation, expressed as: , ; In the formula, Denotes the number of time steps after the convolution operation. Denotes the feature dimension extracted by the convolutional neural network.

[0035] Then, is input into a bidirectional long short-term memory network, and processed by the bidirectional long short-term memory network to model the temporal dependence and generate a temporal feature vector, denoted as: , ; In the formula, Denotes the number of time steps after the convolution operation. Denotes the dimension of the hidden layer of the bidirectional long short-term memory network.

[0036] Finally, the temporal feature vector is input into the attention mechanism, and the temporal features are weighted by the attention mechanism to highlight the importance of critical moments. The specific process is as follows: For each time step , its importance is calculated through a fully connected layer to obtain the attention score , and then the weight of each time step is calculated according to this score, denoted as: ; The temporal feature vector is weighted based on this weight, and the weighted temporal feature vector is denoted as , and this feature representation contains local features, temporal information extracted from the electrocardiogram signal, and highlights important moment information.

[0037] S202. Concatenate all lead feature representations to obtain a high-dimensional feature vector; process the high-dimensional feature vector through a fully connected layer to obtain the first recognition result.

[0038] S3. Conduct a correlation analysis on the 12-lead electrocardiogram signal, and construct a correlation graph using the correlations between leads. It includes: S301. Calculate the Pearson correlation coefficient between the electrocardiogram signals of each pair of leads to obtain the correlation between the electrocardiogram signals of each pair of leads. The calculation formula is as follows: ; In the formula, Denotes the correlation between the electrocardiogram signal of the th lead and the electrocardiogram signal of the th lead. Denotes the voltage value of the th lead at the time point . Denotes the average voltage value of all time points of the th lead. Indicates the voltage value of the th lead at the time point Indicates the average voltage value of all time points of the

[0039] S302. Calculate the Pearson correlation coefficient of the electrocardiogram signals of all leads to form a 12×12 correlation matrix R, which is expressed as: ; In the formula, represents the self-correlation.

[0040] S303. Use the 12 leads as nodes and construct a correlation graph by using the comparison result between the Pearson correlation coefficient between the electrocardiogram signals of each lead in the correlation matrix R and a preset threshold.

[0041] Specifically, each lead in the 12-lead electrocardiogram signal is regarded as a node. If the correlation between the electrocardiogram signals of two leads is greater than the threshold T, it means that there is a strong correlation between the electrocardiogram signals of the two leads. Then, an edge is established between them, which is expressed as: ; In the formula, represents the adjacency relationship between lead and lead indicating that there is a connection between these two leads, indicating that there is no connection between these two leads. In this embodiment, the T value is set to 0.6 according to experience.

[0042] S4. Perform graph structure feature learning on the correlation graph through a trained graph neural network to obtain a second recognition result.

[0043] The graph convolutional network aggregates the information of node neighbors through the adjacency relationship between nodes to update the representation of the nodes, so that the nodes can adjust their feature representations according to the relationship between leads, and then fully learn the correlation relationship between leads. Through the output of the graph neural network, the feature representations of the nodes are processed through a fully connected layer to obtain the arrhythmia recognition result.

[0044] As an implementation manner, S4 is specifically: using the correlation graph (including node features and adjacency matrix) as the input, processing the correlation graph through a graph neural network to determine the second recognition result, where the correlation graph includes node features and an adjacency matrix, the node features represent the electrocardiogram signal information of each lead, and the adjacency matrix represents the correlation between leads.

[0045] Specifically, the correlation graph is processed sequentially through multiple graph convolutional layers, and the node features are updated through information aggregation with neighboring nodes. The rule of the graph convolutional layer is expressed as: ; In the formula, represents the node feature matrix of the -th layer, is the original input feature, is the preprocessed 12-lead electrocardiogram signal, where the signal of each lead is used as the initial node feature, is the normalized adjacency matrix, is the -th trainable weight matrix of the layer, is the activation function.

[0046] In this process, the adjacency matrix determines which nodes have strong correlations, thus ensuring that only the information of nodes with relatively strong correlations is transmitted and aggregated; the feature representation of the nodes is gradually deepened through multiple rounds of graph convolution operations. Through the action of the adjacency matrix in each layer, more neighbor node information is aggregated, so as to capture more complex inter-lead relationships and graph structure information. Specifically: the input signal feature is combined with the structure information in the adjacency matrix by the first graph convolutional layer, and the updated node feature is output; the updated node feature is combined with the adjacency matrix again by the second graph convolutional layer to obtain a deeper feature ; and so on, the feature representation of the nodes is gradually deepened through multiple layers of convolution; finally, the updated node features are processed through a fully connected layer to finally obtain the second recognition result.

[0047] S5. Integrate the first recognition result and the second recognition result to obtain the arrhythmia recognition result.

[0048] Specifically, the corresponding voting weights are determined according to the F1 values obtained after classification by the deep neural network and the graph neural network, and the first recognition result and the second recognition are weighted and voted using the voting weights to obtain the arrhythmia recognition result.

[0049] Exemplarily, the calculation process of the voting weights is as follows: ; ; In the formula, represents the voting weight of the deep neural network, represents the voting weight of the graph neural network, The F1 value obtained after the classification of the deep neural network, representing its classification performance; The F1 value obtained after the classification of the graph neural network, representing its classification performance.

[0050] Under this mechanism, the voting weight of each model is allocated according to the performance of the deep neural network. The neural network with better performance will obtain a higher weight, thus having a greater impact on the final classification result. This strategy effectively improves the accuracy and robustness of the integrated model, ensuring that the classification result is more accurate and reliable.

[0051] Embodiment 2 This embodiment discloses an arrhythmia recognition system based on lead specificity and correlation, including: A specificity information acquisition module, configured to: acquire 12-lead electrocardiogram signals, process the 12-lead electrocardiogram signals through a trained deep learning network, and generate a first recognition result; A correlation information acquisition module, configured to: perform correlation analysis on the 12-lead electrocardiogram signals, construct a correlation graph using the correlation between leads; learn the graph structure features of the correlation graph through a trained graph neural network, and obtain a second recognition result; An integration module, configured to: integrate the first recognition result and the second recognition result to obtain an arrhythmia recognition result; Wherein, the deep learning network includes a plurality of parallel recognition branches, and each recognition branch learns the specificity information of the corresponding electrocardiogram signal based on the lead channel.

[0052] It should be noted here that the above specificity information acquisition module, correlation information acquisition module, and integration module correspond to the steps in Embodiment 1. The examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0053] Embodiment 3 Embodiment 3 of the present invention provides an electronic device, including a memory and a processor, as well as computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps of the above arrhythmia recognition method based on lead specificity and correlation are completed.

[0054] Embodiment 4 Embodiment 4 of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above arrhythmia recognition method based on lead specificity and correlation are completed.

[0055] Example 5 Example 5 of the present invention provides a computer program product, including computer programs / instructions, which when executed by a processor implement the steps of the above-mentioned arrhythmia recognition method based on lead specificity and correlation.

[0056] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the specified functions in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0057] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implement the specified functions in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0058] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the specified functions in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0059] In the above embodiments, the descriptions of each embodiment have their own emphases. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0060] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An arrhythmia recognition method based on lead specificity and correlation, characterized in that Including: Obtain a 12-lead electrocardiogram (ECG) signal, process the 12-lead ECG signal through a trained deep learning network, and generate a first recognition result; Conduct a correlation analysis on the 12-lead ECG signal, and construct a correlation graph using the correlations between leads; Perform graph structure feature learning on the correlation graph through a trained graph neural network to obtain a second recognition result; Integrate the first recognition result and the second recognition result to obtain an arrhythmia recognition result; Wherein, the deep learning network includes multiple parallel recognition branches, and each recognition branch learns specific information of the corresponding ECG signal based on the lead channels.

2. The arrhythmia recognition method based on lead specificity and correlation according to claim 1, characterized in that The process of processing the 12-lead ECG signal through a trained deep learning network to generate a first recognition result includes: Input the 12-lead ECG signal into the corresponding recognition branch according to the lead channels for processing to obtain corresponding lead feature representations; Concatenate all the lead feature representations to obtain a high-dimensional feature vector; process the high-dimensional feature vector through a fully connected layer to obtain a first recognition result.

3. The arrhythmia recognition method based on lead specificity and correlation according to claim 1, wherein The recognition branch includes a convolutional neural network, a bidirectional long short-term memory network, and an attention mechanism connected in sequence; The convolutional neural network is used to extract local features of the ECG signal to generate local feature representations; the bidirectional long short-term memory network is used to extract temporal features of the local feature representations to generate temporal feature vectors; the attention mechanism is used to weight the temporal feature vectors according to the importance of different parts of the ECG signal.

4. The arrhythmia recognition method based on lead specificity and correlation according to claim 1, wherein, The process of conducting a correlation analysis on the 12-lead ECG signal and constructing a correlation graph using the correlations between leads includes: Calculate the Pearson correlation coefficients between the ECG signals of each lead and construct a correlation matrix; Based on the correlation matrix, using the lead channels as nodes, construct a correlation graph using the comparison results between the Pearson correlation coefficients between the ECG signals of each lead and a preset threshold.

5. The arrhythmia recognition method based on lead specificity and correlation according to claim 1, characterized in that The specific process of integrating the first recognition result and the second recognition result to obtain an arrhythmia recognition result is: determine the corresponding voting weights according to the classification performances of the deep neural network and the graph neural network, and perform weighted voting on the first recognition result and the second recognition result using the voting weights to obtain an arrhythmia recognition result.

6. The arrhythmia recognition method based on lead specificity and correlation according to claim 1, characterized in that, When processing the 12-lead ECG signal through a trained deep learning network, it further includes: Filter the 12-lead ECG signal sequentially through an IIR notch filter, a Savitzky-Golay filter, and a Butterworth bandpass filter.

7. An arrhythmia recognition system based on lead specificity and correlation, characterized in that, Including: A specific information acquisition module configured to: obtain a 12-lead ECG signal, process the 12-lead ECG signal through a trained deep learning network, and generate a first recognition result; A correlation information acquisition module configured to: conduct a correlation analysis on the 12-lead ECG signal, construct a correlation graph using the correlations between leads; perform graph structure feature learning on the correlation graph through a trained graph neural network to obtain a second recognition result; An integration module configured to: integrate the first recognition result and the second recognition result to obtain an arrhythmia recognition result; Among them, the deep learning network includes a plurality of recognition branches arranged in parallel, and each recognition branch learns the specific information of the corresponding electrocardiogram signal based on the lead channel.

8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory, characterized in that, The processor executes the computer program to implement the steps of the arrhythmia recognition method based on lead specificity and correlation according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, the steps of the arrhythmia recognition method based on lead specificity and correlation according to any one of claims 1-6 are implemented.

10. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, the steps of the arrhythmia recognition method based on lead specificity and correlation according to any one of claims 1-6 are implemented.

Images