CN114792565A - Health management method and device, wearable device, electronic device and medium - Google Patents

Health management method and device, wearable device, electronic device and medium Download PDF

Info

Publication number
CN114792565A
CN114792565A CN202110114648.1A CN202110114648A CN114792565A CN 114792565 A CN114792565 A CN 114792565A CN 202110114648 A CN202110114648 A CN 202110114648A CN 114792565 A CN114792565 A CN 114792565A
Authority
CN
China
Prior art keywords
static
data
health risk
dynamic
value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110114648.1A
Other languages
Chinese (zh)
Inventor
俞轶
李佳
武超
朱国康
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Anhui Huami Health Technology Co Ltd
Original Assignee
Anhui Huami Health Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Anhui Huami Health Technology Co Ltd filed Critical Anhui Huami Health Technology Co Ltd
Priority to CN202110114648.1A priority Critical patent/CN114792565A/en
Publication of CN114792565A publication Critical patent/CN114792565A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Public Health (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

The invention provides a health management method, a health management device, wearable equipment, electronic equipment and a health management medium, wherein the method comprises the following steps: acquiring dynamic data and static data of a user; acquiring dynamic features corresponding to the dynamic data and static features corresponding to the static data; inputting the dynamic features and the static features into a trained mixed data detection model to obtain a health risk probability value; and inputting the health risk probability value into a trained health risk prediction model to obtain a health risk value so as to predict the health risk value of the disease suffered by the user.

Description

Health management method and device, wearable device, electronic device and medium
Technical Field
The invention relates to the technical field of wearable equipment, in particular to a health management method and device, wearable equipment, electronic equipment and a medium.
Background
Although hypertension is a common chronic cardiovascular disease, and the onset process is relatively slow, patients often show mild clinical symptoms such as palpitation and dizziness, but hypertension is an important factor for serious cardiovascular diseases and organ diseases such as stroke and renal failure. Chinese guidelines for hypertension prevention and treatment (revised 2018) indicate that the prevalence rate of hypertension of 18-year-old and older residents in China in 2012 and 2015 reaches 27.9%, while the awareness rate and control rate of hypertension are only 51.6% and 16.8%, and the prevalence rate of hypertension still tends to increase. Therefore, the widely applicable hypertension detection and prediction has important significance for the prevention and control of hypertension of people in China.
Disclosure of Invention
The present invention is directed to solving, at least in part, one of the technical problems in the related art.
Therefore, the first objective of the present invention is to provide a health management method for predicting a health risk value of a user suffering from a disease.
A second object of the present invention is to provide a health management device.
A third object of the invention is to propose a wearable device.
A fourth object of the invention is to propose an electronic device.
A fifth object of the invention is to propose a computer program product.
In order to achieve the above object, a health management method is provided in an embodiment of a first aspect of the present invention, including: acquiring dynamic data and static data of a user, wherein the dynamic data comprises at least one of heart rate, sleep stage data or activity data of the user, and the static data comprises basic information of the user or comprises the basic information and health information of the user; acquiring dynamic features corresponding to the dynamic data and static features corresponding to the static data; inputting the dynamic features and the static features into a trained mixed data detection model to obtain a health risk probability value; and inputting the health risk probability value into a trained health risk prediction model to obtain a health risk value.
According to an embodiment of the present application, the obtaining the dynamic feature corresponding to the dynamic data includes: obtaining a statistic value of the dynamic data in each time window according to the dynamic data; obtaining the static features corresponding to the static data, including: and acquiring a first static characteristic according to the static data, wherein the first static characteristic comprises a Body Mass Index (BMI) and a body fat rate, and generating a second static characteristic based on a gradient boosting decision tree according to the static data.
According to an embodiment of the present application, the hybrid data detection model includes a dynamic data detection model, a static data detection model, and a migration learning model, the dynamic features and the static features are input into the trained hybrid data detection model to obtain a health risk probability value, including: inputting the dynamic characteristics into the trained dynamic data detection model to obtain a dynamic data detection result; inputting the static characteristics into the trained static data detection model to obtain a static data detection result; inputting the dynamic data detection result and the static data detection result into the trained migration learning model to obtain the health risk probability value; and the migration learning model performs feature fusion on the dynamic data detection result and the static data detection result, and inputs the fused features into a Fully Connected Network (FCN) to obtain a health risk probability value.
According to an embodiment of the present application, the inputting the health risk probability value into a trained health risk prediction model to obtain a health risk value includes: determining a mask value according to the missing condition of the health risk probability value in each time window; and inputting the health risk probability values and the mask values in a plurality of time windows into the trained health risk prediction model to obtain the health risk values.
According to an embodiment of the present application, before the obtaining of the dynamic feature corresponding to the dynamic data and the static feature corresponding to the static data, the method further includes: and preprocessing the dynamic data and the static data, wherein the preprocessing comprises at least one of wildcard representation of the defect field of the static data, screening of the dynamic data in a sliding window mode, conversion from minute level to hour level of the dynamic data or normalization processing of the static data and the dynamic data.
According to an embodiment of the application, if the health risk value is larger than a preset health risk threshold value, outputting early warning information and a target health risk value to the user, obtaining a target dynamic characteristic and a target static characteristic according to the target health risk value and the trained mixed data detection model, and outputting the target dynamic characteristic and the target static characteristic to the user; and if the health risk value is equal to or smaller than the health risk threshold value, outputting the change situation of the health risk value when the dynamic characteristic and the static characteristic fluctuate within a set range near the average value to the user.
According to the health management method, the static data and the dynamic data of the user are analyzed, the health risk value of the user can be accurately obtained, the risk that the user suffers from illness in the future is predicted, and the health risk value prediction result of the user is more objective
To achieve the above object, a second embodiment of the present invention provides a health management device, including: the device comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring dynamic data and static data of a user, the dynamic data comprises at least one of heart rate, sleep stage data or activity data of the user, and the static data comprises basic information of the user or basic information and health information of the user; the second acquisition module is used for acquiring the dynamic features corresponding to the dynamic data and the static features corresponding to the static data; the first prediction module is used for inputting the dynamic characteristics and the static characteristics into a trained mixed data detection model to obtain a health risk probability value; and the second prediction module is used for inputting the health risk probability value into a trained health risk prediction model to obtain a health risk value.
According to an embodiment of the present application, the second obtaining module is further configured to: obtaining a statistic value of the dynamic data in each time window according to the dynamic data; or also for: and acquiring a first static characteristic according to the static data, wherein the first static characteristic comprises a Body Mass Index (BMI) and a body fat rate, and generating a second static characteristic based on a gradient boosting decision tree according to the static data.
According to an embodiment of the present application, the hybrid data detection model includes a dynamic data detection model, a static data detection model, and a migration learning model, and the first prediction module is specifically configured to: inputting the dynamic characteristics into the trained dynamic data detection model to obtain a dynamic data detection result; inputting the static characteristics into the trained static data detection model to obtain a static data detection result; inputting the dynamic data detection result and the static data detection result into the trained migration learning model to obtain the health risk probability value; and the transfer learning model performs feature fusion on the dynamic data detection result and the static data detection result, and inputs the fused features into a full-connection network FCN to obtain a health risk probability value.
According to an embodiment of the present application, the second prediction module is specifically configured to: determining a mask value according to the missing condition of the health risk probability value in each time window; and inputting the health risk probability values and the mask values in a plurality of time windows into the trained health risk prediction model to obtain the health risk values.
According to an embodiment of the present application, the second obtaining module is further configured to: and preprocessing the dynamic data and the static data, wherein the preprocessing comprises at least one of wildcard representation of the defect field of the static data, screening of the dynamic data in a sliding window mode, conversion from minute level to hour level of the dynamic data or normalization processing of the static data and the dynamic data.
According to an embodiment of the application, the second prediction module is further configured to: if the health risk value is larger than a preset health risk threshold value, outputting early warning information and a target health risk value to the user, obtaining a target dynamic characteristic and a target static characteristic according to the target health risk value and the trained mixed data detection model, and outputting the target dynamic characteristic and the target static characteristic to the user; and if the health risk value is equal to or smaller than the health risk threshold value, outputting the change situation of the health risk value when the dynamic characteristic and the static characteristic fluctuate within a set range near the average value to the user.
To achieve the above object, a third aspect of the present invention provides a wearable device, including the health management apparatus as described in the second aspect of the present invention.
In order to achieve the above object, a fourth aspect of the present invention provides an electronic device, including a memory, a processor; wherein the processor executes a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the health management method as described in the embodiment of the first aspect.
In order to achieve the above object, a fifth embodiment of the present invention provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the computer program implements the health management method according to the first embodiment.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The foregoing and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is a flowchart of a health management method according to an embodiment of the present application;
fig. 2 is a flowchart of another health management method according to an embodiment of the present application;
FIG. 3 is a schematic structural diagram of a hybrid data detection model according to an embodiment of the present application;
fig. 4 is a flowchart of another health management method according to an embodiment of the present application;
fig. 5 is a schematic diagram of mask value acquisition according to an embodiment of the present disclosure;
fig. 6 is a flowchart of another health management method according to an embodiment of the present application;
FIG. 7 is a block diagram of a health management device according to an embodiment of the present application;
fig. 8 is a block diagram of a wearable device according to an embodiment of the present disclosure;
fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
Health management methods, apparatuses, wearable devices, electronic devices, and media according to embodiments of the present invention are described below with reference to the accompanying drawings.
Fig. 1 is a flowchart of a health management method according to an embodiment of the present application. It should be noted that the execution subject of the health management method of this embodiment is a health management apparatus, and the health management apparatus may specifically be a hardware device, or software in a hardware device, or the like. Such as a terminal device, a wearable device, a server, etc. As shown in fig. 1, the method specifically comprises the following steps:
step 101, acquiring dynamic data and static data of a user, wherein the dynamic data comprises at least one of heart rate, sleep stage data or activity data of the user, and the static data comprises basic information of the user or comprises basic information and health information of the user.
The dynamic data may specifically include, but is not limited to, data with a dynamically changing characteristic, which is acquired by the wearable device, of at least one of a minute-level heart rate of the wearing user, sleep stage data, activity data and the like, wherein the sleep stage data may be a sleep stage corresponding to each minute of the user, such as a deep sleep stage or a light sleep stage, and the activity data may be acceleration information of the user acquired by the acceleration sensor. The static data may be basic information provided by the user by filling in or the like, and the static data may further include static information such as health information of the user in addition to the basic information, where the basic information may include, for example, sex, age, height, weight and the like, and the health information may specifically include, but is not limited to, at least one of information whether the user drinks alcohol, whether the user smokes smoke and the like.
It should be appreciated that the static data may be obtained by the app matching the wearable device, for example, by the app matching the wearable device sending a questionnaire of the static data to the user, thereby obtaining the static data of the user by identifying the questionnaire.
Optionally, after the dynamic data and the static data are acquired, the dynamic data and the static data may be stored in the form of a Hive table of a data warehouse. The dynamic data of the user can be updated according to the collection condition according to the frequency, for example, every day, and the static data of the user is updated according to the time condition of pushing the questionnaire to the user or modifying the questionnaire by the user.
And 102, acquiring dynamic features corresponding to the dynamic data and static features corresponding to the static data.
As a possible embodiment, acquiring the dynamic feature corresponding to the dynamic data may include: and obtaining the statistic value of the dynamic data in each time window according to the dynamic data.
For example, if the driving dynamics window has a length of 30 days, the dynamics of skewness, kurtosis, maximum value, minimum value, average value, variance, etc. of the dynamics data within 30 days can be obtained.
As another possible embodiment, acquiring static features corresponding to static data includes: and acquiring a first static characteristic according to the static data, wherein the first static characteristic comprises a Body Mass Index (BMI) and a Body fat rate, and generating a second static characteristic based on a gradient boost decision tree according to the static data.
It should be noted that the static features are caused by two aspects, one is a first static feature such as body mass index and body fat rate obtained according to the prior formula, and the other is a second static feature generated by using the gradient boosting decision tree.
For the first static feature, because the static information of the user, such as height, weight and the like, is obtained through the questionnaire, the first static feature belonging to the user, such as body mass index BMI and body fat percentage, can be obtained by using the heuristic feature created by the medical priori knowledge. Wherein, the body mass index BMI is calculated by the formula of body weight (kilogram)/height (meter) 2 The factors of saint ao and weight of the user can be effectively considered. The body fat rate comprehensively considers three types of information of gender, age and body mass index BMI, specifically, the male body fat rate is calculated as 1.2 BMI +0.23 age-5.58, while the female body fat rate is calculated as: 1.2 BMI +0.23 age-5.4.
For the second static feature, a trained gradient boosting decision tree model can be used for obtaining. Specifically, the gradient boost decision tree model may be trained by using a training sample, a new feature may be constructed by using a tree learned by the model, the new feature may be added to the original feature and trained together to obtain a final gradient boost decision tree model, and static data may be input to the final gradient boost decision tree model during application to obtain the second static feature through the final gradient boost decision tree model.
And 103, inputting the dynamic features and the static features into the trained mixed data detection model to obtain a health risk probability value.
It should be noted that the hybrid data detection model includes a dynamic data detection model, a static data detection model, and a transfer learning model.
Further, inputting the dynamic features and the static features into the trained mixed data detection model to obtain a health risk probability value, as shown in fig. 2 and fig. 3, including:
step 201, inputting the dynamic characteristics into the trained dynamic data detection model to obtain a dynamic data detection result.
The dynamic data detection model adopts a structural design of a Long Short-Term Memory network (LSTM), the compensation of the LSTM can be set to 24 and represents the dynamic characteristics of 24 hours in a day, and the input of each step is the statistical characteristics of the dynamic data of the hour of the user, such as skewness, sharpness, maximum value, minimum value, average value, variance and the like of the dynamic data.
Step 202, inputting the static features into the trained static data detection model to obtain a static data detection result.
The static data model adopts the structural design of a Full Connected Network (FCN), static information obtained through questionnaires and/or apps is input into the Full Connected Network (FCN), and a static data detection result can be obtained by using the trained Full Connected Network (FCN).
It should be noted that, in the present application, the dynamic characteristics and the static characteristics are separately trained through the dynamic data detection model and the static data detection model, but are not unified into one network, so that the useful data in the big data can be fully utilized. Specifically, in real life, there may be a case where some users have static data but less dynamic data, or some users have only dynamic data but lack of static data, and if only a comprehensive model of dynamic data and static data is adopted, data loss of some users is easily caused.
It should also be understood that, when the dynamic data detection model and the static data detection model are trained separately, only the data corresponding to the models are still used, so that the training effect of the models is effectively improved, and data loss generated when the trained models are used is avoided.
And 203, inputting the dynamic data detection result and the static data detection result into the trained migration learning model to obtain a health risk probability value, wherein the migration learning model performs feature fusion on the dynamic data detection result and the static data detection result and inputs the fused features into the fully-connected network FCN to obtain the health risk probability value.
That is, the migration learning model migrates the dynamic data detection result and the static data detection result into the migration learning model, wherein the dynamic data detection result may be a hidden layer feature of the dynamic data detection model, and the static data detection result may be a hidden layer feature of the static data detection model. The dynamic data detection result and the static data detection result migrated to the migration learning model need to be subjected to feature fusion first, and then the fused features are input into the fully-connected network FCN to obtain a health risk probability value.
Further, the last layer of the fully-connected network FCN network is a softmax layer, and the softmax layer adopts the following calculation method: suppose that the input to the softmax layer includes k neurons representing k classes of classification task, i respectively 1 -i k Then the probability of softmax layer to nth class outputs Output n Comprises the following steps:
Figure BDA0002917281000000071
from the above formula, it can be seen that the softmax output value ranges between 0 and 1, representing the probability value that the current input is in a different category, i.e., the health risk probability value.
And 104, inputting the health risk probability value into the trained health risk prediction model to obtain a health risk value.
Further, inputting the health risk probability value into the trained health risk prediction model to obtain a health risk value, as shown in fig. 4, including:
step 401, determining a mask value according to the absence condition of the health risk probability value in each time window.
It should be noted that the missing condition of the health risk probability value in each time window means that dynamic data of the user may be missing in the time for predicting the health risk value, for example, when it is set to use data of a year before the day to be predicted to predict the health risk value of the user, the health risk probability value of each day in the past year is needed to determine the health risk value of the day to be predicted, however, the health risk probability value corresponding to the missing date of the dynamic data is also correspondingly missing to affect the prediction of the health risk value of the day to be predicted, so that in order to fully utilize all data, the missing condition of the health risk probability value of the current window is represented by a mask value mask when the health risk prediction is performed.
As shown in fig. 5, if the health risk probability value of the current window is missing, the mask is 0, and if the health risk probability value of the current window is not missing, the mask is 1.
Step 402, inputting the health risk probability values and the mask values in the multiple time windows into the trained health risk prediction model to obtain health risk values.
It should be noted that the health risk prediction model may be a long-term memory network LSTM model, and the health risk value may be obtained by inputting the health risk probability value corresponding to the data volume of the time window when the long-term memory network LSTM is trained and the mask value thereof into the trained health risk prediction model.
In this embodiment, the time windows may be one year, that is, after the day to be detected is determined, the health risk probability value 12 months before the day to be detected and the mask value are input into the trained health risk prediction model, so that the health risk value of the day to be detected can be obtained.
In the application, the long-time and short-time memory network LSTM is adopted to predict the health risk value of whether the user suffers from the hypertension disease in the future, various pathogenic risk factors do not need to be defined manually, and the weight relationship among the factors is provided, so that the prediction result is more objective. In data processing, the long-time and short-time memory network LSTM is a nonlinear model, and a health risk probability value output by the softmax layer of the front-end mixed data detection model is input, so that the long-time and short-time memory network LSTM can obtain a more complex functional relation between input characteristics.
Therefore, the health management method provided by the embodiment of the application can accurately acquire the health risk value of the user by analyzing the static data and the dynamic data of the user, so that the risk of future illness suffering of the user can be predicted, and the health risk value prediction result of the user is more objective.
As a possible embodiment, before obtaining the dynamic features corresponding to the dynamic data and the static features corresponding to the static data, the method further includes: and preprocessing the dynamic data and the static data, wherein the preprocessing comprises at least one of wildcard expression of a defect field of the static data, screening of the dynamic data in a sliding window mode, conversion from minute level to hour level of the dynamic data or normalization processing of the static data and the dynamic data.
It should be noted that, for static data, as can be seen from the contents of a general questionnaire, the main defect fields are discrete variable fields, such as whether to smoke or drink, and the like, and if all fields are required to be filled completely and all data with defect fields are discarded, the data volume is greatly reduced, which seriously affects the later prediction of the health risk value of the user. Therefore, the present application employs a wildcard policy for static data defects, i.e., additional one-dimensional features are used to identify missing fields. A0 for this feature indicates that the field is not missing, and a 1 for this feature indicates that the field is missing.
For dynamic data, a user does not wear the wearable device for various reasons on certain days, so that if the user requires dynamic data recording every day, the data volume is greatly reduced, and the later prediction of the health risk value of the user is seriously influenced. Therefore, in the application, dynamic data is processed in a sliding window mode, one month is taken as a window, each sliding window is required to contain at least 5 days of records, and the month window recorded for less than 5 days does not participate in the prediction of the health risk probability value. In addition, the dynamic data is mainly minute-level data with length 1440, and since the minute-level data may include invalid data such as more noise and defect, the preprocessing may change the minute-level data into hour-level data, and specifically, count the mean and variance of the valid data per minute for 24 hours.
In addition, the method also performs normalization processing on static data and dynamic data, adopts a standard z-fraction normalization method for continuous variables to perform preprocessing, and adopts a one-hot (one-hot) encoding method for discrete variables.
Therefore, the data volume for predicting the health risk value of the user is effectively guaranteed through preprocessing, and the accuracy of the health risk value prediction is greatly improved.
As a feasible embodiment, if the health risk value is greater than a preset health risk threshold, outputting rain information and a target health risk value to the user, obtaining a target dynamic characteristic and a target static characteristic according to the target health risk value and a trained mixed data detection model, and outputting the target dynamic characteristic and the target static characteristic to the user; and if the health risk value is equal to or smaller than the health risk threshold value, outputting the change condition of the health risk value when the dynamic characteristic and the static characteristic fluctuate in a set range near the average value to the user.
Note that the risk threshold θ may be set in advance, wherein the risk threshold may be set in advance for each user.
If the predicted health risk value is larger than the set health risk threshold value, an early warning is sent to the user to inform the user that the future health of the user is possibly influenced by hypertension, namely, the user is likely to suffer from hypertension. Meanwhile, a target health risk value theta ' needs to be provided for the user, wherein theta ' < theta, the target dynamic characteristic and the target static characteristic are obtained by taking theta ' as the output of the trained mixed data detection model and performing reverse thrust, namely, the static characteristic and the dynamic characteristic which should be executed by the user to reach the target health risk value are output to the user, so that the user can adjust according to the target dynamic characteristic and the target static characteristic, and the health risk value is effectively reduced and tends to the target health risk value.
The following description will be given by taking a fully connected network FCN with an input as a two-dimensional eigenvector as an example:
Figure BDA0002917281000000091
where θ' is the target health risk value, assuming X 1 Is based on a target static characteristic and/or a target dynamic characteristic of the user, W is an inter-layer weight in the fully connected network FCN, and X 2 Then it is a parameter other than the feedback parameter. The back-stepping process is to calculate the feedback parameters that can reach the given target health risk value, such as X in the above formula, assuming the other parameters of the user are not changed 2 Deducing X according to the target health risk value under the condition of ensuring invariance 1 The size of (2), derive X 1 The process of (2) is equivalent to solving a one-dimensional equation.
For users with health risk value equal to or less than the health risk threshold value, a feedback report is provided, and the change situation of the risk of suffering diseases, especially hypertension diseases, is provided in the report when certain static characteristics or dynamic characteristics of the users fluctuate within the range of the average level plus 20% of variance, so that the users can make a plan of movement and/or rest more consistent with the physical state of the users after fully knowing the static data and the dynamic data of the users.
Therefore, different health risk threshold values can be set for different users, and the change situation of the health risk value in the fluctuation range can be predicted according to the dynamic characteristics and the static characteristics of the users, so that the health management method is more quantitative and personalized, and the personalized requirements of the users can be met.
As a specific embodiment, as shown in fig. 6, a wearable device and an App matched with the wearable device interact with a user to obtain dynamic data and static data of the user, then obtain dynamic features corresponding to the dynamic data and static features corresponding to the static data by using the dynamic data and the static data, input the dynamic features and the static features into a trained mixed data detection model provided by the present application to obtain a health risk probability value, then input the health risk probability value into a trained health risk prediction model to obtain a health risk value, judge the health risk value, if the health risk value is greater than a health risk threshold, obtain a target health risk value according to the health risk value, and perform a back-stepping to obtain the target dynamic features and the target static features by using the mixed data detection model to feed back to the user, and if the health risk value is equal to or less than the health risk threshold value, acquiring the change situation of the health risk value within the preset range and feeding back the change situation to the user.
In general, the hybrid data detection model can make full use of respective advantages of different state data, and has better recognition performance compared with single state data. The test results using different models under a certain platform data are given below, in this test experiment, 10% of all sports users are randomly selected as a test set, and the rest samples are used as a training set for training the mixed data detection model, and the test contains 2894 normal blood pressure users and 187 hypertension users in total. The recognition accuracy of the platform data respectively adopting a single type of static model and a single type of dynamic model and the hybrid data detection model provided by the application is shown in table 1:
TABLE 1
Degree of specificity Sensitivity of the device Rate of accuracy
Static model 82.31％ 82.89％ 82.34％
Dynamic model 76.50％ 56.15％ 75.27％
Hybrid data detection model 85.59％ 80.21％ 85.26％
It can be seen that the hybrid data detection model provided by the application can achieve higher identification precision compared with a single static model and a single dynamic model, namely, the detection result of the hybrid data detection model provided by the application is superior to that of the single identification model.
In order to implement the above embodiments, the present invention further provides a health management device.
Fig. 7 is a block diagram of a health management apparatus according to an embodiment of the present application. As shown in fig. 7, the health management device 10 includes: a first obtaining module 11, a second obtaining module 12, a first predicting module 13 and a second predicting module 14.
The first obtaining module 11 is configured to obtain dynamic data and static data of a user; the second obtaining module 12 is configured to obtain a dynamic feature corresponding to the dynamic data and a static feature corresponding to the static data, where the dynamic data includes at least one of a heart rate, sleep stage data, or activity amount data of a user, and the static data includes basic information of the user or includes basic information and health information of the user; the first prediction module 13 is configured to input the dynamic features and the static features into a trained mixed data detection model to obtain a health risk probability value; the second prediction module 14 is configured to input the health risk probability value into a trained health risk prediction model to obtain a health risk value.
Further, the second obtaining module 12 is further configured to: acquiring a statistic value of the dynamic data in each time window according to the dynamic data; or also for: and acquiring a first static characteristic according to the static data, wherein the first static characteristic comprises a Body Mass Index (BMI) and a body fat rate, and generating a second static characteristic based on a gradient boosting decision tree according to the static data.
Further, the hybrid data detection model includes a dynamic data detection model, a static data detection model, and a migration learning model, and the first prediction module 13 is specifically configured to: inputting the dynamic characteristics into the trained dynamic data detection model to obtain a dynamic data detection result; inputting the static characteristics into the trained static data detection model to obtain a static data detection result; inputting the dynamic data detection result and the static data detection result into the trained migration learning model to obtain the health risk probability value; and the migration learning model performs feature fusion on the dynamic data detection result and the static data detection result, and inputs the fused features into a Fully Connected Network (FCN) to obtain a health risk probability value.
Further, the second prediction module 14 is specifically configured to: determining a mask value according to the missing condition of the health risk probability value in each time window; and inputting the health risk probability values and the mask values in a plurality of time windows into the trained health risk prediction model to obtain the health risk values.
Further, the second obtaining module 12 is further configured to: and preprocessing the dynamic data and the static data, wherein the preprocessing comprises at least one of wildcard representation of the defect field of the static data, screening of the dynamic data in a sliding window mode, conversion from minute level to hour level of the dynamic data or normalization processing of the static data and the dynamic data.
Further, the second prediction module 14 is further configured to: if the health risk value is larger than a preset health risk threshold value, outputting early warning information and a target health risk value to the user, obtaining a target dynamic characteristic and a target static characteristic according to the target health risk value and the trained mixed data detection model, and outputting the target dynamic characteristic and the target static characteristic to the user; and if the health risk value is equal to or smaller than the health risk threshold value, outputting the change situation of the health risk value when the dynamic characteristic and the static characteristic fluctuate within a set range near the average value to the user.
It should be noted that the foregoing explanation of the embodiment of the health management method is also applicable to the health management apparatus of this embodiment, and is not repeated herein.
Therefore, the health management device provided by the embodiment of the application can accurately acquire the health risk value of the user by analyzing the static data and the dynamic data of the user, so that the risk of future illness suffering of the user can be predicted, and the health risk value prediction result of the user is more objective.
Based on the above embodiments, the embodiment of the present invention further provides a wearable device, as shown in fig. 8, the wearable device 100 includes a health management apparatus 10.
In order to implement the above embodiment, the present invention further provides an electronic device 200, as shown in fig. 9, including a memory 21, a processor 22; wherein, the processor 22 runs the program corresponding to the executable program code by reading the executable program code stored in the memory 21, so as to implement the health management method described in the foregoing first embodiment.
In order to implement the above embodiments, the present invention further proposes a computer-readable storage medium, which stores a computer program, which when executed by a processor implements the health management method as described in the above embodiments of the first aspect.
In the description of the specification, reference to the description of "one embodiment," "some embodiments," "an example," "a specific example," or "some examples" or the like means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or to implicitly indicate the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing steps of a custom logic function or process, and alternate implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present invention.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are well known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.
In addition, functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a separate product, may also be stored in a computer-readable storage medium.
The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc. Although embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are exemplary and not to be construed as limiting the present invention, and that changes, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.

Claims (15)

1. A health management method, comprising:
acquiring dynamic data and static data of a user, wherein the dynamic data comprises at least one of heart rate, sleep stage data or activity data of the user, and the static data comprises basic information of the user or basic information and health information of the user;
acquiring dynamic features corresponding to the dynamic data and static features corresponding to the static data;
inputting the dynamic features and the static features into a trained mixed data detection model to obtain a health risk probability value;
and inputting the health risk probability value into a trained health risk prediction model to obtain a health risk value.
2. The health management method according to claim 1, wherein the obtaining the dynamic characteristics corresponding to the dynamic data comprises: acquiring a statistic value of the dynamic data in each time window according to the dynamic data;
obtaining the static features corresponding to the static data, including: and acquiring a first static characteristic according to the static data, wherein the first static characteristic comprises a Body Mass Index (BMI) and a body fat rate, and generating a second static characteristic based on a gradient boosting decision tree according to the static data.
3. The health management approach of claim 1, wherein the mixed data detection model comprises a dynamic data detection model, a static data detection model and a migration learning model, and the inputting of the dynamic features and the static features into the trained mixed data detection model to obtain the health risk probability value comprises:
inputting the dynamic characteristics into the trained dynamic data detection model to obtain a dynamic data detection result;
inputting the static characteristics into the trained static data detection model to obtain a static data detection result;
inputting the dynamic data detection result and the static data detection result into the trained migration learning model to obtain the health risk probability value; and the migration learning model performs feature fusion on the dynamic data detection result and the static data detection result, and inputs the fused features into a Fully Connected Network (FCN) to obtain a health risk probability value.
4. The health management method of claim 1, wherein the inputting the health risk probability value into a trained health risk prediction model to obtain a health risk value comprises:
determining a mask value according to the missing condition of the health risk probability value in each time window;
and inputting the health risk probability values and the mask values in a plurality of time windows into the trained health risk prediction model to obtain the health risk values.
5. The health management approach of claim 1, wherein before obtaining the dynamic characteristics corresponding to the dynamic data and the static characteristics corresponding to the static data, the method further comprises:
and preprocessing the dynamic data and the static data, wherein the preprocessing comprises at least one of wildcard representation of the defect field of the static data, screening of the dynamic data in a sliding window mode, conversion from minute level to hour level of the dynamic data or normalization processing of the static data and the dynamic data.
6. The health management method according to claim 1, further comprising:
if the health risk value is larger than a preset health risk threshold value, outputting early warning information and a target health risk value to the user, obtaining a target dynamic characteristic and a target static characteristic according to the target health risk value and the trained mixed data detection model, and outputting the target dynamic characteristic and the target static characteristic to the user;
and if the health risk value is equal to or smaller than the health risk threshold value, outputting the change condition of the health risk value when the dynamic characteristic and the static characteristic fluctuate within a set range around the average value to the user.
7. A health management device, comprising:
the device comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring dynamic data and static data of a user, the dynamic data comprises at least one of heart rate, sleep stage data or activity amount data of the user, and the static data comprises basic information of the user or comprises the basic information and health information of the user;
the second acquisition module is used for acquiring the dynamic features corresponding to the dynamic data and the static features corresponding to the static data;
the first prediction module is used for inputting the dynamic characteristics and the static characteristics into a trained mixed data detection model to obtain a health risk probability value;
and the second prediction module is used for inputting the health risk probability value into a trained health risk prediction model to obtain a health risk value.
8. The health management apparatus of claim 7, wherein the second obtaining module is further configured to: obtaining a statistic value of the dynamic data in each time window according to the dynamic data; or
And is further configured to: and acquiring a first static characteristic according to the static data, wherein the first static characteristic comprises a Body Mass Index (BMI) and a body fat rate, and generating a second static characteristic based on a gradient boosting decision tree according to the static data.
9. The health management device of claim 7, wherein the hybrid data detection model comprises a dynamic data detection model, a static data detection model, and a transfer learning model, and the first prediction module is specifically configured to:
inputting the dynamic characteristics into the trained dynamic data detection model to obtain a dynamic data detection result;
inputting the static characteristics into the trained static data detection model to obtain a static data detection result;
inputting the dynamic data detection result and the static data detection result into the trained migration learning model to obtain the health risk probability value; and the transfer learning model performs feature fusion on the dynamic data detection result and the static data detection result, and inputs the fused features into a full-connection network FCN to obtain a health risk probability value.
10. The health management device of claim 7, wherein the second prediction module is specifically configured to:
determining a mask value according to the missing condition of the health risk probability value in each time window;
and inputting the health risk probability values and the mask values in a plurality of time windows into the trained health risk prediction model to obtain the health risk values.
11. The health management device of claim 7, wherein the second obtaining module is further configured to:
and preprocessing the dynamic data and the static data, wherein the preprocessing comprises at least one of wildcard representation of the defect field of the static data, screening of the dynamic data in a sliding window mode, conversion from minute level to hour level of the dynamic data or normalization processing of the static data and the dynamic data.
12. The health management device of claim 7, wherein the second prediction module is further configured to:
if the health risk value is larger than a preset health risk threshold value, outputting early warning information and a target health risk value to the user, obtaining a target dynamic characteristic and a target static characteristic according to the target health risk value and the trained mixed data detection model, and outputting the target dynamic characteristic and the target static characteristic to the user;
and if the health risk value is equal to or smaller than the health risk threshold value, outputting the change condition of the health risk value when the dynamic characteristic and the static characteristic fluctuate within a set range around the average value to the user.
13. A wearable device, characterized in that it comprises a health management apparatus according to any of claims 7-12.
14. An electronic device comprising a memory, a processor;
wherein the processor executes a program corresponding to the executable program code by reading the executable program code stored in the memory for implementing the health management method according to any one of claims 1 to 6.
15. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the health management method according to any one of claims 1-6.
CN202110114648.1A 2021-01-26 2021-01-26 Health management method and device, wearable device, electronic device and medium Pending CN114792565A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110114648.1A CN114792565A (en) 2021-01-26 2021-01-26 Health management method and device, wearable device, electronic device and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110114648.1A CN114792565A (en) 2021-01-26 2021-01-26 Health management method and device, wearable device, electronic device and medium

Publications (1)

Publication Number Publication Date
CN114792565A true CN114792565A (en) 2022-07-26

Family

ID=82459342

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110114648.1A Pending CN114792565A (en) 2021-01-26 2021-01-26 Health management method and device, wearable device, electronic device and medium

Country Status (1)

Country Link
CN (1) CN114792565A (en)

Similar Documents

Publication Publication Date Title
Oh et al. Identifying schizophrenia using structural MRI with a deep learning algorithm
Zhang et al. Automated sleep stage scoring of the Sleep Heart Health Study using deep neural networks
US20210145306A1 (en) Managing respiratory conditions based on sounds of the respiratory system
US20180107798A1 (en) Method for aiding a diagnosis, program and apparatus
CN107785057B (en) Medical data processing method, device, storage medium and computer equipment
US20110077931A1 (en) Methods and apparatus for identifying disease status using biomarkers
CN110570941B (en) System and device for assessing psychological state based on text semantic vector model
CN107430645B (en) System for laboratory value automated analysis and risk notification in intensive care units
US20180211727A1 (en) Automated Evidence Based Identification of Medical Conditions and Evaluation of Health and Financial Benefits Of Health Management Intervention Programs
US20210296001A1 (en) Dementia risk presentation system and method
KR102342770B1 (en) A health management counseling system using the distribution of predicted disease values
CN112967803A (en) Early mortality prediction method and system for emergency patients based on integrated model
CN114943629A (en) Health management and health care service system and health management method thereof
CN112598090B (en) Method, device, equipment and system for health portrait
Balnarsaiah et al. Parkinson’s disease detection using modified ResNeXt deep learning model from brain MRI images
CN117672445A (en) Diabetes mellitus debilitation current situation analysis method and system based on big data
US20220101176A1 (en) System and method for generating a direction inquiry response from biological extractions using machine learning
US20160110516A1 (en) Dynamic monitoring early cancer risk
KR20050050768A (en) Apparatus and method for offering tailored health promotion service for personal
US20230060794A1 (en) Diagnostic Tool
CN114792565A (en) Health management method and device, wearable device, electronic device and medium
Noori et al. A comparative analysis for diabetic prediction based on machine learning techniques
CN114388089A (en) Personal health management method and system based on artificial intelligence
US11887730B2 (en) Systems and methods for unobtrusive digital health assessment
JP2023545704A (en) Systems and methods for exposome clinical applications

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination