Electrocardiosignal visualization method and related equipment

Abstract

The invention discloses an electrocardiosignal visualization method and related equipment, and relates to the field of electrocardiosignal analysis. According to the method, body surface voltage feature vectors of at least one target moment are obtained from an electrocardiogram, a configured potential estimation model is called, each body surface voltage feature vector is processed, and heart surface electrical excitation intensity distribution of each target moment is obtained. And based on the heart surface electrical excitation intensity distribution at each target moment, rendering the heart three-dimensional model to obtain a target heart three-dimensional model corresponding to each target moment. According to the method, the target heart three-dimensional model of the heart electrical excitation state can be displayed, a user can conveniently and visually observe the process of heart electrical excitation from origin to conduction in the time period of electrocardiogram detection, and the abnormal electrical excitation part of the heart is judged in combination with the diagnosis result of the electrocardiogram.

Description

Technical Field

[0001] This application relates to the field of electrocardiogram (ECG) signal processing technology, and more specifically, to an ECG signal visualization method and related equipment. Background Technology

[0002] In the field of cardiac disease diagnosis, electrocardiograms (ECGs) record the changes in electrical activity in different parts of the heart during each cardiac cycle, presenting them as waveforms and providing doctors with important information to assess the heart's health. However, traditional ECG analysis methods still have certain limitations when faced with the diagnosis of some complex cardiac diseases. Taking atrial fibrillation (AF) as an example, normal cardiac excitation follows a specific conduction path and rhythm, while in AF, rapid and irregular fibrillation occurs in various parts of the atria. However, it is difficult to accurately determine the source of AF from multiple waveforms on an ECG alone. Summary of the Invention

[0003] In view of the above problems, this application is proposed to provide a method and related equipment for visualizing electrocardiogram (ECG) signals, so as to map the ECG onto a three-dimensional model of the heart to accurately determine the source of ECG abnormalities. The specific solution is as follows:

[0004] Firstly, this application provides a method for visualizing electrocardiogram (ECG) signals, including:

[0005] Obtain at least one target time point surface voltage feature vector from the electrocardiogram (ECG). The surface voltage feature vector is the set of voltage values ​​of all waveforms in the ECG at the current time.

[0006] The configured potential prediction model is invoked to process each surface voltage feature vector to obtain the heart surface electrical excitation intensity distribution at each target time. The heart surface electrical excitation intensity distribution is the potential prediction value of each network node on the pre-constructed three-dimensional heart model. The potential prediction model is trained based on the sample data of surface voltage feature vectors labeled with the heart surface electrical excitation intensity distribution.

[0007] Based on the distribution of electrical excitation intensity on the surface of the heart at each target time, a three-dimensional model of the heart is rendered to obtain the target three-dimensional model of the heart at each target time.

[0008] In one possible design, in another implementation of the first aspect of the embodiments of this application, the process of rendering a three-dimensional model of the heart based on the distribution of electrical excitation intensity on the surface of the heart at each target time to obtain the target three-dimensional model of the heart at each target time includes:

[0009] Based on the pre-defined correspondence between potential value ranges and RGB values, determine the RGB value corresponding to the potential estimate of each network node;

[0010] Based on the RGB values ​​corresponding to the estimated potential values ​​of all network nodes at each target time, the three-dimensional model of the heart is rendered to obtain the target three-dimensional model of the heart at each target time.

[0011] In one possible design, in another implementation of the first aspect of the embodiments of this application, the method further includes:

[0012] When the number of target times is greater than 1, based on the distribution of cardiac surface electrical excitation intensity corresponding to all target times, frequency domain analysis is performed on the potential prediction of each network node within the time period to which all target times belong, to obtain the excitation frequency of each network node.

[0013] Based on the excitation frequency of each network node, the pre-constructed three-dimensional model of the heart is rendered to obtain a three-dimensional model characterizing the excitation frequency distribution on the surface of the heart.

[0014] In one possible design, in another implementation of the first aspect of the embodiments of this application, the method further includes:

[0015] Based on the target heart three-dimensional model corresponding to all moments in the electrocardiogram, the first cardiac electrical activity state of the electrocardiogram is determined.

[0016] If the second cardiac electrical activity state labeled on the electrocardiogram is different from the first cardiac electrical activity state, the second cardiac electrical activity state is corrected to the first cardiac electrical activity state, and the electrocardiogram labeled with the first cardiac electrical activity state is used as sample data to update the electrocardiogram recognition model.

[0017] In one possible design, another implementation of the first aspect of the embodiments of this application further includes:

[0018] Based on the target heart 3D model corresponding to all moments in the electrocardiogram, the electrical activity state of the target heart in the electrocardiogram is determined and output.

[0019] In one possible design, in another implementation of the first aspect of the embodiments of this application, before obtaining the surface voltage feature vector at least at a target time from the electrocardiogram, the method further includes:

[0020] In response to a user's selection of at least one moment from the electrocardiogram, each selected moment is used as the target moment.

[0021] In one possible design, in another implementation of the first aspect of the embodiments of this application, the method further includes:

[0022] With a target time of 1, display a three-dimensional model of the target heart representing the electrical excitation state of the heart at the target time.

[0023] When the number of target times is greater than 1, the target heart 3D model corresponding to all target times is rendered with time-series animation to obtain a 3D dynamic demonstration animation representing the cardiac electrical excitation state within the time period of all target times, and then displayed.

[0024] Secondly, this application provides an electrocardiogram signal visualization system, including: a processor and a display terminal;

[0025] The display terminal is used to display electrocardiograms and three-dimensional models of the heart, and to capture the user's selection operation at at least one moment in the electrocardiogram and transmit the selection operation to the processor.

[0026] The processor, in response to the selection operation, obtains the body surface voltage feature vector of at least one selected moment from the electrocardiogram, processes each body surface voltage feature vector according to the electrocardiogram signal visualization method described in any of the first aspects of this application, obtains the target heart three-dimensional model at each target moment, and transmits the target heart three-dimensional model to the display terminal.

[0027] Thirdly, this application provides an electronic device, including: a memory and a processor;

[0028] Memory, used to store programs;

[0029] A processor for executing a program to implement the steps of the ECG signal visualization method described in any of the first aspects of this application.

[0030] Fourthly, this application provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the various steps of the electrocardiogram signal visualization method described in any of the first aspects of this application.

[0031] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the various steps of the electrocardiogram signal visualization method described in any of the first aspects of this application.

[0032] By employing the aforementioned technical solution, this application maps the surface voltage feature vector at at least one moment in the electrocardiogram (ECG) to network nodes of a three-dimensional cardiac model using a potential prediction model. It predicts the potential value of each network node, achieving inverse reconstruction from surface signals to the distribution of electrical excitation intensity on the cardiac surface. Furthermore, based on the potential prediction value of each network node, a pre-constructed three-dimensional cardiac model is rendered, generating a target three-dimensional cardiac model that dynamically displays the cardiac electrical excitation state. This allows users to intuitively observe the complete process of cardiac electrical excitation from its origin to conduction within the time period detected by the ECG, and, combined with the diagnostic results of the ECG, determine the location of abnormal electrical excitation in the heart. Attached Figure Description

[0033] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0034] Figure 1 A schematic diagram of an implementation system architecture for the electrocardiogram signal visualization method provided in this application embodiment;

[0035] Figure 2 A flowchart illustrating an electrocardiogram (ECG) signal visualization method provided in this application embodiment;

[0036] Figure 3 Example diagram of a three-dimensional heart model provided in the embodiments of this application;

[0037] Figure 4 An example diagram of an atrial fibrillation load clock is provided for embodiments of this application;

[0038] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0039] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0040] To compensate for the limitations of electrocardiograms in locating the source of cardiac abnormalities, clinical practice often requires patients to undergo complex physical examinations such as CT scans to obtain indirect clues. These clues are then combined with the doctor's experience to make a rough estimate of the location of the cardiac abnormality. This process not only increases the patient's medical costs, including additional time, financial burden, and physical discomfort, but also reduces the doctor's diagnostic efficiency due to its reliance on indirect data and subjective judgment, making the diagnostic process more lengthy and uncertain.

[0041] In response, this application provides a method for visualizing electrocardiogram (ECG) signals, which can be applied to applications such as... Figure 1 The system architecture shown includes a display terminal 10 and a processor 20. The display terminal is used to display an electrocardiogram (ECG) and a 3D model of the heart, and to capture the user's selection operation at at least one moment in the ECG, transmitting the selection operation to the processor.

[0042] The processor, in response to the selection operation, obtains the surface voltage feature vector of at least one selected moment from the electrocardiogram, processes each surface voltage feature vector according to the deployed electrocardiogram signal visualization method, obtains the target heart three-dimensional model at each target moment, and transmits the target heart three-dimensional model to the display terminal.

[0043] In one application scenario, a doctor retrieves a high-density surface electrocardiogram (ECG) of the patient (such as an 18-lead ECG) and a pre-built 3D model of the heart via a display terminal, both displayed side-by-side on the screen. The doctor selects a suspicious point or segment on the ECG waveform using a mouse or touchscreen, and the display terminal transmits the doctor's selection to the processor. The processor processes the surface voltage feature vectors of all moments within the selected time point or time period, ultimately displaying a clear 3D model of the heart at each moment on the terminal screen. This intuitively shows the doctor the 3D model of the heart's electrical excitation state at the selected moment, enabling the doctor to quickly and accurately identify the driving lesion behind the ECG abnormality.

[0044] In this embodiment, the display terminal 10 can be a mobile phone, tablet computer, teaching screen, wearable device, conference terminal, augmented reality (AR) / virtual reality (VR) device, laptop computer, ultra-mobile personal computer (UMPC), netbook, personal digital assistant (PDA), etc. This embodiment does not impose any restrictions on it.

[0045] The processor 20 in this application embodiment may be in the form of, but is not limited to, cloud servers (running as virtual server instances on physical servers), micro servers (suitable for small office or home environments), storage servers (focused on data storage and management), GPU servers (equipped with graphics processing units, suitable for high-performance computing and graphics processing tasks), and modular servers (allowing flexible configuration and expansion of hardware components as needed).

[0046] This application provides a method for visualizing electrocardiogram (ECG) signals. Taking the application of this method to a computer device as an example, the computer device can specifically be... Figure 1 The display terminal 10 or a system consisting of the display terminal 10 and the processor 20. (Refer to...) Figure 2 This application provides a flowchart of an electrocardiogram (ECG) signal visualization method, which may include steps S110 to S130. These steps are described in detail below.

[0047] Step S110: Obtain at least one target time-based surface voltage feature vector from the electrocardiogram. The surface voltage feature vector is the set of voltage values ​​of all waveforms in the electrocardiogram at the current time.

[0048] Among them, 64, 128 or even 256-lead surface electrocardiogram vests can be used to acquire signals and obtain electrocardiograms with extremely high spatial resolution, which is not the only one in this application.

[0049] Furthermore, the voltage signal at each target moment in the electrocardiogram is extracted from multiple waveform lines and converted into a body surface voltage feature vector that can fully characterize the spatial distribution of the electrocardiographic field at that target moment.

[0050] Specifically, for each target time, based on the precise three-dimensional coordinates of each lead recorded in the electrocardiogram on the human body surface, the unstructured voltage values ​​in the electrocardiogram are transformed into structured numerical vectors, namely voltage feature vectors, so that they can be directly processed by computer algorithms and models.

[0051] In one possible implementation, each moment in the electrocardiogram is used as the target moment to map the electrical excitation state of the heart surface throughout the entire electrocardiogram period.

[0052] In another possible implementation, refer to Figure 3The display terminal shows an electrocardiogram (ECG). Through interactive operations between the user and the ECG, a specific moment or several moments within the entire ECG time period can be selected. For example, as shown in the figure, the user selects a time period with moments t1 and t2 as endpoints using the mouse arrow. Based on this, this application uses each moment within the selected t1-t2 time period as the target moment. Alternatively, the user can drag the cursor on the ECG; if the cursor stays on a waveform point for more than one second, the moment corresponding to that waveform point is taken as the target moment. The voltage value at this target moment is obtained from all waveforms in the ECG and converted into a body surface voltage feature vector.

[0053] Step S120: Call the configured potential prediction model to process each body surface voltage feature vector to obtain the heart surface electrical excitation intensity distribution at each target time.

[0054] Among them, the distribution of electrical excitation intensity on the heart surface is the potential prediction value of each network node on the pre-constructed three-dimensional model of the heart. The potential prediction model is trained based on the sample data of body surface voltage feature vector labeled with the distribution of electrical excitation intensity on the heart surface.

[0055] This step utilizes a data-driven machine learning model to learn the complex, nonlinear mapping relationship between body surface features and cardiac surface potential, thereby achieving the inverse solution of the cardiac surface electrical excitation state at the target time. Specifically, the cardiac surface electrical excitation intensity distribution represents the spatial variation of the instantaneous potential amplitude activated by the electrical excitation wave at various points on the cardiac surface (usually the epicardium) at a specific instant during a heartbeat. In this embodiment, it is used to represent the potential values ​​of each network node on the surface of the three-dimensional cardiac model.

[0056] It should be noted that the reason why the surface voltage characteristic vector can be used to inversely calculate the distribution of electrical excitation intensity on the surface of the heart is because the human chest cavity constitutes a definite volumetric conductor system that conforms to the laws of electromagnetic fields. The electrical activity of the heart is the source signal of this system, while the surface potential is the field effect generated after this source signal is conducted through a medium. There is a definite causal relationship between the two, governed by physical laws.

[0057] The potential prediction model used to inversely solve for the surface voltage feature vector can be built using a deep learning-based architecture, such as a Graph Neural Network (GNN) or a Convolutional Neural Network (CNN) with an encoder-decoder. During model training, a large-scale, high-quality training dataset is first constructed. This dataset contains synchronously acquired high-density surface voltage feature vectors (input features) and corresponding real cardiac surface potential distribution data (annotated ground truth values) obtained through invasive intracardiac mapping techniques at the corresponding time points. The model is trained end-to-end using supervised learning, aiming to minimize the error between the predicted and true potential values ​​(such as mean squared error, perceptual similarity loss, etc.). This enables the trained potential prediction model to map surface voltage feature vectors to the electrical excitation intensity distribution on the surface of a pre-constructed cardiac model, thereby obtaining the cardiac surface electrical excitation intensity distribution at the target time.

[0058] Optionally, to improve the physiological rationality and generalization ability of the potential prediction model, physical driving constraints can be introduced during the training process. For example, a forward problem model of electromagnetic field conduction can be embedded as a differentiable hierarchical network, allowing the model to adhere to the basic laws of volumetric conductors while fitting the data. The trained model can quickly and stably deduce the potential distribution on the personalized three-dimensional heart grid nodes at the corresponding time from the new body surface voltage feature vector, achieving a precise and robust inverse mapping from the body surface to the heart surface.

[0059] In the embodiments of this application, the pre-constructed three-dimensional heart model can be a standard three-dimensional heart model or a personalized three-dimensional heart model constructed based on the user's heart scan image; there is no single limitation.

[0060] Step S130: Based on the distribution of electrical excitation intensity on the surface of the heart at each target time, render a three-dimensional model of the heart to obtain the target three-dimensional model of the heart at each target time.

[0061] This step uses computer graphics rendering technology to transfer the potential prediction values ​​of each network node on the three-dimensional heart model obtained in step S120 to the three-dimensional heart model, thereby transforming them into a three-dimensional visual image (target three-dimensional heart model) that users can intuitively and efficiently understand.

[0062] In one possible implementation, step S130 includes: determining the RGB value corresponding to the estimated potential value of each network node based on the pre-defined correspondence between the potential value range and the RGB value; and rendering the three-dimensional heart model based on the RGB values ​​corresponding to the estimated potential values ​​of all network nodes at each target time to obtain the target three-dimensional heart model at each target time.

[0063] During the preprocessing stage, the system will pre-set one or more potential value ranges and correspondence tables with RGB color values. For example, high positive potential is mapped to red, zero potential is mapped to black, and high negative potential is mapped to blue.

[0064] Furthermore, for each network node on the 3D model of the heart, based on the potential prediction at the target time, this potential prediction is compared with a preset potential value range to determine its corresponding range. Based on the matched range, the corresponding RGB value is retrieved from the correspondence table and applied to that node.

[0065] After assigning color values ​​to all nodes, the system starts the graphics rendering engine to render the 3D heart model. For example, color interpolation is performed between adjacent network nodes to ensure smooth and natural color transitions, avoiding harsh color blocks and creating a continuous color cloud across the entire heart model surface. Combined with computer graphics effects such as lighting, shadows, and transparency, the 3D heart model with color information is drawn. This is the final target 3D heart model.

[0066] Based on this, this method directly reproduces the electrical activity (such as premature beats, atrial fibrillation driving foci) at a specific anatomical location (such as the root of the left atrial appendage, the pulmonary vein orifice) at a target time on a patient's personalized heart model using color (such as red spots) according to the user's needs. This provides users with a visualization tool to understand complex electrocardiograms, thereby reducing the patient's examination costs and improving the doctor's diagnostic efficiency.

[0067] In summary, this application utilizes a potential prediction model to map the surface voltage feature vector at at least one moment in an electrocardiogram (ECG) to network nodes of a three-dimensional cardiac model, predicting the potential value of each network node. This achieves inverse reconstruction from surface signals to the distribution of electrical excitation intensity on the cardiac surface. Furthermore, based on the potential prediction value of each network node, a pre-constructed three-dimensional cardiac model is rendered in real time, generating a target three-dimensional cardiac model that dynamically displays the cardiac electrical excitation state. This allows users to intuitively observe the complete process of cardiac electrical excitation from its origin to conduction within the time period detected by the ECG, and, combined with the diagnostic results of the ECG, determine the location of abnormal electrical excitation in the heart.

[0068] Next, we will describe in detail other possible implementations of the ECG signal visualization method provided in this application.

[0069] In one possible design, the ECG signal visualization method may further include: when the number of target times is greater than 1, performing frequency domain analysis on the potential prediction of each network node within the time period to which all target times belong, based on the distribution of cardiac surface electrical excitation intensity corresponding to all target times, to obtain the excitation frequency of each network node; and rendering a pre-constructed three-dimensional model of the heart based on the excitation frequency of each network node to obtain a three-dimensional model characterizing the distribution of cardiac surface excitation frequency.

[0070] In the electrophysiological analysis of atrial fibrillation, the DF value (Dominant Frequency) is the activation frequency of the atrial myocardium per unit time. For details, refer to... Figure 3 This study analyzes the excitation intensity of cardiac surface points at each moment during the entire time period or a specific time period (e.g., t1-t2) of an electrocardiogram. It then counts the frequency at which the potential value of each network node exceeds a preset potential value within that time period. This preset potential value can be determined based on the myocardial tissue electrical excitation threshold corresponding to that network node. Based on this, the excitation frequency (DF value) of each network node is determined.

[0071] Furthermore, based on the DF value corresponding to each network node, the three-dimensional model of the heart is visualized and rendered. Optionally, the DF value is mapped to the network nodes of the three-dimensional model of the heart, and the magnitude of the DF value is associated with the color depth according to a preset color (such as red) mapping rule. The larger the DF value, the darker the color is assigned. The three-dimensional model of the heart is rendered using computer graphics processing technology, and finally a heat map that intuitively reflects the differences in excitation frequency in different areas of the heart is generated.

[0072] Heat maps visually represent the distribution of excitation frequencies in different areas of the heart using varying shades of color. This allows users to quickly identify areas with abnormally high excitation frequencies, which are often the key lesions in atrial fibrillation. This helps to accurately locate the lesion and provides an important basis for developing personalized treatment plans.

[0073] In another possible implementation, the above method further includes: when the number of target times is 1, displaying a target heart 3D model representing the cardiac electrical excitation state at the target time; when the number of target times is greater than 1, performing time-series animation rendering on the target heart 3D models corresponding to all target times to obtain a 3D dynamic demonstration animation representing the cardiac electrical excitation state within the time period of all target times, and displaying it.

[0074] When a user selects only one target moment (e.g., the peak of an atrial premature beat), the system will execute steps S110-S130 only for that target moment, ultimately generating and displaying a single-frame, high-resolution 3D model of the target heart. This model, through color mapping, accurately presents the spatial distribution of electrical excitation intensity across the entire surface of the heart at that instant.

[0075] When a user selects multiple consecutive moments (such as a complete cardiac cycle or a sustained atrial fibrillation segment), the system generates a corresponding 3D model of the target heart for each selected moment. Then, using time-series animation rendering technology, these single-frame target heart 3D models are used as keyframes, with transition frames inserted between them to ensure smooth and continuous movement of the electrical impulse wave, conforming to physiological propagation patterns. Finally, all frames are synthesized into a 3D dynamic demonstration animation that can be looped or played back frame by frame.

[0076] Understandably, the target heart 3D model is a static snapshot of a single target moment, while the 3D dynamic demonstration animation is a higher-level data organization and presentation built upon this, providing a global view. Doctors can first browse the dynamic animation to identify suspicious areas or abnormal patterns of agitation. Then, pausing at a keyframe (i.e., switching to the target heart 3D model at a single target moment) allows for depth measurement and precise localization of that instant.

[0077] An electrocardiogram (ECG) records a sequence of electrical signals showing the changes in cardiac electrical activity over time. At each moment, a target 3D cardiac model is constructed using inverse reconstruction techniques, mapping potential data collected from electrodes on the body surface or within the cardiac cavity onto the surface of the anatomical structures of the heart, forming a potential distribution field at that moment. Once a sequence of 3D models at consecutive time points is built, not only can the dynamic changes in potential on the cardiac surface be observed (such as conduction direction, velocity, and areas of abnormal potential accumulation), but also key features such as the origin, propagation path, and reentry loops of electrical excitation waves can be captured through temporal analysis. This spatiotemporal joint analysis transforms the state of electrical activity from abstract waveform interpretation into a concrete evolution of potential patterns on the cardiac geometry, thus providing direct and quantitative evidence for cardiac diagnosis (such as locating the origin of premature ventricular contractions and identifying the driving foci of atrial fibrillation).

[0078] Based on this, the electrical activity state of the target heart in the electrocardiogram (ECG) can be determined and output using the target heart 3D model corresponding to all time points in the ECG. Specifically, this can be achieved by calculating the point excitation propagation velocity and direction based on the potential values ​​of each network node in the target heart 3D model at all time points, determining the target heart electrical activity state based on the calculation results, and then outputting it. Alternatively, a temporal graph neural network can be used to analyze the sequence of points on the heart surface and the anatomical topology in the target heart 3D model to predict the electrical excitation state of the heart.

[0079] The ECG signal visualization method provided in this application can also guide the training or output of other models. In one application scenario, ECGs can usually be directly input into an ECG recognition model to identify abnormal cardiac activity states (such as atrial fibrillation, sinus rhythm, etc.). However, ECG recognition models are difficult to adapt to long-tail problems. Therefore, the ECG signal visualization method can be used to verify the recognition results of the ECG recognition model.

[0080] Specifically, based on the target heart 3D model corresponding to all moments in the electrocardiogram (ECG), the first cardiac electrical activity state of the ECG is determined; if the second cardiac electrical activity state corresponding to the ECG is different from the first cardiac electrical activity state, the second cardiac electrical activity state is corrected to the first cardiac electrical activity state, and the ECG labeled with the first cardiac electrical activity state is used as sample data to update the ECG recognition model.

[0081] Understandably, long-tail data, due to its small sample size and complex features, is easily influenced by subjective experience or visual errors during the annotation process (such as mislabeling ventricular premature beats as atrial premature beats), leading to the accumulation of label noise. During model training, noisy labels can mislead the gradient update direction, causing the decision boundary to deviate from the true data distribution. Therefore, this application uses this scheme to verify the labels of the sample data before training the ECG recognition model based on ECG sample data labeled with the first cardiac electrical activity state.

[0082] Specifically, according to steps S110-S130, by reversibly reconstructing the distribution of cardiac surface potential and tracking its dynamic evolution, key features such as the origin of electrical excitation (e.g., the location of ectopic pacemakers) and conduction pathways (e.g., reentry circuits) can be quantified, forming a more accurate "description of electrical activity state" than traditional electrocardiogram waveforms.

[0083] Furthermore, the first cardiac electrical activity state derived from the dynamic three-dimensional cardiac model is used to verify the manually labeled second cardiac electrical activity state. If the two are inconsistent, it proves that there is an error in the labeling of the ECG sample. The label of the ECG sample is then corrected from the second cardiac electrical activity state to the first cardiac electrical activity state, overwriting the original label. The corrected sample data is then used to advance the subsequent update and training of the ECG recognition model.

[0084] In another possible implementation, the ECG signal visualization method provided in this application can run in a dual-threaded manner with the ECG recognition model, simultaneously processing the same ECG uploaded by the user to obtain a first cardiac electrical activity state output by the ECG signal visualization method and a second cardiac electrical activity state output by the ECG recognition model.

[0085] This application considers ECG signal visualization methods and achieves physical-level analysis of key features such as the origin of electrical excitation and conduction pathways through reconstruction of the three-dimensional potential field of the heart. The results have higher reliability; therefore, if the first cardiac electrical activity state is inconsistent with the second cardiac electrical activity state, it proves that the ECG recognition model has misidentified the ECG. The first cardiac electrical activity state output by the visualization method is used as a label for the ECG, and the labeled ECG is included in the model training set as an adversarial negative sample to participate in the iterative training of the ECG recognition model. This strengthens the model's ability to identify easily confused cases and long-tailed data, thereby improving the accuracy of diagnosis and treatment.

[0086] Optionally, if the ECG recognition model makes an error in recognizing an ECG, the user can provide feedback to the system via their terminal. The system will then record this ECG as a difficult sample and label it based on the correct recognition result provided by the user. Similarly, the ECG with the corrected label "Second Cardiac Electrical Activity State" can also be recorded as a difficult sample. These difficult samples will be included in the difficult sample pool for subsequent model retraining and evaluation.

[0087] In this embodiment, when the retraining conditions for the ECG recognition model are triggered (e.g., the number of samples in the hard sample pool accumulates to a certain amount), the newly retrained model (ModelV2) will not immediately replace the running old model (ModelV1). At this time, the system enters "shadow mode," meaning the new model performs inference operations on real-time data in the background, but its inference results are not displayed to the user. Simultaneously, the system automatically compares the output differences between ModelV1 and ModelV2 to evaluate the performance of the new model.

[0088] In this embodiment, to ensure the safety of replacing the old model with the new model, a safety switching threshold can be set for the performance of the new model. For example, the system will only automatically complete the hot update and put the new model into practical use when the accuracy of the new model in identifying difficult samples in the "difficult sample pool" is significantly better than that of the old model, and the volatility on regular samples is lower than the preset safety threshold.

[0089] If ModelV2 performs well and meets the safety switching conditions, it will replace ModelV1 as the new front-end model, while ModelV1 will become a historical version, serving as a baseline for future comparisons. When the next update arrives, ModelV3 will be developed, and the system will compare ModelV3 with the current front-end model, ModelV2. If ModelV3 performs better, it will replace ModelV2 as the new front-end model. This process continues, evolving from ModelV1 to ModelV2 to ModelV3 to ... to ModelVn. Unless a major bug is discovered after ModelV2 is launched, requiring a "rollback," the system will not revert to using ModelV1 under normal circumstances.

[0090] In the aforementioned shadow mode, the new model runs in the background alongside the old model. The inference results of the new model do not directly affect business operations, avoiding direct risks caused by problems with the new model and ensuring stable system operation. Even if abnormal output is detected in the new model, the problem can be quickly located, allowing for debugging and repair of the new model without impacting actual business operations, reducing the possibility of the problem escalating. Once the new model has been fully validated in shadow mode and its performance meets or exceeds expected standards, it is smoothly switched to the foreground, replacing the old model. This seamless switching method does not have a significant impact on users, ensuring business continuity and stability.

[0091] In another possible implementation, a time-series analysis approach is employed to visualize the identification results of the cardiac electrical state at each time interval of the electrocardiogram (ECG). Taking an ECG containing 24 hours of surface voltage data as an example, firstly, using an ECG recognition model or ECG signal visualization method, the cardiac electrical activity state, such as atrial fibrillation, is determined for each scanning time interval (30 minutes or 1 hour, etc.) using a sliding window scanning technique. Further, the proportion of the total duration of atrial fibrillation episodes in the overall duration is statistically analyzed to obtain the load value. An atrial fibrillation load clock is generated based on the time intervals of atrial fibrillation episodes and the load values. (Refer to...) Figure 4 The present application provides an example diagram of an atrial fibrillation load clock, which visually presents the time periods of atrial fibrillation in the form of a 24-hour disc, making it easier for users to determine the concentrated time periods of atrial fibrillation.

[0092] Optionally, risk levels can be classified according to preset load standards (such as <5%, 5-10%, >10%) to determine the risk level of the electrocardiogram, triggering an alert for the user and automatically providing doctors with clinical intervention strategy suggestions.

[0093] This application also provides an electronic device in its embodiments. (See reference...) Figure 5The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, tablets, large-screen teaching displays, wearable devices, etc. Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0094] like Figure 5 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 1, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 2 or a program loaded from a storage device 8 into a random access memory (RAM) 3, to implement the electrocardiogram signal visualization method of the foregoing embodiments of this application. When the electronic device is powered on, the RAM 3 also stores various programs and data required for the operation of the electronic device. The processing unit 1, ROM 2, and RAM 3 are interconnected via a bus 4. An input / output (I / O) interface 5 is also connected to the bus 4.

[0095] Typically, the following devices can be connected to I / O interface 5: input devices 6 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 7 including, for example, liquid crystal display terminals (LCDs), speakers, vibrators, etc.; storage devices 8 including, for example, memory cards, hard drives, etc.; and communication devices 9. Communication device 9 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0096] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the electrocardiogram signal visualization methods provided in this application.

[0097] This application also provides a computer-readable storage medium that carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the electrocardiogram signal visualization methods provided in this application.

[0098] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the accompanying drawings of the device embodiments provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0099] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods of the various embodiments of this application.

[0100] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0101] A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0102] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

Claims

1. A method for visualizing electrocardiogram (ECG) signals, characterized in that, include: Obtain at least one target time point surface voltage feature vector from the electrocardiogram (ECG). The surface voltage feature vector is the set of voltage values ​​of all waveforms in the ECG at the current time. The configured potential prediction model is invoked to process each surface voltage feature vector to obtain the heart surface electrical excitation intensity distribution at each target time. The heart surface electrical excitation intensity distribution is the potential prediction value of each network node on the pre-constructed three-dimensional heart model. The potential prediction model is trained based on the sample data of surface voltage feature vectors labeled with the heart surface electrical excitation intensity distribution. Based on the distribution of electrical excitation intensity on the surface of the heart at each target time, a three-dimensional model of the heart is rendered to obtain the target three-dimensional model of the heart corresponding to each target time.

2. The ECG signal visualization method according to claim 1, characterized in that, The process of rendering a three-dimensional model of the heart based on the electrical excitation intensity distribution on the surface of the heart at each target time, to obtain the target three-dimensional model of the heart at each target time, includes: Based on the pre-defined correspondence between potential value ranges and RGB values, determine the RGB value corresponding to the potential estimate of each network node; Based on the RGB values ​​corresponding to the estimated potential values ​​of all network nodes at each target time, the three-dimensional model of the heart is rendered to obtain the target three-dimensional model of the heart at each target time.

3. The ECG signal visualization method according to claim 1, characterized in that, The method further includes: When the number of target times is greater than 1, based on the cardiac surface electrical excitation intensity distribution corresponding to all target times, frequency domain analysis is performed on the potential prediction of each network node within the time period to which all target times belong, to obtain the excitation frequency of each network node. Based on the excitation frequency of each network node, the pre-constructed three-dimensional model of the heart is rendered to obtain a three-dimensional model characterizing the excitation frequency distribution on the surface of the heart.

4. The ECG signal visualization method according to claim 1, characterized in that, The method also includes: Based on the target heart three-dimensional model corresponding to all moments in the electrocardiogram, the first cardiac electrical activity state of the electrocardiogram is determined. If the second cardiac electrical activity state labeled on the electrocardiogram is different from the first cardiac electrical activity state, the second cardiac electrical activity state is corrected to the first cardiac electrical activity state, and the electrocardiogram labeled with the first cardiac electrical activity state is used as sample data to update the electrocardiogram recognition model.

5. The ECG signal visualization method according to claim 1, characterized in that, Also includes: Based on the target heart 3D model corresponding to all moments in the electrocardiogram, the target heart electrical activity state of the electrocardiogram is determined and output.

6. The method for visualizing electrocardiogram signals according to any one of claims 1-5, characterized in that, Before obtaining the surface voltage feature vector at least at a target time from the electrocardiogram, the process also includes: In response to a user's selection of at least one moment in the electrocardiogram, each selected moment is taken as the target moment.

7. The method for visualizing electrocardiogram signals according to any one of claims 1-5, characterized in that, The method also includes: When the number of target times is 1, a three-dimensional model of the target heart representing the electrical excitation state of the heart at the target time is displayed. If the number of target times is greater than 1, perform time-series animation rendering on the target heart 3D model corresponding to all target times to obtain a 3D dynamic demonstration animation representing the cardiac electrical excitation state within the time period of all target times, and then display it.

8. A system for visualizing electrocardiogram (ECG) signals, characterized in that, include: Processor and display terminal; The display terminal is used to display electrocardiograms and three-dimensional models of the heart, and to capture the user's selection operation at at least one moment in the electrocardiogram and transmit the selection operation to the processor. The processor, in response to a selection operation, acquires a body surface voltage feature vector at least one selected moment from the electrocardiogram, and processes each body surface voltage feature vector according to any one of claims 1-7 of the electrocardiogram signal visualization method to obtain a target three-dimensional model of the heart at each target moment, and transmits the target three-dimensional model of the heart to a display terminal.

9. An electronic device, characterized in that, include: Memory and processor; Memory, used to store programs; A processor for executing a program to implement the steps of the ECG signal visualization method as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the various steps of the electrocardiogram signal visualization method as described in any one of claims 1-7.

Images