Intelligent intervention management recommendation method and device based on health portrait

Abstract

The invention discloses an intelligent intervention management recommendation method and device based on a health portrait, and belongs to the technical field of health management. And an intelligent and scientific health management auxiliary tool is provided for primary doctors, so that the problem of insufficient primary medical resources is solved. The health state of an individual is dynamically reflected by using a health portrait, personalized recommendation is generated in combination with an algorithm, and a one-step intervention scheme is avoided. By continuously tracking and evaluating the effectiveness of intervention measures, the recommendation strategy is dynamically adjusted, and the health management efficiency is remarkably improved. And health data acquisition from multiple platforms is supported, heterogeneous data resources are fully utilized, and the health condition of the user is comprehensively described.

Description

Technical Field

[0001] The present application relates to the field of health management technology, and in particular to a method and device for intelligent intervention management recommendation based on health portraits. Background Art

[0002] In modern hospitals, outpatient clinics are the starting point for patients to see a doctor and are also one of the windows for the hospital's service quality. As an important part of the outpatient clinic, the waiting area undertakes the functions of patient reception, triage, nursing guidance and health education. Its service efficiency and quality directly affect the patient's medical experience and the harmony of the doctor-patient relationship. In order to make use of the idle time of patients in the waiting process, many hospitals have installed display screens in the waiting area to play health education videos.

[0003] However, the current mainstream waiting area education management devices have the following technical deficiencies and limitations. Lack of flexibility in playback content: Traditional devices usually use preset playback content and sequence, and can only achieve loop playback of fixed content. It is impossible to flexibly adjust the playback content according to the real-time status of the waiting area (such as whether there are patients waiting, the number and category of waiting patients, etc.), resulting in poor applicability and pertinence of the content. Insufficient personalized education capabilities: The current education devices lack intelligent analysis capabilities and cannot make personalized content recommendations based on the actual situation of waiting patients (such as age, gender, treatment department and specific diseases). There is a large deviation between the playback content and patient needs, making it difficult to achieve accurate health education. High content repetitiveness: The health education content played in the waiting area is mostly pre-uploaded materials. Long-term loop playback of the same content can easily cause patients to be distracted or aesthetically fatigued, making it difficult to achieve the purpose of health education. High dependence on manual intervention: Traditional devices require medical staff or administrators to manually upload education content, arrange playback sequence and adjust playback plans, which increases the management burden of medical institutions and also limits the flexibility and real-time operation of the device.

[0004] With the rapid development of artificial intelligence (AI) technology, it has been gradually applied to multiple links in the medical field, such as image analysis, auxiliary diagnosis and patient management. Combining AI technology to build an intelligent waiting area education device can effectively solve the technical limitations of traditional devices. By collecting waiting area status information and patient information in real time, intelligently matching and pushing personalized education content, it can improve the pertinence, automation and patient satisfaction of health education. Therefore, the development of an intelligent intervention management recommendation device based on health portraits is of great significance to optimizing waiting management and improving the quality of medical services. Summary of the invention

[0005] The embodiment of the present application provides a method and device for intelligent intervention management recommendation based on health portrait. The technical solution is as follows:

[0006] According to one aspect of the present application, a device for intelligent intervention management recommendation based on health portrait is provided, the device comprising:

[0007] A data collection module is used to collect the user's health data from the medical business system, physical examination system, and home health management application program. The data includes the user's physiological indicators, lifestyle information, medical service records, and disease history;

[0008] A health profile building module, used to generate a multi-dimensional dynamic health profile based on the health data, the health profile including the following specific tags: physical health tag, mental health tag, lifestyle tag;

[0009] The risk prediction module analyzes the health profile through a deep learning model, predicts the user's future disease risk, and generates a health risk report containing specific risk scores and risk categories;

[0010] An intervention plan recommendation module, for generating an intervention plan containing specific intervention contents based on the health portrait and the health risk report, wherein the intervention contents include daily diet recommendations, exercise suggestions, mental health counseling measures and drug use management;

[0011] The effect evaluation module is used to track the implementation effect of the user's intervention plan. The implementation effect is evaluated by comparing the changes in the user's health data before and after the intervention, and the health portrait and intervention plan are dynamically optimized according to the evaluation results.

[0012] Optionally, the health portrait construction module constructs the health portrait through a multi-source data fusion algorithm and a label model.

[0013] Optionally, the risk prediction module uses a long short-term memory network (LSTM) model to process the user's time series health data, identify the probability of developing diabetes, hypertension, and cardiovascular disease within a preset time in the future, and output the prediction results in the form of specific scores.

[0014] Optionally, the intervention plan recommendation module combines a collaborative filtering algorithm and a rule-based matching algorithm to generate intervention content described in a specific numerical form.

[0015] Optionally, the effect evaluation module analyzes the effectiveness of the intervention plan based on a decision tree algorithm, and obtains the evaluation result by comparing the baseline value of health data with the data after intervention. The specific comparison content includes the user's blood pressure reduction, weight change and blood sugar control rate.

[0016] On the other hand, a method for intelligent intervention management recommendation based on health portrait is provided, the method is used in the above-mentioned intelligent intervention management recommendation device based on health portrait, and the method comprises:

[0017] Acquire the user's health data from the medical business system, physical examination system, and home health management application program through the data collection module;

[0018] The collected data is labeled and modeled in multiple dimensions through the health portrait building module to generate a health portrait that includes physical health, mental health and lifestyle;

[0019] Analyzing the health profile using a deep learning algorithm through the risk prediction module to generate a health risk report including specific risk scores and risk categories;

[0020] Generate the intervention plan based on the health portrait and the health risk report by the intervention plan recommendation module;

[0021] Provide visual intervention guidance content through the user's health management application and allow users to check in for daily health behaviors;

[0022] The effect evaluation module compares the changes in the user's health data before and after the intervention to evaluate, and dynamically optimizes the health profile and intervention plan based on the evaluation results.

[0023] Optionally, the update frequency of the health portrait is dynamically adjusted based on new data uploaded by the user, and the specific update cycle does not exceed 48 hours.

[0024] Optionally, the intervention plan is generated using a reinforcement learning algorithm, and the recommended content is dynamically adjusted in combination with the user's historical behavior data.

[0025] On the other hand, a computer-readable storage medium is provided, which stores at least one instruction, and the at least one instruction is used to be executed by a processor to implement the intelligent intervention management recommendation method based on health portrait as described in the above aspect.

[0026] In an embodiment of the present application, an intelligent intervention management recommendation device based on health portraits is provided, which is particularly suitable for the field of health management technology. By providing grassroots doctors with intelligent and scientific health management auxiliary tools, the problem of insufficient grassroots medical resources is made up. The health portrait is used to dynamically reflect the health status of the individual, and personalized recommendations are generated in combination with algorithms to avoid "one-size-fits-all" intervention plans. By continuously tracking and evaluating the effectiveness of intervention measures, the recommendation strategy is dynamically adjusted to significantly improve the efficiency of health management. It also supports the collection of health data from multiple platforms, making full use of heterogeneous data resources to comprehensively characterize the user's health status. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a structural diagram of an intelligent intervention management recommendation device based on health portrait provided by an exemplary embodiment of the present application;

[0028] Figure 2 A flowchart of a health portrait-based intelligent intervention management recommendation method provided by an exemplary embodiment of the present application is shown. DETAILED DESCRIPTION

[0029] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0030] The term "multiple" as used herein refers to two or more than two. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the related objects are in an "or" relationship.

[0031] Example 1

[0032] like Figure 1 As shown, a structural schematic diagram of an intelligent intervention management recommendation device based on health portrait is shown, which includes a data collection module, a health portrait construction module, a risk prediction module, an intervention plan recommendation module and an effect evaluation module.

[0033] The data collection module is used to collect users' health data from medical business systems, physical examination systems, and home health management applications. The data includes users' physiological indicators (blood pressure, blood sugar, weight, etc.), lifestyle information (exercise frequency, eating habits, etc.), medical service records, and disease history.

[0034] The data collection module uses multi-source data interface technology to obtain users' health data from medical business systems, physical examination systems, and home health management applications through API, data crawling or ETL (Extract-Transform-Load) technology. The data includes structured data (such as medical records, physical examination reports) and unstructured data (such as exercise information recorded by health equipment). Through data preprocessing technology (such as deduplication, cleaning, and standardization), unified data storage is achieved to provide standardized data input for subsequent analysis.

[0035] The health portrait construction module is used to generate a multi-dimensional dynamic health portrait based on health data. The health portrait includes the following specific labels: physiological health labels (blood pressure status, blood sugar status), mental health labels (psychological stress score), and lifestyle labels (exercise regularity, dietary health score).

[0036] Among them, the health portrait construction module constructs a health portrait through a multi-source data fusion algorithm and a labeling model. Through the multi-source data fusion algorithm, data from different sources are clustered and associated by dimension (physical health, mental health, lifestyle, etc.), and combined with the labeling model, the user's health data is converted into a labeled portrait representation. For example, physiological indicators (such as blood pressure and blood sugar) are mapped to labels such as "normal" and "high" through threshold ranges; mental health scores generate specific values ​​through questionnaire data.

[0037] The risk prediction module analyzes health profiles through deep learning models, predicts users' future disease risks, and generates a health risk report containing specific risk scores and risk categories.

[0038] Among them, the risk prediction module uses the long short-term memory network LSTM model to process the user's time series health data, identify the probability of diabetes, hypertension, and cardiovascular disease within a preset time in the future, and output the prediction results in the form of specific scores. Based on the deep learning algorithm, the long short-term memory network (LSTM) is used to analyze the user's time series health data. LSTM can capture the historical trends and temporal correlations of user health data (such as the association between blood pressure change trends and diabetes risk), and output the probability score and category label of disease risk. The construction and training of the LSTM model utilizes a large amount of historical health data, combined with labeled data for supervised learning; the model input includes multi-dimensional labels in the health portrait, and outputs the disease risk probability score (such as a diabetes risk of 25%).

[0039] The intervention plan recommendation module is used to generate an intervention plan with specific intervention content based on health portraits and health risk reports. The intervention content includes daily diet recommendations (food types, intake), exercise recommendations (exercise type, duration, frequency), mental health counseling measures (specific psychological counseling course recommendations) and drug use management.

[0040] Among them, the intervention plan recommendation module combines collaborative filtering algorithm and rule-based matching algorithm to generate intervention content described in specific numerical form. A method combining collaborative filtering algorithm and rule-based matching algorithm is adopted. The collaborative filtering algorithm generates recommendations based on the historical intervention records of similar users, and the matching algorithm generates specific intervention content based on the labels and risk prediction results in the health portrait. For example, in dietary recommendations, the user's BMI, disease risk, and dietary preferences are considered, and specific food types and intake are output. Prior to this, it is necessary to build a user behavior database and rule base in order to adjust the recommendation accuracy of the intervention content in combination with user characteristics and historical behavior; the reinforcement learning algorithm can dynamically optimize the recommended content so that user behavior feedback can continuously improve the recommendation effect.

[0041] The effect evaluation module is used to track the implementation effect of the user intervention plan. The implementation effect is evaluated by comparing the changes in health data before and after the user's intervention (such as blood pressure, blood sugar, weight and other specific indicators), and dynamically optimizes the health portrait and intervention plan based on the evaluation results.

[0042] Among them, the effect evaluation module analyzes the effectiveness of the intervention plan based on the decision tree algorithm, and obtains the evaluation results by comparing the baseline value of health data with the data after intervention. The specific comparison content includes the user's blood pressure reduction (unit: mmHg), weight change (unit: kilograms) and blood sugar control rate.

[0043] Furthermore, the effect evaluation module conducts a quantitative analysis of the effectiveness of the intervention measures based on the decision tree algorithm. The user's health data before and after the intervention (such as blood pressure, weight, blood sugar, etc.) is used as input features, and the effect of the intervention is evaluated by comparing the baseline value of the health data with the post-intervention value. The evaluation results optimize the health portrait and intervention plan through a feedback mechanism. Specifically, an evaluation indicator system is constructed (such as the extent of blood pressure reduction, weight change, blood sugar control rate) and threshold judgment rules are formulated; the decision tree model provides a basis for adjusting the intervention plan through a classified analysis of the effects of different intervention methods.

[0044] like Figure 1 As shown in the figure, the data collection module is the starting point of the entire device, which is used to collect the user's health data from multiple sources such as medical business systems, physical examination systems, and home health management applications. Specific data includes the user's physiological indicators (such as blood pressure, blood sugar, weight, etc.), lifestyle information (such as exercise frequency, eating habits, etc.), medical service records (such as diagnosis and treatment records), and disease history. After the data is collected, it is uniformly transmitted to the health portrait construction module for further processing.

[0045] The health portrait construction module uses a multi-source data fusion algorithm and a labeling model to analyze and process the collected data to generate a multi-dimensional dynamic health portrait of the user. The health portrait contains the following specific labels: physiological health labels (such as blood pressure status, blood sugar status), mental health labels (such as psychological stress scores), and lifestyle labels (such as exercise regularity and dietary health scores). The generated results of the health portrait are used as input and passed to the risk prediction module.

[0046] The risk prediction module uses deep learning models (such as LSTM) to analyze health profiles and predict the user's future disease risk. The module analyzes time series health data to identify the probability that the user may suffer from diabetes, hypertension, cardiovascular disease, etc. in the future, and generates a health risk report containing risk scores and risk categories. The generated health risk report will be further transmitted to the intervention plan recommendation module.

[0047] The intervention plan recommendation module generates personalized intervention plans with specific intervention content based on health portraits and health risk reports, combined with collaborative filtering algorithms and rule-based matching algorithms. The intervention content specifically includes daily diet recommendations (food types and intake), exercise recommendations (exercise types, duration, frequency), mental health counseling measures (such as recommending specific psychological counseling courses) and drug use management. The intervention plan is presented in a visual form through the user health management application for reference by users or doctors.

[0048] The effect evaluation module is used to track the implementation effect of the user's intervention plan and compare and evaluate the health data before and after the intervention (such as blood pressure, blood sugar, weight and other specific indicators). The module analyzes the effectiveness of the intervention measures based on the decision tree algorithm. The evaluation results include the blood pressure drop (unit: mmHg), weight change (unit: kg) and blood sugar control rate. The evaluation results are used as feedback to dynamically optimize the health profile and intervention plan to achieve a continuous closed loop of health management.

[0049] The entire device adopts a closed-loop working mechanism. The feedback from the effect evaluation module will be sent back to the health profile construction module to update the health profile and ensure that the health profile always reflects the user's latest health status. Through this closed-loop design, the system can continuously optimize the accuracy and scientificity of the intervention plan.

[0050] thus, Figure 1 It demonstrated how the intelligent intervention management recommendation device based on health portrait can achieve a complete closed-loop process from data collection, portrait construction, disease risk prediction, intervention plan recommendation to intervention effect evaluation. Through multi-module collaboration, the system dynamically adjusts health portraits and recommendation content, greatly improving the efficiency and accuracy of personalized health intervention, and providing strong technical support for primary medical workers and users.

[0051] Through the device of the embodiment of the present application, it is possible to realize multi-source collection and integration of user health data and build a dynamic and real-time updated health portrait. The device combines the LSTM deep learning model to achieve high-precision disease risk prediction and generates personalized intervention plans based on user portraits and prediction results. The effect evaluation module can quantify the intervention effect and dynamically optimize the health portrait and recommendation mechanism, significantly improving the scientificity and effectiveness of health intervention.

[0052] Example 2

[0053] On the other hand, Figure 2 As shown, a method for intelligent intervention management recommendation based on health portrait is provided, and the method is used in the above-mentioned intelligent intervention management recommendation device based on health portrait, and the method includes:

[0054] Step 201, obtaining the user's health data from the medical business system, physical examination system, and home health management application program through the data collection module.

[0055] The user health data is collected through multi-source data interfaces. The data collection module supports HTTP API, database connection or batch file import to obtain structured and unstructured data. The data is cleaned, formatted and stored through the standardization module to ensure the consistency of the input data.

[0056] Step 202, multi-dimensionally labeling and modeling the collected data through the health portrait construction module to generate a health portrait that includes physical health, mental health and lifestyle.

[0057] The update frequency of the health portrait is dynamically adjusted based on the new data uploaded by the user, and the specific update cycle does not exceed 48 hours.

[0058] The health portrait construction module uses cluster analysis and labeling modeling technology to map health data into multi-dimensional labels (such as blood pressure status, exercise frequency). The update mechanism is triggered based on user uploaded data or preset update cycles to ensure that the health portrait always reflects the latest user health status.

[0059] Step 203, using a deep learning algorithm through a risk prediction module to analyze the health profile and generate a health risk report containing specific risk scores and risk categories.

[0060] The time series data in the health profile is processed through a deep learning model (such as LSTM), and feature engineering techniques (such as feature scaling and feature selection) are combined to generate disease risk probability scores (such as diabetes risk 25%). The output includes risk category and score.

[0061] Step 204: Generate an intervention plan based on the health profile and health risk report through the intervention plan recommendation module.

[0062] The generation of intervention plans uses reinforcement learning algorithms and dynamically adjusts recommended content based on the user's historical behavior data.

[0063] The intervention plan recommendation module optimizes the recommended content based on the reinforcement learning algorithm. Reinforcement learning uses the user's historical behavior data as feedback to dynamically adjust the recommendation parameters; the collaborative filtering algorithm combines the intervention data of similar users to generate basic recommendations, and the rule matching algorithm generates specific intervention suggestions (such as diet type and exercise duration) based on the tags in the health profile.

[0064] Step 205 : Provide visual intervention guidance content through the user's health management application and allow the user to check in for daily health behaviors.

[0065] The health management application presents the specific content of the intervention plan through front-end visualization technology (such as HTML5 and JavaScript). The daily check-in function uploads user feedback to the data collection module through form data submission.

[0066] Step 206, the effect evaluation module compares the changes in the user's health data before and after the intervention, and dynamically optimizes the health profile and intervention plan based on the evaluation results.

[0067] The effect evaluation module analyzes the effect of the intervention based on the decision tree algorithm by comparing the changes in health indicators before and after the intervention (such as a 5 mmHg drop in blood pressure). The analysis results are fed back to the health profile construction module and the intervention recommendation module to form a closed-loop optimization.

[0068] Through the intelligent intervention management recommendation method of the embodiment of this application, it is possible to achieve efficient collection and processing of health data, dynamically build health portraits in multiple dimensions, and accurately predict the user's health risks through deep learning algorithms. The intervention plan generation technology in the method is combined with a reinforcement learning algorithm to provide personalized intervention content, and the intervention effect is continuously optimized through the evaluation module to achieve a closed-loop process of health management, thereby improving the accuracy and real-time nature of health intervention.

[0069] An embodiment of the present application also provides a computer-readable medium storing at least one instruction, wherein the at least one instruction is loaded and executed by the processor to implement the intelligent intervention management recommendation method based on health portrait as described in the above embodiments.

[0070] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligent intervention management recommendation device based on health portrait, characterized in that: The device comprises: A data collection module is used to collect the user's health data from the medical business system, physical examination system, and home health management application program. The data includes the user's physiological indicators, lifestyle information, medical service records, and disease history; A health profile building module, used to generate a multi-dimensional dynamic health profile based on the health data, the health profile including the following specific tags: physical health tag, mental health tag, lifestyle tag; The risk prediction module analyzes the health profile through a deep learning model, predicts the user's future disease risk, and generates a health risk report containing specific risk scores and risk categories; An intervention plan recommendation module, for generating an intervention plan containing specific intervention contents based on the health portrait and the health risk report, wherein the intervention contents include daily diet recommendations, exercise suggestions, mental health counseling measures and drug use management; The effect evaluation module is used to track the implementation effect of the user's intervention plan. The implementation effect is evaluated by comparing the changes in the user's health data before and after the intervention, and the health portrait and intervention plan are dynamically optimized according to the evaluation results.

2. The device according to claim 1, characterized in that The health portrait construction module constructs the health portrait through a multi-source data fusion algorithm and a label model.

3. The device according to claim 1, characterized in that The risk prediction module uses a long short-term memory network (LSTM) model to process the user's time series health data, identify the probability of developing diabetes, hypertension, and cardiovascular disease within a preset time in the future, and output the prediction results in the form of specific scores.

4. The device according to claim 1, characterized in that The intervention plan recommendation module combines a collaborative filtering algorithm and a rule-based matching algorithm to generate intervention content described in a specific numerical form.

5. The device according to claim 1, characterized in that The effect evaluation module analyzes the effectiveness of the intervention plan based on a decision tree algorithm, and obtains the evaluation result by comparing the baseline value of health data with the data after intervention. The specific comparison content includes the user's blood pressure reduction, weight change and blood sugar control rate.

6. An intelligent intervention management recommendation method based on health portrait, characterized in that: The method is used for the intelligent intervention management recommendation device based on health portrait according to any one of claims 1 to 5, and the method comprises: Acquire the user's health data from the medical business system, physical examination system, and home health management application program through the data collection module; The collected data is labeled and modeled in multiple dimensions through the health portrait building module to generate a health portrait that includes physical health, mental health and lifestyle; Analyzing the health profile using a deep learning algorithm through the risk prediction module to generate a health risk report including specific risk scores and risk categories; Generate the intervention plan based on the health portrait and the health risk report by the intervention plan recommendation module; Provide visual intervention guidance content through the user's health management application and allow users to check in for daily health behaviors; The effect evaluation module compares the changes in the user's health data before and after the intervention to evaluate, and dynamically optimizes the health profile and intervention plan based on the evaluation results.

7. The method according to claim 6, characterized in that The update frequency of the health portrait is dynamically adjusted based on the new data uploaded by the user, and the specific update cycle does not exceed 48 hours.

8. The method according to claim 6, characterized in that The intervention plan is generated using a reinforcement learning algorithm and dynamically adjusts the recommended content in combination with the user's historical behavior data.

Images