A method, device, electronic device, and storage medium for determining electrocardiogram risk.

Abstract

This application provides a method, device, electronic device, and storage medium for judging electrocardiogram (ECG) risk, relating to the field of medical data processing technology. The method includes acquiring ECG and motion signals uploaded by a wearable device on a user within the current time period; determining motion time-series data at a second preset sampling frequency based on the motion signals, and determining the user's motion state type within the current time period; processing the ECG signals using an artifact removal scheme corresponding to the motion state type to obtain ECG time-series data at a first preset sampling frequency; inputting the ECG time-series data, motion time-series data, and motion state type into a lightweight risk judgment model to obtain a first risk judgment result output by the lightweight risk judgment model, the first risk judgment result including the probability value of a preset risk level existing within the user's current time period; and implementing corresponding intervention measures for the user based on the first risk judgment result.

Description

Technical Field

[0001] This application relates to the field of medical data processing technology, and more specifically, to a method, apparatus, electronic device, and storage medium for determining electrocardiogram risk. Background Technology

[0002] With the rapid development of medical and health monitoring technologies, electrocardiogram (ECG) testing has become an important means of screening for cardiovascular diseases and managing daily health.

[0003] In the field of civilian health monitoring, existing portable electrocardiogram (ECG) monitoring devices mainly fall into two categories: one is a simplified medical version of the ECG monitor, which still requires multiple electrodes to be fixed via a chest strap; the other is a single-lead handheld device, which requires both hands to hold or to be held tightly against the chest. Both of these devices have strict requirements on the user's posture during operation, requiring them to remain still to avoid electromyographic interference and motion artifacts. This limits the application scenarios of ECG monitoring devices and prevents them from achieving 24 / 7 ECG monitoring. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, electronic device and storage medium for judging electrocardiogram risk, so as to solve the technical problem that the prior art cannot achieve all-weather electrocardiogram monitoring.

[0005] In a first aspect, the present invention provides a method for determining electrocardiogram (ECG) risk. The method includes acquiring ECG signals and motion signals uploaded by a wearable device on a user within the current time period; determining motion time-series data at a second preset sampling frequency based on the motion signals, and determining the user's motion state type within the current time period; processing the ECG signals using an artifact removal scheme corresponding to the motion state type to obtain ECG time-series data at a first preset sampling frequency; inputting the ECG time-series data, motion time-series data, and motion state type into a lightweight risk discrimination model to obtain a first risk discrimination result output by the lightweight risk discrimination model, the first risk discrimination result including the probability value of the user having a preset risk level within the current time period; and implementing corresponding intervention measures for the user based on the first risk discrimination result.

[0006] In an optional implementation, the method further includes acquiring the ambient noise signal uploaded by the wearable device;

[0007] The environmental noise level is determined based on the environmental noise signal;

[0008] Input ECG time series data, exercise time series data, exercise state type and environmental noise level value into the lightweight risk discrimination model to obtain the first risk discrimination result output by the lightweight risk discrimination model.

[0009] In an optional implementation, the complete motion timing data is determined based on the motion signal;

[0010] Electrocardiogram time series data and motion time series data are input into a CNN-LSTM hybrid model to obtain the second risk discrimination result output by the CNN-LSTM hybrid model;

[0011] Based on the results of the first and second risk assessments, a final risk assessment result is determined in order to implement corresponding intervention measures for the user.

[0012] In an optional implementation, the motion timing data includes X-axis acceleration, Y-axis acceleration, and Z-axis acceleration corresponding to multiple sampling times, and the motion state type is determined by the following method:

[0013] For each sampling time, the resultant acceleration corresponding to that sampling time is calculated based on the X-axis acceleration, Y-axis acceleration and Z-axis acceleration corresponding to that sampling time, forming the resultant acceleration time series data;

[0014] Input the combined acceleration time series data within the current time period and the combined acceleration time series data within the historical time periods into the pre-trained decision tree model to obtain the motion state type output by the decision tree model.

[0015] In an optional implementation, the motion state type includes at least rest, slow movement, walking, running, and postural changes. When the motion state type is rest, slow movement, walking, or running, the following steps are used to remove artifacts from the electrocardiogram signal:

[0016] Based on the motion intensity of the motion state type, the filtering parameters of the adaptive filter are determined;

[0017] The ECG signal input is filtered by an adaptive filter corresponding to the determined filtering parameters to output an ECG signal after artifact removal.

[0018] In an optional implementation, motion time series data and the user's corresponding life data are input into a CNN-LSTM hybrid model to obtain a second risk discrimination result output by the CNN-LSTM hybrid model;

[0019] Based on the results of the first and second risk assessments, a final risk assessment result is determined in order to implement corresponding intervention measures for the user.

[0020] In an optional implementation, the corresponding intervention measures are performed in the following manner:

[0021] When the risk assessment result is low risk, a corresponding health log is generated for the user on the APP and uploaded to the cloud at the same time;

[0022] When the risk assessment result is medium risk, a short vibration action is executed on the device to generate a health reminder message on the APP and push it to the user;

[0023] When the risk assessment result is high risk, the device will perform a long vibration and flashing light action, generate a warning message on the APP, and send a warning message to the pre-bound emergency contacts.

[0024] Secondly, the present invention provides a device for determining electrocardiographic risk, the device comprising:

[0025] The acquisition module is used to acquire electrocardiogram (ECG) signals and motion signals uploaded by the wearable device on the user within the current time period.

[0026] The sampling module is used to determine the motion time sequence data at a second preset sampling frequency based on the motion signal, and to determine the type of motion state of the user in the current time period;

[0027] The noise reduction module is used to process the electrocardiogram (ECG) signal using an artifact removal scheme corresponding to the motion state type, so as to obtain ECG timing data at a first preset sampling frequency.

[0028] The discrimination module is used to input ECG time series data, exercise time series data, and exercise state type into the lightweight risk discrimination model to obtain the first risk discrimination result output by the lightweight risk discrimination model. The first risk discrimination result includes the probability value of the user's current time period containing a preset risk level.

[0029] The intervention module is used to implement corresponding intervention measures for users based on the first risk assessment result.

[0030] Thirdly, the present invention provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus, and the processor executes the machine-readable instructions to perform the steps of any of the ECG risk determination methods described in the foregoing embodiments.

[0031] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of any of the electrocardiogram risk determination methods described in the foregoing embodiments.

[0032] This application provides a method, device, electronic device, and storage medium for judging electrocardiogram (ECG) risk. The method includes acquiring ECG and motion signals uploaded by a wearable device on a user within the current time period; determining motion time-series data at a second preset sampling frequency based on the motion signals, and determining the user's motion state type within the current time period; processing the ECG signals using an artifact removal scheme corresponding to the motion state type to obtain ECG time-series data at a first preset sampling frequency; inputting the ECG time-series data, motion time-series data, and motion state type into a lightweight risk judgment model to obtain a first risk judgment result output by the lightweight risk judgment model, the first risk judgment result including the probability value of a preset risk level existing within the user's current time period; and executing corresponding intervention measures for the user based on the first risk judgment result. By fusing continuous ECG waveforms with multimodal information such as motion and body position, all-weather, individualized ECG risk judgment and intelligent intervention are achieved. It can acquire high-quality ECG signals without requiring the user to remain still, and can identify transient abnormalities and exercise-induced risks in real time, and generate graded and actionable intervention suggestions, which can significantly improve the timeliness of cardiac abnormality warnings, the accuracy of interventions, and the sustainability of individualized health management. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 A flowchart illustrating the steps of a method for determining electrocardiogram risk provided in this application embodiment;

[0035] Figure 2 A flowchart illustrating the steps of another method for determining electrocardiogram risk provided in this application embodiment;

[0036] Figure 3 This application also provides a schematic diagram of the structure of a device for determining electrocardiogram risk in an embodiment of the present application;

[0037] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0038] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0039] Example 1

[0040] In one embodiment of this application, a system for identifying and intervening in electrocardiogram (ECG) risks is provided, comprising a device, an app, and a cloud platform. The data acquisition unit can be deployed in a wearable device. This wearable device integrates an ECG sensor, an accelerometer, a microphone, etc.

[0041] The ECG sensor here can be a three-dry electrode, model ADS1292R, to collect single-lead (lead II) or three-lead ECG signals. The accelerometer sensor can be a triaxial accelerometer (±8g range) paired with a gyroscope, model MPU6050, for collecting motion signals. The microphone is used to collect ambient sound signals to identify electromagnetic interference.

[0042] Figure 1 This is a flowchart illustrating the steps of a method for determining electrocardiogram (ECG) risk, as provided in an embodiment of this application. Figure 1 As shown, in one feasible implementation, a method for determining electrocardiogram risk is provided, the method comprising:

[0043] S10: The APP acquires the ECG and motion signals uploaded by the wearable device (device) on the user's body within the current time period, as well as the environmental noise signals uploaded by the wearable device.

[0044] This also includes determining the environmental noise level based on the environmental noise signal.

[0045] S11. The APP determines the motion timing data of the second preset sampling frequency based on the motion signal, and determines the type of motion state of the user in the current time period.

[0046] In step S11, the second preset sampling frequency can be 250Hz.

[0047] Motion time series data includes X-axis acceleration, Y-axis acceleration, and Z-axis acceleration corresponding to multiple sampling times. The motion state type is determined by the following method:

[0048] For each sampling moment, the resultant acceleration is calculated based on the X-axis, Y-axis, and Z-axis accelerations corresponding to that sampling moment, forming the resultant acceleration time series data. The resultant acceleration time series data within the current time period and the resultant acceleration time series data within historical time periods are input into a pre-trained decision tree model to obtain the motion state type output by the decision tree model.

[0049] The resultant acceleration Vm can be calculated using the following formula:

[0050] Vm = √(Vx² +Vy² + Vz²) - 1g;

[0051] Where Vx, Vy, and Vz are the X-axis acceleration, Y-axis acceleration, and Z-axis acceleration, respectively.

[0052] The historical time periods here are preceding the current time period and are continuous. For the resultant acceleration time series data with a preset step size as the window, features such as resultant acceleration mean, resultant acceleration standard deviation, resultant acceleration peak value, resultant acceleration RMS, zero crossing rate, peak-to-valley difference, dominant frequency, low-frequency energy ratio (0.5-2Hz), mid-frequency energy ratio (2-5Hz), frequency domain entropy, X-axis contribution, Y-axis contribution, Z-axis contribution, XY-axis correlation, XZ-axis correlation, pitch angle, roll angle, body position type (coded as 0-5), motion direction stability, periodicity index (autocorrelation peak value), and current body position duration can be extracted.

[0053] Based on the above characteristics, the decision tree can first perform a primary classification to determine the activity level. It determines whether the mean resultant acceleration is less than 0.05; if so, it is a static activity. It then determines whether the standard deviation of the roll angle is greater than 5°; if so, it is a postural change. Otherwise, it proceeds to static subdivision. Static subdivision can be based on pitch angle, further subdividing into standing, sitting, supine, and lateral recumbent positions.

[0054] If the mean resultant acceleration is not less than 0.05, it is a dynamic subdivision. Determine if the periodicity index is greater than 0.7. If it is, it enters the periodic motion classification; if not, it is determined to be a non-periodic activity.

[0055] Periodic activities can be determined based on the dominant frequency. When the dominant frequency is low (1-1.8Hz), the type of exercise is walking; when the dominant frequency is medium (1.8-2.5Hz), the type of exercise is brisk walking; and when the dominant frequency is high (2.5-4Hz), the type of exercise is running.

[0056] Finally, the results can be further verified based on other features.

[0057] Other deep learning models or matching algorithms can also be used to classify motion state types; there are no restrictions.

[0058] S12. The APP uses an artifact removal scheme corresponding to the motion state type to process the ECG signal to obtain ECG timing data at the first preset sampling frequency.

[0059] In step S12, different artifact removal schemes can be selected based on different motion state types to more accurately remove artifact noise from the electrocardiogram (ECG) signal. Simultaneously, environmental noise can also be removed from the ECG signal.

[0060] In one feasible implementation, baseline filtering based on morphology can be performed when the signal is stationary, while RLS adaptive filtering can be used when walking or running. The filtered signal is then decomposed into wavelet components to remove approximation coefficients and retain detail coefficients to reconstruct the signal. Finally, a 50Hz / 60Hz adaptive notch filter is used, with the bandwidth adjusted according to the ambient noise level.

[0061] In one feasible implementation, the types of motion states include at least stationary, slow movement, walking, running, and positional changes.

[0062] When the type of motion is stationary, slow movement, walking, or running, remove artifacts from the electrocardiogram signal using the following steps:

[0063] Based on the motion intensity of the motion state type, the filtering parameters of the adaptive filter are determined. The ECG signal is input and filtered by the adaptive filter corresponding to the determined filtering parameters to output an ECG signal with artifacts removed.

[0064] The filtering parameters here may include, but are not limited to, the filter's frequency, bandwidth, etc.

[0065] The first preset frequency here can be 250Hz. After removing noise, the ECG signal is sampled to generate ECG timing data at multiple ECG sampling points.

[0066] The app can also acquire environmental noise signals uploaded by wearable devices and determine the environmental noise level based on these signals.

[0067] S13. The APP inputs ECG time series data, exercise time series data, exercise state type and environmental noise level value into the lightweight risk discrimination model to obtain the first risk discrimination result output by the lightweight risk discrimination model.

[0068] The first risk assessment result includes the probability value of the user's presence at a preset risk level within the current time period.

[0069] In step S13, preliminary risk assessment can be performed via an app located closer to the user. To further improve timeliness, a lightweight model is used, such as a Lightweight Gradient Boosting Machine (LightGBM) or a Random Forest. Preset risk levels can include low risk, medium risk, and high risk. Low risk indicates occasional premature beats and mild baseline drift. Medium risk indicates frequent premature beats and slight ST-segment depression. High risk indicates persistent ventricular tachycardia and significant ST-segment elevation.

[0070] The input data to the model can be sampled data within the current time period, or it can include historical data for a certain duration based on the motion state type to ensure recognition accuracy.

[0071] In one specific embodiment, the lightweight risk discrimination model can extract ECG microstructural features, exercise load features, and physiological state features, respectively. ECG microstructural features can include RR interval variability (SDNN, RMSSD), QT interval, ST segment slope, P wave width, and waveform morphology features. Exercise load features can include metabolic equivalents (METs), the deviation of percentage heart rate reserve (%HRR) from exercise intensity, etc. Physiological state features can include individualized resting heart rate, nocturnal heart rate variability trends, and real-time body position.

[0072] S14. Based on the first risk assessment result, the APP will implement corresponding intervention measures for the user.

[0073] In step S14, the corresponding intervention measures can be implemented as follows: When the risk assessment result is low risk, a health log corresponding to the user is generated on the APP and uploaded to the cloud. When the risk assessment result is medium risk, a short vibration is executed on the device, a health reminder message is generated on the APP and pushed to the user. When the risk assessment result is high risk, a long vibration and light flashing are executed on the device, a warning message is generated on the APP, and a warning message is sent to the pre-bound emergency contact.

[0074] The present application provides a method for judging ECG risk by integrating the identification of exercise type with ECG risk judgment, thereby realizing the perception of the user environment. This not only removes the limitations of ECG signal acquisition scenarios, but also improves the accuracy of ECG risk judgment.

[0075] Example 2

[0076] Figure 2 A flowchart illustrating the steps of another method for determining electrocardiogram risk provided in this application embodiment. Figure 2 As shown, in one embodiment of this application, a method for determining electrocardiogram risk is provided, applicable to both APP and cloud platforms. The method includes:

[0077] S20: The APP acquires the ECG and motion signals uploaded by the wearable device on the user's body within the current time period, as well as the environmental noise signals uploaded by the wearable device.

[0078] This also includes determining the environmental noise level based on the environmental noise signal.

[0079] S21. Based on the motion signal, the APP determines the motion time sequence data of the second preset sampling frequency and determines the motion state type of the user in the current time period.

[0080] S22. The APP uses an artifact removal scheme corresponding to the motion state type to process the electrocardiogram signal to obtain electrocardiogram timing data at the first preset sampling frequency.

[0081] S23. The APP inputs ECG time series data, exercise time series data, exercise state type and environmental noise level value into the lightweight risk discrimination model to obtain the first risk discrimination result output by the lightweight risk discrimination model.

[0082] In a feasible embodiment, hardware triggering can be used to ensure microsecond-level time alignment between ECG and motion signals. Data is packaged into a data frame every 250ms, containing 62 ECG sampling points (250Hz × 0.25s), 25 motion sampling points (100Hz × 0.25s), 1 motion state type identifier, and 1 ambient noise level value (threshold of 0-100 dB).

[0083] In one feasible implementation, Z-score standardization can be performed using the moving average and standard deviation of the user's historical data before sampling the ECG and cloud signals. The sampling duration can also be adjusted based on seasonal variations, such as nighttime heart rate.

[0084] In one feasible implementation, when the first risk assessment result is a high-risk type, complete data for the first 10 minutes and the last 5 minutes can be uploaded to the cloud.

[0085] S24. The cloud inputs ECG time series data, exercise time series data and the user's corresponding life data into the CNN-LSTM hybrid model to obtain the second risk discrimination result output by the CNN-LSTM hybrid model, and feeds it back to the APP to verify the first risk discrimination result.

[0086] S25. Based on the first risk assessment result and the second risk assessment result, the APP determines the final risk assessment result and implements corresponding intervention measures for the user.

[0087] In this embodiment, to further improve the accuracy of ECG interpretation, full-scale data recognition can be performed in the cloud, while lightweight data recognition can be performed via an app. Finally, the two risk assessment results are combined to determine the final risk assessment result for intervention. When the two analysis results differ, the second risk assessment result shall prevail.

[0088] Furthermore, CNN-LSTM hybrid models can be deployed in the cloud here.

[0089] In this embodiment, when the final risk assessment result is low risk, there is no real-time notification on the device, but the event is recorded in the health log on the app. The event is then accumulated and statistically analyzed in the cloud, and mentioned in the weekly report.

[0090] When the final risk assessment result is medium risk, the device will vibrate briefly (1 second). The app will send a push notification with messages such as "Occasional premature beats detected; suggestions for rest or deep breathing, and reminders to drink water," etc.

[0091] When the final risk assessment result is high risk, the device will vibrate continuously (for 3 seconds) and the LED red light will flash. The APP will display a full-screen warning interface and provide a voice prompt: "An abnormal heart rhythm has been detected. Please rest immediately." An alert message with location information will be automatically sent.

[0092] The cloud here can also store long-term data for each user, as well as be used for personalized updates and optimizations, and to fine-tune the model. For example, based on user feedback, the risk assessment threshold for the user can be adjusted, and personalized parameters such as the user's normal heart rate range and ST segment baseline value can be updated.

[0093] Example 3

[0094] like Figure 3 As shown, based on the same inventive concept, this application also provides an electrocardiogram risk determination device 30, which includes:

[0095] The acquisition module 310 is used to acquire the electrocardiogram signal and motion signal uploaded by the wearable device on the user within the current time period;

[0096] The sampling module 320 is used to determine the motion time sequence data of the second preset sampling frequency based on the motion signal, and to determine the motion state type of the user in the current time period;

[0097] The noise reduction module 330 is used to process the electrocardiogram signal through an artifact removal scheme corresponding to the motion state type to obtain electrocardiogram timing data at a first preset sampling frequency;

[0098] The discrimination module 340 is used to input ECG time series data, exercise time series data, and exercise state type into the lightweight risk discrimination model to obtain the first risk discrimination result output by the lightweight risk discrimination model. The first risk discrimination result includes the probability value of the user's current time period containing a preset risk level.

[0099] The intervention module 350 is used to implement corresponding intervention measures for users based on the first risk assessment result.

[0100] In a preferred embodiment, the discrimination module 340 is further configured to acquire the environmental noise signal uploaded by the wearable device;

[0101] The environmental noise level is determined based on the environmental noise signal;

[0102] Input ECG time series data, exercise time series data, exercise state type and environmental noise level value into the lightweight risk discrimination model to obtain the first risk discrimination result output by the lightweight risk discrimination model.

[0103] In a preferred embodiment, the discrimination module 340 is further configured to determine complete motion timing data based on the motion signal;

[0104] Electrocardiogram time series data and motion time series data are input into a CNN-LSTM hybrid model to obtain the second risk discrimination result output by the CNN-LSTM hybrid model;

[0105] Based on the results of the first and second risk assessments, a final risk assessment result is determined in order to implement corresponding intervention measures for the user.

[0106] In a preferred embodiment, the motion timing data includes X-axis acceleration, Y-axis acceleration, and Z-axis acceleration corresponding to multiple sampling times. The sampling module 320 determines the motion state type in the following manner:

[0107] For each sampling time, the resultant acceleration corresponding to that sampling time is calculated based on the X-axis acceleration, Y-axis acceleration and Z-axis acceleration corresponding to that sampling time, forming the resultant acceleration time series data;

[0108] Input the combined acceleration time series data within the current time period and the combined acceleration time series data within the historical time periods into the pre-trained decision tree model to obtain the motion state type output by the decision tree model.

[0109] In a preferred embodiment, the motion state type includes at least stationary, slow movement, walking, running, and postural changes. When the motion state type is stationary, slow movement, walking, or running, the noise reduction module 330 removes artifacts from the electrocardiogram signal through the following steps:

[0110] Based on the motion intensity of the motion state type, the filtering parameters of the adaptive filter are determined;

[0111] The ECG signal input is filtered by an adaptive filter corresponding to the determined filtering parameters to output an ECG signal after artifact removal.

[0112] In a preferred embodiment, the discrimination module 340 is further configured to input motion time series data and user-corresponding life data into the CNN-LSTM hybrid model to obtain a second risk discrimination result output by the CNN-LSTM hybrid model;

[0113] Based on the results of the first and second risk assessments, a final risk assessment result is determined in order to implement corresponding intervention measures for the user.

[0114] In a preferred embodiment, the intervention module 350 performs the corresponding intervention measures in the following manner:

[0115] When the risk assessment result is low risk, a corresponding health log is generated for the user on the APP and uploaded to the cloud at the same time;

[0116] When the risk assessment result is medium risk, a short vibration action is executed on the device to generate a health reminder message on the APP and push it to the user;

[0117] When the risk assessment result is high risk, the device will perform a long vibration and flashing light action, generate a warning message on the APP, and send a warning message to the pre-bound emergency contacts.

[0118] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, and a bus 430.

[0119] The memory 420 stores machine-readable instructions that can be executed by the processor 410. When the electronic device 400 is running, the processor 410 and the memory 420 communicate via the bus 430. When the machine-readable instructions are executed by the processor 410, the steps of an ECG risk determination method as described in the above method embodiment can be executed. For specific implementation details, please refer to the method embodiment, which will not be repeated here.

[0120] This application also provides a computer-readable storage medium storing a computer program. When the computer program is run by a processor, it can execute the steps of an electrocardiogram risk determination method as described in the above method embodiments. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0121] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0122] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0123] Furthermore, the units described as separate components may or may not be physically separate. 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0124] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0125] It should be noted that if the function is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0126] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0127] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for determining electrocardiographic risk, characterized in that, The method includes: Acquire ECG and motion signals from wearable devices on the user within the current time period; Based on motion signals, motion time-series data with a second preset sampling frequency is determined, and the user's motion state type within the current time period is determined. The motion time-series data includes X-axis acceleration, Y-axis acceleration, and Z-axis acceleration corresponding to multiple sampling moments. The motion state type is determined as follows: For each sampling moment, the resultant acceleration corresponding to that sampling moment is calculated based on the X-axis acceleration, Y-axis acceleration, and Z-axis acceleration, forming resultant acceleration time-series data; the resultant acceleration time-series data within the current time period and the resultant acceleration time-series data within historical time periods are input into a pre-trained decision tree model to obtain the motion state type output by the decision tree model. The ECG signal is processed using an artifact removal scheme corresponding to the motion state type to obtain ECG timing data at a first preset sampling frequency. ECG time series data, exercise time series data, and exercise state type are input into a lightweight risk discrimination model to obtain a first risk discrimination result output by the lightweight risk discrimination model. The first risk discrimination result includes the probability value of the user's current time period containing a preset risk level. The motion time series data and the user's corresponding life data are input into the CNN-LSTM hybrid model to obtain the second risk discrimination result output by the CNN-LSTM hybrid model; Based on the results of the first and second risk assessments, the final risk assessment result is determined so that corresponding intervention measures can be implemented for the user. Specifically, when the motion type is stationary, slow movement, walking, or running, the following steps are used to remove artifacts from the electrocardiogram signal: Based on the motion intensity of the motion state type, the filtering parameters of the adaptive filter are determined; The ECG signal input is filtered by an adaptive filter corresponding to the determined filtering parameters to output an ECG signal after removing artifacts. The filtering parameters include the filter's frequency and bandwidth.

2. The method according to claim 1, characterized in that, It also includes acquiring environmental noise signals uploaded by wearable devices; The environmental noise level is determined based on the environmental noise signal; Input ECG time series data, exercise time series data, exercise state type and environmental noise level value into the lightweight risk discrimination model to obtain the first risk discrimination result output by the lightweight risk discrimination model.

3. The method according to claim 1, characterized in that, The corresponding intervention measures shall be implemented in the following ways: When the risk assessment result is low risk, a corresponding health log is generated for the user on the APP and uploaded to the cloud at the same time; When the risk assessment result is medium risk, a short vibration action is executed on the device to generate a health reminder message on the APP and push it to the user; When the risk assessment result is high risk, the device will perform a long vibration and flashing light action, generate a warning message on the APP, and send a warning message to the pre-bound emergency contacts.

4. A device for determining electrocardiographic risk, characterized in that, The apparatus applicable to the electrocardiogram risk assessment method according to any one of claims 1 to 3 comprises: The acquisition module is used to acquire electrocardiogram (ECG) signals and motion signals uploaded by the wearable device on the user within the current time period. The sampling module is used to determine the motion time sequence data at a second preset sampling frequency based on the motion signal, and to determine the type of motion state of the user in the current time period; The noise reduction module is used to process the electrocardiogram (ECG) signal using an artifact removal scheme corresponding to the motion state type, so as to obtain ECG timing data at a first preset sampling frequency. The discrimination module is used to input ECG time series data, exercise time series data, and exercise state type into the lightweight risk discrimination model to obtain the first risk discrimination result output by the lightweight risk discrimination model. The first risk discrimination result includes the probability value of the user's current time period containing a preset risk level. The intervention module is used to implement corresponding intervention measures for users based on the first risk assessment result.

5. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus, and the processor executes the machine-readable instructions to perform the steps of the ECG risk assessment method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the ECG risk assessment method as described in any one of claims 1 to 3.

Images