Health assessment methods and systems

Abstract

The present application relates to the technical field of health assessment, and provides a health assessment method and an assessment system. The method includes: obtaining the basic health data of the target user, including the user's personal information and current health characteristics; configuring a preset monitoring data source set, and training a two-way health analysis model; collecting the health monitoring data of the target user according to the preset monitoring data source set; inputting the current health characteristics into the model to perform an inverse operation of generating the target data distribution, and generating predicted associated abnormal data; setting the predicted associated abnormal data as a control variable, and inputting it into the model together with the health monitoring data to perform a forward generation operation, and generating the target health assessment result of the target user. The present application solves the technical problem of inaccurate analysis results due to the instability of some health indicators when conducting a health assessment of a user, and realizes the effect of improving the accuracy of the health assessment results by predicting associated abnormal data and performing a forward generation operation through a two-way health analysis model.

Description

Technical Field

[0001] The present application relates to the field of health assessment, and specifically to health assessment methods and systems. Background Art

[0002] With rising health awareness, more and more people are relying on health monitoring devices and applications to track and manage their health. These devices can collect multidimensional health data from users in real time, such as heart rate, blood pressure, and body temperature. Through data analysis, they generate health assessment reports, helping users understand their physical condition and make appropriate adjustments. However, existing health assessment methods still have many shortcomings in practical applications. First, when a user is diagnosed with a disease, some health indicators may fluctuate significantly due to the disease, resulting in inaccurate assessment results. For example, symptoms such as difficulty sleeping and slow reaction time can affect the stability of multiple health data sets, ultimately distorting the analysis results. Second, existing assessment methods often rely on universal health standards, fail to fully consider individual differences, and lack personalized health assessment solutions. Furthermore, existing health assessments often ignore the complex relationships between multidimensional data and fail to effectively address potential anomalies in health data, resulting in reduced reliability of analysis results. Therefore, there is an urgent need for health management technologies that can handle data fluctuations, correlate anomalies, and provide personalized assessments to better reflect users' actual health status and provide effective health management solutions. Summary of the Invention

[0003] This application provides a health assessment method and system to solve the technical problem of inaccurate analysis results due to the instability of some health indicators when conducting user health assessments.

[0004] In view of the above problems, this application provides a health assessment method and system.

[0005] The first aspect disclosed in the present application provides a health assessment method, which includes: obtaining basic health data of a target user, wherein the basic health data includes user personal information and current health characteristics; configuring a preset monitoring data source set and training a two-way health analysis model; collecting health monitoring data of the target user according to the preset monitoring data source set; inputting the current health characteristics into the two-way health analysis model to perform an inverse operation of generating a target data distribution, thereby generating predicted associated abnormal data; setting the predicted associated abnormal data as a control variable, and inputting it into the two-way health analysis model together with the health monitoring data for a forward generation operation, thereby generating a target health assessment result for the target user.

[0006] Another aspect disclosed in the present application provides a health assessment system, which includes: a basic health data acquisition module, wherein the basic health data acquisition module is used to acquire the basic health data of the target user, wherein the basic health data includes user personal information and current health characteristics; a two-way health analysis model training module, wherein the basic health data acquisition module is used to configure a preset monitoring data source set and train a two-way health analysis model; a health monitoring data acquisition module, wherein the basic health data acquisition module is used to collect the health monitoring data of the target user according to the preset monitoring data source set; a predicted associated abnormal data generation module, wherein the basic health data acquisition module is used to input the current health characteristics into the two-way health analysis model to perform an inverse operation of generating a target data distribution, thereby generating predicted associated abnormal data; a health assessment result generation module, wherein the basic health data acquisition module is used to set the predicted associated abnormal data as a control variable, and input it into the two-way health analysis model together with the health monitoring data to perform a forward generation operation, thereby generating a target health assessment result for the target user.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The aforementioned health assessment method obtains basic health data of a target user, including their personal information and current health characteristics. This data provides the foundation for subsequent health assessments. Subsequently, a pre-set set of monitoring data sources is configured to collect the user's health monitoring data and train a bidirectional health analysis model. After model training is complete, the user's current health characteristics are input into the bidirectional health analysis model based on the collected health monitoring data, and a reverse operation is performed. This reverse operation generates predicted, associated abnormal data related to the user's health characteristics, revealing potential health risks or abnormalities. By generating this abnormal data, unstable factors in the user's health can be effectively predicted. Subsequently, this predicted abnormal data is used as control variables and, combined with the actual collected health monitoring data, is re-input into the bidirectional health analysis model for a forward generation operation. This forward generation operation aims to generate the final health assessment result for the target user based on the complete data. This bidirectional model training and operation enables a more accurate analysis of the user's health status. Especially when the user's health data fluctuates or is abnormal, it can still provide an accurate and personalized health assessment, thereby helping the user prevent potential health risks.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1 Schematic diagram of a health assessment method in one embodiment.

[0012] Figure 2 This is an architecture diagram of a health assessment system in one embodiment.

[0013] Explanation of the accompanying symbols: basic health data acquisition module 1, two-way health analysis model training module 2, health monitoring data collection module 3, prediction and associated abnormal data generation module 4, health assessment result generation module 5. DETAILED DESCRIPTION

[0014] The embodiments of the present application provide a health assessment method and system to solve the technical problem of inaccurate analysis results caused by the instability of some health indicators when performing a health assessment on a user.

[0015] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0016] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0017] Example 1, as Figure 1 As shown, the present application provides a health assessment method, the method comprising:

[0018] Obtain the target user's basic health data, wherein the basic health data includes the user's personal information and current health characteristics.

[0019] In an embodiment of the present application, during the health assessment process, the target user's basic health data is first obtained. This data includes the user's personal information and current health characteristics. The user's personal information includes basic data such as age, gender, height, and weight. This information can be manually entered by the user or automatically collected by a smart device. This basic information provides important background support for subsequent health analysis, helping the system terminal to perform more accurate health assessments based on individual differences. In addition, the user's current health characteristics are collected, especially information on confirmed diseases. This confirmed information may involve chronic diseases, acute diseases, mental health conditions, etc., and has a significant impact on the volatility of health data. For example, a confirmed disease may cause abnormal fluctuations in certain health indicators. By obtaining this characteristic information, potential health risks can be effectively identified and these abnormal data can be processed in the subsequent assessment. By combining the user's basic health data and confirmed disease information, fluctuations and abnormalities related to health characteristics can be better processed during the assessment process, thereby improving the accuracy and personalization of the assessment. This process ensures that users can be provided with more personalized and accurate health assessment results, helping them better manage their health.

[0020] Configure a preset set of monitoring data sources and train a bidirectional health analysis model.

[0021] In one embodiment, the health assessment process first requires configuring a set of pre-set monitoring data sources to collect the target user's health data and retrieve historical health data from historical logs. This data source set includes multiple health monitoring devices and sensors, such as heart rate monitors, blood pressure monitors, and blood glucose monitors, which can obtain the user's physical health indicators in real time. By integrating this multi-dimensional health monitoring data, comprehensive basic information is provided for subsequent analysis. After configuring the monitoring data sources, a large amount of historical health data is obtained from historical logs based on these monitoring data sources to train a bidirectional health analysis model. The core function of this model is that it not only performs forward analysis based on the input health data to generate the user's current health status, but also has the ability to perform reverse inference. Specifically, during the forward analysis process, the user's overall health status is assessed by inputting the user's health data (such as heart rate, blood pressure, blood glucose, etc.), and a comprehensive health assessment result is generated by combining the correlations between the multi-dimensional data. During the reverse inference process, the model can reversely infer the input data that may have caused the user's health status based on the user's known health status, such as confirmed diseases or certain health characteristics, and thus predict potential health abnormalities. Through repeated training and optimization, the bidirectional health analysis model gradually improves its accuracy in processing health data, ensuring that even when faced with complex health data, it can still accurately identify abnormal data and make reasonable assessments. The model's bidirectional capabilities give system terminals greater flexibility and accuracy in health monitoring and anomaly prediction, providing users with more personalized and comprehensive health assessment results.

[0022] Furthermore, the present application provides for configuring a preset set of monitoring data sources and training a bidirectional health analysis model, including:

[0023] According to the preset monitoring data source set, multiple groups of historical monitoring data and corresponding multiple groups of health status identification information are collected; a generative adversarial network is initialized, wherein the generative adversarial network includes a first subnetwork and a second subnetwork; based on the multiple groups of historical monitoring data and the multiple groups of health status identification information, the first subnetwork and the second subnetwork are subjected to generative adversarial training using a cycle consistency mechanism to generate the bidirectional health analysis model that meets the preset convergence conditions.

[0024] Preferably, during the health assessment process, multiple sets of historical monitoring data and corresponding health status identification information are first collected from historical logs based on a preset set of monitoring data sources. The historical monitoring data includes the user's multi-dimensional health indicators, such as heart rate, blood pressure, and body temperature, while the health status identification information reflects the user's health status. This data serves as the basis for model training. Subsequently, a generative adversarial network (GAN) is initialized, which consists of two main subnetworks: a first subnetwork and a second subnetwork. Each subnetwork contains a generator and a discriminator. The generator of the first subnetwork learns how to generate health status based on the input monitoring data, while its discriminator is responsible for determining whether the generated health status is authentic and valid. Conversely, the generator of the second subnetwork learns how to reversely generate possible corresponding monitoring data based on the known health status, while its discriminator determines the rationality of this generated monitoring data. Through this two-way generative adversarial mechanism, the model can not only generate health status in a forward direction, but also generate potential abnormal monitoring data through reverse deduction. In particular, for certain specific health issues, such as blood sugar instability in diabetic patients, possible blood sugar fluctuation data can be reversely deduced based on the known health status. This flexibility enables the system terminal to selectively use or derive relevant health data based on the needs and specific health conditions of different users. During the generative adversarial training process, a cycle consistency mechanism is used to ensure the convergence of the model. This mechanism ensures that the conversion between the first subnetwork and the second subnetwork is consistent, that is, the health status generated by the first subnetwork and the data reversely generated by the second subnetwork can be restored to the original monitoring data, and vice versa. Through such iterative training, the model can continuously optimize the generation process based on historical monitoring data and health status identification information, and finally generate a bidirectional health analysis model that meets the preset convergence conditions. This model can not only generate health status based on real-time monitoring data, but also reversely deduce possible monitoring data based on the health status, thereby providing more comprehensive and accurate support for health assessment.

[0025] Furthermore, the present application provides a method for performing generative adversarial training on the first sub-network and the second sub-network based on the multiple sets of historical monitoring data and the multiple sets of health status identification information using a cycle consistency mechanism to generate the bidirectional health analysis model that meets the preset convergence conditions, including:

[0026] A first mapping domain is established with the multiple groups of historical monitoring data, and a second mapping domain is established with the multiple groups of health status identification information; forward generative adversarial training is performed on the first sub-network according to the first mapping domain and the second mapping domain, and reverse generative adversarial training is performed on the second sub-network according to the first mapping domain and the second mapping domain, and iterative training is performed through the cycle consistency mechanism to generate the bidirectional health analysis model; wherein, the forward generative adversarial training converts and maps the data in the first mapping domain into the data in the second mapping domain through a generator, and the reverse generative adversarial training converts and maps the data in the second mapping domain into the data in the first mapping domain through a generator.

[0027] Optionally, during the training process of the bidirectional health analysis model, two mapping domains are first established based on multiple sets of historical monitoring data and corresponding health status identification information. The first mapping domain contains multiple sets of historical health monitoring data (such as heart rate, blood pressure, body temperature, etc.), and the second mapping domain contains corresponding health status identification information (such as health characteristics, etc.). Subsequently, the first sub-network and the second sub-network in the initialized generative adversarial network are trained in a bidirectional generative adversarial manner based on the established first mapping domain and the second mapping domain. In forward training, the monitoring data in the first mapping domain is input into the generator of the first sub-network. The task of the generator is to map these monitoring data into the health status in the second mapping domain. The generator attempts to generate realistic health status data, and the discriminator is responsible for judging whether the generated health status is consistent with the real health status identification. During the training process, the generator and the discriminator continue to compete, and the generator continuously optimizes by adjusting its parameters so that the health status it generates is closer and closer to the real data. For example, given historical heart rate and blood pressure data, the generator generates corresponding heart rhythm and blood pressure profiles. The discriminator evaluates the authenticity of these scores and provides feedback, prompting the generator to further adjust its generation strategy in the next training cycle. Reverse generative training involves inputting health status identifiers from the second mapping domain into the generator of the second subnetwork, which then reverse-maps the data into the corresponding monitoring data from the first mapping domain. During this process, the discriminator determines whether the generated data matches the actual monitoring data. For example, based on the diabetic status, health monitoring data related to the disease (such as blood sugar fluctuations) can be derived. Through repeated training of generation and discrimination, the model learns to better derive monitoring data from health status, making the generated monitoring data more consistent with actual health characteristics. To ensure consistency in the conversion process between forward and reverse generation, a cycle consistency mechanism is introduced. This mechanism requires that the health status obtained through forward generation can be reconstructed into the original monitoring data through reverse generation. Conversely, monitoring data generated through reverse generation should also be able to reconstruct the original health status through forward generation. This is achieved through the cycle consistency loss function within the cycle consistency mechanism. The purpose of this loss function is to calculate the difference between the generated data and the original data, using this as a feedback signal during the training process. By minimizing this loss function, the consistency of data conversion during the forward and backward generation processes can be ensured, ensuring that the generated data remains consistent with the original input data. During the training process, multiple iterations of historical monitoring data and health status information are input into the model. Through forward and backward generative adversarial training and the constraints of the cycle consistency mechanism, the model gradually converges, meaning that the loss value meets the preset requirements. Convergence means that the generator can produce realistic health status or monitoring data, and the discriminator can accurately distinguish between real and fake data. The resulting bidirectional health analysis model can achieve high-precision health status prediction and data generation.Through this training process, the model can not only generate the user's health status based on monitoring data, but also reversely deduce possible monitoring data based on the health status, achieving a comprehensive and accurate health assessment.

[0028] Furthermore, this application provides a cycle consistency mechanism, including:

[0029] The cycle consistency mechanism includes a cycle consistency loss function, and the expression of the cycle consistency loss function is: L(G, F) = ||F(G(x))-x||+||G(F(y))-y||; wherein, L(G, F) is the cycle consistency loss function; x is the data in the first mapping domain; G(x) is the data of the second mapping domain generated by the first sub-network according to the first mapping domain data x; F(G(x)) is the data converted back to the first mapping domain generated by the second sub-network according to the data G(x); y is the data in the second mapping domain; F(y) is the data of the first mapping domain generated by the second sub-network according to the second mapping domain data y; G(F(y)) is the data converted back to the second mapping domain generated by the first sub-network according to the data F(y).

[0030] Optionally, during training, a cycle consistency mechanism is used to ensure consistency between the forward generation from monitoring data to healthy states, and the reverse generation from healthy states to monitoring data. The core of this mechanism lies in the internal cycle consistency loss function. The cycle consistency loss function is as follows: L(G, F) = ||F(G(x))-x|| + ||G(F(y))-y||; where L(G, F) is the cycle consistency loss function, which is used to ensure that the data remains consistent with the original data after being converted between different mapping domains through the transformation of the forward generator and the reverse generator. x is the data in the first mapping domain, which is the user's health monitoring data; G(x) is the data in the second mapping domain generated by the first subnetwork based on the first mapping domain data x; F(G(x)) is the data converted back to the first mapping domain generated by the second subnetwork based on the data G(x); y is the data in the second mapping domain, which is the user's health status identification information; F(y) is the data in the first mapping domain generated by the second subnetwork based on the second mapping domain data y; G(F(y)) is the data converted back to the second mapping domain generated by the first subnetwork based on the data F(y).

[0031] The current health features are input into the bidirectional health analysis model to perform an inverse operation of generating target data distribution, thereby generating predicted associated abnormal data.

[0032] In one embodiment, in the health assessment, the current health characteristics are input into the bidirectional health analysis model for processing, and the reverse operation is performed through the model. During this reverse operation, the trained model is used to reversely deduce the input data. The purpose of the reverse operation is to reversely generate the distribution of the corresponding health monitoring data from the current health status and identify possible abnormal points. For example, if the current health characteristics show that the user has high blood pressure, the model will reversely generate monitoring data that may be associated with high blood pressure, such as heart rate, blood pressure, etc., and use this to identify data with abnormal fluctuations as predicted associated abnormal data. These predicted associated abnormal data can reveal that the user's current health characteristics may lead to abnormal monitoring data, which serves as an important reference for further analysis of the user's health status.

[0033] Furthermore, the present application provides a method of inputting the current health features into the bidirectional health analysis model to perform an inverse operation of generating a target data distribution, thereby generating predicted associated abnormal data, including:

[0034] The second sub-network in the bidirectional health analysis model is called to perform an inverse operation of generating target data distribution on the current health feature to generate the predicted associated abnormal data.

[0035] Optionally, in the bidirectional health analysis model, the second subnetwork is a subnetwork used to reversely generate data. When the user's current health characteristics are input into the model, the internal second subnetwork is called to perform the inverse operation. The core of the inverse operation is to reversely infer the possible monitoring data distribution from the health status based on the current health characteristics. Specifically, the second subnetwork will try to generate possible data related to these health characteristics from the user's current health status information through the learned mapping ability to obtain a target data distribution. Subsequently, data with abnormal identification is extracted from the target data distribution, and these data are summarized to output predicted associated abnormal data, providing a variable basis for subsequent user health status analysis.

[0036] Furthermore, this application provides a two-way health analysis model, including:

[0037] The bidirectional health analysis model also includes generating an optimization reminder network, which is used to perform prediction accuracy and information supplement reminders based on the number of data sources contained in the health monitoring data. The steps of constructing the optimization reminder network include: recording the types and quantities of data sources contained in the multiple sets of historical monitoring data, generating missing data source types, and configuring corresponding health monitoring accuracy; training the optimization reminder network with the missing data source types and corresponding health monitoring accuracy, wherein the optimization reminder network includes triggering health status features.

[0038] Optionally, the bidirectional health analysis model also includes a generation optimization reminder network, which is primarily used to provide prediction accuracy and information supplementation reminders when data sources are incomplete or monitoring data is insufficient. First, multiple sets of the user's historical monitoring data are recorded, including the types and number of data sources involved in each set. For example, a user's historical health data may include multiple types such as heart rate and blood pressure, and each type of data source may have different monitoring frequencies in different time periods. The type and quantity of each set of monitoring data are categorized and stored to ensure a comprehensive understanding of the historical data. This record of data types provides a reference basis for subsequent analysis. When analyzing the user's historical data, missing data source types are identified. If a user does not collect certain key health data sources within a certain period of time (for example, a user's blood pressure data is missing), these missing data sources are marked. For each missing data source, a corresponding health monitoring accuracy is assigned, which is configured by professional personnel. Professional personnel will assess the impact of the missing data source on the prediction results based on the specific application and needs of the health data and provide corresponding accuracy adjustments to the system terminal. For example, if a user only provides heart rate data but lacks blood pressure data, the overall prediction accuracy will be set to a lower level. Health monitoring accuracy quantifies the reliability of prediction results when different data sources are missing. The accuracy of the user's health assessment is adjusted based on the completeness of historical data. After generating and identifying the types of missing data sources, these missing data sources and their corresponding health monitoring accuracy are used as input to train the optimization reminder network. The purpose of the optimization reminder network is to remind users to supplement their data when data is insufficient, thereby improving the accuracy of prediction results. During training, through steps such as forward propagation, loss calculation, backpropagation, and parameter optimization, the network learns the impact of different data sources on prediction accuracy based on the accuracy of historical data and configuration. Furthermore, the optimization reminder network is configured with a triggering health status feature, which is a key function of the optimization reminder network. If a user currently provides only a small amount of monitoring data (for example, only heart rate data and no other relevant data), a comprehensive analysis is performed based on the missing historical data sources and the current health status. Specifically, if the limited data currently available indicates health anomalies (such as an elevated heart rate), the health status feature is triggered, and the user is notified of the health anomaly and a data supplementation reminder, prompting the user to supplement the missing health data sources (such as blood pressure) to ensure the accuracy of the health assessment. In this way, the optimized reminder network can not only provide reminders when data is insufficient, but also help users conduct more comprehensive health monitoring when potential health risks arise.

[0039] The predicted associated abnormal data is set as a control variable, and is input into the bidirectional health analysis model together with the health monitoring data to perform a forward generation operation to generate a target health assessment result of the target user.

[0040] In one embodiment, the health assessment uses the predicted associated abnormal data as a control variable, filtering out standard health data corresponding to the predicted associated abnormal data from a pre-set standard health database. This filtered standard health data is then combined with the user's health monitoring data and fed into a bidirectional health analysis model for forward generation. This generates a health assessment result for the target user, helping them better manage their health and prevent potential problems.

[0041] Furthermore, the present application provides a method of setting the predicted associated abnormal data as a control variable and inputting the data together with the health monitoring data into the bidirectional health analysis model for a forward generation operation to generate a target health assessment result of the target user, including:

[0042] Establish a preset standard health database; based on the preset standard health database, use the predicted associated abnormal data as a control variable to perform an abnormality elimination operation to generate abnormality elimination data; use the abnormality elimination data to replace the health monitoring data with homologous data and then input the data into the bidirectional health analysis model for a forward generation operation to generate the target health assessment result.

[0043] Preferably, a preset standard health database is first established. This database contains normal health monitoring data and corresponding health statuses, serving as a standard reference for analyzing the user's health status. This standard data is derived from health data statistics of different populations and covers a variety of health characteristics, such as normal heart rate ranges, blood pressure fluctuations, and blood sugar levels. This database provides a health benchmark for the system terminal, which is used for comparison and anomaly elimination during subsequent evaluations. After obtaining the user's health monitoring data and generating predicted associated abnormal data, this abnormal data reflects the user's current health characteristics and possible future abnormalities, such as abnormal blood pressure or heart rate fluctuations. These predicted associated abnormal data are used as control variables and compared with the preset standard health database to perform anomaly elimination. The purpose of anomaly elimination is to use the control variables to filter out standard health data of the same type as the predicted associated abnormal data from the preset standard health database, thereby eliminating abnormal portions of the data. This allows the system terminal to remove abnormal data caused by the user's current health characteristics. After the anomaly elimination operation is completed, anomaly elimination data is generated, i.e., standard health data of the same type as the predicted associated abnormal data. Subsequently, the anomaly elimination data is used to replace the abnormal data in the health monitoring data, ensuring that the user's health data is more reliable in subsequent analysis. This health monitoring data, after replacing the original data, is then fed into the bidirectional health analysis model for a forward generation operation. By combining this health data with the removed anomalies, the model can generate more accurate health assessment results. The forward generation operation uses the corrected health data to generate a health assessment result that aligns with the user's current status, known as the target health assessment result. This result considers the impact of current health status on certain monitoring data, ensuring a more accurate and personalized assessment result, helping users better understand their health status and providing personalized health management recommendations.

[0044] In summary, the embodiments of the present application have at least the following technical effects:

[0045] The embodiment of the present application obtains the basic health data of the target user and configures a preset set of monitoring data sources to train a two-way health analysis model. After collecting the user's health monitoring data, the model is used to perform inverse operations to generate predicted associated abnormal data, and then these abnormal data are used as control variables and input into the model together with the health monitoring data to perform forward generation operations, and finally generate the health assessment results of the target user. The training of the model is based on multiple sets of historical data and health status identification information, and forward and reverse generation operations are performed through a generative adversarial network, and a cycle consistency mechanism is used for optimization. In addition, it also includes an optimized reminder network that can provide prediction accuracy reminders and data supplementation suggestions based on the integrity of the health monitoring data source. These technical effects jointly solve the technical problem of inaccurate analysis results due to the instability of some health indicators when conducting user health assessments, and achieve the effect of improving the accuracy of health assessment results by predicting associated abnormal data and performing forward generation operations through a two-way health analysis model.

[0046] Example 2, based on the same inventive concept as the health assessment method in the above example, Figure 2 As shown, the present application provides a health assessment system, which includes: a basic health data acquisition module 1, wherein the basic health data acquisition module 1 is used to acquire the basic health data of the target user, wherein the basic health data includes user personal information and current health characteristics; a two-way health analysis model training module 2, wherein the basic health data acquisition module 2 is used to configure a preset monitoring data source set and train a two-way health analysis model; a health monitoring data acquisition module 3, wherein the basic health data acquisition module 3 is used to collect the health monitoring data of the target user according to the preset monitoring data source set; a predicted associated abnormal data generation module 4, wherein the basic health data acquisition module 4 is used to input the current health characteristics into the two-way health analysis model to perform an inverse operation of generating a target data distribution, and generate predicted associated abnormal data; a health assessment result generation module 5, wherein the basic health data acquisition module 5 is used to set the predicted associated abnormal data as a control variable, and input it into the two-way health analysis model together with the health monitoring data to perform a forward generation operation to generate a target health assessment result for the target user.

[0047] Furthermore, the basic health data acquisition module 2 is also used to perform the following method:

[0048] According to the preset monitoring data source set, multiple groups of historical monitoring data and corresponding multiple groups of health status identification information are collected; a generative adversarial network is initialized, wherein the generative adversarial network includes a first subnetwork and a second subnetwork; based on the multiple groups of historical monitoring data and the multiple groups of health status identification information, the first subnetwork and the second subnetwork are subjected to generative adversarial training using a cycle consistency mechanism to generate the bidirectional health analysis model that meets the preset convergence conditions.

[0049] Furthermore, the basic health data acquisition module 2 is also used to perform the following method:

[0050] A first mapping domain is established with the multiple groups of historical monitoring data, and a second mapping domain is established with the multiple groups of health status identification information; forward generative adversarial training is performed on the first sub-network according to the first mapping domain and the second mapping domain, and reverse generative adversarial training is performed on the second sub-network according to the first mapping domain and the second mapping domain, and iterative training is performed through the cycle consistency mechanism to generate the bidirectional health analysis model; wherein, the forward generative adversarial training converts and maps the data in the first mapping domain into the data in the second mapping domain through a generator, and the reverse generative adversarial training converts and maps the data in the second mapping domain into the data in the first mapping domain through a generator.

[0051] Furthermore, the basic health data acquisition module 2 is also used to perform the following method:

[0052] The cycle consistency mechanism includes a cycle consistency loss function, and the expression of the cycle consistency loss function is: L(G, F) = ||F(G(x))-x||+||G(F(y))-y||; wherein, L(G, F) is the cycle consistency loss function; x is the data in the first mapping domain; G(y) is the data in the second mapping domain generated by the first sub-network according to the first mapping domain data x; F(G(X)) is the data converted back to the first mapping domain generated by the second sub-network according to the data G(x); y is the data in the second mapping domain; F(y) is the data in the first mapping domain generated by the second sub-network according to the second mapping domain data y; G(F(y)) is the data converted back to the second mapping domain generated by the first sub-network according to the data F(y).

[0053] Furthermore, the basic health data acquisition module 4 is also used to perform the following method:

[0054] The second sub-network in the bidirectional health analysis model is called to perform an inverse operation of generating target data distribution on the current health feature to generate the predicted associated abnormal data.

[0055] Furthermore, the basic health data acquisition module 4 is also used to perform the following method:

[0056] Record the data source types and quantities included in the multiple sets of historical monitoring data, generate missing data source types, and configure corresponding health monitoring accuracy; train the optimization reminder network with the missing data source types and corresponding health monitoring accuracy, wherein the optimization reminder network includes triggering health status features.

[0057] Furthermore, the basic health data acquisition module 5 is also used to perform the following method:

[0058] Establish a preset standard health database; based on the preset standard health database, use the predicted associated abnormal data as a control variable to perform an abnormality elimination operation to generate abnormality elimination data; use the abnormality elimination data to replace the health monitoring data with homologous data and then input the data into the bidirectional health analysis model for a forward generation operation to generate the target health assessment result.

[0059] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0060] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0061] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A health assessment method, characterized in that: include: Obtaining basic health data of the target user, wherein the basic health data includes user personal information and current health characteristics; Configure a preset set of monitoring data sources and train a bidirectional health analysis model; Collecting the health monitoring data of the target user according to the preset monitoring data source set; Inputting the current health features into the bidirectional health analysis model to perform an inverse operation of generating target data distribution, thereby generating predicted associated abnormal data; The predicted associated abnormal data is set as a control variable, and inputted into the bidirectional health analysis model together with the health monitoring data to perform a forward generation operation to generate a target health assessment result of the target user; The configuration of the preset monitoring data source set and the training of the two-way health analysis model include: Collect multiple sets of historical monitoring data and corresponding multiple sets of health status identification information according to the preset monitoring data source set; Initializing a generative adversarial network, wherein the generative adversarial network includes a first sub-network and a second sub-network; Based on the multiple sets of historical monitoring data and the multiple sets of health status identification information, a cycle consistency mechanism is used to perform generative adversarial training on the first sub-network and the second sub-network to generate the bidirectional health analysis model that meets the preset convergence condition; According to the multiple sets of historical monitoring data and the multiple sets of health status identification information, a cycle consistency mechanism is used to perform generative adversarial training on the first sub-network and the second sub-network to generate the bidirectional health analysis model that meets the preset convergence conditions, including: Establishing a first mapping domain with the multiple sets of historical monitoring data, and establishing a second mapping domain with the multiple sets of health status identification information; Performing forward generative adversarial training on the first sub-network according to the first mapping domain and the second mapping domain, performing reverse generative adversarial training on the second sub-network according to the first mapping domain and the second mapping domain, and iteratively training through the cycle consistency mechanism to generate the bidirectional health analysis model; The forward generative adversarial training converts the data in the first mapping domain into the data in the second mapping domain through a generator, and the reverse generative adversarial training converts the data in the second mapping domain into the data in the first mapping domain through a generator.

2. The health assessment method according to claim 1, wherein: The cycle consistency mechanism includes a cycle consistency loss function, and the expression of the cycle consistency loss function is: ； in, is the cycle consistency loss function; x is the data in the first mapping domain; The data of the second mapping domain is generated by the first sub-network according to the first mapping domain data x; For the second sub-network according to the data The generated data is converted back to the first mapping domain; y is the data in the second mapping domain; The data of the first mapping domain is generated by the second sub-network according to the second mapping domain data y; For the first sub-network according to the data The generated data is converted back to the second mapping domain.

3. The health assessment method according to claim 2, wherein: Inputting the current health features into the bidirectional health analysis model to perform an inverse operation of generating target data distribution to generate predicted associated abnormal data, including: The second sub-network in the bidirectional health analysis model is called to perform an inverse operation of generating target data distribution on the current health feature to generate the predicted associated abnormal data.

4. The health assessment method according to claim 1, wherein: The predicted associated abnormal data is set as a control variable, and inputted into the bidirectional health analysis model together with the health monitoring data to perform a forward generation operation to generate a target health assessment result of the target user, including: Establish a preset standard health database; Based on the preset standard health database, using the predicted associated abnormal data as a control variable, performing an abnormality elimination operation to generate abnormality elimination data; The health monitoring data is replaced with homologous data using the abnormality elimination data, and then input into the bidirectional health analysis model for a forward generation operation to generate the target health assessment result.

5. The health assessment method according to claim 3, wherein: The bidirectional health analysis model further includes generating an optimization reminder network, which is used to perform prediction accuracy and information supplement reminders based on the number of data sources included in the health monitoring data. The steps of constructing the optimization reminder network include: Record the data source types and quantities included in the multiple sets of historical monitoring data, generate missing data source types, and configure corresponding health monitoring accuracy; The optimized reminder network is trained with the missing data source type and the corresponding health monitoring accuracy, wherein the optimized reminder network includes triggering health status features.

6. A health assessment system, characterized in that: The steps for implementing the health assessment method according to any one of claims 1 to 5 include: Basic health data acquisition module: obtains the basic health data of the target user, wherein the basic health data includes the user's personal information and current health characteristics; Bidirectional health analysis model training module: configures a preset set of monitoring data sources and trains a bidirectional health analysis model; Health monitoring data collection module: collects the health monitoring data of the target user according to the preset monitoring data source set; Prediction-related abnormal data generation module: inputs the current health features into the bidirectional health analysis model to perform an inverse operation of generating target data distribution, thereby generating prediction-related abnormal data; Health assessment result generation module: sets the predicted associated abnormal data as a control variable, inputs it together with the health monitoring data into the two-way health analysis model for a forward generation operation, and generates the target health assessment result of the target user.

Images