WO2021121226A1 - 一种心电信号的预测方法、装置、终端以及存储介质 - Google Patents

一种心电信号的预测方法、装置、终端以及存储介质 Download PDF

Info

Publication number
WO2021121226A1
WO2021121226A1 PCT/CN2020/136538 CN2020136538W WO2021121226A1 WO 2021121226 A1 WO2021121226 A1 WO 2021121226A1 CN 2020136538 W CN2020136538 W CN 2020136538W WO 2021121226 A1 WO2021121226 A1 WO 2021121226A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
atrial fibrillation
ecg
classification model
training
Prior art date
Application number
PCT/CN2020/136538
Other languages
English (en)
French (fr)
Inventor
李露平
陈茂林
韩羽佳
贾淼
郭光明
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to EP20903369.5A priority Critical patent/EP4062824A4/en
Priority to US17/757,671 priority patent/US20230036193A1/en
Publication of WO2021121226A1 publication Critical patent/WO2021121226A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A61B5/349—Detecting specific parameters of the electrocardiograph cycle
    • A61B5/361—Detecting fibrillation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A61B5/1101—Detecting tremor
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7221—Determining signal validity, reliability or quality
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor

Definitions

  • This application belongs to the technical field of vital signs recognition, and in particular relates to a method, device, terminal, and storage medium for predicting ECG signals.
  • Atrial fibrillation is the most common persistent chronic arrhythmia, often due to disordered atrial activity and irregular atrial compression. As people pay more attention to health and the incidence of atrial fibrillation gradually increases, how to identify whether a user has an atrial fibrillation event becomes particularly important.
  • the existing atrial fibrillation detection technology can only determine that the user is an atrial fibrillation patient when an atrial fibrillation event occurs. However, when an atrial fibrillation event occurs, the user's life safety may be endangered and the user's health may be affected. It can be seen that the existing atrial fibrillation detection technology can only inform the user whether an atrial fibrillation event has occurred, and cannot predict the user's atrial fibrillation event, and the detection effect is poor.
  • the embodiments of the present application provide an ECG signal prediction method, device, terminal, and storage medium, which can solve the problem that the existing atrial fibrillation detection technology cannot predict the user's atrial fibrillation event and the detection effect is poor.
  • an ECG signal prediction method including:
  • the risk degree of atrial fibrillation about to occur is calculated, and whether the target user is about to have atrial fibrillation is predicted.
  • the signal category of the ECG signal output by the atrial fibrillation signal classification model is obtained, where Before the atrial fibrillation signal classification model is obtained by training the atrial fibrillation patient as a model training sample, it further includes:
  • the training signal set consisting of a plurality of historical signals in continuous acquisition order; the training signal set includes at least one atrial fibrillation signal;
  • a preset native classification model is trained to obtain the atrial fibrillation signal classification model.
  • the training a preset native classification model through the risk signal of the training signal set to obtain the atrial fibrillation signal classification model includes:
  • Training is performed according to the signal characteristic parameters and the signal category to obtain the atrial fibrillation signal classification model.
  • the acquiring a training signal set composed of a plurality of ECG signals in sequential acquisition order includes:
  • all the effective information is encapsulated to obtain the training signal set.
  • said importing the ECG signal into a preset atrial fibrillation signal classification model to obtain the signal category of the ECG signal output by the atrial fibrillation signal classification model wherein:
  • the atrial fibrillation signal classification model is obtained by training atrial fibrillation patients as a model training sample, and includes:
  • the calculating the risk of atrial fibrillation about to occur according to the signal type of the ECG signal, and predicting whether the target user is about to have atrial fibrillation includes:
  • the risk level of the onset of atrial fibrillation is calculated according to the signal type of the ECG signal, and the risk of the occurrence of atrial fibrillation is predicted Whether the target user is about to have atrial fibrillation, also includes:
  • a path to a doctor is generated.
  • the method further includes : If a newly-added ECG signal is received, the newly-added signal category of the newly-added ECG signal is identified through the atrial fibrillation signal classification model;
  • an atrial fibrillation probability curve is generated.
  • an ECG signal prediction device including:
  • the ECG signal acquisition unit is used to acquire the ECG signal of the target user
  • the signal category identification unit is used to import the ECG signal into a preset atrial fibrillation signal classification model to obtain the signal category of the ECG signal output by the atrial fibrillation signal classification model, wherein the atrial fibrillation signal classification model is Trained with patients with atrial fibrillation as a model training sample;
  • the atrial fibrillation transmission probability calculation unit is used to calculate the risk of atrial fibrillation about to occur according to the signal type of the electrocardiogram signal, and predict whether the target user is about to have atrial fibrillation.
  • the embodiments of the present application provide a terminal device, a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the The computer program implements the ECG signal prediction method described in any one of the above-mentioned first aspects.
  • an embodiment of the present application provides a computer-readable storage medium that stores a computer program, and is characterized in that, when the computer program is executed by a processor, any of the above-mentioned aspects of the first aspect is implemented.
  • a method for predicting the ECG signal is described in detail below.
  • the embodiments of the present application provide a computer program product, which when the computer program product runs on a terminal device, causes the terminal device to execute the ECG signal prediction method described in any one of the above-mentioned first aspects.
  • the collected ECG signals are imported into the preset atrial fibrillation signal classification model to determine the signal category of each ECG signal, and the ECG signals of the user in multiple collection periods are obtained.
  • the probability of occurrence of atrial fibrillation of the target user through the signal types of multiple ECG signals, and according to the numerical value of the probability of occurrence of atrial fibrillation, the prediction of atrial fibrillation events can be realized, which is convenient for users to determine their own body Status, improve the detection effect and user experience.
  • FIG. 1 is a block diagram of a part of the structure of a mobile phone provided by an embodiment of the present application
  • FIG. 2 is a structural block diagram of an ECG signal prediction system provided by an embodiment of the present application.
  • Fig. 3 is a structural block diagram of an ECG signal prediction system provided by another embodiment of the present application.
  • FIG. 4 is an implementation flowchart of an ECG signal prediction method provided by the first embodiment of the present application.
  • FIG. 5 is a schematic diagram of the output of the probability of atrial fibrillation provided by an embodiment of the present application.
  • Fig. 6 is a schematic diagram of a comparison of relevant markers between atrial fibrillation patients and normal patients provided by an embodiment of the present application;
  • FIG. 7 is a specific implementation flowchart of an ECG signal prediction method provided by the second embodiment of the present application.
  • FIG. 8 is a schematic diagram of a training signal set provided by an embodiment of the present application.
  • FIG. 10 is a schematic diagram of marking of signal types provided by an embodiment of the present application.
  • FIG. 11 is a specific implementation flowchart of an ECG signal prediction method S4022 provided by the fourth embodiment of the present application.
  • FIG. 12 is a specific implementation flowchart of an ECG signal prediction method S402 provided by the fifth embodiment of the present application.
  • FIG. 13 is a specific implementation flowchart of an ECG signal prediction method S403 provided by the sixth embodiment of the present application.
  • FIG. 15 is a schematic diagram of outputting alarm information provided by an embodiment of the present application.
  • FIG. 16 is a schematic diagram of output of a visit path provided by an embodiment of the present application.
  • FIG. 17 is a specific implementation flowchart of an ECG signal prediction method provided by the eighth embodiment of the present application.
  • FIG. 18 is a schematic diagram of the output of atrial fibrillation probability curve provided by an embodiment of the present application.
  • FIG. 19 is a structural block diagram of an ECG signal prediction device provided by an embodiment of the present application.
  • FIG. 20 is a schematic diagram of a terminal device provided by another embodiment of the present application.
  • the term “if” can be construed as “when” or “once” or “in response to determination” or “in response to detecting “.
  • the phrase “if determined” or “if detected [described condition or event]” can be interpreted as meaning “once determined” or “in response to determination” or “once detected [described condition or event]” depending on the context ]” or “in response to detection of [condition or event described]”.
  • the ECG signal prediction method provided by the embodiments of this application can be applied to mobile phones, tablet computers, wearable devices, in-vehicle devices, augmented reality (AR)/virtual reality (VR) devices, notebook computers, and super Mobile personal computers (ultra-mobile personal computers, UMPC), netbooks, personal digital assistants (personal digital assistants, PDAs) and other terminal devices can also be applied to databases, servers, and service response systems based on terminal artificial intelligence.
  • This application is implemented The example does not impose any restrictions on the specific types of terminal equipment.
  • the terminal device may be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a wireless local loop (Wireless Local Loop, WLL) station, Personal Digital Assistant (PDA) devices, handheld devices with wireless communication capabilities, computing devices or other processing devices connected to wireless modems, computers, laptops, handheld communication devices, handheld computing devices, and /Or other devices used to communicate on the wireless system and next-generation communication systems, for example, mobile terminals in 5G networks or mobile terminals in the future evolved Public Land Mobile Network (PLMN) network, etc.
  • STAION, ST station
  • WLAN Wireless Local Loop
  • PDA Personal Digital Assistant
  • the wearable device can also be a general term for applying wearable technology to intelligently design daily wear and develop wearable devices, such as glasses, gloves, Watches, clothing and shoes, etc.
  • a wearable device is a portable device that is directly worn on the body or integrated into the user's clothes or accessories, and is attached to the user's body to collect the user's atrial fibrillation signal. Wearable devices are not only a kind of hardware device, but also realize powerful functions through software support, data interaction, and cloud interaction.
  • wearable smart devices include full-featured, large-sized, complete or partial functions that can be implemented without relying on smart phones, such as smart watches or smart glasses, and only focus on a certain type of application function, and need to be used in conjunction with other devices such as smart phones. , Such as all kinds of smart bracelets and smart jewelry for physical sign monitoring.
  • Fig. 1 shows a block diagram of a part of the structure of a mobile phone provided in an embodiment of the present application.
  • the mobile phone includes: a radio frequency (RF) circuit 110, a memory 120, an input unit 130, a display unit 140, a sensor 150, an audio circuit 160, a near field communication module 170, a processor 180, and a power supply 190.
  • RF radio frequency
  • FIG. 1 does not constitute a limitation on the mobile phone, and may include more or fewer components than those shown in the figure, or a combination of some components, or different component arrangements.
  • the RF circuit 110 can be used for receiving and sending signals during information transmission or communication. In particular, after receiving the downlink information of the base station, it is processed by the processor 180; in addition, the designed uplink data is sent to the base station.
  • the RF circuit includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (LNA), a duplexer, and the like.
  • the RF circuit 110 can also communicate with the network and other devices through wireless communication.
  • the above-mentioned wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile Communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (Code Division) Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE)), Email, Short Messaging Service (SMS), etc., through RF circuits 110 receives the user's atrial fibrillation signal fed back by other terminals.
  • GSM Global System of Mobile Communication
  • GPRS General Packet Radio Service
  • CDMA Code Division Multiple Access
  • WCDMA Wideband Code Division Multiple Access
  • LTE Long Term Evolution
  • Email Short Messaging Service
  • SMS Short Messaging Service
  • the memory 120 can be used to store software programs and modules.
  • the processor 180 executes various functional applications and data processing of the mobile phone by running the software programs and modules stored in the memory 120, such as storing the received atrial fibrillation signal in the memory 120 .
  • the memory 120 may mainly include a program storage area and a data storage area.
  • the program storage area may store an operating system, an application program required by at least one function (such as a sound playback function, an image playback function, etc.), etc.; Data created by the use of mobile phones (such as audio data, phone book, etc.), etc.
  • the memory 120 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.
  • the input unit 130 may be used to receive inputted numeric or character information, and generate key signal input related to user settings and function control of the mobile phone 100.
  • the input unit 130 may include a touch panel 131 and other input devices 132.
  • the touch panel 131 also known as a touch screen, can collect user touch operations on or near it (for example, the user uses any suitable objects or accessories such as fingers, stylus, etc.) on the touch panel 131 or near the touch panel 131. Operation), and drive the corresponding connection device according to the preset program.
  • the display unit 140 can be used to display information input by the user or information provided to the user and various menus of the mobile phone, such as outputting the received atrial fibrillation signal of the user, and outputting the identification result after the type of the atrial fibrillation signal is determined.
  • the display unit 140 may include a display panel 141.
  • the display panel 141 may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), etc.
  • the touch panel 131 can cover the display panel 141. When the touch panel 131 detects a touch operation on or near it, it transmits it to the processor 180 to determine the type of the touch event, and then the processor 180 responds to the touch event.
  • the type provides corresponding visual output on the display panel 141.
  • the touch panel 131 and the display panel 141 are used as two independent components to realize the input and input functions of the mobile phone, but in some embodiments, the touch panel 131 and the display panel 141 can be integrated. Realize the input and output functions of the mobile phone.
  • the mobile phone 100 may also include at least one sensor 150, such as a light sensor, a motion sensor, and other sensors.
  • the light sensor may include an ambient light sensor and a proximity sensor.
  • the ambient light sensor can adjust the brightness of the display panel 141 according to the brightness of the ambient light.
  • the proximity sensor can close the display panel 141 and/or when the mobile phone is moved to the ear. Or backlight.
  • the accelerometer sensor can detect the magnitude of acceleration in various directions (usually three-axis), and can detect the magnitude and direction of gravity when it is stationary.
  • the terminal device can be used to collect the user's atrial fibrillation signal, the terminal device can also be equipped with an electrocardiographic sensor, and the user's electrocardiographic signal can be obtained through the electrocardiographic sensor.
  • the audio circuit 160, the speaker 161, and the microphone 162 can provide an audio interface between the user and the mobile phone.
  • the audio circuit 160 can transmit the electrical signal converted from the received audio data to the speaker 161, which is converted into a sound signal for output by the speaker 161; on the other hand, the microphone 162 converts the collected sound signal into an electrical signal, which is then output by the audio circuit 160.
  • the terminal device may play the prediction result of the ECG signal through the audio circuit 160, and notify the user through a voice signal.
  • the terminal device can receive atrial fibrillation signals sent by other devices through the near field communication module 170.
  • the near field communication module 170 is integrated with a Bluetooth communication module, establishes a communication connection with the wearable device through the Bluetooth communication module, and receives feedback from the wearable device Signal of atrial fibrillation.
  • FIG. 1 shows the near field communication module 170, it can be understood that it is not a necessary component of the mobile phone 100, and can be omitted as needed without changing the essence of the application.
  • the processor 180 is the control center of the mobile phone. It uses various interfaces and lines to connect various parts of the entire mobile phone. It executes by running or executing software programs and/or modules stored in the memory 120 and calling data stored in the memory 120. Various functions and processing data of the mobile phone can be used to monitor the mobile phone as a whole.
  • the processor 180 may include one or more processing units; preferably, the processor 180 may integrate an application processor and a modem processor, where the application processor mainly processes the operating system, user interface, application programs, etc. , The modem processor mainly deals with wireless communication. It can be understood that the foregoing modem processor may not be integrated into the processor 180.
  • the mobile phone 100 also includes a power source 190 (such as a battery) for supplying power to various components.
  • a power source 190 such as a battery
  • the power source may be logically connected to the processor 180 through a power management system, so that functions such as charging, discharging, and power consumption management can be managed through the power management system.
  • Fig. 2 shows a structural block diagram of an ECG signal prediction system provided by an embodiment of the present application.
  • the ECG signal prediction system includes a mobile terminal 210 and a wearable device 220.
  • the mobile terminal 210 and the wearable device 220 may establish a communication connection through a near field communication method.
  • the ECG signal prediction device is specifically a mobile terminal 210 used by the user.
  • the mobile terminal 210 can receive the ECG signal sent by the wearable device 220, for example, establish a communication connection with the wearable device 220 through a near field communication method such as Bluetooth communication or WIFI communication, and receive the ECG signal sent by the wearable device 220, It can also establish communication with other remote communication terminals storing target users through wired communication or wireless communication, and receive the ECG signals sent by the remote communication terminals, and configure by receiving the ECG signals of multiple different users.
  • the atrial fibrillation signal classification model, and the signal type of the ECG signal fed back by the wearable device 220 in multiple collection periods is identified according to the atrial fibrillation signal classification model. And calculate the probability of occurrence of atrial fibrillation of the target user according to the signal types of multiple ECG signals, so as to predict the user's atrial fibrillation event.
  • the wearable device 220 is specifically used to collect the user's biometric signal.
  • the biometric signal may be an original collected electrocardiogram signal, or an atrial fibrillation signal generated by processing an atrial fibrillation recognition algorithm. If the collected ECG signals are obtained, the original ECG signals can be sent to the mobile terminal 210, and the ECG signals can be recognized by the mobile terminal 210 for atrial fibrillation and converted into atrial fibrillation signals; it can also be built in by the wearable device 220 After converting the ECG signal into the atrial fibrillation signal, the processing module of the A-fibrillation signal is sent to the mobile terminal 210.
  • the ECG signal prediction system further includes a cloud server 230.
  • the cloud server may receive the ECG signals fed back by various other electronic devices 240.
  • the ECG signals fed back by the above-mentioned electronic devices 240 are the ECG signals of patients with atrial fibrillation, and the above-mentioned atrial fibrillation signals are constructed based on the ECG signals of the above-mentioned patients with atrial fibrillation.
  • the terminal device 210 can download the classification model of the cloud server 230, without the need to build the model locally.
  • Fig. 3 shows a structural block diagram of an ECG signal prediction system provided by another embodiment of the present application.
  • the ECG signal prediction system includes a server 310, a terminal device 320 and a wearable device 330.
  • the server 310 can communicate with the terminal device 320 and the wearable device 330 through a wired and/or wireless network.
  • the server 310 can receive atrial fibrillation signals sent by multiple wearable devices 330, that is, the server belongs to a cloud device.
  • the wearable device 330 can establish a communication connection with the terminal device 320 through a near field communication method.
  • the terminal device 320 can be installed with the server 310 associated
  • the wearable device 330 sends the ECG signal to the terminal device 320 for storage, and when the terminal device 320 runs the above client program, the ECG signal is encapsulated by the client program, and the encapsulated data The packet is sent to the server 310.
  • the signal type of the ECG signal can be identified through the built-in atrial fibrillation signal classification model, and the signal type is fed back to the terminal device 320 corresponding to the target user or sent to the wearable device 330 ;
  • the server 310 configures the corresponding database for different users, the ECG signal can be stored in the database associated with the target user according to the user ID of the ECG signal, according to all historical signals of the target user in the database Calculate the probability of occurrence of atrial fibrillation of the target user, and return the probability of occurrence of atrial fibrillation to the wearable device 330 or the terminal device 320 for display output.
  • the terminal device 320 may determine the occurrence probability of atrial fibrillation of the target user according to the signal types of all the ECG signals that have been collected, and output the above-mentioned atrial fibrillation through the display interface. Probability of atrial fibrillation.
  • the terminal device 320 may also be configured with a communication module, such as a Bluetooth communication module or a WIFI communication module, through the communication module to receive the ECG signal fed back by the wearable device 330 used by the user, and pass the atrial fibrillation signal to the local client.
  • the program is sent to the server 310.
  • the client program of the terminal device 320 is equipped with a conversion model of the probability of occurrence of atrial fibrillation, and the terminal device 320 generates a signal category sequence based on the signal categories of all collected ECG signals according to the sequence of acquisition time, and the signal The category sequence is imported into the above-mentioned conversion model, and the probability of occurrence of atrial fibrillation of the target user is calculated.
  • the corresponding relationship between the wearable device 330 and the terminal device 320 may be stored in the server 310.
  • the wearable device 330 and the terminal device 320 that have a corresponding relationship may belong to the same entity user, that is, the user can collect the ECG signal through the wearable device 330 and upload it to the server 310, and the server 310 is calculating the signal corresponding to the ECG signal After the probability of occurrence of atrial fibrillation is calculated based on the category or signal categories of multiple ECG signals, the output result is fed back to the terminal device 320 for display.
  • the server 310 may send the aforementioned output result to the target user and the terminal device 320 of the associated user.
  • the wearable device 330 may be a smart wearable device, such as a smart bracelet or a smart watch.
  • the wearable device may be installed with a client program matching the server 310, and after the user’s ECG signals are collected, the client program The program sends the collected signal to the server 310.
  • the wearable device 330 can send the ECG signal to the terminal device 320 in the same environment as the wearable device 330, and upload the ECG signal to the server 310 through the terminal device 320.
  • the target user is an elderly person, and the target user's associated user is the child of the elderly person.
  • the wearable device 330 is worn on the wrist of the elderly. After the wearable device 330 collects the elderly’s ECG signals, it can send the ECG signals directly to the server 310 through the built-in communication module, or send the ECG signals to the elderly’s
  • the terminal device 320 forwards to the server 310 through the terminal device 320.
  • the server 310 After the server 310 receives the ECG signal, it can identify the signal category of the ECG signal through the atrial fibrillation signal classification model, and can directly feed the signal category back to the terminal device 320 of the elderly, or after calculating the probability of the elderly’s atrial fibrillation , And send the aforementioned probability of occurrence of atrial fibrillation to the terminal device 320 of the elderly and the terminal device 320 of the children of the elderly.
  • the server 310 detects that the value of the occurrence probability of atrial fibrillation of the elderly is greater than the preset probability threshold, the server 310 sends the occurrence probability of the atrial fibrillation of the elderly to the terminal device 320 of the children of the elderly.
  • the terminal device 320 After receiving the probability of atrial fibrillation, the terminal device 320 can output it through an interactive model to inform the elderly about the current physical condition.
  • the execution subject of the process is a device installed with a prediction program of an ECG signal.
  • the device of the ECG signal prediction program may specifically be a terminal device, which may be a smart phone, a tablet computer, a notebook computer, etc. used by the user to classify the obtained target user's ECG signal Identify and determine the probability of occurrence of atrial fibrillation for the target user.
  • Fig. 1 shows an implementation flowchart of the ECG signal prediction method provided by the first embodiment of the present application, and the details are as follows:
  • the electrocardiogram signal may be an electrocardiogram signal obtained by a wearable device or an electrocardiogram acquisition device.
  • the wearable device and the electrocardiogram acquisition device may be equipped with an electrocardiogram sensor, such as an electrocardiogram (ECG) sensor Or a Photoplethysmograph (PPG) sensor, which can be used to obtain the ECG signal of the wearing user or the detected user.
  • ECG electrocardiogram
  • PPG Photoplethysmograph
  • the wearable device can be a device that can be in contact with the user’s skin, such as a wristband or a watch. By detecting the expansion of blood vessels in the contact area, the user’s heart rate can be obtained, and the user’s heart rate can be generated based on the heart rate corresponding to each collection moment. ECG signal.
  • the terminal device may preprocess the ECG signal through a preset signal optimization algorithm, so that the accuracy of subsequent type recognition can be improved.
  • the optimization method includes but is not limited to one or a combination of the following: signal amplification, signal filtering, abnormal detection, signal repair, etc.
  • anomaly detection specifically refers to the extraction of multiple waveform characteristic parameters, such as the maximum continuous duration of the signal, the number of waveform interruptions, the duration of acquisition interruptions, and the signal-to-noise ratio of the waveform, according to the signal waveform of the original ECG signal acquired, and the acquisition is based on the above
  • the waveform feature value calculates the signal quality of the ECG signal. If it is detected that the signal quality is lower than the effective signal threshold, the ECG signal is identified as an invalid signal, and the subsequent signal type identification operation is not performed on the invalid signal. Conversely, if the signal quality is higher than the effective signal threshold, the ECG signal is identified as a valid signal, and the operations of S402 and S403 are executed.
  • the signal repair is specifically performing waveform fitting on the interrupted area in the process of collecting the ECG signal through a preset waveform fitting algorithm to generate a continuous ECG waveform.
  • the waveform fitting algorithm can be a neural network.
  • the parameters in the waveform fitting algorithm are adjusted so that the waveform trend of the fitted ECG signal matches the waveform trend of the target user , Thereby improving the waveform fitting effect.
  • the signal repair operation is performed after the above-mentioned abnormality detection operation, because when the missing waveform of the ECG signal is modified by the signal, the ECG signal acquisition instruction will be increased, thereby affecting the operation of abnormality detection, and thus the acquisition quality cannot be poor.
  • the terminal device can first determine whether the ECG signal is a valid signal through the abnormal detection algorithm; if the ECG signal is a valid signal, the ECG signal is repaired through the signal repair algorithm; otherwise, If the ECG signal is an abnormal signal, there is no need to perform signal repair, thereby reducing unnecessary repair operations.
  • the wearable device or the ECG acquisition device may be provided with a storage unit, and the above two devices store the ECG signals collected by the user in the storage unit, and the stored data in the storage unit is detected.
  • the ECG signal meets the preset upload threshold, for example, the data volume of the ECG signal is greater than the preset data volume threshold, or when the acquisition time of the ECG signal meets the corresponding upload threshold, the stored ECG signal can be packaged , And send it to the terminal device.
  • the terminal device can receive other devices or obtain the target user's ECG signal through a built-in sensor. If it is obtained through other devices, such as a wearable device or an ECG collection device, the target user's ECG signal is obtained. Signal, the corresponding ECG signal feedback cycle can be configured for the above-mentioned different collection devices.
  • the collection device can periodically send the user's atrial fibrillation signal to the terminal device according to the feedback cycle. For example, the length of the collection cycle is 48 seconds.
  • the collection device can use 48s as a collection period to generate a signal segment from the collected ECG signal, and send an ECG signal segment to the terminal device every 48s, and the terminal device can identify the signal corresponding to each ECG signal segment. Types of.
  • the collection device when the user starts the client program, the collected data can be obtained from the time when the ECG signal is fed back last time to the start time of the current client program
  • the ECG signal is encapsulated by the client program, and the ECG signal between the above two moments is sent to the terminal device, and the terminal device is used to identify all the ECG signals in a unified manner.
  • the device establishes a long connection, and the communication link between the above two is established only when the client program is started, thereby reducing the energy consumption of the ECG acquisition device and the terminal device, and improving the endurance of the device.
  • the ECG signal is imported into a preset atrial fibrillation signal classification model, and the signal category of the ECG signal output by the atrial fibrillation signal classification model is obtained, wherein the atrial fibrillation signal classification model is atrial fibrillation signal classification model.
  • the patient is trained as a model training sample.
  • the terminal device is preset with an atrial fibrillation signal classification model, and the collected ECG signals are imported into the atrial fibrillation signal classification model to determine the signal category corresponding to the ECG signal.
  • This signal category can be used to indicate the probability of an atrial fibrillation event.
  • the terminal device may divide the ECG signal into two categories, namely the first type signal and the second type signal. Among them, the first type of signal indicates that there is a greater risk of an atrial fibrillation event; and the second type of signal indicates that the user has a low probability of an atrial fibrillation event.
  • the terminal device can be divided into different levels according to the probability of occurrence of atrial fibrillation events, and different levels correspond to a signal category.
  • the probability of occurrence of fibrillation is divided into N levels.
  • the Nth level corresponds to the lowest probability of occurrence of atrial fibrillation; and the 1st level corresponds to atrial fibrillation
  • the occurrence probability is the highest, and the terminal device can recognize the hierarchical category described in the ECG signal according to the aforementioned atrial fibrillation classification model.
  • the embodiment of the present application does not determine whether the ECG signal is an atrial fibrillation signal when an atrial fibrillation event occurs, but can compare the non-atrial fibrillation ECG signal when a non-atrial fibrillation event occurs.
  • the signal category of the signal is further marked, so as to realize the prediction of the user's atrial fibrillation event, realize the purpose of informing the user in advance, prevent the user from having an atrial fibrillation event without warning, and improve the detection effect and detection range.
  • the atrial fibrillation signal classification model can also be used to identify atrial fibrillation signals, that is, the ECG signal corresponding to the user's atrial fibrillation event.
  • the signal category corresponding to the ECG signal of the atrial fibrillation event is the atrial fibrillation signal category.
  • different types of standard signals can be set in the atrial fibrillation signal classification model.
  • the terminal device can directly use the ECG signal as the input parameter of the model, and the atrial fibrillation signal classification model can match the input ECG signal with the standard signal of each category, and identify the signal of the ECG signal based on the matching result category.
  • the terminal device can separately calculate and identify the matching degree between each standard signal and the ECG signal, and select the signal type of the standard signal with the largest matching value as the signal type of the ECG signal; if there are two or The above standard signal is consistent with the ECG signal and the matching degree value is the largest.
  • the ECG signal can be divided into multiple signal segments through the preset signal segmentation algorithm, and each signal segment is calculated separately with the highest matching degree above.
  • the terminal device recognizes that the matching degree between the ECG signal and the standard signal of the first category is the same as the value of the matching degree between the ECG signal and the standard signal of the second category, and both are the largest.
  • the terminal equipment can divide the ECG signal into three signal segments, namely the first signal segment, the second signal segment, and the third signal segment, and calculate the difference between the above three signal segments and the two types of standard signals.
  • the matching degree between each signal segment and determine the sub-category corresponding to each signal segment.
  • the recognition results are as follows: the first signal segment (first category), the second signal segment (first category), and the third signal segment (second category), Therefore, it can be determined that the number of signal segments matching the first type is greater than the signal segments matching the second type, and the signal type of the ECG signal is determined to be the first type.
  • the terminal device can standardize the cardiac signal before calculating the degree of matching between the ECG signal and the standard signal.
  • the terminal device adjusts the signal duration of the ECG signal according to the standard duration of the standard signal so that the standard duration is the same as the adjusted signal duration. If the signal duration of the ECG signal is longer than the standard duration, the ECG signal can be intercepted, and the intercepted ECG signal is consistent with the signal duration of the standard signal; if the signal duration of the ECG signal is less than the standard duration, it can be pulled through the signal
  • the insufficient area is filled in ways such as extension and cyclic extension, so that the signal duration of the adjusted ECG signal is consistent with the standard duration of the standard signal.
  • the matching degree calculation between the signals can be performed, and the signal type of the ECG signal can be identified based on the matching degree.
  • a dynamic time warping algorithm can be used.
  • the specific implementation method is as follows:
  • the terminal device can convert the ECG signal according to the heart rate values collected by each collection time node. Convert to a heart rate sequence.
  • the corresponding coordinate grid is generated, and the difference between the elements corresponding to the intersection of each coordinate grid is used as the element corresponding to the coordinate grid Distance value.
  • the element distance values corresponding to the intersection points of each grid are superimposed to obtain the total distance value corresponding to the path.
  • the path with the smallest total distance value is selected as the difference between the heart rate sequence and the standard sequence.
  • the distance value corresponding to the characteristic path is used as the distance value between the ECG signal and the standard signal, and the matching degree between the two signals is calculated based on the distance value.
  • the ECG signal fed back by the target user has M collection moments.
  • a heart rate sequence containing M elements is generated.
  • the standard sequence corresponding to the standard signal contains N elements, then According to the above two sequences, an M*N coordinate grid can be generated, and the coordinate distance value of the coordinate (m, n) is the distance value between the m-th element in the heart rate sequence and the n-th element in the standard sequence.
  • the terminal device can calculate the characteristic parameters of the ECG signal in multiple characteristic dimensions, convert the ECG signal into a signal characteristic sequence, and use the signal characteristic sequence as the input and output of the atrial fibrillation signal classification model.
  • the signal type of the ECG signal can also be provided with a feature extraction layer, and the terminal device inputs the ECG signal into the atrial fibrillation signal classification model, and the ECG signal can be converted into a feature sequence through the feature extraction layer in the model.
  • the signal characteristics of the above-mentioned ECG signal include, but are not limited to, one or a combination of the following: the maximum heart rate value, the minimum heart rate value, the first heart rate duration (that is, the duration exceeding the first heart rate threshold), and the second heart rate duration ( That is, the duration less than the second heart rate threshold) and so on.
  • the terminal device may also determine the joint characteristic value according to the ECG signal and the associated signal adjacent to the acquisition time of the ECG signal, such as the average heart rate.
  • the rate of change, etc., the signal characteristic value determined by the single signal and the joint characteristic value described above constitute the signal characteristic sequence described above.
  • the risk degree of the about to occur atrial fibrillation is calculated, and it is predicted whether the target user is about to have atrial fibrillation.
  • the terminal device after the terminal device has identified the signal type of the ECG signal, it can predict the probability of the target user having an atrial fibrillation event based on the signal types of the multiple ECG signals of the target user, that is, the aforementioned atrial fibrillation Probability of occurrence.
  • the vital signs of the human body Before the atrial fibrillation time is triggered, the vital signs of the human body have certain signs, that is, the ECG signal closer to the atrial fibrillation event has a certain commonality.
  • the signal type of the ECG signal it can be determined whether the ECG signal conforms to the proximity
  • the common characteristics of the ECG signal of the atrial fibrillation time so that according to the signal category of the ECG signal, the probability of the occurrence of the atrial fibrillation event of the user can be determined.
  • the signal types of multiple ECG signals it can be determined that the ECG signal of the target user and the ECG signal before the occurrence of atrial fibrillation are in common, whether it is sporadic or frequent, and based on the judgment result
  • the probability of occurrence of atrial fibrillation of the target user can be determined more accurately, and the user's atrial fibrillation event can be predicted, without the user needing to alert the user when an atrial fibrillation event occurs.
  • the terminal device may sequentially combine the signal categories of the various ECG signals according to the sequence of the collection time of the various ECG signals to generate a signal category sequence.
  • the signal category sequence is imported into the probability conversion function, and the probability of the occurrence of atrial fibrillation of the target user is calculated.
  • the terminal device may generate a corresponding probability value according to the signal type of each ECG signal, and perform weighted superposition of multiple probability values to calculate the occurrence probability of atrial fibrillation of the target user.
  • the weighted value of each ECG signal is determined according to the difference with the current time. The smaller the difference with the current time, the larger the corresponding weighting weight; conversely, if the difference with the current time is larger, Then the corresponding weighted weight is smaller.
  • the aforementioned algorithm for calculating the probability of occurrence of atrial fibrillation may specifically be:
  • Probability is the probability of occurrence of the aforementioned atrial fibrillation
  • SignalType i is the probability value corresponding to the signal type of the i-th ECG signal
  • CurrentTime is the current time
  • CollectTime i is the collection time of the i-th ECG signal
  • N is the above-obtained The total number of ECG signals.
  • the terminal device can output the probability of occurrence of atrial fibrillation to the user through a built-in interactive module.
  • the output method includes but is not limited to: output through notification, output through broadcast, or add a component for real-time feedback of the occurrence probability of atrial fibrillation on a preset interface, and adjust the corresponding parameter value in the component Perform output.
  • the terminal device can obtain the explanatory speech segment associated with the occurrence probability of atrial fibrillation and the atrial fibrillation consultation, and generate an atrial fibrillation report by generating the atrial fibrillation probability, the explanatory speech segment and the atrial fibrillation consultation.
  • the user can obtain more information related to the trigger probability of the atrial fibrillation through the atrial fibrillation report, which improves the readability of the probability of atrial fibrillation and facilitates the user to understand his physical state.
  • the terminal device can push the corresponding atrial fibrillation consultation for the user from the cloud database according to the occurrence probability of the current user's atrial fibrillation, thereby improving the matching degree between the pushed information and the user, and realizing the purpose of accurately pushing information.
  • Fig. 5 shows an output schematic diagram of the probability of occurrence of atrial fibrillation provided by an embodiment of the present application.
  • the terminal device can determine the probability of occurrence of atrial fibrillation of the target user according to the signal types of multiple ECG signals.
  • the value shown in the figure is 88, and the corresponding explanation is configured under the probability of occurrence of atrial fibrillation.
  • the segment that is, "You may have atrial fibrillation in a short time in the future", and atrial fibrillation information is added below the explanatory segment.
  • the user can click the UI control corresponding to the atrial fibrillation information to read the specific information content.
  • the existing atrial fibrillation technology can predict the probability of a user's atrial fibrillation event by detecting the content of markers related to atrial fibrillation in the user's body, for example, by detecting the user's BNP content or FGF-23 Content to determine the probability of the user’s atrial fibrillation event.
  • Fig. 6 shows a schematic diagram of a comparison of relevant markers between atrial fibrillation patients and normal patients provided by an embodiment of the present application. As shown in Figure 6, the levels of FGF-23 and BNP in patients with atrial fibrillation are higher than those of normal users.
  • the corresponding content threshold can be set to determine whether the user is a patient with atrial fibrillation and predict whether the user will trigger an atrial fibrillation event. .
  • markers related to atrial fibrillation is not only related to atrial fibrillation, but also related to other diseases, it is not possible to directly determine the user's atrial fibrillation behavior through the markers, that is, the recognition accuracy of the above methods is low.
  • the existing atrial fibrillation technology can configure a corresponding risk score table based on multiple risk factors that have a greater relationship with the occurrence of atrial fibrillation events, and configure a corresponding contribution value for each risk factor, through Obtain user information and compare each risk factor, determine the risk items included in the user, calculate the user's total score value based on the contribution values corresponding to all risk items, and determine whether the user is a patient with atrial fibrillation based on the total score value , So as to predict the user's atrial fibrillation event.
  • Table 1 shows the atrial fibrillation risk score table provided by an embodiment of the present application.
  • the risk scale for atrial fibrillation uses the CHADS scoring method and the CHA2DS2-VASc scoring method, and includes the following risk factors: congestive heart failure/left ventricular dysfunction, hypertension, age 75 or older, diabetes, stroke/TIA / History of thromboembolism, vascular disease, age 65 to 74 years, gender (female) 8 items. Each project has a corresponding contribution value.
  • the terminal device can determine the user's score value by comparing the number of risk factor items matched by the user's information and the corresponding score, so as to predict the probability of the user's atrial fibrillation event.
  • This scoring method is mainly for the elderly and cannot cover all groups of people, and the scope of application is small.
  • the embodiment of this application can collect the user's ECG signal in the daily process, and determine the signal category of each user's ECG signal separately, and calculate the signal category from multiple ECG signals.
  • the probability of the occurrence of atrial fibrillation of the user can be accurately predicted by observing the ECG signal of the user for a period of time, which improves the accuracy of the prediction, and the collection of the ECG signal is applicable to all groups of people, thus expanding Scope of application.
  • the ECG signal prediction method imports the collected ECG signal into a preset atrial fibrillation signal classification model to determine the signal category of each ECG signal, and After acquiring the signal types of the user’s ECG signals in multiple acquisition periods, calculate the target user’s atrial fibrillation probability by using the multiple ECG signal types, and according to the numerical value of the probability of atrial fibrillation, Realize the prediction of atrial fibrillation events, which facilitates users to determine their own physical conditions, and improves the detection effect and user experience.
  • FIG. 7 shows a specific implementation flowchart of an ECG signal prediction method provided by the second embodiment of the present application.
  • the ECG signal is imported into the preset atrial fibrillation signal classification model to obtain the atrial fibrillation signal classification model.
  • a training signal set consisting of a plurality of historical signals in continuous acquisition order is obtained; the training signal set includes at least one atrial fibrillation signal.
  • the terminal device before the terminal device recognizes the ECG signal, it can train the preset native training model through the training signal set, and the atrial fibrillation classification that can be used to identify the signal category of the ECG signal has been obtained. model. Based on this, the terminal device needs to be able to obtain multiple training signal sets.
  • the method of obtaining can be: if the terminal device is a smart device such as a smart phone or a tablet computer used by the user, the smart device can establish a communication connection with the cloud database to download multiple training signal sets from the cloud database; if the terminal device As a server, the server can be used to receive ECG signals fed back by various electronic devices and store them in a local database. In this case, the server can directly extract the above-mentioned training signal set from the local database.
  • each training signal set contains multiple historical signals, and the order of each historical signal in the signal set matches the sequence of its acquisition time.
  • the above-mentioned training signal set is specifically a training signal set composed of electrocardiogram signals of patients with atrial fibrillation. Since the atrial fibrillation signal classification model is specifically used to determine whether the signal features of the ECG signal have the precursor features of the atrial fibrillation event, when training the atrial fibrillation signal classification model, it is necessary to collect the signal before the atrial fibrillation event occurs.
  • the above-mentioned precursor features are determined from the signal before the occurrence of atrial fibrillation events, so as to realize the prediction of atrial fibrillation events.
  • the terminal device when the terminal device obtains the training signal set, it needs to collect the signal containing the atrial fibrillation event, that is, the atrial fibrillation signal, and the user with the atrial fibrillation signal is the atrial fibrillation patient. Based on this, the terminal device can select the target training user according to the atrial fibrillation patient identification of each user, and obtain the signal set corresponding to each target training user as the above-mentioned training signal set. Among them, each training signal set contains at least one atrial fibrillation signal.
  • S701 may specifically be: the terminal device can obtain user information of the target user, and the user information can be used to reflect the user's biological characteristics, including but It is not limited to one or a combination of the following: user age, user gender, current health status, illness record, etc.
  • the terminal device can identify the training object matching the user information from the cloud database, and obtain the ECG signal set of the training object as the training signal set, so as to realize the customized atrial fibrillation signal classification model for the user. , Thereby improving the accuracy of subsequent training.
  • S701 may specifically be: the terminal device can be divided according to a preset rule of crowd division , Divide the existing users in the database into multiple user groups, and configure corresponding atrial fibrillation signal classification models for different user groups. Based on this, when training the atrial fibrillation signal classification model, the training signal set in the user group associated with the model is used to train it, thus achieving the purpose of configuring different atrial fibrillation signal classification models for different characteristic groups .
  • the terminal device will obtain the user information of the target user, determine the user group to which the target user belongs based on the user information, and identify the ECG signal of the target user through the atrial fibrillation signal classification model associated with the user group.
  • risk signals other than the atrial fibrillation signal are extracted from the training signal set.
  • the terminal device can perform atrial fibrillation signal determination operations on each historical signal in the training signal set, and identify the corresponding atrial fibrillation signal when an atrial fibrillation event occurs, and mark the atrial fibrillation signal to remove the atrial fibrillation signal Other historical signals other than atrial fibrillation are identified as risk signals.
  • the terminal device may be configured with characteristic parameters of atrial fibrillation, and the terminal device may extract the signal characteristic value of each historical signal separately, and match the signal characteristic value of each historical signal with the aforementioned characteristic parameter of atrial fibrillation, and determine whether the historical signal is based on the matching result. Belongs to atrial fibrillation signal.
  • FIG. 8 shows a schematic diagram of a training signal set provided by an embodiment of the present application.
  • the training signal set contains 10 historical signals, and each historical signal is identified by a rectangular area, that is, a PPG column.
  • the height of the PPG column can be determined according to the average heart rate value during the collection time. If the average heart rate value during the collection time is higher, the height of the PPG column is higher; on the contrary, if the PPG column is lower, it means The average heart rate value corresponding to the historical signal during the acquisition time is relatively low. Since the atrial fibrillation event is extremely rapid in a certain period of time, sometimes as high as 200 beats per minute, the average heart rate is higher in the characterization of the ECG signal.
  • the ECG signal of an atrial fibrillation event can be identified by the height of the PPG column, that is, the atrial fibrillation signal marked in the figure.
  • the terminal device After the terminal device recognizes the atrial fibrillation signal, it can recognize historical signals other than the aforementioned atrial fibrillation signal as a risk signal.
  • a preset native classification model is trained through the risk signal of the training signal set to obtain the atrial fibrillation signal classification model.
  • the terminal device may be trained by acquiring the risk signal, and the terminal device may mark the corresponding signal category for each risk signal in advance, and construct a training sample based on the marking information and the risk signal.
  • the terminal device imports multiple risk signals into the native classification model, calculates multiple prediction categories, and respectively identifies whether each predicted category matches the preset signal category of the corresponding risk signal, and calculates the prediction corresponding to the native classification model Loss value, and whether the native classification model converges and the predicted loss value is less than the preset loss threshold.
  • the native classification model has been adjusted, and the adjusted native classification model is recognized as the atrial fibrillation signal classification model; otherwise, If it does not converge, it is necessary to adjust the learning parameters in the native classification model so that the aforementioned native classification model meets the two conditions of convergence and the predicted loss value is less than the preset loss threshold.
  • the terminal device may be configured with multiple native classification models of different types, and the multiple native classification models can be trained and learned at the same time through the above risk signals, and based on the corresponding convergence of the multiple native classification models. For time and loss value, a preferred native classification model is selected based on the above two parameters, and an atrial fibrillation signal classification model is constructed according to the preferred native classification model.
  • the terminal device can determine the training weight of the risk signal according to the acquisition time difference between the risk signal and its associated atrial fibrillation signal.
  • the atrial fibrillation signal associated with the risk signal is specifically the atrial fibrillation signal that has the smallest acquisition time difference with the risk signal among all atrial fibrillation signals whose acquisition time is after the risk signal, that is, the atrial fibrillation signal that is closest to the risk signal .
  • the associated atrial fibrillation signal is historical signal 3.
  • the risk signal is extracted, training and learning are performed, and the atrial fibrillation signal classification model is obtained, which can realize the prediction of atrial fibrillation events and improve Improved forecast accuracy.
  • FIG. 9 shows a specific implementation flowchart of an ECG signal prediction method S703 provided by the third embodiment of the present application.
  • S703 in an ECG signal prediction method provided in this embodiment includes: S901 to S903, and the details are as follows:
  • the training a preset native classification model through the risk signal of the training signal set to obtain the atrial fibrillation signal classification model includes:
  • the signal category corresponding to each risk signal is determined according to the time difference between the acquisition time of the risk signal and the trigger time of the associated atrial fibrillation signal.
  • the terminal device can identify the acquisition time corresponding to each risk signal, and select the atrial fibrillation signal associated with the risk signal from the training signal set according to the acquisition time, where the associated atrial fibrillation signal may specifically be: the acquisition time is Among all atrial fibrillation signals after the risk signal, the atrial fibrillation signal that has the smallest acquisition time difference with the risk signal is the atrial fibrillation signal that is closest to the risk signal.
  • the terminal device can configure the corresponding distance time range for different signal types, and identify the distance time range corresponding to the risk signal according to the time difference between the acquisition time of the risk signal and the trigger time of the associated atrial fibrillation signal. Signal category.
  • the terminal device may divide the signal category into two types, namely a type 0 signal and a type 1 signal.
  • the type 0 signal is that the time difference between the collection and the associated atrial fibrillation signal is within a preset time threshold
  • the type 1 signal is that the time difference between the collection of the associated atrial fibrillation signal is outside the preset time threshold.
  • the time threshold can be 2 hours.
  • FIG. 10 shows a schematic diagram of marking of signal types provided by an embodiment of the present application. As shown in FIG.
  • each historical signal is 1 hour, that is, the time difference between two adjacent atrial fibrillation signals is 1 hour.
  • the time threshold is 2 hours, it can be determined that the time difference between historical signal 1 and historical signal 2 and the associated atrial fibrillation signal (ie historical signal 3) are all within 2 hours, and the above two risk signals are all type 0 Signal; and the time difference between historical signal 4, historical signal 5, historical signal 6 and the associated atrial fibrillation signal (ie historical signal 9) is greater than 2 hours, then the above three risk signals are all type 1, and so on, Determine the signal type of each risk signal.
  • the terminal device can also be divided into N types, and configure corresponding time thresholds for different signal types, for example, the first type of signal and the associated atrial fibrillation signal
  • the acquisition time difference is between t0 and t1; the acquisition time difference between the second type of signal and the associated atrial fibrillation signal is between t1 and t2,..., the difference between the Nth type of signal and the associated atrial fibrillation signal
  • the difference in acquisition time is between tN-1 and tN.
  • the terminal device After the terminal device generates the atrial fibrillation signal classification model trained in the above-mentioned manner, it imports the actually collected ECG signal into the above-mentioned model and determines the signal category of the risk signal. In addition to determining the probability of an atrial fibrillation event, you can also determine the predicted time between the next atrial fibrillation event based on the signal type of the ECG signal. Since the time difference between the aforementioned signal category and the occurrence of the atrial fibrillation signal is one-to-one correspondence, it is possible that each signal category corresponds to a predicted time. The terminal device can query the predicted time associated with it according to the signal type of the ECG signal, and prompt the user according to the predicted time.
  • the characteristic value of the risk signal in each preset signal characteristic dimension is calculated to obtain the signal characteristic parameter of the risk signal.
  • the terminal device may be configured with multiple signal characteristic dimensions, and different signal characteristic dimensions are used to represent different signal characteristics of the ECG signal.
  • the terminal device can analyze the risk signal, determine the characteristic value of the risk signal in each signal characteristic dimension, and import each characteristic value into the parameter template according to the corresponding position of each signal characteristic dimension in the parameter template to generate the Signal characteristic parameter of risk signal.
  • the signal characteristic dimension includes, but is not limited to, one or a combination of the following: average heart rate, maximum heart rate, minimum heart rate, duration exceeding the first heart rate threshold, duration below the second heart rate threshold, and so on.
  • the terminal device After the terminal device has established the signal characteristic parameters and signal categories of each risk signal, it generates a training sample, trains through multiple training samples, and adjusts the parameters in the native classification model until the result is converged.
  • the adjusted native classification model is identified as the atrial fibrillation signal classification model.
  • the risk type is marked according to the time difference between the risk signal and the associated atrial fibrillation signal, so that the risk signal with the precursory features of the atrial fibrillation event can be extracted, so as to be able to be based on the identified ECG
  • the signal type predicts atrial fibrillation events and improves the detection effect.
  • FIG. 11 shows a specific implementation flowchart of an ECG signal prediction method S701 provided by the fourth embodiment of the present application.
  • S701 in an ECG signal prediction method provided in this embodiment includes: S1101 to S1104, which are detailed as follows:
  • the acquiring a training signal set consisting of a plurality of ECG signals in continuous acquisition order includes:
  • the terminal device before importing the historical signals in the training signal set into the native classification model for training, can filter the historical signals in the training signal set to filter out the historical signals with poor collection quality, so as to be able to Improve the accuracy of follow-up training. Based on this, when the wearable device obtains the user's ECG signal, in addition to feeding back the ECG waveform, it can also add the associated exercise parameter to the ECG signal and feed back the above two data to the terminal device.
  • the terminal device may determine the exercise state of the training user when the historical signal of the training user is collected according to the exercise parameters associated with each historical signal.
  • the terminal device can determine the validity of the historical signal according to the above-mentioned motion parameters.
  • the motion parameter can be obtained through a motion sensing module such as an acceleration sensor and a gyroscope in the wearable device, and the terminal device can determine the historical user's motion state through the sensing value fed back by the motion sensing module.
  • a motion sensing module such as an acceleration sensor and a gyroscope in the wearable device
  • the jitter duration of the historical signal is determined.
  • the terminal device can analyze the signal waveform of the historical signal, determine the waveform segment with jitter, and use the duration of the aforementioned waveform segment as the jitter duration of the historical signal.
  • the jittered waveform segment can be the waveform segment of the interrupted area during the acquisition process, or the waveform segment corresponding to the waveform change frequency higher than the normal value.
  • the terminal device may calculate the signal quality of the historical signal using the above two parameters, and compare the calculated signal quality with a preset quality threshold to determine whether the historical signal is a valid signal. Wherein, if the signal quality is greater than or equal to the quality threshold, the historical signal is identified as a valid signal; conversely, if the signal quality is less than the quality threshold, the historical signal is identified as an invalid signal.
  • the way of calculating the signal quality may be: determining the first quality factor according to the ratio between the jitter duration and the signal duration of the historical signal, where the longer the jitter duration, the greater the value of the first quality factor The smaller the value; the second quality factor is determined according to the ratio between the motion parameter and the static motion parameter, where the larger the value of the motion parameter, the greater the user's motion amplitude, and the smaller the corresponding second quality factor value .
  • the weighted sum of the above two quality factors is performed to calculate the signal quality of the historical signal.
  • the terminal device determines the signal sequence of each effective signal in the training signal set according to the sequence of the acquisition time of each effective signal, thereby encapsulating multiple effective signals, filtering invalid signals, and forming the aforementioned training signal Set, improve the accuracy of the follow-up training process.
  • the historical signals are filtered to filter out invalid signals, so that the accuracy of subsequent training operations can be improved.
  • FIG. 12 shows a specific implementation flowchart of an ECG signal prediction method S402 provided by the fifth embodiment of the present application.
  • S402 in an ECG signal prediction method provided in this embodiment includes: S1201 to S1203, which are detailed as follows:
  • the ECG signal is imported into a preset atrial fibrillation signal classification model to obtain the signal category of the ECG signal output by the atrial fibrillation signal classification model, wherein the atrial fibrillation signal classification model is based on the atrial fibrillation signal classification model.
  • Patients with tremor are trained as model training samples, including:
  • the vital sign parameters of the target user are determined according to the ECG signal.
  • the terminal device before the terminal device uses the atrial fibrillation signal classification model to classify the target user's ECG signal, it can adjust the atrial fibrillation signal classification model according to the user's vital signs, which is the model warm-up phase, so that The subsequent identification process of the signal category is more accurate.
  • the terminal device can obtain the vital sign parameters of the target user through user input or wearable device feedback.
  • the vital characteristic parameter may be the static heart rate of the target user, such as the heart rate value corresponding to the user in a state of not moving for a long time (including sitting and sleeping processes), and the heart rate value during movement.
  • the target user's heart rate change range and the reference value of the heart rate can be determined, and the above values are used as the vital sign parameters of the target user.
  • the classification threshold of the atrial fibrillation signal classification model is adjusted based on the vital sign parameters.
  • the terminal device can adjust the classification threshold in the atrial fibrillation signal classification model through the collected vital sign parameters of the target user, so that the classification process can be matched with the physical state of the target user, and personalized customization is realized.
  • the purpose of the classification model For example, the terminal device can determine the heart rate fluctuation amplitude of the target user according to the difference between the static heart rate and the dynamic heart rate in the vital sign parameters, and adjust the atrial fibrillation trigger threshold based on the heart rate fluctuation amplitude, so as to accurately determine the user Whether the current heart rate value is close to the aforementioned atrial fibrillation trigger threshold is used to determine the occurrence probability of atrial fibrillation corresponding to the ECG signal.
  • the signal category of the ECG signal is identified through the adjusted atrial fibrillation signal classification model.
  • the terminal device can perform the signal category identification operation on the target user's ECG signal.
  • the specific identification operation reference may be made to the related description of the foregoing embodiment, which will not be repeated here.
  • the atrial fibrillation signal classification model is adjusted based on the vital sign parameters of the target user, so that the atrial fibrillation signal classification model matches the target user, and the identification of atrial fibrillation signals is improved accuracy.
  • FIG. 13 shows a specific implementation flowchart of an ECG signal prediction method S403 provided by the sixth embodiment of the present application.
  • an ECG signal prediction method S403 provided in this embodiment includes: S4031 to S4032, which are detailed as follows:
  • the calculating the risk of atrial fibrillation about to occur according to the signal category of the ECG signal, and predicting whether the target user is about to have atrial fibrillation includes:
  • the signal categories preset by the terminal device include at least two categories, namely risk categories and non-risk categories.
  • the signal characteristics of the ECG signal of the risk category include more signs of pre-atrial fibrillation events.
  • the atrial fibrillation signal classification model can identify the above-mentioned types of ECG signals as the ECG signals of the risk category; vice versa. , If the signal characteristics of the ECG signal do not exist or match with fewer precursor features before the atrial fibrillation event, in this case, the atrial fibrillation signal classification model can identify the above-mentioned types of ECG signals as non-risk types of ECG signal.
  • the risk category there can be multiple sub-categories cascaded under the risk category, such as the first level risk category, the second level risk category, etc.; the non-risk category can also be cascaded
  • the number of specific cascaded categories is determined according to the specific AF signal classification model, which is not limited here.
  • the terminal device may be divided into N signal categories according to the time difference between the atrial fibrillation signals associated with the risk signal distance during the training process.
  • the first signal category is the closest to the trigger time of the atrial fibrillation event, and the first signal category can be identified as a risk category; while the second signal category to the Nth signal category are farther away from the trigger time of the atrial fibrillation event, you can The second signal category to the Nth signal category are identified as non-risk categories.
  • the terminal device can acquire the signal types of all collected ECG signals within a preset time period, and count the number of signals whose risk types are ECG signals.
  • the preset time period It can be 1 day or a week.
  • the terminal device counts the number of ECG signals whose signal type is the risk type among all the ECG signals in the preset time period between the acquisition time difference with the ECG signal, that is, the above-mentioned The number of signals. For example, if the collection time of a certain ECG signal is 18:00 on December 13th, and the preset time period is 1 day, all the ECG signals collected between 18:00 on December 12 and 18:00 on December 13 will be acquired. It also counts the number of ECG signals whose signal category is the risk category in the obtained ECG signals.
  • the probability of occurrence of the atrial fibrillation is calculated according to the number of signals.
  • the terminal device can be provided with a conversion function for the probability of occurrence of atrial fibrillation.
  • the terminal device imports the number of signals mentioned above into the conversion function, and calculates the collected heart rate. The probability of atrial fibrillation corresponding to the target user at the time of the electrical signal.
  • the terminal device may calculate the frequency of occurrence of the risk signal of the target user according to the number of signals and the preset time period, and determine the probability of occurrence of atrial fibrillation based on the frequency.
  • the terminal device may divide the above-mentioned time period into multiple sub-time periods, and respectively count the appearance frequency of the corresponding risk signal in each sub-time period, thereby generating the appearance of the risk signal corresponding to the above-mentioned time period Frequency change curve, based on the above-mentioned appearance change curve to determine the probability of occurrence of atrial fibrillation.
  • FIG. 14 shows a specific implementation flowchart of an ECG signal prediction method provided by the seventh embodiment of the present application.
  • the method for predicting an ECG signal provided by this embodiment is based on the signal type of the ECG signal, After calculating the risk of atrial fibrillation about to occur and predicting whether the target user is about to have atrial fibrillation, it also includes: S1401 to S1403, which are detailed as follows:
  • the probability of occurrence of atrial fibrillation is greater than a preset probability threshold, the risk of atrial fibrillation will be calculated according to the signal type of the electrocardiogram signal, and whether the target user is about to have atrial fibrillation is predicted After that, it also includes:
  • S1401 determine the associated user preset by the target user, and send alarm information to the terminal of the associated user.
  • the target user if it is detected that the probability of occurrence of atrial fibrillation of the target user is greater than the preset probability threshold, it means that the target user is likely to have an atrial fibrillation event in the near future, and the target user needs to be warned at this time.
  • the way of early warning can notify users through pop-ups, information prompts, voice broadcasts, and output prompt sounds.
  • users can also be notified through S1401 to S1403.
  • the target user may be preset with the associated user and the communication address of the associated user.
  • the terminal device sends alarm information to the terminal of the associated user according to the communication address of the associated user pre-configured by the target user.
  • the alarm information may include the aforementioned probability of occurrence of atrial fibrillation and an explanatory segment corresponding to the probability of occurrence of atrial fibrillation.
  • the terminal device may mark a shortcut call button associated with the user on the display interface.
  • the target user can directly initiate a call request to the terminal of the associated user through the quick call button of the motor.
  • the associated user may be the contact number of the guardian, relative or related hospital of the target user.
  • FIG. 15 shows a schematic diagram of outputting alarm information provided by an embodiment of the present application.
  • the target user s atrial fibrillation occurrence probability is 88%
  • the preset probability threshold is 80%.
  • the terminal device will send an alarm message to the terminal of the associated user.
  • the content of the alarm message can be "User A The current probability of atrial fibrillation is 88%. Please pay close attention to user A’s physical condition.”
  • a quick call button for contacting the associated user can be set on the preset display interface to facilitate the user to quickly notify the associated user.
  • the address of the hospital closest to the location information is obtained according to the current location information of the target user.
  • the terminal device when the terminal device detects that the probability of occurrence of atrial fibrillation of the target user is greater than the preset probability threshold, the terminal device can provide the target user with medical consultation, reduce user operations, and improve consultation acquisition efficiency.
  • the terminal device can obtain the current location information of the target user. If the terminal device is a smart device used by the user, it can obtain the positioning signal through the built-in positioning module, and determine the location information of the target user based on the positioning signal; if If the terminal device is a server, it can send a location acquisition request to the user terminal of the target user. For the user terminal, you can obtain the location information through the built-in positioning module and feed it back to the server.
  • the terminal device may mark the location information of the target user on the preset map interface.
  • the terminal device can access the third-party server through the program call interface API, and mark the above-mentioned location information on the preset map interface through the map application provided by the third-party server, and use the location information as Baseline, find a hospital closest to the user's location, and determine the above-mentioned hospital address.
  • the terminal device after the terminal device obtains the current location information of the target user and the hospital address, it can generate a treatment path with the location information as the starting point and the hospital address as the terminal.
  • the terminal device can call a third-party map application, import the above two locations into the third-party map application, and output the above-mentioned path to a doctor through the path generation algorithm of the third-party map application.
  • FIG. 16 shows a schematic diagram of output of a visit path provided by an embodiment of the present application.
  • the terminal device can add a prompt button of "Query Hospital Path" on the display interface.
  • the terminal device can jump to the corresponding treatment path display page by clicking the prompt button.
  • the terminal device can obtain the user's current location information, determine the hospital address according to the location information, and generate the corresponding treatment path.
  • the alarm information is sent to the associated user and the path to see a doctor is provided, which reduces the operations that the target user needs to perform, improves the operation efficiency, and realizes the automatic alarm information.
  • the purpose of the release is not limited to the probability threshold.
  • FIG. 17 shows a specific implementation flowchart of an ECG signal prediction method provided by the eighth embodiment of the present application.
  • an ECG signal prediction method provided in this embodiment further includes: S1701 to S1703, which are detailed as follows:
  • the method further includes:
  • the newly added signal category of the newly added ECG signal is identified through the atrial fibrillation signal classification model.
  • the wearable device can send the collected new ECG signal to the terminal device during the user's use of the wearable device in a preset feedback period.
  • the terminal device receives the new ECG signal each time,
  • the corresponding new signal category can be determined through the above-mentioned trained atrial fibrillation signal classification model.
  • the method for determining the newly added signal category is as described in the above embodiment, and the implementation process is completely the same, and will not be repeated here.
  • the terminal device can recalculate the target user’s atrial fibrillation probability based on the previously identified probability of atrial fibrillation and the newly acquired new signal category of the newly acquired ECG signal, and calculate the aforementioned atrial fibrillation Update probability.
  • the specific calculation method can be as follows: the terminal equipment according to the signal category of all the ECG signals corresponding to the probability of occurrence of atrial fibrillation calculated last time, and the new signal category of the new ECG signal collected this time, through preset
  • the conversion algorithm for the probability of occurrence of atrial fibrillation re-determines the probability of occurrence of atrial fibrillation of the target user to obtain the above-mentioned atrial fibrillation update probability.
  • an atrial fibrillation probability curve is generated according to all the identified occurrence probabilities of atrial fibrillation.
  • the terminal device can record the probability of occurrence of atrial fibrillation of the target user each time it is calculated, so as to obtain the historical probability of each determination and the current calculated probability of atrial fibrillation, which can generate atrial fibrillation.
  • Probability curve which can facilitate the target user to determine the trend change of the probability of atrial fibrillation.
  • FIG. 18 shows a schematic output diagram of the atrial fibrillation probability curve provided by an embodiment of the present application. As shown in FIG. 18, the terminal device can output the corresponding display interface after calculating the user's atrial fibrillation probability each time, and the user can determine the current atrial fibrillation probability on the display interface.
  • the target user’s wearable device After the target user’s wearable device collects a new ECG signal, it can send the ECG signal to the terminal device, and the terminal device can identify the signal category of the newly added ECG signal, and based on the new signal after identification The category recalculates the probability of occurrence of atrial fibrillation for the target user.
  • the user clicks the "View Probability Trend" button it will jump to the corresponding display page, output the target user's atrial fibrillation probability curve, and determine the corresponding atrial fibrillation occurrence probability at each update time.
  • the target user’s atrial fibrillation probability is recalculated, which can improve the real-time performance of the alarm operation, and output the corresponding atrial fibrillation probability curve, which is convenient for the user Determine the probability trend, so that users can understand their physical condition.
  • FIG. 19 shows a structural block diagram of the ECG signal prediction device provided by an embodiment of the present application. For ease of description, only the same as the embodiment of the present application is shown. The relevant part.
  • the ECG signal prediction device includes:
  • the ECG signal acquiring unit 191 is used to acquire the ECG signal of the target user
  • the signal category identification unit 192 is configured to import the ECG signal into a preset atrial fibrillation signal classification model to obtain the signal category of the ECG signal output by the atrial fibrillation signal classification model, wherein the atrial fibrillation signal classification model It is obtained by training patients with atrial fibrillation as the model training sample;
  • the atrial fibrillation transmission probability calculation unit 193 is configured to calculate the risk of atrial fibrillation about to occur according to the signal type of the electrocardiogram signal, and predict whether the target user is about to have atrial fibrillation.
  • the ECG signal prediction device further includes:
  • a training signal set acquisition unit configured to acquire a training signal set consisting of a plurality of historical signals in continuous acquisition order; the training signal set includes at least one atrial fibrillation signal;
  • a risk signal extraction unit configured to extract risk signals other than the atrial fibrillation signal from the training signal set
  • the atrial fibrillation signal classification model training unit is configured to train a preset native classification model through the risk signal of the training signal set to obtain the atrial fibrillation signal classification model.
  • the atrial fibrillation signal classification model training unit includes:
  • the signal category labeling unit is configured to determine the signal category corresponding to each risk signal according to the time difference between the acquisition time of the risk signal and the trigger time of the associated atrial fibrillation signal;
  • a signal characteristic parameter generating unit configured to calculate the characteristic value of the risk signal in each preset signal characteristic dimension to obtain the signal characteristic parameter of the risk signal
  • the sample training unit is used for training according to the signal characteristic parameters and the signal category to obtain the atrial fibrillation signal classification model.
  • the training signal set acquisition unit includes:
  • An exercise parameter acquisition unit configured to acquire the exercise parameter of the training user when the historical signal is collected
  • a jitter duration determining unit configured to determine the jitter duration of the historical signal
  • a valid signal identification unit configured to determine whether the historical signal is a valid signal according to the jitter duration and the motion parameter
  • the effective signal screening unit is configured to encapsulate all the effective information according to the sequence of the acquisition time of each effective signal to obtain the training signal set.
  • the signal type identification unit 192 includes:
  • a vital sign parameter acquisition unit configured to determine the vital sign parameters of the target user according to the electrocardiogram signal
  • a classification threshold adjustment unit configured to adjust the classification threshold of the atrial fibrillation signal classification model based on the vital sign parameters
  • the atrial fibrillation signal classification model calling unit is used to identify the signal category of the ECG signal through the adjusted atrial fibrillation signal classification model.
  • the atrial fibrillation transmission probability calculation unit 193 includes:
  • a signal number counting unit configured to count the number of signals of the ECG signal whose signal type is the risk type in the target user within a preset time period
  • the signal number conversion unit is configured to calculate the occurrence probability of the atrial fibrillation according to the number of signals.
  • the ECG signal prediction device further includes:
  • the alarm information sending unit is used to determine the associated user preset by the target user, and send alarm information to the terminal of the associated user; and/or
  • the location information acquiring unit is configured to acquire the address of the hospital closest to the location information according to the current location information of the target user;
  • a treatment path generation unit is used to generate a treatment path according to the location information and the hospital address.
  • the ECG signal prediction device further includes:
  • the newly added signal category determining unit is configured to identify the newly added signal category of the newly added ECG signal through the atrial fibrillation signal classification model if the newly added ECG signal is received;
  • An atrial fibrillation occurrence probability update unit configured to recalculate the atrial fibrillation occurrence probability of the target user based on the newly added signal category
  • the atrial fibrillation probability curve generating unit is configured to generate an atrial fibrillation probability curve based on all the identified occurrence probabilities of atrial fibrillation.
  • the ECG signal prediction device provided by the embodiment of the present application can also import the collected ECG signal into the preset atrial fibrillation signal classification model to determine the signal category of each ECG signal, and obtain the The signal category of the user’s ECG signal in multiple collection periods.
  • the target user’s atrial fibrillation probability is calculated through the multiple ECG signal categories, and according to the numerical value of the probability of atrial fibrillation, it can be achieved.
  • the prediction of tremor events facilitates users to determine their own physical conditions, and improves the detection effect and user experience.
  • FIG. 20 is a schematic structural diagram of a terminal device provided by an embodiment of this application.
  • the terminal device 20 of this embodiment includes: at least one processor 200 (only one is shown in FIG. 20), a processor, a memory 201, and a processor stored in the memory 201 and capable of being processed in the at least one processor.
  • the terminal device 20 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server.
  • the terminal device may include, but is not limited to, a processor 200 and a memory 201.
  • FIG. 20 is only an example of the terminal device 20, and does not constitute a limitation on the terminal device 20. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. , For example, can also include input and output devices, network access devices, and so on.
  • the so-called processor 200 may be a central processing unit (Central Processing Unit, CPU), and the processor 200 may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSPs), and application specific integrated circuits (Application Specific Integrated Circuits). , ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the memory 201 may be an internal storage unit of the terminal device 20, such as a hard disk or a memory of the terminal device 20. In other embodiments, the memory 201 may also be an external storage device of the ** device/terminal device 20, for example, a plug-in hard disk equipped on the terminal device 20, a smart memory card (Smart Media Card, SMC). ), Secure Digital (SD) card, Flash Card, etc. Further, the memory 201 may also include both an internal storage unit of the terminal device 20 and an external storage device. The memory 201 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program. The memory 201 can also be used to temporarily store data that has been output or will be output.
  • BootLoader boot loader
  • An embodiment of the present application also provides a network device, which includes: at least one processor, a memory, and a computer program stored in the memory and running on the at least one processor, and the processor executes The computer program implements the steps in any of the foregoing method embodiments.
  • the embodiments of the present application also provide a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in each of the foregoing method embodiments can be realized.
  • the embodiments of the present application provide a computer program product.
  • the steps in the foregoing method embodiments can be realized when the mobile terminal is executed.
  • the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium.
  • the computer program can be stored in a computer-readable storage medium.
  • the computer program can be stored in a computer-readable storage medium.
  • the steps of the foregoing method embodiments can be implemented.
  • the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate forms.
  • the computer-readable medium may at least include: any entity or device capable of carrying the computer program code to the photographing device/terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), and random access memory (RAM, Random Access Memory), electric carrier signal, telecommunications signal and software distribution medium.
  • ROM read-only memory
  • RAM random access memory
  • electric carrier signal telecommunications signal and software distribution medium.
  • U disk mobile hard disk, floppy disk or CD-ROM, etc.
  • computer-readable media cannot be electrical carrier signals and telecommunication signals.
  • the disclosed apparatus/network equipment and method may be implemented in other ways.
  • the device/network device embodiments described above are only illustrative.
  • the division of the modules or units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units.
  • components can be combined or integrated into another system, or some features can be omitted or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Veterinary Medicine (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Cardiology (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

一种基于人工智能（Artificial Intelligence，AI）心电信号的预测方法、装置、终端（210、320、20）以及存储介质，适用于生命特征识别技术领域，该方法包括：获取目标用户的心电信号（S401）；将心电信号导入预设的房颤信号分类模型，获得房颤信号分类模型输出的心电信号的信号类别（S402），其中，房颤信号分类模型是以房颤患者作为模型训练样本训练得到的；根据心电信号的信号类别，计算房颤将要发作的风险程度，预测目标用户是否即将要发生房颤（S403）。能够实现对房颤事件的预测，方便了用户确定自身的身体状况，提高了检测效果以及用户的使用体验。

Description

一种心电信号的预测方法、装置、终端以及存储介质
本申请要求于2019年12月18日提交国家知识产权局、申请号为201911307688.7、申请名称为“一种心电信号的检测方法、装置、终端以及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请属于生命特征识别技术领域，尤其涉及一种心电信号的预测方法、装置、终端以及存储介质。
背景技术
心房颤动，简称房颤，是最常见的持续性慢性心律失常，常源于无序的心房活动和不规则的心房压缩。随着人们对健康重视程度的不断提高，以及房颤的发病率逐渐上升，如何识别用户是否存在房颤事件则显得尤为重要。现有的房颤检测技术，只能在用户发生房颤事件时，才能够判定该用户为房颤患者，但房颤事件发生时有可能会危急到用户的生命安全，影响用户的健康。由此可见，现有的房颤检测技术，只能告知用户是否发生房颤事件，无法对用户的房颤事件进行预测，检测效果差。
发明内容
本申请实施例提供了一种心电信号的预测方法、装置、终端以及存储介质，可以解决现有的房颤检测技术，无法对用户的房颤事件进行预测，检测效果差的问题。
第一方面，本申请实施例提供了一种心电信号的预测方法，包括：
获取目标用户的心电信号；
将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的；
根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤。
在第一方面的一种可能的实现方式中，在所述将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的之前，还包括：
获取由多个采集次序连续的历史信号组成的训练信号集；所述训练信号集至少包含一个房颤信号；
从所述训练信号集中提取除所述房颤信号外的风险信号；
通过所述训练信号集的所述风险信号，对预设的原生分类模型进行训练，得到所述房颤信号分类模型。
在第一方面的一种可能的实现方式中，所述通过所述训练信号集的所述风险信号，对预设的原生分类模型进行训练，得到所述房颤信号分类模型，包括：
根据所述风险信号的采集时间与关联的房颤信号的触发时间之间的时间差值，确定每个所述风险信号对应的信号类别；
计算所述风险信号在各个预设的信号特征维度的特征值，得到所述风险信号的信号特征参量；
根据所述信号特征参量以及所述信号类别，进行训练，得到所述房颤信号分类模型。
在第一方面的一种可能的实现方式中，所述获取由多个采集次序连续的心电信号组成的训练信号集，包括：
获取采集所述历史信号时所述训练用户的运动参量；
确定所述历史信号的抖动时长；
根据所述抖动时长以及所述运动参量，判断所述历史信号是否为有效信号；
根据各个所述有效信号的采集时间的先后次序，对所有所述有效信息进行封装，得到所述训练信号集。
在第一方面的一种可能的实现方式中，所述将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的，包括：
根据所述心电信号，确定所述目标用户的生命体征参数；
基于所述生命体征参数调整所述房颤信号分类模型的分类阈值；
通过调整后的所述房颤信号分类模型，识别所述心电信号的信号类别。
在第一方面的一种可能的实现方式中，所述根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤，包括：
统计所述目标用户在预设的时间段内存在所述信号类别为风险类别的所述心电信号的信号个数；
根据所述信号个数计算所述房颤发生概率。
在第一方面的一种可能的实现方式中，若所述房颤发生概率大于预设的概率阈值，在所述根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤之后，还包括：
确定所述目标用户预设的关联用户，并向所述关联用户的终端发送告警信息；和/或
根据所述目标用户当前的位置信息，获取与所述位置信息距离最近的医院地址；
根据所述位置信息以及所述医院地址，生成就诊路径。
在第一方面的一种可能的实现方式中，在所述根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤之后，还包括：若接收到新增心电信号，则通过所述房颤信号分类模型识别所述新增心电信号的新增信号类别；
基于所述新增信号类别，重新计算所述目标用户的所述房颤发生概率；
根据已识别的所有所述房颤发生概率，生成房颤概率曲线。
第二方面，本申请实施例提供了一种心电信号的预测装置，包括：
心电信号获取单元，用于获取目标用户的心电信号；
信号类别识别单元，用于将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以 房颤患者作为模型训练样本训练得到的；
房颤发送概率计算单元，用于根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤。
第三方面，本申请实施例提供了一种终端设备，存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时实现上述第一方面中任一项所述心电信号的预测方法。
第四方面，本申请实施例提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现上述第一方面中任一项所述心电信号的预测方法。
第五方面，本申请实施例提供了一种计算机程序产品，当计算机程序产品在终端设备上运行时，使得终端设备执行上述第一方面中任一项所述心电信号的预测方法。
可以理解的是，上述第二方面至第五方面的有益效果可以参见上述第一方面中的相关描述，在此不再赘述。
本申请实施例与现有技术相比存在的有益效果是：
本申请实施例通过将采集得到的心电信号导入到预设的房颤信号分类模型，分别确定每一个心电信号的信号类别，并在获取了该用户在多个采集周期内的心电信号的信号类别，通过多个心电信号的信号类别，计算目标用户的房颤发生概率，并根据该房颤发生概率的数值大小，能够实现对房颤事件的预测，方便了用户确定自身的身体状况，提高了检测效果以及用户的使用体验。
附图说明
图1是本申请实施例提供的手机的部分结构的框图；
图2是本申请一实施例提供的心电信号的预测***的结构框图；
图3是本申请另一实施例提供的心电信号的预测***的结构框图；
图4是本申请第一实施例提供的一种心电信号的预测方法的实现流程图；
图5是本申请一实施例提供的房颤发生概率的输出示意图；
图6是本申请一实施例提供的房颤患者与正常患者的相关标志物的比对示意图；
图7是本申请第二实施例提供的一种心电信号的预测方法具体实现流程图；
图8是本申请一实施例提供的训练信号集的示意图；
图9是本申请第三实施例提供的一种心电信号的预测方法S703具体实现流程图；
图10是本申请一实施例提供的信号类别的标记示意图；
图11是本申请第四实施例提供的一种心电信号的预测方法S4022具体实现流程图；
图12是本申请第五实施例提供的一种心电信号的预测方法S402具体实现流程图；
图13是本申请第六实施例提供的一种心电信号的预测方法S403的具体实现流程图；
图14是本申请第七实施例提供的一种心电信号的预测方法具体实现流程图；
图15是本申请一实施例提供的告警信息的输出示意图；
图16是本申请一实施例提供的就诊路径的输出示意图；
图17是本申请第八实施例提供的一种心电信号的预测方法具体实现流程图；
图18是本申请一实施例提供的房颤概率曲线的输出示意图；
图19是本申请一实施例提供的一种心电信号的预测设备的结构框图；
图20是本申请另一实施例提供的一种终端设备的示意图。
具体实施方式
以下描述中，为了说明而不是为了限定，提出了诸如特定***结构、技术之类的具体细节，以便透彻理解本申请实施例。然而，本领域的技术人员应当清楚，在没有这些具体细节的其它实施例中也可以实现本申请。在其它情况中，省略对众所周知的***、装置、电路以及方法的详细说明，以免不必要的细节妨碍本申请的描述。
应当理解，当在本申请说明书和所附权利要求书中使用时，术语“包括”指示所描述特征、整体、步骤、操作、元素和/或组件的存在，但并不排除一个或多个其它特征、整体、步骤、操作、元素、组件和/或其集合的存在或添加。
还应当理解，在本申请说明书和所附权利要求书中使用的术语“和/或”是指相关联列出的项中的一个或多个的任何组合以及所有可能组合，并且包括这些组合。
如在本申请说明书和所附权利要求书中所使用的那样，术语“如果”可以依据上下文被解释为“当...时”或“一旦”或“响应于确定”或“响应于检测到”。类似地，短语“如果确定”或“如果检测到[所描述条件或事件]”可以依据上下文被解释为意指“一旦确定”或“响应于确定”或“一旦检测到[所描述条件或事件]”或“响应于检测到[所描述条件或事件]”。
另外，在本申请说明书和所附权利要求书的描述中，术语“第一”、“第二”、“第三”等仅用于区分描述，而不能理解为指示或暗示相对重要性。
在本申请说明书中描述的参考“一个实施例”或“一些实施例”等意味着在本申请的一个或多个实施例中包括结合该实施例描述的特定特征、结构或特点。由此，在本说明书中的不同之处出现的语句“在一个实施例中”、“在一些实施例中”、“在其他一些实施例中”、“在另外一些实施例中”等不是必然都参考相同的实施例，而是意味着“一个或多个但不是所有的实施例”，除非是以其他方式另外特别强调。术语“包括”、“包含”、“具有”及它们的变形都意味着“包括但不限于”，除非是以其他方式另外特别强调。
本申请实施例提供的心电信号的预测方法可以应用于手机、平板电脑、可穿戴设备、车载设备、增强现实(augmented reality，AR)/虚拟现实(virtual reality，VR)设备、笔记本电脑、超级移动个人计算机(ultra-mobile personal computer，UMPC)、上网本、个人数字助理(personal digital assistant，PDA)等终端设备上，还可以应用于数据库、服务器以及基于终端人工智能的服务响应***，本申请实施例对终端设备的具体类型不作任何限制。
例如，所述终端设备可以是WLAN中的站点(STAION，ST)，可以是蜂窝电话、无绳电话、会话启动协议(Session InitiationProtocol，SIP)电话、无线本地环路(Wireless Local Loop，WLL)站、个人数字处理(Personal Digital Assistant，PDA)设备、具有无线通信功能的手持设备、计算设备或连接到无线调制解调器的其它处理设备、电脑、膝上型计算机、手持式通信设备、手持式计算设备、和/或用于在无线***上进行通信的其它设备以及下一代通信***，例如，5G网络中的移动终端或者未来演进的公共陆地移动网络(Public Land Mobile Network，PLMN)网络中的移动终端等。
作为示例而非限定，当所述终端设备为可穿戴设备时，该可穿戴设备还可以是应用穿戴式技术对日常穿戴进行智能化设计、开发出可以穿戴的设备的总称，如眼镜、手套、手表、服饰及鞋等。可穿戴设备即直接穿在身上，或是整合到用户的衣服或配件的一种便携式设备，通过附着与用户身上，采集用户的房颤信号。可穿戴设备不仅仅是一种硬件设备，更是通过软件支持以及数据交互、云端交互来实现强大的功能。广义穿戴式智能设备包括功能全、尺寸大、可不依赖智能手机实现完整或者部分的功能，如智能手表或智能眼镜等，以及只专注于某一类应用功能，需要和其它设备如智能手机配合使用，如各类进行体征监测的智能手环、智能首饰等。
以所述终端设备为手机为例。图1示出的是与本申请实施例提供的手机的部分结构的框图。参考图1，手机包括：射频(Radio Frequency，RF)电路110、存储器120、输入单元130、显示单元140、传感器150、音频电路160、近场通信模块170、处理器180、以及电源190等部件。本领域技术人员可以理解，图1中示出的手机结构并不构成对手机的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。
下面结合图1对手机的各个构成部件进行具体的介绍：
RF电路110可用于收发信息或通话过程中，信号的接收和发送，特别地，将基站的下行信息接收后，给处理器180处理；另外，将设计上行的数据发送给基站。通常，RF电路包括但不限于天线、至少一个放大器、收发信机、耦合器、低噪声放大器(Low Noise Amplifier，LNA)、双工器等。此外，RF电路110还可以通过无线通信与网络和其他设备通信。上述无线通信可以使用任一通信标准或协议，包括但不限于全球移动通讯***(Global System of Mobile communication，GSM)、通用分组无线服务(General Packet Radio Service，GPRS)、码分多址(Code Division Multiple Access，CDMA)、宽带码分多址(Wideband Code Division Multiple Access,WCDMA)、长期演进(Long Term Evolution,LTE))、电子邮件、短消息服务(Short Messaging Service，SMS)等，通过RF电路110接收其他终端反馈的关于用户的房颤信号。
存储器120可用于存储软件程序以及模块，处理器180通过运行存储在存储器120的软件程序以及模块，从而执行手机的各种功能应用以及数据处理，例如将接收到的房颤信号存储于存储器120内。存储器120可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作***、至少一个功能所需的应用程序(比如声音播放功能、图像播放功能等)等；存储数据区可存储根据手机的使用所创建的数据(比如音频数据、电话本等)等。此外，存储器120可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。
输入单元130可用于接收输入的数字或字符信息，以及产生与手机100的用户设置以及功能控制有关的键信号输入。具体地，输入单元130可包括触控面板131以及其他输入设备132。触控面板131，也称为触摸屏，可收集用户在其上或附近的触摸操作(比如用户使用手指、触笔等任何适合的物体或附件在触控面板131上或在触控面板131附近的操作)，并根据预先设定的程式驱动相应的连接装置。
显示单元140可用于显示由用户输入的信息或提供给用户的信息以及手机的各种菜单，例如输出接收到的用户的房颤信号，以及确定了房颤信号的类别后，输出的识 别结果。显示单元140可包括显示面板141，可选的，可以采用液晶显示器(Liquid Crystal Display，LCD)、有机发光二极管(Organic Light-Emitting Diode,OLED)等形式来配置显示面板141。进一步的，触控面板131可覆盖显示面板141，当触控面板131检测到在其上或附近的触摸操作后，传送给处理器180以确定触摸事件的类型，随后处理器180根据触摸事件的类型在显示面板141上提供相应的视觉输出。虽然在图1中，触控面板131与显示面板141是作为两个独立的部件来实现手机的输入和输入功能，但是在某些实施例中，可以将触控面板131与显示面板141集成而实现手机的输入和输出功能。
手机100还可包括至少一种传感器150，比如光传感器、运动传感器以及其他传感器。具体地，光传感器可包括环境光传感器及接近传感器，其中，环境光传感器可根据环境光线的明暗来调节显示面板141的亮度，接近传感器可在手机移动到耳边时，关闭显示面板141和/或背光。作为运动传感器的一种，加速计传感器可检测各个方向上(一般为三轴)加速度的大小，静止时可检测出重力的大小及方向，可用于识别手机姿态的应用(比如横竖屏切换、相关游戏、磁力计姿态校准)、振动识别相关功能(比如计步器、敲击)等；至于手机还可配置的陀螺仪、气压计、湿度计、温度计、红外线传感器等其他传感器，在此不再赘述。进一步地，若终端设备可以用于采集用户的房颤信号，则终端设备上还可以配置有心电传感器，通过心电传感器获取用户的心电信号。
音频电路160、扬声器161，传声器162可提供用户与手机之间的音频接口。音频电路160可将接收到的音频数据转换后的电信号，传输到扬声器161，由扬声器161转换为声音信号输出；另一方面，传声器162将收集的声音信号转换为电信号，由音频电路160接收后转换为音频数据，再将音频数据输出处理器180处理后，经RF电路110以发送给比如另一手机，或者将音频数据输出至存储器120以便进一步处理。例如，终端设备可以通过音频电路160，播放心电信号的预测结果，通过语音信号的方式通知用户。
终端设备可以通过近场通信模块170可以接收其他设备发送的房颤信号，例如该近场通信模块170集成有蓝牙通信模块，通过蓝牙通信模块与可佩戴设备建立通信连接，并接收可佩戴设备反馈的房颤信号。虽然图1示出了近场通信模块170，但是可以理解的是，其并不属于手机100的必须构成，完全可以根据需要在不改变申请的本质的范围内而省略。
处理器180是手机的控制中心，利用各种接口和线路连接整个手机的各个部分，通过运行或执行存储在存储器120内的软件程序和/或模块，以及调用存储在存储器120内的数据，执行手机的各种功能和处理数据，从而对手机进行整体监控。可选的，处理器180可包括一个或多个处理单元；优选的，处理器180可集成应用处理器和调制解调处理器，其中，应用处理器主要处理操作***、用户界面和应用程序等，调制解调处理器主要处理无线通信。可以理解的是，上述调制解调处理器也可以不集成到处理器180中。
手机100还包括给各个部件供电的电源190(比如电池)，优选的，电源可以通过电源管理***与处理器180逻辑相连，从而通过电源管理***实现管理充电、放电、 以及功耗管理等功能。
图2示出了本申请一实施例提供的心电信号的预测***的结构框图。参见图2所示，该心电信号的预测***包括移动终端210以及可穿戴设备220。其中，移动终端210与可穿戴设备220可以通过近场通信方式建立通信连接。
其中，本申请提供的心电信号的预测装置具体为用户使用的移动终端210。移动终端210可以接收可穿戴设备220发送的心电信号，例如通过蓝牙通信方式或WIFI通信方式等近场通信方式与可穿戴设备220建立通信连接，并接收可穿戴设备220发送的心电信号，还可以通过有线通信方式或无线通信方式等与其他存储有目标用户的远端通信终端建立通信，并接收远端通信终端发送的心电信号，并通过接收多个不同用户的心电信号，配置房颤信号分类模型，并根据房颤信号分类模型识别可穿戴设备220在多个采集周期反馈的心电信号的信号类型。并根据多个心电信号的信号类型计算目标用户的房颤发生概率，以便对用户的房颤事件进行预测。
可穿戴设备220具体用于采集用户的生物特征信号，该生物特征信号可以为原始采集到的心电信号，也可以经过房颤识别算法处理生成的房颤信号。若采集得到为心电信号，则可以将原始的心电信号发送给移动终端210，通过移动终端210对心电信号进行房颤识别，并转换为房颤信号；也可以通过可穿戴设备220内置的处理模块，将心电信号转换为房颤信号后，将房颤信号发送给移动终端210。
优选地，心电信号的预测***还包括云端服务器230。其中，云端服务器可以接收各个其他电子设备240反馈的心电信号，特别地，上述电子设备240反馈的为房颤患者的心电信号，基于上述房颤患者的心电信号构建上述的房颤信号分类模型，并向各个终端设备210发布上述房颤信号分类模型，终端设备210通过下载云端服务器230的分类模型即可，无需在本地进行模型构建。
图3示出了本申请另一实施例提供的心电信号的预测***的结构框图。参见图3所示，该心电信号的预测***包括服务器310、终端设备320以及可穿戴设备330。其中，服务器310与终端设备320以及可穿戴设备330之间可以通过有线和/或无线网络进行通信。
其中，服务器310可以接收多个可穿戴设备330发送的房颤信号，即服务器属于云端设备。可选地，若可穿戴设备330无法连接互联网，在该情况下，可穿戴设备330可以通过近场通信方式与终端设备320建立通信连接，在该情况下，终端设备320可以安装有服务器310关联的客户端程序，可穿戴设备330将心电信号发送给终端设备320进行存储，并在终端设备320运行上述客户端程序时，通过客户端程序对心电信号进行封装，并将封装后的数据包发送给服务器310。在服务器接收到可穿戴设备发送的心电信号后，可以通过内置的房颤信号分类模型识别心电信号的信号类型，并信号类型反馈给目标用户对应的终端设备320或者发送给可穿戴设备330；当然，若服务器310为不同的用户配置对应的数据库，则可以根据心电信号的用户标识，将该心电信号存储到该目标用户关联的数据库内，根据该数据库内目标用户的所有历史信号的信号类别，计算目标用户的房颤发生概率，并将房颤发生概率返回给可穿戴设备330或者终端设备320进行显示输出。
终端设备320可以在接收到目标用户本次采集的心电信号的信号类别后，可以根 据已采集得到所有的心电信号的信号类别，确定目标用户的房颤发生概率，并通过显示界面输出上述房颤发生概率。其中，终端设备320还可以配置有通信模块，例如蓝牙通信模块或WIFI通信模块，通过通信模块接收用户使用的可穿戴设备330反馈的心电信号，并将房颤信号通过安装于本地的客户端程序发送给服务器310。可选地，终端设备320的客户端程序安装有房颤发生概率的转换模型，终端设备320将所有已采集的心电信号的信号类别根据采集时间的先后次序生成一个信号类别序列，将该信号类别序列导入到上述的转换模型，计算出目标用户的房颤发生概率。
需要说明的是，服务器310内可以存储有可穿戴设备330与终端设备320之间的对应关系。其中，存在对应关系的可穿戴设备330以及终端设备320可以属于同一实体用户，即用户可以通过可穿戴设备330采集心电信号，并上传至服务器310，服务器310在计算出心电信号对应的信号类别或者通过多个心电信号的信号类别计算出房颤发生概率后，将输出结果反馈给终端设备320进行显示。当然，若目标用户存在关联用户，则服务器310可以将上述的输出结果发送给目标用户以及关联用户的终端设备320。
可穿戴设备330可以为一智能可穿戴设备，例如智能手环或智能手表，该可穿戴设备可以安装有与服务器310匹配的客户端程序，并在采集得到用户的心电信号后，通过客户端程序将采集得到的信号发送给服务器310。当然，可穿戴设备330可以将心电信号发送给与其存在同一环境下的终端设备320，并通过终端设备320上传给服务器310。
作为示例而非限定，目标用户为老人，而目标用户的关联用户的老人的子女。可穿戴设备330佩戴与老人的手腕上，在可穿戴设备330采集到老人的心电信号后，可以通过内置的通信模块将心电信号直接发送给服务器310，或者将心电信号发送给老人的终端设备320，通过终端设备320转发给服务器310。服务器310在接收到心电信号后，可以通过房颤信号分类模型识别该心电信号的信号类别，可以直接将信号类别反馈老人的终端设备320，也可以在计算到老人的房颤发生概率后，将上述的房颤发生概率发送给老人的终端设备320以及老人的子女的终端设备320。优选地，服务器310在检测到老人的房颤发生概率的数值大于预设的概率阈值时，将老人的房颤发生概率发送给老人的子女的终端设备320。终端设备320接收到房颤发生概率后，可以通过交互模型进行输出，用以告知老人当前的身体情况。
在本申请实施例中，流程的执行主体为安装有心电信号的预测程序的设备。作为示例而非限定，心电信号的预测程序的设备具体可以为终端设备，该终端设备可以为用户使用的智能手机、平板电脑、笔记本电脑等，对获取得到的目标用户的心电信号进行类别识别，并确定该目标用户的房颤发生概率。图1示出了本申请第一实施例提供的心电信号的预测方法的实现流程图，详述如下：
在S401中，获取目标用户的心电信号。
在本实施例中，心电信号可以为通过可穿戴设备或心电采集设备获取得到的心电信号，上述可穿戴设备以及心电采集设备可以配置有心电传感器，例如心电图(Electrocardiogram，ECG)传感器或光电容描记(Photoplethysmograph，PPG)传感 器，可以用于获取佩戴用户或所检测用户的心电信号。该可穿戴设备可以为手环或手表等能够与用户皮肤接触的设备，通过检测接触区域的血管扩张情况，获取所佩戴用户的心率值，并基于各个采集时刻所对应的心率值，生成用户的心电信号。当然，若终端设备配置有心电传感器，则可以通过心电传感器采集用户的心电信号。可选地，终端设备在获取得到心电信号后，可以通过预设的信号优化算法对心电信号进行预处理，从而能够提高后续类型识别的准确性。其中，优化的方式包括但不限于以下一种或多种的组合：信号放大、信号滤波、异常检测、信号修复等。
其中，异常检测具体为根据采集得到的原始心电信号的信号波形，提取多个波形特征参数，例如信号最大连续时长、波形中断次数、采集中断时长、波形信噪比等，并根据上述采集得到波形特征值计算心电信号的信号质量，若检测到该信号质量低于有效信号阈值，则识别心电信号为无效信号，不对无效信号执行后续信号类别的识别操作。反之，若该信号质量高于有效信号阈值，则识别心电信号为有效信号，执行S402以及S403的操作。
其中，信号修复具体为通过预设的波形拟合算法对采集心电信号过程中的中断区域进行波形拟合，生成连续的心电波形。该波形拟合算法可以为一神经网络，通过采集目标用户的历史信号，对波形拟合算法中的参数进行调整，以使得拟合后的心电信号的波形走向与目标用户的波形走向相匹配，从而提高了波形拟合效果。优选地，该信号修复操作在上述异常检测操作之后执行，由于通过信号修改心电信号缺失的波形时，会提高心电信号的采集指令，从而影响异常检测的操作，从而无法对采集质量较差的异常信号进行识别，基于此，终端设备可以先通过异常检测算法，判断心电信号是否有效信号；若该心电信号为有效信号，则通过信号修复算法对心电信号进行信号修复；反之，若心电信号为异常信号，则无需进行信号修复，从而减少了不必要的修复操作。
在一种可能的实现方式中，该可穿戴设备或心电采集设备可以设置有存储单元，上述两个设备将采集到用户的心电信号存储于存储单元内，在检测到存储单元内存储的心电信号满足预设的上传阈值时，例如心电信号的数据量大于预设的数据量阈值，或者心电信号的采集时长满足对应的上传阈值时，可以将已存储的心电信号进行封装，并发送给终端设备。
可选地，在本实施例中，终端设备可以接收其他设备或通过内置的传感器获取目标用户的心电信号，若是通过其他设备，例如可穿戴设备或心电采集设备等获取目标用户的心电信号，则可以为上述不同的采集设备配置对应的心电信号反馈周期，采集设备可以根据该反馈周期，定期向终端设备发送用户的房颤信号，举例性地，该采集周期的长度为48秒，采集设备可以以48s为一个采集周期，将采集到的心电信号生成一个信号段，并且每48s向终端设备发送一个心电信号段，终端设备可以根据分别识别各个心电信号段对应的信号类型。可选地，若采集设备配置有与终端设备对应的客户端程序，在用户启动该客户端程序时，则可以将上次反馈心电信号时刻至当前客户端程序的启动时刻之间所采集得到的心电信号，通过客户端程序进行封装，并将上述两个时刻之间的心电信号发送给终端设备，通过终端设备统一对所有心电信号进行类别识别，无需可穿戴设备长时间与终端设备建立长连接，只有在客户端程序启动时， 才建立上述两者之间的通信链接，从而减少了心电采集装置以及终端设备的能耗，提高了设备的续航能力。
在S402中，将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的。
在本实施例中，终端设备预设有房颤信号分类模型，将采集到的心电信号导入到房颤信号分类模型内，可以确定心电信号对应的信号类别。该信号类别可以用于指示房颤事件的发生概率。作为示例而非限定，终端设备可以将心电信号划分为两个类别，分别为第一型信号以及第二型信号。其中，第一型信号表示存在较大风险发生房颤事件；而第二型信号则表示用户发生房颤事件的概率较低。当然，终端设备可以根据发生房颤事件的概率大小，划分为不同的层级，不同的层级对应一个信号类别，例如将发颤发生概率划分为N个层级，其中层级越高，则对应的房颤事件的发生概率越小；而层级越低，则对应的房颤事件的发声概率越高，在该情况下，第N个层级对应的房颤发生概率最低；而第1个层级对应的房颤发生概率最高，终端设备可以根据上述的房颤分类模型，识别心电信号所述的层级类别。
与现有的房颤技术相比，本申请实施例并非在发生房颤事件时，判定心电信号是否为房颤信号，而是可以在非房颤事件发生时，对非房颤的心电信号的信号类别进一步标记，从而实现对用户的房颤事件进行预测，实现了提早告知用户的目的，避免用户在没有预警的情况下发生房颤事件，提高了检测效果以及检测范围。
需要说明的是，该房颤信号分类模型也可以用于识别房颤信号，即用户发生房颤事件时对应的心电信号。此时，发生房颤事件的心电信号对应的信号类别为房颤信号类别。
在一种可能的实现方式中，房颤信号分类模型内可以设置有不同类别的标准信号。在该情况下，终端设备可以直接将心电信号作为模型的输入参量，房颤信号分类模型可以将输入的心电信号与各个类别的标准信号进行匹配，并基于匹配结果识别心电信号的信号类别。作为示例而非限定，终端设备可以分别计算识别各个标准信号与心电信号的匹配度，并选取匹配度的数值最大的一个标准信号的信号类别作为心电信号的信号类型；若存在两个或以上的标准信号与心电信号的匹配度数值一致且最大，此时可以通过预设的信号分割算法，将心电信号划分为多个信号段，并分别计算各个信号段与上述匹配度最高的多个候选类别的标准信号之间匹配度，识别各个信号段关联的子类别，基于所有信号段的子类别识别整体心电信号的信号类别。
举例性地，终端设备识别到心电信号与第一类别的标准信号之间的匹配度，与心电信号与第二类别的标准信号之间的匹配度的数值相同，且均最大。此时终端设备可以将心电信号划分为3个信号段，分别为第一信号段、第二信号段以及第三信号段，并分别计算上述三个信号段与上述两个类别的标准信号之间的匹配度，并确定每个信号段对应的子类别，识别结果如下：第一信号段(第一类别)、第二信号段(第一类别)、第三信号段(第二类别)，因此可以确定与第一类别匹配的信号段的个数大于第二类型匹配的信号段，则判定心电信号的信号类别为第一类别。
可选地，终端设备在计算心电信号与标准信号之间的匹配度之前，可以对心电信 号进行标准化处理。终端设备根据标准信号的标准时长，调整心电信号的信号时长，以使标准时长与调整后的信号时长相同。若心电信号的信号时长大于标准时长，则可以对心电信号进行截取，截取后的心电信号与标准信号的信号时长一致；若心电信号的信号时长小于标准时长，则可以通过信号拉伸、循环延伸等方式对不足的区域进行填充，以使调整后的心电信号的信号时长与标准信号的标准时长一致。在进行标准化处理后，则可以执行信号之间的匹配度计算，并基于匹配度识别心电信号的信号类别。
可选地，在计算心电信号与标准信号之间的匹配度时，可以采用动态时间规整算法，实现的方式具体如下：终端设备可以根据各个采集时间节点采集到的心率数值，将心电信号转换为心率序列。根据心率序列以及标准信号对应的标准序列内包含的元素个数，生成对应的坐标网格，并将各个坐标网格相交点所对应的元素之间的差值，作为该坐标网格对应的元素距离值，在计算路径的总距离时，将经过的各个网格交点对应的元素距离值进行叠加，则得到该路径对应的总距离值，选取总距离值最小的路径作为心率序列与标准序列之间的特征路径，并将该特征路径对应的距离值作为心电信号与标准信号之间的距离值，基于距离值计算上述两个信号之间的匹配度。
举例性地，目标用户反馈的心电信号具有M个采集时刻，根据每个采集时刻对应的心率值，则生成包含M个元素的心率序列，标准信号对应的标准序列包含有N个元素，则根据上述两个序列可以生成一个M*N的坐标网格，而坐标(m,n)的坐标距离值即为心率序列中第m个元素与标准序列中第n个元素之间的距离值，并计算到达目标点(M,N)的所有路径中，总距离值最小的一个路径作为特征路径，将特征路径对应的总距离值作为心电信号与标准信号之间的距离值。
在一种可能的实现方式中，终端设备可以计算心电信号在多个特征维度的特征参数，将心电信号转换为信号特征序列，并将信号特征序列作为房颤信号分类模型的输入，输出心电信号的信号类型。可选地，该房颤信号分类模型还可以设置有特征提取层，终端设备将心电信号输入到房颤信号分类模型内，可以通过模型内的特征提取层将心电信号转换为特征序列，并将特征序列发送给与特征提取层串联的信号分类层，计算该特征序列与各个候选类别之间的匹配度，并基于所述匹配度作为该候选类别对应的概率值，并将所有概率值导入到全连接层，输出心电信号的信号类别。其中，上述心电信号的信号特征包括但不限于以下一种或多种的组合：最大心率值、最小心率值、第一心率时长(即超过第一心率阈值的时长)、第二心率时长(即小于第二心率阈值的时长)等。可选地，终端设备除了根据当前反馈的心电信号确定信号特征值外，还可以根据心电信号以及与该心电信号的采集时间相邻的关联信号，确定联合特征值，例如平均心率的变化率等，将上述单个信号确定的信号特征值以及上述的联合特征值构成上述的信号特征序列。
在S403中，根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤。
在本实施例中，终端设备在识别了心电信号的信号类别后，可以根据该目标用户的多个心电信号的信号类别，预测目标用户发生房颤事件的概率值，即上述的房颤发生概率。由于房颤时间触发之前，人体的生命体征有一定的征兆，即越靠近房颤事件的心电信号具有一定的共性，通过判定心电信号的信号类别，则可以确定该心电信号 是否符合靠近房颤时间的心电信号的共性特征，从而根据心电信号的信号类别，可以确定该用户发生房颤事件的发生概率。而通过多个心电信号的信号类别，可以确定该目标用户的心电信号与发生房颤事件之前的心电信号存在共性的这一现象，是偶发性或是频发性，并根据判断结果能够更为准确地确定目标用户的房颤发生概率，实现对用户的房颤事件进行预测，无需用户在发生房颤事件时才对用户进行告警。
在一种可能的实现方式中，终端设备可以根据各个心电信号采集时间的先后次序，将各个心电信号的信号类别依次组合，生成信号类别序列。将信号类别序列导入到概率转换函数，计算出目标用户的房颤发生概率。
在一种可能的实现方式中，终端设备可以根据各个心电信号的信号类别，生成对应的概率值，将多个概率值进行加权叠加，计算出该目标用户的房颤发生概率。其中，每个心电信号的加权值根据与当前时间之间的差值确定，与当前时间的差值越小，则对应的加权权重越大；反之，若与当前时间的差值越大，则对应的加权权重越小。可选地，上述的计算房颤发生概率的算法具体可以为：
Figure PCTCN2020136538-appb-000001
其中，Probability为上述房颤发生概率；SignalType i为第i个心电信号的信号类型对应的概率值；CurrentTime为当前时间；CollectTime i为第i个心电信号的采集时间，N为上述获取的心电信号的总个数。
在本实施例中，终端设备可以通过内置的交互模块将房颤发生概率输出给用户。其中，输出的方式包括但不限于：通过通知方式进行输出、通过播报的方式进行输出，或者在预设的界面上添加用于实时反馈房颤发生概率的组件，通过调整组件内对应的参数值进行输出。
在一种可能的实现方式中，终端设备可以获取与房颤发生概率关联的解释语段以及房颤咨询，并将房颤发生概率、解释语段以及房颤咨询生成房颤报告。用户可以通过房颤报告获取更多与该房颤触发概率的相关的信息，提高了房颤发生概率的可读性，方便用户了解自己的身体状态。其中，终端设备可以根据当前用户的房颤发生概率，从云端数据库中为用户推送对应的房颤咨询，从而提高了推送信息与用户之间的匹配度，实现了精准推送信息的目的。
图5示出了本申请一实施例提供的房颤发生概率的输出示意图。参见图5所示，终端设备可以根据多个心电信号的信号类型，确定目标用户的房颤发生概率，在图中显示的数值为88，并在房颤发生概率的下配置有对应的解释语段，即“您在未来短时间内可能发生房颤”，并在解释语段下方添加有房颤资讯，用户可以通过点击房颤资讯对应的UI控件，阅读具体的资讯内容。
在一种可能的实现方式中，现有房颤技术可以通过检测用户体内与房颤相关的标志物含量，预测用户发生房颤事件的概率，例如可以通过检测用户的BNP含量或者FGF-23的含量，来判定用户的房颤事件发生概率。图6示出了本申请一实施例提供的房颤患者与正常患者的相关标志物的比对示意图。参见图6所示，房颤患者的FGF-23以及BNP的含量均高于正常用户，可以通过设置对应的含量阈值，以确定该用户是否为房颤患者，从而预测用户是否会触发房颤事件。而由于与房颤相关的标志物的含量 高低，不但与房颤相关，还会与其他疾病相关，并不能通过标志物直接判定用户存在房颤行为，即上述方式的识别准确率较低。
在一种可能的实现方式中，现有房颤技术可以根据多个对发生房颤事件存在较大关系的风险因素配置对应的风险评分表，并为每个风险因素配置对应的贡献值，通过获取用户信息与各个风险因素进行比对，判断该用户包含的风险项，基于所有风险项对应的贡献值，计算该用户的总评分值，根据总评分值确定用户判定该用户是否为房颤患者，从而对用户的房颤事件进行预测。作为示例而非限定，表1示出了本申请一实施例提供的房颤风险评分表。该房颤风险评分表采用了CHADS评分方式以及CHA2DS2-VASc评分方式，并包含有以下风险因素：充血性心力衰竭/左心室功能障碍、高血压、年龄75岁或以上、糖尿病、脑卒中/TIA/血栓栓塞病史、血管性疾病、年龄65～74岁、性别(女性)共8个项目。每个项目有对应的贡献值。终端设备可以通过比对用户信息匹配的风险因素的项数以及对应的分值，确定该用户的评分值，从而预测用户发生房颤事件的概率。然而上述方式没有考虑个体的个性化差异，从而导致了评分准确性较低，而且通过风险因素可以确定，该评分方式主要针对老年人，无法覆盖所有人群，适用范围小。
Figure PCTCN2020136538-appb-000002
表1
与上述两种技术不同的是，本申请实施例可以通过在日常过程中采集用户的心电信号，并分别确定各个用户的心电信号的信号类别，并通过多个心电信号的信号类别计算用户的房颤发生概率，通过一段时间对用户的心电信号观察，能够对用户的房颤发生概率进行精准预测，提高了预测准确性，而且心电信号的采集适用于所有人群，从而扩大了适用范围。
以上可以看出，本申请实施例提供的一种心电信号的预测方法通过将采集得到的心电信号导入到预设的房颤信号分类模型，分别确定每一个心电信号的信号类别，并在获取了该用户在多个采集周期内的心电信号的信号类别，通过多个心电信号的信号类别，计算目标用户的房颤发生概率，并根据该房颤发生概率的数值大小，能够实现对房颤事件的预测，方便了用户确定自身的身体状况，提高了检测效果以及用户的使用体验。
图7示出了本申请第二实施例提供的一种心电信号的预测方法的具体实现流程图。 参见图7，相对于图4所述实施例，本实施例提供的一种心电信号的预测方法中在所述将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的之前，还包括：S701～S703，具体详述如下：
在S701中，获取由多个采集次序连续的历史信号组成的训练信号集；所述训练信号集至少包含一个房颤信号。
在本实施例中，终端设备在对心电信号进行识别之前，可以通过训练信号集对预设的原生训练模型进行训练，已得到可以用于进行心电信号的信号类别进行识别的房颤分类模型。基于此，终端设备需要可以获取多个训练信号集。其中，获取的方式可以为：若终端设备为用户使用的智能手机或平板电脑等智能设备，则智能设备可以通过与云端数据库建立通信连接，从云端数据库处下载多个训练信号集；若终端设备为服务器，服务器可以用于接收各个电子设备反馈的心电信号，并存储于本地的数据库内，在该情况下，服务器可以直接从本地数据库提取上述训练信号集。其中，每个训练信号集包含有多个历史信号，并且每个历史信号在信号集中的次序与其采集时间的先后次序相匹配。
在一种可能的实现方式中，上述训练信号集具体为房颤患者的心电信号构成的训练信号集。由于房颤信号分类模型具体用于判断心电信号的信号特征是否具有房颤事件发生的前兆特征，因此，在对房颤信号分类模型进行训练时，需要采集包含房颤事件发生之前的信号，从房颤事件发生之前的信号中确定上述的前兆特征，从而实现对房颤事件的预测。基于上述原因，终端设备在获取训练信号集时，需要采集包含房颤事件的信号，即房颤信号，而存在房颤信号的用户即为房颤患者。基于此，终端设备可以根据各个用户的房颤患者标识，选取出目标训练用户，并获取各个目标训练用户对应的信号集，作为上述的训练信号集。其中，每一个训练信号集内至少包含一个房颤信号。
在一种可能的实现方式中，若终端设备具体为一用户的智能终端，则S701具体可以为：终端设备可以获取目标用户的用户信息，该用户信息可以用于体现用户的生物特征，包括但不限于以下一种或多种的组合：用户年龄、用户性别、当前的健康状态、患病记录等。终端设备可以从云端数据库中识别与所述用户信息相匹配的训练对象，并获取上述训练对象的心电信号集作为训练信号集，从而能够实现为用户定制化与其相匹配的房颤信号分类模型，从而提高了后续训练的准确性。
在一种可能的实现方式中，若终端设备具体为一云端服务器，即用于响应各个目标用户上传的心电信号的信号类别，则S701具体可以为：终端设备可以根据预设的人群划分规则，将数据库内的已有用户划分为多个用户组，并为不同的用户组配置对应的房颤信号分类模型。基于此，在对房颤信号分类模型进行训练时，则采用与该模型关联的用户组内的训练信号集对其进行训练，从而实现了为不同特征人群配置不同的房颤信号分类模型的目的。在响应的过程中，终端设备会获取目标用户的用户信息，基于用户信息确定目标用户所属的用户组，并通过该用户组关联的房颤信号分类模型，识别目标用户的心电信号。
在S702中，从所述训练信号集中提取除所述房颤信号外的风险信号。
在本实施例中，终端设备可以对训练信号集中的各个历史信号进行房颤信号判定操作，并识别出发生房颤事件时对应的房颤信号，并标记出房颤信号，将除房颤信号外的其他非房颤的历史信号识别为风险信号。其中，终端设备可以配置有房颤特征参数，终端设备可以分别提取各个历史信号的信号特征值，并将各个历史信号的信号特征值与上述房颤特征参量进行匹配，基于匹配结果确定历史信号是否属于房颤信号。
作为示例而非限定，图8示出了本申请一实施例提供的训练信号集的示意图。参见图8所示，该训练信号集内包含有10个历史信号，每个历史信号通过一个矩形区域标识，即PPG柱。该PPG柱的高度可以根据采集时间内的平均心率值的大小确定，若在采集时间内的平均心率值较高，则该PPG的高度越高；反之，若该PPG柱越低，则表示在采集时间内该历史信号对应的平均心率值较低。由于房颤事件是在一定时间内心房跳动极为迅速，有时候高达每分钟200次以上，因此在心电信号的表征上即为平均心率较高。因此，可以通过PPG柱的高度识别出存在房颤事件的心电信号，即图中标记的房颤信号。终端设备在识别得到房颤信号后，则可以将除上述房颤信号外的历史历史信号识别为风险信号。
在S703中，通过所述训练信号集的所述风险信号，对预设的原生分类模型进行训练，得到所述房颤信号分类模型。
在本实施例中，终端设备可以通过获取得到的风险信号进行训练，终端设备可以预先为各个风险信号标记相应的信号类别，根据标记信息以及风险信号构建得到一个训练样本。终端设备将多个风险信号导入到原生分类模型内，计算得到多个预测类别，并分别识别各个预测类别与其对应的风险信号预设的信号类别是否匹配，从而计算得到该原生分类模型对应的预测损失值，并原生分类模型是否收敛且预测损失值小于预设的损失阈值，若是，则识别该原生分类模型已调整完毕，并将调整后的原生分类模型识别为房颤信号分类模型；反之，若不收敛，则需要调整原生分类模型内的学习参量，以使上述的原生分类模型满足收敛且预测损失值小于预设的损失阈值这两个条件。
在一种可能的实现方式中，终端设备可以配置有多个不同类型的原生分类模型，并通过上述的风险信号同时对多个原生分类模型训练学习，并基于上述多个原生分类模型对应的收敛时间以及损失值，基于上述两个参量选取优选的原生分类模型，并根据该优选的原生分类模型构建房颤信号分类模型。
在一种可能的实现方式中，终端设备可以根据风险信号与其关联的房颤信号之间的采集时间差，确定该风险信号的训练权重，该采集时间的差值越小，则对应的训练权重越大。其中，与风险信号关联的房颤信号具体为采集时间在风险信号之后的所有房颤信号中，与该风险信号之间的采集时间差最小发生的房颤信号，即距离风险信号最近的房颤信号。继续参见图8所示，对于历史信号1和历史信号2而言，其关联的房颤信号为历史信号3，而对于历史信号4而言，由于历史信号3的采集时间先于历史信号4，即便该历史信号3与历史信号4之间的采集时间差最小，但并不认为其为关联房颤信号，即历史信号4关联的房颤信号为历史信号9。根据各个风险信号的训练权重进行训练学习，能够提高与房颤事件发生时间较近的风险信号的训练贡献。由于采集时间越近，则包含的房颤事件的前兆特征越多，预测越准确，反之，若采集时间越远，则包含的前兆特征越小。
在本申请实施例中，通过获取训练信号集并对训练信号集内的历史信号进行筛选，提取出风险信号，进行训练学习，得到房颤信号分类模型，能够实现对房颤事件进行预测，提高了预测准确性。
图9示出了本申请第三实施例提供的一种心电信号的预测方法S703的具体实现流程图。参见图9，相对于图7所述实施例，本实施例提供的一种心电信号的预测方法中S703包括：S901～S903，具体详述如下：
进一步地，所述通过所述训练信号集的所述风险信号，对预设的原生分类模型进行训练，得到所述房颤信号分类模型，包括：
在S901中，根据所述风险信号的采集时间与关联的房颤信号的触发时间之间的时间差值，确定每个所述风险信号对应的信号类别。
在本实施例中，终端设备可以识别各个风险信号对应的采集时间，并根据采集时间从训练信号集中选取出风险信号关联的房颤信号，其中，关联的房颤信号具体可以为：采集时间在风险信号之后的所有房颤信号中，与该风险信号之间的采集时间差最小发生的房颤信号，即距离风险信号最近的房颤信号。终端设备可以为不同的信号类别配置对应的相距时间范围，根据风险信号的采集时间与关联的房颤信号的触发时间之间的时间差值所落入的相距时间范围，识别该风险信号对应的信号类别。
作为示例而非限定，终端设备可以将信号类别划分为两个类型，分别为0型信号以及1型信号。其中，0型信号为与关联房颤信号之间采集的时间差在预设的时间阈值范围内，而1型信号为关联房颤信号之间采集的时间差在预设的时间阈值范围外。该时间阈值可以为2小时。例如，某一风险信号的采集时间为15点，而关联的房颤信号的采集时间为16点，上述两个信号的采集时间差为1小时，则可以确定上述风险信号的信号类型为0型；而另一风险信号的采集时间为10点，关联的房颤信号的采集时间为16点，则上述两个采集时间的时间差为6小时，则可以确定上述风险信号的信号类型为1型。图10示出了本申请一实施例提供的信号类别的标记示意图。参见图10所示，每个历史信号的采集周期为1小时，即两个相邻的房颤信号之间的时间差为1小时。其中，时间阈值为2小时，则可以确定历史信号1以及历史信号2与关联的房颤信号(即历史信号3)之间的时间差均在2小时内，则上述两个风险信号均为0型信号；而历史信号4、历史信号5以及历史信号6与关联的房颤信号(即历史信号9)之间的时间差均大于2小时，则上述三个风险信号均为1型，以此类推，确定各个风险信号的信号类型。
当然，终端设备除了可以将房颤类型分为2个类型外，还可以划分为N个类型，并为不同的信号类型配置对应的时间阈值，例如，第一类型的信号与关联的房颤信号之间的采集时间差在t0～t1之间；而第二类型的信号与关联的房颤信号之间的采集时间差在t1～t2之间，…，第N类型的信号与关联的房颤信号之间的采集时间差在tN-1～tN之间。在该情况下，终端设备在生成了通过上述方式训练后的房颤信号分类模型后，在将实际采集到的心电信号导入到上述模型内，判定风险信号的信号类别后，可以终端设备除了确定房颤事件的发生概率外，还可以根据心电信号的信号类别，确定距离下一次发生房颤事件之间的预测时间。由于上述的信号类别与房颤信号的发生事件之间的时间差是一一对应的，因此可以即每个信号类别与对应一个预测时间。终端设备 可以根据心电信号的信号类型，查询与之关联的预测时间，并根据预测时间对用户进行提示。
在S902中，计算所述风险信号在各个预设的信号特征维度的特征值，得到所述风险信号的信号特征参量。
在本实施例中，终端设备可以配置有多个信号特征维度，不同的信号特征维度用于表示心电信号的不同信号特征。终端设备可以对风险信号进行解析，确定该风险信号在各个信号特征维度上的特征值，并根据各个信号特征维度的在参量模板中对应的位置，将各个特征值导入到参量模板内，生成该风险信号的信号特征参量。其中，该信号特征维度包括但不限于以下一种或多种的组合：平均心率、心率最大值、心率最小值、超出第一心率阈值的持续时长、低于第二心率阈值的持续时长等。
在S903中，根据所述信号特征参量以及所述信号类别，进行训练，得到所述房颤信号分类模型。
在本实施例中，终端设备在建立了各个风险信号的信号特征参量以及信号类别后，生成一个训练样本，通过多个训练样本进行训练，调整原生分类模型内的参量，直到结果收敛，并将调整好的原生分类模型识别为房颤信号分类模型。
在本申请实施例中，根据风险信号与关联的房颤信号的之间的时间差进行风险类型的标记，从而能够提取出存在房颤事件的前兆特征的风险信号，从而能够根据识别得到的心电信号的信号类型对房颤事件进行预测，提高了检测效果。
图11示出了本申请第四实施例提供的一种心电信号的预测方法S701的具体实现流程图。参见图11，相对于图7所述实施例，本实施例提供的一种心电信号的预测方法中S701包括：S1101～S1104，具体详述如下：
进一步地，所述获取由多个采集次序连续的心电信号组成的训练信号集，包括：
在S1101中，获取采集所述历史信号时所述训练用户的运动参量。
在本实施例中，终端设备在将训练信号集内的历史信号导入到原生分类模型进行训练之前，可以对训练信号集内的历史信号进行筛选，过滤掉采集质量较差的历史信号，从而能够提高后续训练的准确性。基于此，可穿戴设备在获取用户的心电信号时，除了反馈心电波形外，还可以将关联的运动参量添加到心电信号内，并将上述两个数据反馈给终端设备。终端设备可以根据各个历史信号关联的运动参量，确定采集该训练用户的历史信号时，该训练用户的运动状态。由于用户在进行剧烈运动时，心率值会升高，而该升高并非由于房颤事件引起的，是处于正常运动反应，并且运动过程中，可穿戴设备与用户皮肤之间接触不紧密，会出现例如晃动的现象，从而影响心电信号的采集质量，基于此，终端设备可以根据上述的运动参量，确定该历史信号的有效性。
其中，该运动参量可以通过可穿戴设备内的加速度传感器以及陀螺仪等运动感应模块获取得到，终端设备可以通过上述运动感应模块反馈的感应值，确定历史用户的运动状态。
在S1102中，确定所述历史信号的抖动时长。
在本实施例中，终端设备可以对历史信号的信号波形进行解析，确定存在抖动的波形段，并基于上述波形段的持续时长，作为历史信号的抖动时长。其中，抖动的波形段可以为采集过程中中断区域的波形段，可以为波形变化频率高于正常值所对应的 波形段。
在S1103中，根据所述抖动时长以及所述运动参量，判断所述历史信号是否为有效信号。
在本实施例中，终端设备可以通过上述两个参量计算历史信号的信号质量，并将计算得到的信号质量与预设的质量阈值进行比对，判断该历史信号是否为有效信号。其中，若信号质量大于或等于质量阈值，则识别该历史信号为有效信号；反之，若该信号质量小于质量阈值，则识别该历史信号为无效信号。
在一种可能的实现方式中，计算信号质量的方式可以为：根据抖动时长与历史信号的信号时长之间的比值，确定第一质量因子，其中，抖动时长越长，该第一质量因子的数值越小；根据运动参量与静态运动参量之间的比值，确定第二质量因子，其中，运动参量的数值越大，则表示用户的运动幅度越大，对应的第二质量因子的数值越小。对上述两个质量因子进行加权求和，计算得到该历史信号的信号质量。
在S1104中，根据各个所述有效信号的采集时间的先后次序，对所有所述有效信息进行封装，得到所述训练信号集。
在本实施例中，终端设备根据各个有效信号的采集时间的先后次序，确定各个有效信号在训练信号集内的信号次序，从而将多个有效信号进行封装，过滤无效信号，并构成上述训练信号集，提高后续训练过程的准确性。
在本申请实施例中，通过对历史信号进行筛选，筛选出无效的信号，从而能够提高后续训练操作的准确性。
图12示出了本申请第五实施例提供的一种心电信号的预测方法S402的具体实现流程图。参见图12，相对于图4、图7、图9以及图11任一所述实施例，本实施例提供的一种心电信号的预测方法中S402包括：S1201～S1203，具体详述如下：
进一步地，所述将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的，包括：
在S1201中，根据所述心电信号，确定所述目标用户的生命体征参数。
在本实施例中，终端设备在使用房颤信号分类模型对目标用户的心电信号进行分类之前，可以根据用户的生命体征对房颤信号分类模型进行调整，即为模型预热阶段，从而使得后续的信号类别的识别过程更为准确。
在本实施例中，终端设备可以通过用户输入的方式或者可穿戴设备反馈等方式获取目标用户的生命体征参数。举例性地，该生命特征参数可以为目标用户的静态心率，例如用户在长时间不移动的状态下(包括静坐以及睡眠过程)对应的心率值，以及移动过程中的心率值。通过获取在不同状态下的心率值，可以确定目标用户的心率变化幅度以及心率的基准值，将上述数值作为目标用户的生命体征参数。
在S1202中，基于所述生命体征参数调整所述房颤信号分类模型的分类阈值。
在本实施例中，终端设备可以通过采集得到的目标用户生命体征参数对房颤信号分类模型内的分类阈值进行调整，从而能够使得分类过程与目标用户的身体状态相匹配，实现了个性化定制分类模型的目的。举例性地，终端设备可以根据生命体征参数中静态心率与动态心率之间的差值，确定目标用户的心率浮动幅值，并基于该心率浮 动幅值调整房颤触发阈值，从而能够准确判定用户当前的心率值是否接近上述的房颤触发阈值，以判定心电信号对应的房颤发生概率。
在S1203中，通过调整后的所述房颤信号分类模型，识别所述心电信号的信号类别。
在本实施例中，终端设备在通过目标用户的生命体征参数对房颤信号分类模型的分类阈值进行调整后，可以对目标用户的心电信号进行信号类别的识别操作。具体的识别操作可以参见前述实施例的相关描述，在此不再赘述。
在本申请实施例中，通过获取目标用户的生命体征参数，并基于生命体征参数对房颤信号分类模型进行调整，从而使得房颤信号分类模型与目标用户相匹配，提高了房颤信号的识别准确性。
图13示出了本申请第六实施例提供的一种心电信号的预测方法S403的具体实现流程图。参见图13，相对于图4、图7、图9以及图11任一所述实施例，本实施例提供的一种心电信号的预测方法S403包括：S4031～S4032，具体详述如下：
进一步地，所述根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤，包括：
在S4031中，统计所述目标用户在预设的时间段内存在所述信号类别为风险类别的所述心电信号的信号个数。
在本实施例中，终端设备预设的信号类别中至少包含有两种分类，分别为风险类别以及非风险类别。其中，风险类别的心电信号的信号特征包含较多房颤事件发生前的预兆特征，在该情况下，房颤信号分类模型可以识别上述类型的心电信号为风险类别的心电信号；反之，若心电信号的信号特征并不存在或匹配有较少房颤事件发生前的前兆特征，在该情况下，房颤信号分类模型可以识别上述类型的心电信号为非风险类别的心电信号。当然，根据房颤分类模型的可识别类型个数的不同，风险类别下可以级联有多个子类别，例如第一级风险类别、第二级风险类别等；而非风险类别下也可以级联有多个子类别，具体级联的类别数根据具体的房颤信号分类模型所决定，在此不做限定。
作为示例而非限定，如上述实施例中举例的，终端设备在训练的过程中，可以根据风险信号距离关联的房颤信号之间的时间差划分为N个信号类别。其中，第一信号类别与房颤事件的触发时间距离最近，可以将第一信号类别识别为风险类别；而第二信号类别至第N信号类别与房颤事件的触发时间距离较远，可以将第二信号类别至第N信号类别识别为非风险类别。
在本实施例中，终端设备可以获取预设时间段内所有采集得到的心电信号的信号类别，并统计风险类别为心电信号的信号个数，作为示例而非限定，该预设时间段可以为1天或一周。终端设备在接收到新的心电信号后，统计与该心电信号之间采集时间差在预设时间段内所有心电信号中，信号类别为风险类别的心电信号的个数，即上述的信号个数。例如，某一心电信号的采集时间为12月13日18点，预设时间段为1天，则获取12月12日18时至12月13日18时之间所有采集到的心电信号，并统计获取得到的心电信号中信号类别为风险类别的心电信号的信号个数。
在S4032中，根据所述信号个数计算所述房颤发生概率。
在本实施例中，终端设备根据上述风险类别的心电信号的信号个数，终端设备可以设置有房颤发生概率转换函数，终端设备将上述的信号个数导入转换函数内，计算出采集心电信号时刻目标用户对应的房颤发生概率。
在一种可能的实现方式中，终端设备可以根据上述的信号个数以及预设的时间段，计算目标用户的风险信号的出现频率，并基于上述出现频率确定上述的房颤发生概率。
在一种可能的实现方式中，终端设备可以将上述时间段划分为多个子时间段，并分别统计各个子时间段内对应的风险信号的出现频率，从而生成上述时间段对应的风险信号的出现频率变化曲线，基于上述出现变化曲线确定房颤发生概率。
在本申请实施例中，通过统计在预设时间段内存在风险类别的心电信号的信号个数，并基于信号个数确定房颤发生概率，能够根据心电信号的信号类别，对房颤事件进行预测，提高了检测效果。
图14示出了本申请第七实施例提供的一种心电信号的预测方法的具体实现流程图。参见图14，相对于图4、图7、图9以及图11任一所述实施例，本实施例提供的一种心电信号的预测方法在所述根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤之后，还包括：S1401～S1403，具体详述如下：
进一步地，若所述房颤发生概率大于预设的概率阈值，在所述根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤之后，还包括：
在S1401中，确定所述目标用户预设的关联用户，并向所述关联用户的终端发送告警信息。
在本实施例中，若检测到目标用户的房颤发生概率大于预设的概率阈值，则表示目标用户近期很大可能会出现房颤事件，此时需要对目标用户进行预警。其中，预警的方式可以通过弹框、信息提示、语音播报、输出提示音等方式通知用户。除了上述方式外，还可以通过S1401至S1403的方式通知用户。
在本实施例中，目标用户可以预设有关联用户以及关联用户的通信地址。终端设备根据目标用户预先配置的关联用户的通信地址，向关联用户的终端发送告警信息，该告警信息可以包含有上述房颤发生概率以及与该房颤发生概率对应的解释语段。
在一种可能的实现方式中，终端设备可以在显示界面上标记出关联用户的快捷通话按钮。目标用户可以通过电机该快捷通话按钮直接向关联用户的终端发起通话请求。在该情况下，该关联用户可以为目标用户的监护人、亲人或相关医院的联系电话。示例性地，图15示出了本申请一实施例提供的告警信息的输出示意图。参见图15所示，目标用户的房颤发生概率为88％，而预设的概率阈值为80％，此时终端设备会向关联用户的终端发送告警信息，该告警信息内容可以为“用户A当前房颤发生概率为88％，请密切关注用户A的身体情况。”并且可以在预设的显示界面上设置有联系关联用户的快捷通话按钮，方便用户快速通知关联用户。
在S1402中，根据所述目标用户当前的位置信息，获取与所述位置信息距离最近的医院地址。
在本实施例中，终端设备在检测到目标用户的房颤发生概率大于预设的概率阈值 时，可以为目标用户提供就诊咨询，减少用户的操作，提高咨询获取效率。在该情况下，终端设备可以获取目标用户当前的位置信息，若终端设备为用户使用的智能设备，则可以通过内置的定位模块，获取定位信号，并基于定位信号确定目标用户的位置信息；若终端设备为服务器，则可以向目标用户的用户终端发送位置获取请求，用户终端你可以通过内置的定位模块，获取位置信并反馈给服务器。
在本实施例中，终端设备可以在预设的地图界面上标记出目标用户的位置信息。在一种可能的实现方式中，终端设备可以通过程序调用接口API访问第三方服务器，通过第三方服务器提供的地图应用程序在预设的地图界面上标记出上述的位置信息，并以位置信息为基准，查找与用户位置最近的一个医院，从而确定过上述的医院地址。
在S1403中，根据所述位置信息以及所述医院地址，生成就诊路径。
在本实施例中，终端设备在获取了目标用户当前的位置信息以及医院地址后，可以生成一条以位置信息为起点，医院地址为终端的就诊路径。其中，终端设备可以调用第三方地图应用，将上述两个位置导入到第三方地图应用，通过第三方地图应用的路径生成算法，输出上述的就诊路径。
示例性地，图16示出了本申请一实施例提供的就诊路径的输出示意图。参见图16所示，终端设备可以在显示界面上添加有“查询医院路径”的提示按钮。终端设备可以通过点击该提示按钮跳转到对应的就诊路径显示页面，在用户执行点击操作时，终端设备可以获取用户当前的位置信息，并根据位置信息确定医院地址，并生成对应的就诊路径。
在本申请实施例中，在目标用户的房颤发生概率大于概率阈值时，向关联用户发送告警信息以及提供就诊路径，减少目标用户所需执行的操作，提高了操作效率，实现了告警信息自动发布的目的。
图17示出了本申请第八实施例提供的一种心电信号的预测方法的具体实现流程图。参见图17，相对于图4、图7、图9以及图11任一所述实施例，本实施例提供的一种心电信号的预测方法还包括：S1701～S1703，具体详述如下：
进一步地，在所述分别识别各个所述房颤聚类组对应的房颤类型之后，还包括：
在S1701中，若接收到新增心电信号，则通过所述房颤信号分类模型识别所述新增心电信号的新增信号类别。
在本实施例中，可穿戴设备在用户使用的过程可以以预设的反馈周期将采集到的新增心电信号发送给终端设备，终端设备在每次接收到的新增心电信号时，均可通过上述训练好的房颤信号分类模型确定对应的新增信号类别。其中，确定新增信号类别的方式如上实施例所述，实现过程完全相同，在此不再赘述。
在S1702中，基于所述新增信号类别，重新计算所述目标用户的所述房颤发生概率。
在本实施例中，终端设备可以根据上一次识别的房颤发生概率以及新采集得到的新增心电信号的新增信号类别，重新计算该目标用户的房颤发生概率，计算上述的房颤更新概率。具体计算的方式可以为：终端设备根据上一次计算得到的房颤发生概率所对应的所有心电信号的信号类别，以及本次采集得到的新增心电信号的新增信号类别，通过预设的房颤发生概率转换算法，重新确定目标用户的房颤发生概率，得到上 述的房颤更新概率。
在S1703中，根据已识别的所有所述房颤发生概率，生成房颤概率曲线。
在本实施例中，终端设备在每一次计算得到目标用户的房颤发生概率后可以对其进行记录，从而得到每一次判定的历史概率以及当前计算得到的房颤更细概率，可以生成房颤概率曲线，从而能够方便目标用户确定房颤发生概率的趋势变化。示例性地，图18示出了本申请一实施例提供的房颤概率曲线的输出示意图。参见图18所示，终端设备在每一次计算得到用户的房颤发生概率后，可以输出对应的显示界面，用户可以在显示界面确定当前的房颤发生概率。由于目标用户的可穿戴设备采集完成一个新的心电信号后，可以将心电信号发送给终端设备，终端设备可以对新增心电信号进行信号类别的识别，并根据识别后的新增信号类别重新计算目标用户的房颤发生概率。当用户点击“查看概率走势”的按钮时，则跳转到对应的显示页面，输出目标用户的房颤概率曲线，确定在各个更新时刻对应的房颤发生概率。
在本申请实施例中，在每次接收到目标用户的心电信号后，均重新计算目标用户的房颤发生概率，能够提高告警操作的实时性，并输出对应的房颤概率曲线，方便用户确定概率走势，便于用户了解自身的身体状况。
应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
对应于上文实施例所述的心电信号的预测方法，图19示出了本申请实施例提供的心电信号的预测装置的结构框图，为了便于说明，仅示出了与本申请实施例相关的部分。
参照图19，该心电信号的预测装置包括：
心电信号获取单元191，用于获取目标用户的心电信号；
信号类别识别单元192，用于将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的；
房颤发送概率计算单元193，用于根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤。
可选地，所述心电信号的预测装置还包括：
训练信号集获取单元，用于获取由多个采集次序连续的历史信号组成的训练信号集；所述训练信号集至少包含一个房颤信号；
风险信号提取单元，用于从所述训练信号集中提取除所述房颤信号外的风险信号；
房颤信号分类模型训练单元，用于通过所述训练信号集的所述风险信号，对预设的原生分类模型进行训练，得到所述房颤信号分类模型。
可选地，所述房颤信号分类模型训练单元包括：
信号类别标记单元用于根据所述风险信号的采集时间与关联的房颤信号的触发时间之间的时间差值，确定每个所述风险信号对应的信号类别；
信号特征参量生成单元，用于计算所述风险信号在各个预设的信号特征维度的特征值，得到所述风险信号的信号特征参量；
样本训练单元，用于根据所述信号特征参量以及所述信号类别，进行训练，得到所述房颤信号分类模型。
可选地，所述训练信号集获取单元包括：
运动参量获取单元，用于获取采集所述历史信号时所述训练用户的运动参量；
抖动时长确定单元，用于确定所述历史信号的抖动时长；
有效信号识别单元，用于根据所述抖动时长以及所述运动参量，判断所述历史信号是否为有效信号；
有效信号筛选单元，用于根据各个所述有效信号的采集时间的先后次序，对所有所述有效信息进行封装，得到所述训练信号集。
可选地，所述信号类别识别单元192包括：
生命体征参数获取单元，用于根据所述心电信号，确定所述目标用户的生命体征参数；
分类阈值调整单元，用于基于所述生命体征参数调整所述房颤信号分类模型的分类阈值；
房颤信号分类模型调用单元，用于通过调整后的所述房颤信号分类模型，识别所述心电信号的信号类别。
可选地，所述房颤发送概率计算单元193包括：
信号个数统计单元，用于统计所述目标用户在预设的时间段内存在所述信号类别为风险类别的所述心电信号的信号个数；
信号个数转换单元，用于根据所述信号个数计算所述房颤发生概率。
可选地，所述心电信号的预测装置还包括：
告警信息发送单元，用于确定所述目标用户预设的关联用户，并向所述关联用户的终端发送告警信息；和/或
位置信息获取单元，用于根据所述目标用户当前的位置信息，获取与所述位置信息距离最近的医院地址；
就诊路径生成单元，用于根据所述位置信息以及所述医院地址，生成就诊路径。
可选地，所述心电信号的预测装置还包括：
新增信号类别确定单元，用于若接收到新增心电信号，则通过所述房颤信号分类模型识别所述新增心电信号的新增信号类别；
房颤发生概率更新单元，用于基于所述新增信号类别，重新计算所述目标用户的所述房颤发生概率；
房颤概率曲线生成单元，用于根据已识别的所有所述房颤发生概率，生成房颤概率曲线。
因此，本申请实施例提供的心电信号的预测装置同样可以通过将采集得到的心电信号导入到预设的房颤信号分类模型，分别确定每一个心电信号的信号类别，并在获取了该用户在多个采集周期内的心电信号的信号类别，通过多个心电信号的信号类别，计算目标用户的房颤发生概率，并根据该房颤发生概率的数值大小，能够实现对房颤事件的预测，方便了用户确定自身的身体状况，提高了检测效果以及用户的使用体验。
图20为本申请一实施例提供的终端设备的结构示意图。如图20所示，该实施例 的终端设备20包括：至少一个处理器200(图20中仅示出一个)处理器、存储器201以及存储在所述存储器201中并可在所述至少一个处理器200上运行的计算机程序202，所述处理器200执行所述计算机程序202时实现上述任意各个心电信号的预测方法实施例中的步骤。
所述终端设备20可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。该终端设备可包括，但不仅限于，处理器200、存储器201。本领域技术人员可以理解，图20仅仅是终端设备20的举例，并不构成对终端设备20的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如还可以包括输入输出设备、网络接入设备等。
所称处理器200可以是中央处理单元(Central Processing Unit，CPU)，该处理器200还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
所述存储器201在一些实施例中可以是所述终端设备20的内部存储单元，例如终端设备20的硬盘或内存。所述存储器201在另一些实施例中也可以是所述**装置/终端设备20的外部存储设备，例如所述终端设备20上配备的插接式硬盘，智能存储卡(Smart Media Card,SMC)，安全数字(Secure Digital,SD)卡，闪存卡(Flash Card)等。进一步地，所述存储器201还可以既包括所述终端设备20的内部存储单元也包括外部存储设备。所述存储器201用于存储操作***、应用程序、引导装载程序(BootLoader)、数据以及其他程序等，例如所述计算机程序的程序代码等。所述存储器201还可以用于暂时地存储已经输出或者将要输出的数据。
需要说明的是，上述装置/单元之间的信息交互、执行过程等内容，由于与本申请方法实施例基于同一构思，其具体功能及带来的技术效果，具体可参见方法实施例部分，此处不再赘述。
所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，仅以上述各功能单元、模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能单元、模块完成，即将所述装置的内部结构划分成不同的功能单元或模块，以完成以上描述的全部或者部分功能。实施例中的各功能单元、模块可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中，上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。另外，各功能单元、模块的具体名称也只是为了便于相互区分，并不用于限制本申请的保护范围。上述***中单元、模块的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
本申请实施例还提供了一种网络设备，该网络设备包括：至少一个处理器、存储器以及存储在所述存储器中并可在所述至少一个处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现上述任意各个方法实施例中的步骤。
本申请实施例还提供了一种计算机可读存储介质，所述计算机可读存储介质存储 有计算机程序，所述计算机程序被处理器执行时实现可实现上述各个方法实施例中的步骤。
本申请实施例提供了一种计算机程序产品，当计算机程序产品在移动终端上运行时，使得移动终端执行时实现可实现上述各个方法实施例中的步骤。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请实现上述实施例方法中的全部或部分流程，可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一计算机可读存储介质中，该计算机程序在被处理器执行时，可实现上述各个方法实施例的步骤。其中，所述计算机程序包括计算机程序代码，所述计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述计算机可读介质至少可以包括：能够将计算机程序代码携带到拍照装置/终端设备的任何实体或装置、记录介质、计算机存储器、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、电载波信号、电信信号以及软件分发介质。例如U盘、移动硬盘、磁碟或者光盘等。在某些司法管辖区，根据立法和专利实践，计算机可读介质不可以是电载波信号和电信信号。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述或记载的部分，可以参见其它实施例的相关描述。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
在本申请所提供的实施例中，应该理解到，所揭露的装置/网络设备和方法，可以通过其它的方式实现。例如，以上所描述的装置/网络设备实施例仅仅是示意性的，例如，所述模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个***，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通讯连接可以是通过一些接口，装置或单元的间接耦合或通讯连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
以上所述实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围，均应包含在本申请的保护范围之内。

Claims (11)

  1. 一种心电信号的预测方法，其特征在于，包括：
    获取目标用户的心电信号；
    将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的；
    根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤。
  2. 根据权利要求1所述的预测方法，其特征在于，在所述将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的之前，还包括：
    获取由多个采集次序连续的历史信号组成的训练信号集；所述训练信号集至少包含一个房颤信号；
    从所述训练信号集中提取除所述房颤信号外的风险信号；
    通过所述训练信号集的所述风险信号进行训练，得到所述房颤信号分类模型。
  3. 根据权利要求2所述的预测方法，其特征在于，所述通过所述训练信号集的所述风险信号进行训练，得到所述房颤信号分类模型，包括：
    根据所述风险信号的采集时间与关联的房颤信号的触发时间之间的时间差值，确定每个所述风险信号对应的信号类别；
    计算所述风险信号在各个预设的信号特征维度的特征值，得到所述风险信号的信号特征参量；
    根据所述信号特征参量以及所述信号类别进行训练，得到所述房颤信号分类模型。
  4. 根据权利要求2所述的预测方法，其特征在于，所述获取由多个采集次序连续的心电信号组成的训练信号集，包括：
    获取采集所述历史信号时所述训练用户的运动参量；
    确定所述历史信号的抖动时长；
    根据所述抖动时长以及所述运动参量，判断所述历史信号是否为有效信号；
    根据各个所述有效信号的采集时间的先后次序，对所有所述有效信息进行封装，得到所述训练信号集。
  5. 根据权利要求1-4任一项所述的预测方法，其特征在于，所述将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的，包括：
    根据所述心电信号，确定所述目标用户的生命体征参数；
    基于所述生命体征参数调整所述房颤信号分类模型的分类阈值；
    通过调整后的所述房颤信号分类模型，识别所述心电信号的信号类别。
  6. 根据权利要求1-4任一项所述的预测方法，其特征在于，所述根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤，包括：
    统计所述目标用户在预设的时间段内存在所述信号类别为风险类别的所述心电信号的信号个数；
    根据所述信号个数计算所述房颤发生概率。
  7. 根据权利要求1-4任一项所述的预测方法，其特征在于，若所述房颤发生概率大于预设的概率阈值，在所述根据所述心电信号的信号类别，计算房颤将要发作的风 险程度，预测所述目标用户是否即将要发生房颤之后，还包括：
    确定所述目标用户预设的关联用户，并向所述关联用户的终端发送告警信息；和/或
    根据所述目标用户当前的位置信息，获取与所述位置信息距离最近的医院地址；
    根据所述位置信息以及所述医院地址，生成就诊路径。
  8. 根据权利要求1-4任一项所述的预测方法，其特征在于，在所述根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤之后，还包括：若接收到新增心电信号，则通过所述房颤信号分类模型识别所述新增心电信号的新增信号类别；
    基于所述新增信号类别，重新计算所述目标用户的所述房颤发生概率；
    根据已识别的所有所述房颤发生概率，生成房颤概率曲线。
  9. 一种心电信号的预测装置，其特征在于，包括：
    心电信号获取单元，用于获取目标用户的心电信号；
    信号类别识别单元，用于将所述心电信号导入预设的房颤信号分类模型，获得所述房颤信号分类模型输出的心电信号的信号类别，其中，所述房颤信号分类模型是以房颤患者作为模型训练样本训练得到的；
    房颤发送概率计算单元，用于根据所述心电信号的信号类别，计算房颤将要发作的风险程度，预测所述目标用户是否即将要发生房颤。
  10. 一种终端设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时实现如权利要求1至8任一项所述的方法。
  11. 一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现如权利要求1至8任一项所述的方法。
PCT/CN2020/136538 2019-12-18 2020-12-15 一种心电信号的预测方法、装置、终端以及存储介质 WO2021121226A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP20903369.5A EP4062824A4 (en) 2019-12-18 2020-12-15 ELECTROCARDIOSIGNAL PREDICTION METHOD AND DEVICE, TERMINALS AND STORAGE MEDIUM
US17/757,671 US20230036193A1 (en) 2019-12-18 2020-12-15 Electrocardiosignal Prediction Method and Apparatus, Terminal, and Storage Medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911307688.7A CN112971790A (zh) 2019-12-18 2019-12-18 一种心电信号的检测方法、装置、终端以及存储介质
CN201911307688.7 2019-12-18

Publications (1)

Publication Number Publication Date
WO2021121226A1 true WO2021121226A1 (zh) 2021-06-24

Family

ID=75264815

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/136538 WO2021121226A1 (zh) 2019-12-18 2020-12-15 一种心电信号的预测方法、装置、终端以及存储介质

Country Status (4)

Country Link
US (1) US20230036193A1 (zh)
EP (1) EP4062824A4 (zh)
CN (2) CN112971790A (zh)
WO (1) WO2021121226A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114386479B (zh) * 2021-12-09 2023-02-03 首都医科大学附属北京友谊医院 医学数据处理方法、装置、存储介质及电子设备
WO2022179645A2 (zh) * 2022-06-13 2022-09-01 合肥心之声健康科技有限公司 心电图分析方法、装置、电子设备和存储介质
CN117717351B (zh) * 2024-02-08 2024-05-14 南京鼓楼医院集团宿迁医院有限公司 一种基于蓝牙传输的心电图机快速辅助匹配方法和***

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160022162A1 (en) * 2013-03-08 2016-01-28 Singapore Health Services Pte Ltd System and method of determining a risk score for triage
US20190125273A1 (en) * 2017-09-05 2019-05-02 Medtronic, Inc. Differentiation of heart failure risk scores for heart failure monitoring
CN109745041A (zh) * 2017-11-03 2019-05-14 三星电子株式会社 事件检测方法及设备、房颤检测方法和非暂时性存储介质
CN110403600A (zh) * 2019-07-26 2019-11-05 武汉海星通技术股份有限公司 基于差值时间散点图的阵发性房颤智能分析方法及***
CN110547792A (zh) * 2019-07-31 2019-12-10 平安科技（深圳）有限公司 房颤的检测方法、装置、计算机设备及存储介质
CN110573211A (zh) * 2017-03-07 2019-12-13 心脏起搏器股份公司 事后心房颤动检测

Family Cites Families (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8155735B2 (en) * 2006-09-19 2012-04-10 The Cleveland Clinic Foundation Prediction and prevention of postoperative atrial fibrillation in cardiac surgery patients
US8676303B2 (en) * 2008-05-13 2014-03-18 The Regents Of The University Of California Methods and systems for treating heart instability
US9420956B2 (en) * 2013-12-12 2016-08-23 Alivecor, Inc. Methods and systems for arrhythmia tracking and scoring
CN103919544A (zh) * 2014-04-30 2014-07-16 苏州博思长健电子科技有限公司 基于穿戴式心电监测设备的冠心病和心肌梗死概率性预警***
US20160106330A1 (en) * 2014-10-17 2016-04-21 ECM Foresights, LLC Cardiac monitoring device and method
CN108601544B (zh) * 2016-07-29 2021-02-19 华为技术有限公司 用于心电测量的装置和方法
WO2018119316A1 (en) * 2016-12-21 2018-06-28 Emory University Methods and systems for determining abnormal cardiac activity
EP3565463B1 (en) * 2017-01-09 2021-02-24 Cardiac Pacemakers, Inc. Atrial fibrillation discrimination using heart rate clustering
US11179046B2 (en) * 2017-02-03 2021-11-23 Koninklijke Philips N.V. Method and system for detection of atrial fibrillation
CN107391900B (zh) * 2017-05-05 2020-07-14 陈一昕 房颤检测方法、分类模型训练方法及终端设备
AU2017203433A1 (en) * 2017-05-22 2018-12-06 Neuroanalytics Pty Ltd System and method for identification of neuro-cardiological disturbances
CN107550486A (zh) * 2017-09-11 2018-01-09 南京大学 一种利用去除qrs波形后的体表心电预测房颤发生的方法
CN109480819A (zh) * 2017-09-11 2019-03-19 南京大学 一种利用多路体表心电预测房颤发生的方法
TWI688371B (zh) * 2017-12-06 2020-03-21 高雄醫學大學 心房顫動信號型態擷取及輔助診斷智能裝置
CN108186011B (zh) * 2017-12-13 2021-06-01 深圳竹信科技有限公司 房颤检测方法、装置及可读存储介质
CN110226928B (zh) * 2018-03-06 2022-07-01 深圳市理邦精密仪器股份有限公司 房颤伴室早和房颤伴差传的识别方法和装置
US20190333639A1 (en) * 2018-04-26 2019-10-31 Vektor Medical, Inc. Generating a model library of models of an electromagnetic source
CN109199343A (zh) * 2018-07-17 2019-01-15 北京邮电大学 一种心房颤动的评估方法、***和设备
US10463314B1 (en) * 2018-07-19 2019-11-05 Toyota Motor Engineering & Manufacturing North America, Inc. Systems and methods for predicting and detecting a cardiac event
CN109009073B (zh) * 2018-07-20 2021-07-20 广州视源电子科技股份有限公司 房颤检测装置及存储介质
CN109288515B (zh) * 2018-11-14 2021-06-22 东南大学 基于穿戴式心电信号中早搏信号的周期性监测方法及装置
CN109770862B (zh) * 2019-03-29 2022-03-08 广州视源电子科技股份有限公司 心电信号分类方法、装置、电子设备和存储介质
TWI701681B (zh) * 2019-05-16 2020-08-11 中國醫藥大學附設醫院 心房顫動預測模型及其預測系統
CN110384495B (zh) * 2019-06-18 2021-06-04 华为技术有限公司 一种ecg检测方法与穿戴设备
CN110575161B (zh) * 2019-08-27 2021-12-28 复旦大学 基于心脏标测激动序列图的房颤分析预测方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160022162A1 (en) * 2013-03-08 2016-01-28 Singapore Health Services Pte Ltd System and method of determining a risk score for triage
CN110573211A (zh) * 2017-03-07 2019-12-13 心脏起搏器股份公司 事后心房颤动检测
US20190125273A1 (en) * 2017-09-05 2019-05-02 Medtronic, Inc. Differentiation of heart failure risk scores for heart failure monitoring
CN109745041A (zh) * 2017-11-03 2019-05-14 三星电子株式会社 事件检测方法及设备、房颤检测方法和非暂时性存储介质
CN110403600A (zh) * 2019-07-26 2019-11-05 武汉海星通技术股份有限公司 基于差值时间散点图的阵发性房颤智能分析方法及***
CN110547792A (zh) * 2019-07-31 2019-12-10 平安科技（深圳）有限公司 房颤的检测方法、装置、计算机设备及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4062824A4

Also Published As

Publication number Publication date
CN112971790A (zh) 2021-06-18
US20230036193A1 (en) 2023-02-02
EP4062824A4 (en) 2023-01-25
CN112603327A (zh) 2021-04-06
CN112603327B (zh) 2022-03-11
EP4062824A1 (en) 2022-09-28

Similar Documents

Publication Publication Date Title
WO2021208902A1 (zh) 一种睡眠报告的生成方法、装置、终端以及存储介质
WO2021121226A1 (zh) 一种心电信号的预测方法、装置、终端以及存储介质
US11116425B2 (en) Pacing activity data of a user
KR102446811B1 (ko) 복수의 디바이스들로부터 수집된 데이터 통합 및 제공 방법 및 이를 구현한 전자 장치
CN107773214B (zh) 一种最佳唤醒策略的方法、计算机可读介质和***
EP3268881B1 (en) Method and wearable device for obtaining audio data for diagnosis
CN107997767A (zh) 用于识别用户活动的方法及其电子设备
CN108369808A (zh) 电子设备和用于控制该电子设备的方法
CN109074867A (zh) 用于提供汇总和连续学习以改善健康结果的***和方法
US20210015415A1 (en) Methods and systems for monitoring user well-being
CN108877939A (zh) 一种具有智能特征提取功能的健康管理***
CN108968920A (zh) 一种数据检测方法及电子设备
JP2023504398A (ja) 連続血糖モニタリングに基づく提案
US20140012593A1 (en) Apparatuds and method for lifestyle management based on model
CN114724727A (zh) 一种疫情预测的方法及电子设备
WO2021098488A1 (zh) 一种房颤信号的分类方法、装置、终端以及存储介质
US20180366024A1 (en) Providing suggested behavior modifications for a correlation
WO2019104779A1 (zh) 一种压力检测的方法及终端
US20210401295A1 (en) System and methods utilizing artificial intelligence algorithms to analyze wearable activity tracker data
CN109067830A (zh) 消息推送方法、装置及终端
Santos et al. Context inference for mobile applications in the UPCASE project
US20220392590A1 (en) Granular data update sharing
US20230290502A1 (en) Machine learning framework for detection of chronic health conditions
CN111681761B (zh) 一种面向情境的健康风险识别方法及***
WO2016036486A1 (en) Pacing activity data of a user

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20903369

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020903369

Country of ref document: EP

Effective date: 20220622

NENP Non-entry into the national phase

Ref country code: DE