CN113130024A

CN113130024A - Medical event prediction method, wearable device and computer-readable storage medium

Info

Publication number: CN113130024A
Application number: CN201911403659.0A
Authority: CN
Inventors: 李露平; 陈茂林
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2019-12-31
Filing date: 2019-12-31
Publication date: 2021-07-16
Also published as: WO2021136030A1

Abstract

The application provides a medical event prediction method, wearable equipment and a computer-readable storage medium, which belong to the technical field of big data processing, and the calculation method comprises the following steps: acquiring life signal data of a user; setting windows based on a predetermined rule, wherein each window comprises a plurality of pieces of vital signal data; extracting features from the multiple sections of vital signal data in the window; and calculating the characteristics extracted based on the window and a pre-trained prediction model to obtain a calculation result. This computational result is accurate, the state of reaction user that can be accurate to user's timely understanding self state, discovery problem early prevention.

Description

Medical event prediction method, wearable device and computer-readable storage medium

Technical Field

The present application relates to the field of medical event prediction technologies, and in particular, to a method for predicting a medical event, a wearable device, and a computer-readable storage medium.

Background

Atrial Fibrillation (AF), also known as atrial Fibrillation, is a common type of arrhythmia. Prolonged atrial fibrillation can cause serious complications such as heart failure, hypertension, and the most serious stroke.

In the prior art, a detected person goes to a hospital for detection, medical instruments, devices and the like acquire data of a user, such as PPG (photoplethysmography), in a valid time, and a doctor or a mentor judges whether the current event is atrial fibrillation according to the data. Is suitable for discovering individuals suffering from atrial fibrillation diseases from the population. However, the individual proportion of the atrial fibrillation disease is low, most people are in a potential state of possibly suffering from atrial fibrillation in the future, and the potential risk of the attack of the atrial fibrillation is difficult to predict by the existing detection method because the atrial fibrillation event does not occur yet.

Disclosure of Invention

In view of this, the present application provides a method and an apparatus for predicting a medical event, a wearable device, an electronic device, and a computer-readable storage medium, which can provide an accurate calculation result and facilitate a user to predict the medical event.

Some embodiments of the present application provide a method of training a predictive model. The present application is described below in terms of several aspects, embodiments and advantages of which are mutually referenced.

In a first aspect, the present application provides a method for training a prediction model, which is applied to an electronic device, and the method includes: acquiring multiple segments of time-based historical vital signal data of at least one user, that is, multiple segments of time-based historical vital signal data of one user can be acquired, or multiple segments of time-based historical vital signal data of multiple users can be acquired, wherein the multiple segments of historical vital signal data refer to multiple segments of time-based historical vital signal data corresponding to multiple segments of different time, and the vital signal data of the users can be in a unit of a 'bar' or a unit of a 'byte' (data size); at least one window for the vital signal data is set for each user based on a predetermined rule, which may be a limit on a time period, e.g., defining the time period in units of hours, days, months or years, or a predetermined amount of vital signal data, e.g., a certain number or bytes of vital signal data. And each window comprises a plurality of segments of historical vital signal data; extracting features from the multiple sections of historical vital signal data in the window for each user, labeling the window, and determining attribute categories corresponding to the multiple sections of historical vital signal data in the window, wherein the attribute categories are used for representing medical events, such as atrial fibrillation, corresponding to the window and associated with the user; and taking the window as a unit, taking the characteristics extracted based on the window and the attribute category corresponding to the window as a training sample set, and training by a semi-supervised training method or a fully-supervised training method to obtain the prediction model. The prediction model can calculate the calculation result of the user more accurately, the calculation result can be used for indicating the analysis prediction of the trend and probability of the medical events which do not occur currently, such as atrial fibrillation, but the atrial fibrillation which possibly occurs in a certain period of time in the future, and provides some visual strong reference.

In one possible implementation of the first aspect, the predetermined rule includes: the predetermined time period and/or the predetermined amount of the plurality of pieces of historical vital signal data is taken as the length of the window, and specifically, a predetermined number of the plurality of pieces of historical vital signal data (e.g., 50 pieces of historical vital signal data) or a predetermined total size of the plurality of pieces of historical vital signal data (e.g., 50 bytes of historical vital signal data) may be taken as the length of the window. The window length is set so as to store rich training sample sets conveniently, and the accuracy of the prediction model is improved.

In a possible implementation of the first aspect, the setting of the window further requires an update process of data in the window, where the update process of data in the window includes: and acquiring newly acquired historical vital signal data of the user, and storing the newly acquired historical vital signal data of the user in a window to update the window, wherein the newly acquired historical vital signal data of the user refers to a set of historical vital signal data which does not participate in training. By updating the window, the characteristics extracted based on the window and the attribute category corresponding to the window can be updated in time, so that the prediction model is trained continuously, the calculation result of the prediction model reflects the current state of the user more accurately, and the accuracy of the prediction model is improved.

In a possible implementation of the first aspect, the update processing of the data in the window includes two situations, that is, the window has an idle position (the window is not full of the set length) and the window has no idle position (the window is full of the set length), specifically, after the historical vital signal data of the newly acquired user is acquired, the window has an idle position, and the newly acquired historical vital signal data of the user is directly stored in the idle position in the window; after the newly acquired historical vital signal data of the user is acquired, no idle position exists in the window, two modes can be adopted, and only the data with the latest acquisition time is reserved or the historical vital signal data in the window and the newly acquired historical vital signal data of the user which is not stored in the window are selected or rejected.

In a possible implementation of the first aspect, the historical vital signal data in the window and the newly acquired historical vital signal data of the user that is not stored in the window may be discarded by, specifically, scoring the historical vital signal data of each user in the window and the newly acquired historical vital signal data of each user according to a predetermined state of the user, for example, a health state and a disease state of the user, and the vital signal data of the health state may be used as a first excluded object, and the more the score of the vital signal data closer to the health state is, the less the vital signal data closer to the disease state is, that is, the vital signal data with the score in the predetermined state is deleted until the length of the window is satisfied.

Further, according to a classification rule from near to far from the current time when the historical vital signal data of the newly collected user is in time, the historical vital signal data stored in the window is classified into corresponding classes; and sequentially checking various classes from far to near according to the current time until the class with the historical vital signal data is checked, scoring the historical vital signal data in the class, and deleting the vital signal data with scores in a preset state of the user until the length of a window is met. Obtaining vital signal data within the window by the classification rule is more reflective of the risk that the user may have a medical event.

In a possible implementation of the first aspect, at least one piece of vital signal data is retained in each class, different weight values are set for historical vital signal data in each class according to the degree of contribution to the prediction model, and a lower weight value is given to vital signal data that contributes less accurately to the prediction model. The higher the weighted value is, the more the accuracy of the prediction model can be influenced, and because the degree that the vital signal data with longer time reflects the current state of the user is weaker, the weighted value is set from low to high according to the sequence from far to near from the current time, so that the influence degree of the vital signal data with longer time to the accuracy of the prediction model is reduced.

In a possible implementation of the first aspect, in the fully supervised training method, labeling a window, and determining an attribute category corresponding to historical vital signal data in the window includes: and carrying out manual linear labeling or automatic labeling on the window corresponding to the attribute category in the intermediate state, wherein the automatic labeling adopts a radial basis function and a nearest neighbor classification algorithm for labeling. The labeling mode is simple, and meanwhile, the accuracy of the prediction model can be improved compared with a partial labeling mode.

In a second aspect, the present application provides a method for predicting a medical event, applied to an electronic device, the method including: acquiring vital signal data of a user, which may be in units of "bars" or in units of "bytes" (data size); setting windows based on a predetermined rule, each window including a plurality of pieces of vital signal data; extracting features from the multiple sections of vital signal data in the window; and calculating the characteristics extracted based on the window and a pre-trained prediction model by taking the window as a unit to obtain a calculation result.

According to the data calculation method, the calculation result is more accurate, the state of the user can be accurately reflected, the user can timely know the state of the user, and problems can be found and prevented early.

In one possible implementation of the above second aspect, the predetermined time period and/or the predetermined amount of the plurality of segments of historical vital signal data is taken as the length of the window, and specifically, a predetermined number of the plurality of segments of historical vital signal data (e.g., 50 pieces of historical vital signal data) or a predetermined total size of the plurality of segments of historical vital signal data (e.g., 50 bytes of historical vital signal data) may be taken as the length of the window. The window length is set so as to store rich training sample sets conveniently, and the accuracy of the prediction model is improved.

In a possible implementation of the second aspect, the setting the window based on a predetermined rule further includes: the updating process of the data in the window specifically comprises the following steps: and acquiring newly acquired vital signal data of the user, and storing the newly acquired vital signal data into the window to update the window, wherein the newly acquired vital signal data is current vital signal data which is acquired at the current time and does not participate in calculation. By updating the window, the features extracted based on the window can be updated in time, so that the calculation result is continuously updated and the current state of the user is reflected in time.

In a possible implementation of the second aspect, the update processing of the data in the window includes two situations, that is, the window has enough idle positions (the window is not full of the set length) and the window does not have enough idle positions (the window is full of the set length), specifically, for the newly acquired vital signal data of the user, the window has enough idle positions, and then the current vital signal data is directly stored in the idle positions according to the acquisition sequence; and for newly acquired life signal data of the user, if the window does not have enough idle positions, only the data with the latest acquisition time is reserved or the life signal data in the window and the newly acquired life signal data of the user which is not stored in the window are selected or rejected.

In a possible implementation of the second aspect, the process of taking or rejecting the vital signal data in the window and the newly acquired vital signal data of the user not stored in the window includes: and scoring each piece of vital signal data in the window and newly acquired vital signal data according to a preset state, and deleting the vital signal data with the score in the preset state until the set size of the window is met.

Further, according to a classification rule from near to far from the current time when the newly acquired vital signal data is acquired, dividing the vital signal data stored in the window into corresponding classes; and sequentially checking the classes according to the sequence from far to near from the current time until the class of the vital signal data is checked, scoring the vital signal data in the class, and deleting the vital signal data with the score in a preset state of the user. Obtaining vital signal data within the window by the classification rule is more reflective of the risk that the user may have a medical event.

In one possible implementation of the second aspect, at least one vital signal data is retained in each class, and different weight values are set for the vital signal data in each class according to the degree of contribution to the calculation result. The vital signal data that contributes less to the calculation result is given a lower weight value. The higher the weighted value is, the more the accuracy of the calculation result can be influenced, and because the degree that the vital signal data with longer time reflects the current state of the user is weaker, the weighted value is set from low to high according to the sequence from far to near from the current time, so that the influence degree of the vital signal data with longer time to the accuracy of the calculation result is reduced. Therefore, the vital signal data in the window can reflect the current state of the user more truly, and the accuracy of the calculation result is improved.

In a possible implementation of the second aspect, the data calculation method further includes: counting attribute categories of other users in the same preset range with the calculation result based on data in the big data platform, and states of the other users corresponding to the attribute categories respectively; and evaluating the states of the users corresponding to the calculation results according to the states of the other users corresponding to the attribute categories respectively. Prediction of the user's probability of likely occurrence of a medical event (e.g., atrial fibrillation) over a future period of time may be facilitated to facilitate early prevention.

In one possible implementation of the second aspect, the method for predicting a medical event further includes: before the vital signal data of the user is acquired, noise reduction processing is required on the vital signal data acquired each time, for example, by filtering, signal quality analysis and other processing modes, so as to improve the reliability of the data.

In a third aspect, the present application provides an electronic device for generating a predictive model, comprising:

the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring multiple segments of historical vital signal data of at least one user based on time; a first window setting module for setting at least one window for vital signal data based on a predetermined rule for each user, each window including a plurality of pieces of historical vital signal data; the first extraction module is used for extracting features from multiple sections of historical vital signal data in a window aiming at each user; the marking module is used for marking the window and determining attribute categories corresponding to the multiple sections of historical vital signal data in the window, wherein the attribute categories are used for representing medical events corresponding to the window and associated with the user; and the training module is used for taking the window as a unit, taking the characteristics based on the window and the attribute category corresponding to the window as a training sample set, and training by a semi-supervised training method or a fully-supervised training method to obtain the prediction model.

The electronic device for generating a prediction model according to the embodiment of the present application, which is configured to execute the method of the first aspect, may train to obtain a prediction model, and the prediction model may accurately calculate a calculation result of the user, and the calculation result may be used to indicate an analysis prediction of a trend and a probability of a medical event that does not occur currently, such as atrial fibrillation, but may occur in a future period of time, and provide some visually strong references.

In one possible implementation of the third aspect, the predetermined rule includes: the predetermined time period and/or the predetermined amount of the plurality of segments of historical vital signal data is taken as the length of the window, specifically, a predetermined number of the plurality of segments of historical vital signal data is taken as the length of the window, or a predetermined total size of the plurality of segments of historical vital signal data is taken as the length of the window. The window length is set so as to store rich training sample sets conveniently, and the accuracy of the prediction model is improved.

In a possible implementation of the third aspect, the first window setting module is further configured to perform update processing on data in a window, and includes: and acquiring newly acquired historical vital signal data of the user, and storing the newly acquired historical vital signal data of the user in a window to update the window, wherein the newly acquired historical vital signal data of the user refers to a set of historical vital signal data which does not participate in training. By updating the window, the characteristics extracted based on the window and the attribute category corresponding to the window can be updated in time, so that the prediction model is trained continuously, the calculation result of the prediction model reflects the current state of the user more accurately, and the accuracy of the prediction model is improved.

In a possible implementation of the third aspect, the first window setting module is specifically configured to: when a free position exists in the window after the newly acquired historical vital signal data of the user is acquired, directly storing the newly acquired historical vital signal data of the user to the free position in the window; and after the newly acquired historical vital signal data of the user is acquired, no idle position exists in the window, and only the data with the latest acquisition time is reserved or the historical vital signal data in the window and the newly acquired historical vital signal data of the user which is not stored in the window are selected or rejected.

In a possible implementation of the third aspect, the first window setting module is specifically configured to: and scoring the historical vital signal data of each user in the window and the newly acquired historical vital signal data of each user according to preset states of the users, wherein the preset states comprise health states and disease onset states of the users, and deleting the vital signal data with the scores in the preset states until the length of the window is met.

In a possible implementation of the third aspect, the first window setting module is specifically configured to: classifying historical vital signal data stored in a window into corresponding classes according to a classification rule from near to far from the current time when the historical vital signal data of a newly acquired user is acquired; and sequentially checking various classes from far to near according to the current time until the class with the historical vital signal data is checked, scoring the historical vital signal data in the class, and deleting the vital signal data with scores in a preset state of the user until the length of a window is met. Obtaining vital signal data within the window by the classification rule is more reflective of the risk that the user may have a medical event.

In a possible implementation of the third aspect, the first window setting module is further specifically configured to: at least one vital signal data is retained in each class, and different weight values are set for historical vital signal data in each class according to the contribution degree of the historical vital signal data to the prediction model. The vital signal data which contributes less to the accuracy of the prediction model is given a lower weight value. The higher the weighted value is, the more the accuracy of the prediction model can be influenced, and because the degree that the vital signal data with longer time reflects the current state of the user is weaker, the weighted value is set from low to high according to the sequence from far to near from the current time, so that the influence degree of the vital signal data with longer time to the accuracy of the prediction model is reduced.

In a possible implementation of the third aspect, in the full-supervised training method, the labeling module is specifically configured to: and carrying out manual linear labeling or automatic labeling on the window corresponding to the attribute category in the intermediate state, wherein the automatic labeling adopts a radial basis function and a nearest neighbor classification algorithm for labeling. The labeling mode is simple, and meanwhile, the accuracy of the prediction model can be improved compared with a partial labeling mode.

In a fourth aspect, the present application provides an electronic device for computing, comprising: the second acquisition module is used for acquiring vital signal data of a user; the second window setting module is used for setting windows based on a preset rule, and each window comprises a plurality of sections of vital signal data; the second extraction module is used for extracting characteristics of the multiple sections of life signal data in the window; and the calculation module is used for calculating the characteristics extracted based on the window and the pre-trained prediction model to obtain a calculation result by taking the window as a unit.

According to the electronic device for calculating data, which is used for executing the method of the second aspect, the calculation result based on the vital signal data of the user is obtained, the calculation result is more accurate, and the state of the user can be accurately reflected, so that the user can timely know the state of the user, and problems can be found and prevented early.

In one possible implementation of the fourth aspect, the predetermined rule includes: the predetermined time period and/or the predetermined amount of the pieces of vital signal data is taken as the length of the window. Specifically, the predetermined number of pieces of vital signal data is included as the length of the window, or the predetermined total size of the pieces of vital signal data is included as the length of the window. The window length is set so as to store rich training sample sets conveniently, and the accuracy of the prediction model is improved.

In a possible implementation of the fourth aspect, the second windowing module is further configured to perform an update process on data in a window, and includes: and acquiring newly acquired vital signal data of the user, and storing the newly acquired vital signal data into the window to update the window, wherein the newly acquired vital signal data is current vital signal data which is acquired at the current time and does not participate in calculation. By updating the window, the features extracted based on the window can be updated in time, so that the calculation result is continuously updated and the current state of the user is reflected in time.

In a possible implementation of the fourth aspect, the second window setting module is specifically configured to: when a window has an idle position after newly acquired life signal data of a user are acquired, directly storing the current life signal data to the idle position according to the acquisition sequence; and when no idle position exists in the window after the newly acquired life signal data of the user is acquired, only the data with the latest acquisition time is reserved or the life signal data in the window and the newly acquired life signal data of the user which is not stored in the window are selected or rejected.

In a possible implementation of the fourth aspect, the second window setting module is specifically configured to: and scoring each piece of vital signal data in the window and newly acquired vital signal data according to a preset state, and deleting the vital signal data with the score in the preset state until the set size of the window is met.

Further, the second window setting module is specifically configured to: classifying the vital signal data stored in the window into corresponding classes according to a classification rule from near to far from the current time of newly acquired vital signal data; and sequentially checking the classes according to the sequence from far to near from the current time until the class of the vital signal data is checked, scoring the vital signal data in the class, and deleting the vital signal data with the score in a preset state of the user. Obtaining vital signal data within the window by the classification rule is more reflective of the risk that the user may have a medical event.

In a possible implementation of the fourth aspect, the second window setting module is further specifically configured to: at least one vital signal data is retained in each class, and different weight values are set for the vital signal data in each class according to the contribution degree to the calculation result. The vital signal data that contributes less to the calculation result is given a lower weight value. The higher the weighted value is, the more the accuracy of the calculation result can be influenced, and because the degree that the vital signal data with longer time reflects the current state of the user is weaker, the weighted value is set from low to high according to the sequence from far to near from the current time, so that the influence degree of the vital signal data with longer time to the accuracy of the calculation result is reduced. Therefore, the vital signal data in the window can reflect the current state of the user more truly, and the accuracy of the calculation result is improved.

In a possible implementation of the fourth aspect, the electronic device further includes: the evaluation module is used for counting the attribute categories of other users in the same preset range with the calculation result based on the data in the big data platform and the states of the other users corresponding to the attribute categories respectively; and the evaluation module evaluates the state of the user corresponding to the calculation result according to the states of other users corresponding to the attribute categories respectively. The calculation of the probability that the user predicts the atrial fibrillation possibility in the future period of time can be facilitated, and early prevention is facilitated.

In a possible implementation of the fourth aspect, the electronic device further includes: and the processing module is used for performing noise reduction processing on the acquired vital signal data in a filtering, signal quality analysis and other modes before acquiring the vital signal data of the user so as to improve the reliability of the data. In a possible implementation of the fourth aspect, the electronic device further includes: and the display module is used for displaying the calculation result. So that the user can intuitively know the self condition and take corresponding preventive measures in time.

In a fifth aspect, the present application provides a data computing system comprising: the electronic device for generating a prediction model of the embodiment of the third aspect described above; and an electronic device for computing of an embodiment of the fourth aspect described above.

In a sixth aspect, the present application provides a wearable device, comprising: the collector is used for periodically collecting life signal data of a user; a memory to store instructions; a processor for reading instructions stored in the memory to perform the steps of: acquiring life signal data of a user; setting a window based on a predetermined rule, the window including a plurality of pieces of vital signal data; extracting features from the multiple sections of vital signal data in the window; calculating the characteristics extracted based on the window and a prediction model to obtain a calculation result; and the display screen is used for displaying the calculation result so that a user can observe the calculation result intuitively. According to the accurate calculation result that wearing equipment can obtain of this application embodiment, this calculation result more can accurate reaction user's health state to user's timely understanding self state, discovery problem early prevention.

In a possible implementation of the above sixth aspect, the predetermined rule includes: at least one of the predetermined time period and the predetermined amount of the pieces of vital signal data is taken as the length of the window.

In one possible implementation of the above-mentioned sixth aspect, the determining the predetermined amount of the pieces of vital signal data as the length of the window comprises: and taking the preset number of the multiple sections of vital signal data as the length of the window, or taking the preset total size of the multiple sections of vital signal data as the length of the window. The window length is set so as to store rich training sample sets conveniently, and the accuracy of the prediction model is improved.

In a possible implementation of the sixth aspect, the processor is specifically configured to set a window based on a predetermined rule, and perform an update process on data in the window, where the update process on data in the window includes: for newly acquired life signal data of a user, if enough idle positions exist in the window, directly storing the current life signal data to the idle positions according to the acquisition sequence; and for newly acquired life signal data of the user, if the window does not have enough free positions, accepting or rejecting the life signal data in the window and the newly acquired life signal data of the user which is not stored in the window.

In a possible implementation of the above sixth aspect, the processor is specifically configured to: according to the sequence of the acquisition time, only the data with the latest acquisition time is reserved under the condition that the window is full.

In a possible implementation of the above sixth aspect, the processor is specifically configured to: and scoring each piece of vital signal data in the window and newly acquired vital signal data according to a preset state, and deleting the vital signal data with the score in the preset state until the set size of the window is met.

In a possible implementation of the above sixth aspect, the processor is specifically configured to: classifying the vital signal data stored in the window into corresponding classes according to a classification rule from near to far from the current time when the newly acquired vital signal data is acquired; and sequentially checking various classes from far to near according to the current time until the class of the vital signal data is checked, scoring the vital signal data in the class, and deleting the vital signal data with the score in a preset state of the user. Obtaining vital signal data within the window by the classification rule is more reflective of the risk that the user may have a medical event.

In one possible implementation of the above-described sixth aspect, at least one vital signal data is retained in each class, and different weight values are set for the vital signal data in each class according to the degree of contribution to the calculation result. The higher the weighted value is, the more the accuracy of the calculation result can be influenced, and because the degree that the vital signal data with longer time reflects the current state of the user is weaker, the weighted value is set from low to high according to the sequence from far to near from the current time, so that the influence degree of the vital signal data with longer time to the accuracy of the calculation result is reduced. Therefore, the vital signal data in the window can reflect the current state of the user more truly, and the accuracy of the calculation result is improved.

In a possible implementation of the above sixth aspect, the processor further includes: counting attribute categories of other users in the same preset range with the calculation result and states of other users corresponding to the attribute categories respectively based on data in a big data platform; and evaluating the states of the users corresponding to the calculation results according to the states of the other users corresponding to the attribute categories respectively. The calculation of the probability that the user predicts the atrial fibrillation possibility in the future period of time can be facilitated, and early prevention is facilitated.

In a seventh aspect, the present application provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the computer program causes the processor to execute the method of the first aspect and the method of the second aspect.

Drawings

FIG. 1a is a schematic diagram of a prediction system according to an embodiment of the present application;

FIG. 1b is a schematic structural diagram of a prediction system of a processor on a cloud server according to an embodiment of the present application;

FIG. 1c is a schematic diagram of a prediction system of a processor on a wearable device according to an embodiment of the present application;

FIG. 1d is a block diagram of a prediction system of a processor on a computing device according to an embodiment of the present application;

FIG. 2 is a flow chart of a method for predicting atrial fibrillation according to the present application;

FIG. 3 is a graph illustrating probability distribution of atrial fibrillation according to some embodiments of the present disclosure;

FIG. 4 is a flow chart of a method of training a predictive model according to some embodiments of the present application;

FIG. 5 is a schematic diagram of an electronic device for generating a predictive model according to an embodiment of the present application;

FIG. 6 is a schematic diagram of an electronic device for computing according to an embodiment of the present application;

FIG. 7 is a schematic structural view of a wearable device according to some embodiments of the present application;

FIG. 8a is a graph of calculated results of a medical event according to some embodiments of the present application;

FIG. 8b is a schematic illustration of a UI interface displaying risk values output by a module according to some embodiments of the present application;

FIG. 8c is a schematic illustration of a UI interface displaying results of calculations output by a module according to some embodiments of the present application;

FIG. 9 is a block diagram of an apparatus of some embodiments of the present application;

fig. 10 is a block diagram of a system on a chip (SoC) in accordance with some embodiments of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It is understood that a medical event in the present application refers to an adverse cardiac event that may occur to a user in relation to physical health, and non-limiting examples may include Atrial Fibrillation (AF) or cardiac infarction, etc. The medical event is an application scenario implemented by the computing method provided by the application. By collecting data relating to medical events and performing calculations, the results obtained can be used to indicate an analytical prediction of trends and probabilities of medical events that refer to medical events that the user is not currently taking place, but are likely to take place within some future period of time, as well as to provide some visually strong reference. It will be appreciated by those skilled in the art that the present application is intended to protect a method of data processing (computing) and corresponding apparatus and system, and is not intended to protect a method of medical diagnosis.

It can be understood that, for the calculation method provided by the present application, the higher the calculation result, the higher the probability that the medical event occurs in a certain period of time in the future for the user, and it can also be understood that the time when the medical event is likely to occur for the user is closer to the current predicted time.

According to an embodiment of the present application, the prediction model may be obtained based on historical vital signal data (such as Photo ply graph, PPG for short) in a big data platform and by using machine learning training, where the historical vital signal data may be a set of historical vital signal data of medical events similar to the subject.

It can be understood that the window of the present application may be a predetermined time period as the length of one window, or may be a predetermined amount of user vital signal data as the length of one window, or may be a combination of the two to set the length of the window, and the vital signal data of multiple segments of users may be obtained by setting the window, so as to improve the accuracy of the calculation.

It will be appreciated that as used herein, the term module may refer to or include an Application Specific Integrated Circuit (ASIC), an electronic circuit, a processor (shared, dedicated, or group) and/or memory that executes one or more software or firmware programs, a combinational logic circuit, and/or other suitable hardware components that provide the described functionality, or may be part of such hardware components.

It will be appreciated that in the various embodiments of the present application, the processor may be a microprocessor, a digital signal processor, a microcontroller, the like, and/or any combination thereof. According to another aspect, the processor may be a single-core processor, a multi-core processor, the like, and/or any combination thereof.

Embodiments of the present application will be further described with reference to the accompanying drawings.

According to some embodiments of the present application, a schematic diagram of a prediction system is disclosed, which may be used for prediction of a medical event, wherein non-limiting examples of the medical event may include cardiac infarction or atrial fibrillation, etc., and fig. 1a shows a schematic diagram of the structure of the prediction system. As shown in fig. 1a, the system includes a wearable device 101, a user wearing the wearable device, and a cloud server 102 wirelessly connected to the wearable device.

Non-limiting examples of wearable device 101 include a portable device that may be worn directly on a person's body, such as on the wrist, chest, etc., or integrated into the user's clothing or accessories. A device presented in the form of a bracelet or a watch and worn on the wrist is a typical example of the wearable device that is easier to understand. The wearable device 101 can be worn on a user for a long time and is used for periodically collecting vital signal data (such as PPG signal data) of the user based on a specific rule, the collected vital signal data of the user is calculated through a prediction model to obtain a calculation result of the user, and the calculation result is transmitted to the wearable device for the user to consult.

Cloud server 102 may be connected to one or more wearable devices. This is only the case where it is connected to one wearable device as an exemplary illustration. In addition, the cloud server 102 may collect and integrate vital sign data of the user from a plurality of wearable devices, which are used for the calculations disclosed herein and to obtain final calculation results. In addition, the cloud server 102 may also be connected to other necessary devices, such as a router, and a server with other auxiliary functions, but for simplicity of illustration, they are not listed in fig. 1 a.

The collected vital signal data of the user is subjected to prediction calculation through the prediction model to obtain a calculation result of the user, the user can visually obtain the calculation result through the wearable device 101, and the calculation result is used for evaluating the probability of the medical event of the user in a future period of time, so that the user can take preventive measures according to the condition of the user in time. The predictive computation based on vital signal data of the user is implemented in a specific processor, for example in the processor 103 according to the present application.

The processor 103 for invoking instructions to implement the computation may be disposed in the cloud server 102, incorporated into the wearable device 101, or disposed in a computing device 104, such as a mobile phone, a personal computer, etc., which is independent of the wearable device 101 and the cloud server 102, but can be connected to and synchronize data with the wearable device 101 and the cloud server 102, as shown in fig. 1a to 1 d.

According to one embodiment of the application, the calculation based on vital sign data of the user includes training of a predictive model and calculation of a result of the user. The establishment of the predictive model and the calculation of the outcome for the vital signal data of the specific user based on the predictive model may, as described above, be implemented by the same processor 103, e.g. the processor 103 is in the wearable device, that is the wearable device 101 may enable the training of the predictive model and the calculation of the outcome of the specific user. It may also be implemented by two separate processors, for example, a first processor responsible for training of the predictive model and a second processor responsible for computation of the results for a particular user based on the predictive model, where the two processors are separate, they may be present in any two of the cloud server, the computing device (if present), and the wearable device, respectively.

Based on the above description of the embodiments, according to some embodiments of the present application, the following describes a prediction method of a medical event implemented by a processor of a prediction system according to the present application with specific embodiments, and fig. 2 shows a flowchart of the prediction method of a medical event implemented by a processor of a prediction system according to the present application, and as shown in fig. 2, the prediction method specifically includes:

step S210, vital signal data of the user are acquired, wherein the vital signal data of the user are acquired by the wearable device. The vital sign data of the user may be in units of "bars" or in units of "bytes" (data size). In the following exemplary description, an example in which the unit of "bar" is the unit of vital sign data of the user is described. It will be understood by those skilled in the art that the case where "bytes" or other units are used as the unit of vital sign data of the user is similar. The vital signal data of the user of the present application may be PPG signal data.

According to an embodiment of the present application, in order to improve the reliability of the data, each time the acquired vital signal data needs to be subjected to noise reduction processing, for example, by filtering, signal quality analysis, and other processing manners.

Step S220, setting windows based on a predetermined rule, each window including a plurality of segments of vital signal data, wherein the plurality of segments are a plurality of segments of time, each segment of time may have at least one or more pieces of vital signal data, and the plurality of segments of vital signal data are a plurality of pieces of vital signal data in the plurality of segments of time. Here, the predetermined rule may include one or both of the predetermined time period and the predetermined amount of the pieces of vital signal data as the length of the window. For example, the predetermined time period may be a period of time, for example, a time period in units of hours, days, months, and years, as a length of one window, or a total predetermined amount of the plurality of pieces of vital signal data as a length of the window, where the predetermined amount as the length of the window may be a predetermined number of or a predetermined total size of data storage spaces of the plurality of pieces of vital signal data, for example, a storage space for storing a certain number of pieces of user vital signal data in units of bars or in units of data size (bytes) as a length of one window, or a combination of the predetermined time period and the predetermined amount of the plurality of pieces of vital signal data as a length of the window, such as 100 pieces of data in 6 months as a length of the window, which is not limited herein.

For a user, after a plurality of measurements, a plurality of segments of vital signal data can be obtained. These multiple segments of vital signal data can be fit into one or more windows as desired. In case there are multiple windows, the vital signal data contained in these different windows may also be partially overlapping. According to an embodiment of the application, the plurality of segments of vital signal data include a plurality of segments of historical vital data of the user and newly acquired vital signal data of the user, and the vital signal data of the user in the window needs to be updated and processed over time, that is, the window is used as a unit, the vital signal data in the window is updated, the window is updated accordingly, and a result calculated based on the vital signal data in the window is continuously updated, so that the continuous detection of the physical condition of the user is realized.

It can be understood that, in order to improve the accuracy of the calculation result, in consideration of the problems that the vital signal data may be unevenly distributed in time, the time error exists between the signal acquisition point and the real event, the signal acquisition point and the event may be missed and the like in the process of acquiring the vital signal data of the user, the method and the device prolong the data acquisition time by increasing the amount of the vital signal data of the user so as to reduce the error as much as possible, and improve the reliability of the calculation result.

The update processing of vital signal data within a window, according to one embodiment of the present application, is further described below, taking PPG signal data as an example.

S211, setting an initial size of the window, for example, 100 positions for storing the PPG signal data, and storing 100 pieces of PPG signal data.

And S212, when the wearable device completes one successful acquisition, storing the newly acquired PPG signal data into a window, wherein the newly acquired PPG signal data is the current PPG signal data acquired at the current time.

During updating, if the window has an idle position (namely the window is not full of 100 pieces of PPG signal data), the current PPG signal data is directly stored to the idle position according to the sequence of acquisition.

If the window is full, data may need to be discarded. How to cut data will be described in detail in the following embodiments.

According to an implementation manner of the embodiment of the application, only the first 100 pieces of data with the latest acquisition time can be retained under the condition that the window is full according to the sequence of the acquisition time.

According to another implementation manner of the embodiment of the application, since it is considered that the PPG signal data is sometimes in the normal range and sometimes in the abnormal range before the medical event occurs to the user, collecting the PPG signal data in the abnormal range as much as possible is more beneficial for predicting the future occurrence of the medical event of the user so as to remind the user of the effect of early prevention. Therefore, in the present application, by scoring each piece of PPG signal data within the window and the newly acquired PPG signal data, the PPG signal data is divided into health state PPG signal data and abnormal PPG signal data (unhealthy state PPG signal data), and the score is higher the closer the PPG signal data is to the health state (the more regular the beat is). For example, taking the heart rate value of an adult as an example, the heart rate value of an adult in a healthy state is 60 to 100 times/minute, and the heart rate values other than the heart rate values are abnormal PPG signal data (heart rate values less than 60 times/minute or more than 100 times/minute), and the more the abnormal PPG signal data is less than 60 times/minute, the closer the abnormal PPG signal data is to the 60 times/minute, the higher the abnormal PPG signal data is to the 60 times/minute, and the more the abnormal PPG signal data is to the 100 times/minute. And if the score value of the newly acquired PPG signal data is not the highest value compared with the PPG signal data in the window, namely the newly acquired PPG signal data is not the PPG signal data closest to the health state, deleting the PPG signal data with the highest score in the window, storing the newly acquired PPG signal data instead, and obtaining the updated window so that the PPG signal data in the window can reflect the physical condition of the user in time. Further, prior to scoring each piece of PPG signal data within the window and the newly acquired PPG signal data, the stored PPG signal data within the window may be classified by time, e.g., into 5 classes from near to far from the current time when the newly acquired PPG signal data was taken, the stored PPG signal data is time-classified into 5 classes: the PPG signal data at the same hour as the current time of the newly acquired PPG signal data is of type 1, the PPG signal data on the same day except for type 1 is of type 2, the PPG signal data in the same week except for types 1 and 2 is of type 3, the PPG signal data in the same month except for types 1, 2, and 3 is of type 4, and the PPG signal data at other times except for the first four is of type 5. Assume that the classification results are: of the 100 pieces of PPG signal data, 30 pieces of PPG signal data exist in class 1, 0 pieces of PPG signal data exist in class 2, 50 pieces of PPG signal data exist in class 3, 20 pieces of PPG signal data exist in class 4, and 0 piece of PPG signal data exist in class 5. When data is chosen or rejected, the classes are checked sequentially from far to near from the current time, namely from the 5 th class to the 1 st class until a class with more than 1 piece of PPG signal data is found, for example, when the 4 th class is found, and the fourth class has 20 pieces of data, each piece of PPG signal data in the 4 th class is scored, the piece of PPG signal data with the highest score is selected, and the PPG signal data is deleted. Thereby obtaining space in the window to store the newly acquired PPG signal data. By the alternative mode, the PPG signal data far away from the current time can be preferentially deleted, so that the data in the window can express the recent state of the user.

Furthermore, in order to reflect the physical state of the user for a long time, the vital signal data in the window are classified according to the order from far to near from the current time, and at least 1 piece of vital signal data is kept in each class during updating, that is, according to the above-mentioned scoring rule, the PPG signal data in each class is continuously deleted, and when the PPG signal data in a certain class only remains or only 1 piece of PPG signal data in the class itself exists, the PPG signal data is kept no matter whether the PPG signal data is the highest value or not. And the weight value is preset for the vital signal data in each class, and the lower weight value is given to the vital signal data with less influence on the calculation result. The higher the weight value is, the more the calculation result can be influenced. The vital signal data which contributes less to the accuracy of the calculation result may be vital signal data which is a little longer than the current time, for example, the PPG vital signal data which is half a year ago is a vital signal data which is a little longer than the current time, and considering that the physical state of the user may have a large change, such as a state from a healthy state to an intermediate state (a state between the healthy state and a diseased state) and a state from the intermediate state to the healthy state, the data which is too long may not correctly express the current state of the user, and therefore, the weight values are set from low to high in the order from far to near from the current time, so as to reduce the influence of the vital signal data which is farther from the current time on the calculation result. It may also be vital signal data collected by the user in a special state, for example, by sensing the user in motion by a motion detection sensor, or by the user actively selecting or inputting whether the user is in motion or in a sick state by the device, and considering that these data contribute little or nothing to the accuracy of the calculation result, they may be given a lower weight, or directly removed. Therefore, the vital signal data in the window can reflect the current state of the user more truly, and the accuracy of the calculation result is improved.

Step S230, performing feature extraction on the multiple sections of vital signal data in the window. It can be understood that, since the vital signal data of the user is in a plurality of time periods, the extraction of the features is more diverse, and not only the data feature extraction of each vital signal data segment, but also the fusion of the features of the data features changing with time is needed. For example, the feature extraction may be data meeting set requirements in a plurality of pieces of vital signal data, such as values (maximum values, average values, or minimum values) of the vital signal data, and taking a heart rate as an example, data in a period of time (e.g., 5 minutes, 1 hour, etc.) of each day of the user are respectively obtained, for example, a heart rate (times/minute) obtained on the first day is 62, 63, 65, 68, a heart rate (times/minute) obtained on the second day is 65, 68, 69, 65, and a heart rate (times/minute) obtained on the third day is 69, 68, 69, 70. When the setting requirement is the highest value, 68 on the first day, 69 on the second day and 70 on the third day are extracted, respectively. And relationships and regularity existing between the data, such as a characteristic of gradually increasing regularity appearing in the highest values 68, 69, 70 from the first day to the third day, or a weight value of the data, and the like. Not only the data features in each segment (time period) of vital signal data, such as the highest values of heart rate on the first day, the second day, and the third day, can be extracted. Regular sequence features existing between these data features, such as the highest ascending sequence feature over three days, are also extracted. The features of the highest value in the present application are merely exemplary descriptions, and other setting requirements such as the lowest value or the highest value may be taken, and are not limited herein.

According to the embodiment of the application, the characteristics are diversified due to the characteristic extraction of the multi-section vital signal data in the window, and the multi-section vital signal data in the window can be calculated from multiple dimensions due to the complicated characteristics, so that the reliability of the calculation result is improved.

And step S240, taking a window as a unit, and calculating the characteristics extracted based on the window and a pre-trained prediction model to obtain a calculation result.

The prediction model is obtained through a full-supervision or semi-supervision training mode based on the characteristics of historical vital signal data in a big data platform and attribute categories corresponding to windows formed by the historical vital signal data. The quantified calculation may characterize a probability of predicting the occurrence of a medical event for the user. Specifically, the probability (0-100%) that a medical event will occur within a certain period of time in the future (e.g., one month, 2 months, 3 months, and/or 4 months, etc.) can be evaluated based on the calculation. The higher the value of the calculation result is, the higher the probability that the user will have a medical event in a certain period of time in the future is set to be, and it can also be understood that the time when the user is likely to have a medical event is closer to the current predicted time.

Taking a medical event of atrial fibrillation as an example, and describing an evaluation process of the medical event by combining a specific embodiment, fig. 3 shows a probability distribution graph of the occurrence of atrial fibrillation corresponding to a calculation result, as shown in fig. 3, when the current calculation result of a user is 12%, people with calculation results equal to 12% or close to 12% are found out based on a big data platform, people with the occurrence of atrial fibrillation are further found out from similar people, and statistics is made on the time from the time when the calculation result of the people with atrial fibrillation is 12% or close to 12%, for example, 10% -20% of the people with the occurrence of atrial fibrillation, to the time from the occurrence of atrial fibrillation. For example, the proportion of the number of people with atrial fibrillation in one month to the total population is calculated, so that the calculation of the probability that the user can intuitively predict the atrial fibrillation possibly occurring in the future period of time can be facilitated, and the early prevention can be facilitated.

Based on the above description of the embodiments, according to some embodiments of the present application, the following describes an exemplary training method of a predictive model in a specific embodiment, and fig. 4 shows a flowchart of the training method of the predictive model based on vital signal data, and as shown in fig. 4, the training method of the predictive model specifically includes:

step S410, obtaining a plurality of sections of historical vital signal data of at least one user based on time, wherein the historical vital signal data refers to vital signal data which belongs to a plurality of users and is based on a big data platform, and the historical vital signal data can be obtained by collecting wearing equipment of different users. The multiple pieces of historical vital signal data refer to multiple pieces of vital signal data corresponding to multiple pieces of different time. Because the big data platform has more historical vital signal data, the model obtained by training based on the historical vital signal data in the big data platform has higher credibility, and the calculation result is more accurate.

At step S420, at least one window for vital signal data is set based on a predetermined rule for each user, each window including a plurality of pieces of historical vital signal data. Where the historical vital signal data may be vital signal data of an existing user within the big data platform.

The setting of the window during the training of the model may be the same as in the calculation method described above.

According to an embodiment of the application, at least one window for vital signal data is set based on a predetermined rule. Further comprising updating the vital signal data within the window for each user.

In this application, for the update processing of the historical vital signal data in the window, reference may be made to the update processing method of the vital signal data in the window of the user in the above embodiment, which is not described herein again. It should be noted that, because the training of the prediction model is based on historical vital signal data of a plurality of users in the big data platform, in order to improve the accuracy of the prediction model, the window may be set to a time period of half a year or one year, and the frequency of updating the window may be lower than the frequency of updating the vital signal data in the window of the user in the calculation method, so that the historical vital data of the users in the big data platform is accumulated to a certain amount to improve the accuracy of the prediction model.

Step S430, for each user, extracting features from the multiple pieces of historical vital signal data within the window. It can be understood that, because the larger the data volume is based on multiple pieces of historical vital signal data in a large data platform, the more complex the relationship is, and the more diverse the feature extraction, not only the data feature extraction for each piece of vital signal data in the window of each user, but also the fusion of the features of the data features changing with time is required. For example, the feature extraction may be data meeting set requirements in a plurality of pieces of vital signal data, and relations and rules existing among the data. Not only the data features in each segment (time segment) of vital signal data, such as the highest value, the average value, or the lowest value of the data in each segment of vital signal data, can be extracted. And extracting regular sequence features existing among the data features, for example, extracting features from the highest value of data in each segment of vital signal data, wherein the highest values of the multiple segments of vital signal data follow a rule that gradually increases with time, the features are combined with each other, and the like, and are not limited herein.

Step S440, labeling the window, and determining the attribute type corresponding to the multiple sections of historical vital signal data in the window. Wherein the attribute categories are used for characterizing user-associated medical events, such as risk of occurrence of atrial fibrillation, e.g. high risk, low risk.

And such attribute classes may correspond to calculated values. The physical state of the user, e.g. the probability that the user may suffer from atrial fibrillation, or so-called risk value (0-100%) is determined based on the calculated values. Wherein, the closer the risk value is to 0, the more the user has no risk of atrial fibrillation, and the closer the risk value is to 100%, the more the user has a high possibility of risk occurrence, even the risk occurrence.

The attribute class may also characterize the physical state of the user in another form, for example, taking the vital signal data as the PPG signal data, the corresponding attribute class may include a normal heart rate, atrial fibrillation, or an intermediate heart rate therebetween, etc.

And step S450, taking a window as an example, taking the extracted features based on the window and the attribute categories corresponding to the window as a training sample set, and training on a machine by a semi-supervised training method and a fully-supervised training method to obtain a prediction model. That is, the predictive model may be a semi-supervised predictive model or a fully supervised predictive model.

It can be understood that the semi-supervised predictive model refers to an attribute category corresponding to the vital signal data of a part of people obtained by a part of labeling mode. Specifically, the partial labeling refers to labeling a part of people in the big data platform, the rest of people may not be labeled, and in the labeling, it is usually considered that the attribute categories corresponding to the vital signal data of people in the intermediate state are difficult to label, so that only the attribute categories corresponding to the vital signal data of people in the normal state and the diseased state are labeled. For example, in a population in a large data platform, in a normal state, for example, the PPG signal data in a window is normal, the attribute class corresponding to the window is a risk value of 0 or a normal heart rate, in a diseased state, for example, the value of the PPG signal data in the window is already a value at the time of atrial fibrillation, and the corresponding attribute class is a risk value of 100% or atrial fibrillation, a semi-supervised prediction model of the target is trained based on the labeled result.

It can be understood that the fully supervised prediction model refers to a fully labeled way for obtaining attribute categories corresponding to vital signal data of all people in a big data platform. That is to say, the fully supervised prediction model labels the vital signal data and the corresponding attribute types of the user in the normal state and the diseased state, and also labels the attribute types corresponding to the vital signal data of the user in the intermediate state between the normal state and the diseased state, for example, the result corresponding to the PPG signal data value in the window is between the normal state and the diseased state value, and the corresponding attribute types are a certain range of values (not including 0 and 100%) or an intermediate heart rate (sub-health) with a risk value of 0-100%. The labeling of the intermediate state may be performed automatically by using strategies such as Radial Basis Function (Radial Basis Function) and KNN nearest neighbor classification algorithm (k-nearest neighbors).

In addition, in some embodiments of the application, for the labeling of the user in the intermediate state in the fully supervised prediction model, the difficulty degree of the labeling of the intermediate state is considered to be higher, and the user in the intermediate state can be artificially and linearly labeled by assuming that the user follows a linear development law from a normal state to a diseased state. The labeling mode is simple, and meanwhile, the accuracy of the prediction model can be improved compared with a partial labeling mode.

In a preferred embodiment of the present application, when the number of people who have medical events is small and the number of people who are in a normal state is large, an over-fitting phenomenon occurs, and in order to reduce the over-fitting phenomenon, the over-fitting phenomenon can be reduced by adjusting parameters for multiple times and comparing the effects of multiple machine learning models.

According to the embodiment of the application, the prediction model trained by the method can be used for calculating the calculation result of the user more accurately, the calculation result can be used for indicating the analysis and prediction of the trend and probability of the medical event which does not occur currently, such as atrial fibrillation, but the atrial fibrillation which may occur in a certain period of time in the future, and providing some visually strong references.

An electronic device 500 for generating a predictive model is disclosed according to some embodiments of the present application, and fig. 5 illustrates a schematic structural diagram of the electronic device 500 for generating a predictive model. As shown in fig. 5, the electronic device 500 includes:

a first obtaining module 510 for obtaining a plurality of segments of time-based historical vital signal data of at least one user. The historical vital signal data refers to vital signal data based on the existing user in the big data platform, and the historical vital signal data can be acquired by the wearable device and the first acquisition module.

A first window setting module 520 for setting at least one window for vital signal data based on a predetermined rule for each user, each window including a plurality of pieces of historical vital signal data. Wherein the predetermined rule may comprise a predetermined time period and/or a predetermined amount of the pieces of vital signal data as the length of the window. The predetermined amount as the window length may be a predetermined number of pieces of vital signal data or a predetermined total size of data storage space, for example, a length of a window of storage space for storing a certain number of pieces of user vital signal data in units of bars or in units of data size (bytes), or a length of a window in combination of a predetermined time period and a predetermined amount of pieces of vital signal data.

A first extraction module 530 for extracting features from the plurality of segments of historical vital signal data within the window for each user.

And the labeling module 540 is configured to label the window and determine an attribute category corresponding to the multiple pieces of historical vital signal data in the window, where the attribute category is used to characterize a medical event corresponding to the window and associated with the user. The labeling of the window comprises the window labeling of all users in the big data platform, namely the full labeling, and the labeling of the window of part of the users, namely the part labeling.

And a training module 550, configured to use the window-based features and the attribute categories corresponding to the window as a training sample set, and train the training sample set by using a semi-supervised training method or a fully-supervised training method to obtain a prediction model.

The first window setting module 520 is further configured to update the data in the window, where the update process of the data in the window includes: and acquiring newly acquired historical vital signal data of the user, and storing the newly acquired historical vital signal data of the user in a window to update the window, wherein the newly acquired historical vital signal data of the user refers to a set of historical vital signal data which does not participate in training.

In the present application, the functions and the work flow of each component of the prediction model generation device 500 have been described in detail in the foregoing embodiments, and specific reference may be made to the training method of the prediction model in the foregoing embodiments, which is not described herein again.

According to the electronic device 500 for generating the prediction model of the embodiment of the application, the prediction model with accurate calculation can be obtained, and the calculation result of the vital signal data of the user calculated by the prediction model has higher accuracy and reliability.

The electronic device for generating the prediction model according to the embodiment of the present application may be implemented as a wearable device, a server connected to the wearable device, and a terminal device such as a smart phone, a tablet PC, a desktop PC, and a notebook PC that can communicate with the wearable device.

According to some embodiments of the present application, an electronic device 600 for computing is disclosed, and fig. 6 shows a schematic structural diagram of the electronic device 600 for computing. As shown in fig. 6, the electronic device 600 includes a processor therein, and is operable to implement the method for predicting a medical event according to the foregoing embodiment, and the electronic device 600 includes:

and a second obtaining module 610 for obtaining the vital signal data of the user, wherein the vital signal data of the user can be collected by the wearable device and obtained by the second obtaining module. Where the vital sign data of the user may be in units of bars or in units of data size (bytes). May be PPG signal data, such as frequency of heart beats.

And a second window setting module 620 for setting windows based on a predetermined rule, each window including a plurality of pieces of vital signal data. Wherein the predetermined rule may comprise a predetermined time period and/or a predetermined amount of the pieces of vital signal data as the length of the window. The predetermined amount as the window length may be a predetermined number of pieces of vital signal data or a predetermined total size of data storage space, for example, a length of a window of storage space for storing a certain number of pieces of user vital signal data in units of bars or in units of data size (bytes), or a length of a window in combination of a predetermined time period and a predetermined amount of pieces of vital signal data. The second extraction module 630 is configured to extract features from multiple segments of vital signal data in the window.

And the calculating module 640 is configured to calculate, by taking a window as a unit, a calculation result based on the features extracted by the window and the pre-trained prediction model.

According to an embodiment of the application, the electronic device 600 for computing further comprises: the evaluation module is used for counting the attribute categories of other users in the same preset range with the calculation result based on the data in the big data platform and the states of the other users corresponding to the attribute categories respectively;

and the evaluation module evaluates the state of the user corresponding to the calculation result according to the states of other users corresponding to the attribute categories respectively.

According to an embodiment of the present application, the electronic device for computing further comprises:

the data processing module can be used for processing abnormal vital signal data and carrying out noise reduction processing on the vital signal data of the user acquired each time, for example, noise reduction processing is carried out on the vital signal data through filtering, signal quality analysis and the like so as to obtain a more accurate calculation result.

In the present application, the functions and the workflow of each component of the electronic device have been described in detail in the foregoing embodiments, and for details, reference may be made to the method for predicting a medical event in the foregoing embodiments, and details are not described herein again.

According to the electronic device 600 for calculating in the embodiment of the application, the calculation result of the medical event that can be obtained can be used for evaluating the probability of the medical event occurring for the user. Further, the probability (0-100%) that a medical event will occur within a certain period of time in the future (e.g., one month, 2 months, 3 months, and/or 4 months, etc.) can be evaluated based on the calculation. The higher the calculation result is, the higher the probability that the user is labeled as having a medical event in a certain period of time in the future is, and it can also be understood that the time when the user is likely to have the medical event is closer to the current predicted time. The electronic device 600 for computing according to the embodiment of the present application may be implemented as a wearable device, a server connected to the wearable device, and a terminal device such as a smart phone, a tablet PC, a desktop PC (Personal Computer), a notebook PC, and the like that can communicate with the wearable device.

The electronic device for computing according to the embodiment of the application may further include a display module for presenting the result of the computing and a graph or an image obtained based on the result of the computing to a user.

According to some embodiments of the present application, a wearable device is disclosed, and fig. 7 shows a schematic structural view of a wearable device 700. As shown in fig. 7, in particular, the wearable device 700 may include:

the collector 710 is configured to periodically collect the vital signal data of the user, and it is understood that the collector 710 may actively collect the vital signal data of the user, or may be controlled by the user and collect the vital signal data at a suitable time.

A memory 720 for storing instructions; a processor 730 for reading the instructions stored in the memory to execute the prediction method of the medical event of the above embodiment; and

a display device 740 implemented as a display screen for displaying the calculation result, so that the user can visually know the calculation result through the wearable device 700, and further can know the probability of the occurrence of the medical event in a certain period of time in the future through the calculation result.

Although the wearable device 700 itself includes the display screen for displaying information such as a calculation result in the above embodiments, it will be understood by those skilled in the art that some wearable devices may not have the display screen themselves, and according to the wearable device of the embodiments of the present application, data may be synchronized to a terminal device having a display screen, such as a smartphone, a tablet PC, a desktop PC, and a notebook PC, which is communicably connected to the wearable device, so as to present a corresponding numerical value, graphic, or image to a user.

As an exemplary application scenario according to an embodiment of the present application, after a user wears the wearable device according to the present application, the wearable device may periodically or manually initiate collection of vital signal data by the user. When the vital signal data is collected, the vital signal data is immediately or accumulated to a certain extent in the wearable device, and then the vital signal data is transmitted to the cloud server side. The cloud server side includes an electronic device for calculating a probability of a medical event based on vital signal data of a user. After receiving the vital signal data of the user, the cloud server performs calculation by using the calculation method according to the embodiment of the application to obtain a calculation result.

The calculation result may be a simple display of the result directly or may be a result of more complicated analysis based on the calculation result.

According to one embodiment of the application, in order to facilitate the user to know the physical condition of the user, a trend graph of the change of the calculation result along with the time can be drawn, and the user can intuitively obtain the change trend of the calculation result through the trend graph. It is understood that the representation of the probability of the medical event can also be visually represented by a trend graph, so that the user can visually obtain the probability of the medical event occurring in the future. Wherein, the trend chart can be a chart, a bar chart or a graph. In other embodiments of the application, the user can be prompted through a voice broadcast mode and the like, so that the user can visually know the self condition and take corresponding preventive measures in time.

Fig. 8a shows an exemplary graph of the calculated results of a medical event, which are exemplarily listed as risk values, and as shown in fig. 8a, from 2.3 days to 4.9 days, a window is defined, wherein the window may include vital signal data within one hour of 2.3 days, vital signal data within two hours of 2.4 days, and a set of vital signal data within a certain period of 2.8 days and 3.10 days, and it can be seen that the values of the vital signal data within the window are not continuous in time. When the vital sign data of the user in the window is accumulated to a certain amount, the calculation result on the day of 4.9 is calculated to be 0.2 based on the vital sign data of the user in the window. The updated window is from 2.4 to 4.16 days, that is, the vital sign data of the user of the number 4.16 is added, and the calculation result of the day of 4.16 days is calculated to be 0.19 based on the vital sign data of the user in the updated window.

The calculation result can be sent to the wearable device from the cloud server to be displayed, so that the user can know the relevant state of the medical event from the wearable device very easily. Fig. 8b and 8c illustrate schematic views of UI interfaces output by a display module according to a specific embodiment of the present application. The calculation method provided by the application has the advantages that besides the accurate calculation result of the medical event is provided for the user, the analysis result can be displayed to the user in an intuitive form to help the user to know the risk value of the medical event in the future, and the calculation method provided by the application also has one of the advantages.

As shown in the UI interface diagram in fig. 8b, according to the continuously updated calculation result data, a trend chart of the calculation result of the user (e.g., the risk value of the medical event shown in fig. 8 b) can be drawn, and through the trend chart, the user can intuitively observe the trend of the calculation result to further determine the probability that the medical event may occur.

As shown in the UI interface diagram in fig. 8c, in combination with the big data platform, the calculation result of the user may be matched and compared with the result samples of numerous people on the big data platform, so as to predict the calculation of the probability that atrial fibrillation may occur in the future in combination with the comparison result, and visually display the calculation result for early prevention. Such as the time of atrial fibrillation at the distance shown in fig. 8b, but the application is not so limited and other probabilities will occur as understood by those skilled in the art.

In addition, as will be understood by those skilled in the art, although the wearable device includes its own display device for processing display data and a display screen for displaying related information, the display interface as shown in fig. 8b and 8c may be implemented on the display screen of the wearable device itself, or may be implemented on other terminal devices after completing data synchronization with the wearable device, such as a smart phone, a desktop computer, a smart television, and the like, and the application is not limited herein.

According to some embodiments of the present application, the wearable device 700 may further include a data processing module, and the data processing module may be configured to process abnormal vital signal data and perform noise reduction processing on the vital signal data of the user collected each time, for example, perform noise reduction processing on the vital signal data through filtering, signal quality analysis, and the like, so as to obtain a more accurate calculation result.

It will be appreciated that abnormal vital signal data in the present application may comprise vital signal data of a characteristic, characteristic or value that is outside a predetermined threshold range or level that is expected, healthy or normal by the user, for example, in the case of PPG signal data in which the user is in motion, the exceeding of which may be understood to constitute a vital signal data abnormality. In some other cases, PPG signal data that is not regular by the user may be considered abnormal vital signal data.

In some embodiments of the application, the processing module may further sense a motion state of the user, and during a period of the user's motion, when the vital signal data exceeds a certain range, it may be considered that the vital signal data does not constitute abnormal vital signal data, for example, when the PPG signal data of the user is elevated due to the motion in the motion state, the PPG signal data does not constitute abnormal PPG signal data. Therefore, the obtained vital signal data of the users in the window are not uniformly distributed in time, the vital signal data of the users possibly appearing in the window under the condition of setting the same time duration is more or less, the values in some time periods are more dense, and the values in some time periods are less. Therefore, the vital signal data of the user in the window can be accumulated to a certain amount, and the accuracy and the reliability of the calculation result can be effectively improved when the result is calculated.

In other embodiments of the present application, the wearable device may be connected to a terminal device, such as a mobile phone and a computer, and one or more of the calculation result, a trend graph of the change of the calculation result at any time, and a probability trend graph of the occurrence of the medical event corresponding to the calculation result may be displayed through the terminal device.

According to an embodiment of the application, the wearable device may further include an evaluation module, after the receiving module receives the calculation result obtained through cloud computing, the evaluation module searches for a crowd similar to the calculation result according to the cloud big data platform, that is, searches for a crowd with the calculation result equal to or close to the calculation result, and evaluates the probability of the medical event occurring in a future period of time by comparing the medical event occurring in the future period of time by the similar crowd.

Referring now to FIG. 9, shown is a block diagram of an apparatus 1200 in accordance with one embodiment of the present application. The device 1200 may include one or more processors 1201 coupled to a controller hub 1203. For at least one embodiment, the controller hub 1203 communicates with the processor 1201 via a multi-drop Bus such as a Front Side Bus (FSB), a point-to-point interface such as a Quick Path Interconnect (QPI), or similar connection 1206. The processor 1201 executes instructions that control general types of data processing operations. In one embodiment, Controller Hub 1203 includes, but is not limited to, a Graphics Memory Controller Hub (GMCH) (not shown) and an Input/Output Hub (IOH) (which may be on separate chips) (not shown), where the GMCH includes a Memory and a Graphics Controller and is coupled to the IOH.

The device 1200 may also include a coprocessor 1202 and a memory 1204 coupled to the controller hub 1203. Alternatively, one or both of the memory and GMCH may be integrated within the processor (as described herein), with the memory 1204 and coprocessor 1202 being directly coupled to the processor 1201 and to the controller hub 1203, with the controller hub 1203 and IOH being in a single chip. The Memory 1204 may be, for example, a Dynamic Random Access Memory (DRAM), a Phase Change Memory (PCM), or a combination of the two. In one embodiment, coprocessor 1202 is a special-Purpose processor, such as, for example, a high-throughput MIC processor (MIC), a network or communication processor, compression engine, graphics processor, General Purpose Graphics Processor (GPGPU), embedded processor, or the like. The optional nature of coprocessor 1202 is represented in FIG. 9 by dashed lines.

Memory 1204, as a computer-readable storage medium, may include one or more tangible, non-transitory computer-readable media for storing data and/or instructions. For example, the memory 1204 may include any suitable non-volatile memory, such as flash memory, and/or any suitable non-volatile storage device, such as one or more Hard-Disk drives (Hard-Disk drives, hdd (s)), one or more Compact Discs (CD) drives, and/or one or more Digital Versatile Discs (DVD) drives.

In one embodiment, device 1200 may further include a Network Interface Controller (NIC) 1206. Network interface 1206 may include a transceiver to provide a radio interface for device 1200 to communicate with any other suitable device (e.g., front end module, antenna, etc.). In various embodiments, the network interface 1206 may be integrated with other components of the device 1200. The network interface 1206 may implement the functions of the communication unit in the above-described embodiments.

The device 1200 may further include an Input/Output (I/O) device 1205. I/O1205 may include: a user interface designed to enable a user to interact with the device 1200; the design of the peripheral component interface enables peripheral components to also interact with the device 1200; and/or sensors may be configured to determine environmental conditions and/or location information associated with device 1200.

It is noted that fig. 9 is merely exemplary. That is, although fig. 9 shows that the apparatus 1200 includes a plurality of devices, such as the processor 1201, the controller hub 1203, the memory 1204, etc., in practical applications, an apparatus using the methods of the present application may include only a part of the devices of the apparatus 1200, for example, only the processor 1201 and the NIC1206 may be included. The nature of the alternative device in fig. 9 is shown in dashed lines.

According to some embodiments of the present application, the memory 1204 serving as a computer-readable storage medium stores instructions, which when executed on a computer, cause the system 1200 to perform the calculation method according to the above embodiments, which may specifically refer to the method of the above embodiments, and will not be described herein again.

Referring now to fig. 10, shown is a block diagram of a SoC (System on Chip) 1300 in accordance with an embodiment of the present application. In fig. 10, like parts have the same reference numerals. In addition, the dashed box is an optional feature of more advanced socs. In fig. 10, SoC1300 includes: an interconnect unit 1350 coupled to the application processor 1310; a system agent unit 1380; a bus controller unit 1390; an integrated memory controller unit 1340; a set or one or more coprocessors 1320 which may include integrated graphics logic, an image processor, an audio processor, and a video processor; a Static Random Access Memory (SRAM) unit 1330; a Direct Memory Access (DMA) unit 1360. In one embodiment, the coprocessor 1320 includes a special-purpose processor, such as, for example, a network or communication processor, compression engine, GPGPU, a high-throughput MIC processor, embedded processor, or the like.

Included in Static Random Access Memory (SRAM) unit 1330 may be one or more computer-readable media for storing data and/or instructions. A computer-readable storage medium may have stored therein instructions, in particular, temporary and permanent copies of the instructions. The instructions may include: when executed by at least one unit in the processor, the Soc1300 may execute the calculation method according to the foregoing embodiment, which may specifically refer to the method of the foregoing embodiment and will not be described herein again.

Embodiments of the mechanisms disclosed herein may be implemented in hardware, software, firmware, or a combination of these implementations. Embodiments of the application may be implemented as computer programs or program code executing on programmable systems comprising at least one processor, a storage system (including volatile and non-volatile memory and/or storage elements), at least one input device, and at least one output device.

Program code may be applied to input instructions to perform the functions described herein and generate output information. The output information may be applied to one or more output devices in a known manner. For purposes of this Application, a processing system includes any system having a Processor such as, for example, a Digital Signal Processor (DSP), a microcontroller, an Application Specific Integrated Circuit (ASIC), or a microprocessor.

The program code may be implemented in a high level procedural or object oriented programming language to communicate with a processing system. The program code can also be implemented in assembly or machine language, if desired. Indeed, the mechanisms described in this application are not limited in scope to any particular programming language. In any case, the language may be a compiled or interpreted language.

In some cases, the disclosed embodiments may be implemented in hardware, firmware, software, or any combination thereof. The disclosed embodiments may also be implemented as instructions carried by or stored on one or more transitory or non-transitory machine-readable (e.g., computer-readable) storage media, which may be read and executed by one or more processors. For example, the instructions may be distributed via a network or via other computer readable media. Thus, a machine-readable medium may include any mechanism for storing or transmitting information in a form readable by a machine (e.g., a computer), including, but not limited to, floppy diskettes, optical disks, Compact disk Read Only memories (CD-ROMs), magneto-optical disks, Read Only Memories (ROMs), Random Access Memories (RAMs), Erasable Programmable Read Only Memories (EPROMs), Electrically Erasable Programmable Read Only Memories (EEPROMs), magnetic or optical cards, flash Memory, or a tangible machine-readable Memory for transmitting information (e.g., carrier waves, infrared signals, digital signals, etc.) using the Internet in electrical, optical, acoustical or other forms of propagated signals. Thus, a machine-readable medium includes any type of machine-readable medium suitable for storing or transmitting electronic instructions or information in a form readable by a machine (e.g., a computer).

In the drawings, some features of the structures or methods may be shown in a particular arrangement and/or order. However, it is to be understood that such specific arrangement and/or ordering may not be required. Rather, in some embodiments, the features may be arranged in a manner and/or order different from that shown in the figures. In addition, the inclusion of a structural or methodical feature in a particular figure is not meant to imply that such feature is required in all embodiments, and in some embodiments, may not be included or may be combined with other features.

It should be noted that, in the embodiments of the apparatuses in the present application, each unit/module is a logical unit/module, and physically, one logical unit/module may be one physical unit/module, or may be a part of one physical unit/module, and may also be implemented by a combination of multiple physical units/modules, where the physical implementation manner of the logical unit/module itself is not the most important, and the combination of the functions implemented by the logical unit/module is the key to solve the technical problem provided by the present application. Furthermore, in order to highlight the innovative part of the present application, the above-mentioned device embodiments of the present application do not introduce units/modules which are not so closely related to solve the technical problems presented in the present application, which does not indicate that no other units/modules exist in the above-mentioned device embodiments.

It is noted that, in the examples and descriptions of this patent, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, the use of the verb "comprise a" to define an element does not exclude the presence of another, same element in a process, method, article, or apparatus that comprises the element.

While the present application has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present application.

Claims

1. A method of predicting a medical event, the method comprising:

the electronic equipment acquires life signal data of a user;

the electronic equipment sets a window based on a preset rule, wherein the window comprises a plurality of sections of vital signal data;

the electronic equipment extracts features from the multi-section vital signal data in the window;

and the electronic equipment calculates a calculation result based on the characteristics extracted from the window and a prediction model.

2. The method of claim 1, wherein the predetermined rule comprises:

at least one of the predetermined time period and the predetermined amount of the pieces of vital signal data is taken as the length of the window.

3. The method of claim 2, wherein taking the predetermined amount of the plurality of pieces of vital signal data as the length of the window comprises:

using a predetermined number of the plurality of pieces of vital signal data as the length of the window, or

And taking the preset total size of the plurality of pieces of vital signal data as the length of the window.

4. The method according to claim 1, wherein the window is set based on a predetermined rule, and the update process is performed on data within the window,

the updating process of the data in the window comprises the following steps:

for newly acquired life signal data of a user, if enough idle positions exist in the window, directly storing the current life signal data to the idle positions according to the acquisition sequence;

and for newly acquired life signal data of the user, if the window does not have enough free positions, accepting or rejecting the life signal data in the window and the newly acquired life signal data of the user which is not stored in the window.

5. The method of claim 4, wherein the rounding off the vital signal data in the window with the newly acquired vital signal data of the user not stored in the window comprises:

according to the sequence of the acquisition time, only the data with the latest acquisition time is reserved under the condition that the window is full.

6. The method of claim 4, wherein the rounding off the vital signal data in the window with the newly acquired vital signal data of the user not stored in the window comprises:

and scoring each piece of vital signal data in the window and newly acquired vital signal data according to a preset state, and deleting the vital signal data with the score in the preset state until the set size of the window is met.

7. The method of claim 6, wherein the rounding off the vital signal data in the window with the newly acquired vital signal data of the user not stored in the window comprises:

classifying the vital signal data stored in the window into corresponding classes according to a classification rule from near to far from the current time when the newly acquired vital signal data is acquired;

and sequentially checking various classes from far to near according to the current time until the class of the vital signal data is checked, scoring the vital signal data in the class, and deleting the vital signal data with the score in a preset state of the user.

8. The method of claim 7,

at least one vital signal data is retained in each class, and different weight values are set for the vital signal data in each class according to the contribution degree to the calculation result.

9. The method of claim 1, wherein the pre-trained predictive model is obtained by machine learning training, and the training method comprises:

obtaining time-based historical vital signal data for at least one user;

setting at least one window for vital signal data based on a predetermined rule, each of the windows including a plurality of pieces of historical vital signal data;

extracting features from the plurality of segments of historical vital signal data within the window;

labeling the window, and determining attribute categories corresponding to the multiple pieces of historical vital signal data in the window, wherein the attribute categories are used for representing medical events corresponding to the window and associated with the user;

and taking the features extracted based on the window and the attribute categories corresponding to the window as a training sample set, and training by a semi-supervised training method or a fully-supervised training method to obtain a prediction model.

10. The method of claim 9, wherein labeling the window and determining the attribute class corresponding to the historical vital signal data within the window in the fully supervised training approach comprises:

and carrying out manual linear labeling or automatic labeling on the window corresponding to the attribute category in the intermediate state, wherein the automatic labeling adopts a radial basis function and a nearest neighbor classification algorithm for labeling.

11. The method of claim 1, further comprising:

counting attribute categories of other users in the same preset range with the calculation result and states of other users corresponding to the attribute categories respectively based on data in a big data platform;

and evaluating the states of the users corresponding to the calculation results according to the states of the other users corresponding to the attribute categories respectively.

12. The method of claim 1, further comprising:

before acquiring vital signal data of a user, carrying out noise reduction processing on the vital signal data of the user.

13. A wearable device, comprising:

the collector is used for periodically collecting life signal data of a user;

a memory to store instructions; and

a processor for reading the instructions stored in the memory to perform the steps of:

acquiring life signal data of a user;

setting a window based on a predetermined rule, the window including a plurality of pieces of vital signal data;

extracting features from the multiple sections of vital signal data in the window;

calculating the characteristics extracted based on the window and a prediction model to obtain a calculation result;

and the display screen is used for displaying the calculation result.

14. The wearable device of claim 13, wherein the predetermined rule comprises:

15. The wearable device of claim 14, wherein taking the predetermined amount of the pieces of vital signal data as the length of the window comprises:

16. The wearable device of claim 13, wherein the processor is specifically configured to: setting a window based on a predetermined rule, and updating data in the window,

the updating process of the data in the window comprises the following steps:

17. The wearable device of claim 16, wherein the processor is specifically configured to:

18. The method of claim 16, wherein the processor is specifically configured to:

19. The method of claim 18, wherein the processor is specifically configured to:

20. The wearable device of claim 19,

21. The wearable device of claim 1, wherein the processor further comprises:

22. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program which, when executed by a processor, causes the processor to perform the method of any of claims 1-12.