Personal health record dynamic management method and equipment combined with virtual technology

Abstract

The present invention discloses a method and device for dynamic management of personal health records in combination with virtual technology, and relates to the technical field of health care. The method includes: after obtaining user authorization, reading and encrypting personal health records for offline storage; calling local channels to parse records and establish indicator attention; activating wearable devices, collecting user indicators through scene perception, and establishing a time series indicator data set; configuring the minimum granularity window to extract data features and cluster analysis; configuring trigger channels through records and scene perception, and performing identification to establish results; updating virtual data according to records and trigger results and visually presenting them. The method solves the technical problem that it is difficult to dynamically synchronize user health status and scene perception data in existing personal health record management, resulting in delayed health record updates and inability to accurately reflect the user's real-time health status, and realizes accurate health indicator collection and analysis combined with scene perception, and improves the technical effect of the intelligent and personalized level of health management.

Description

Technical Field

[0001] The present invention relates to the field of health and medical technology, and in particular to a method and device for dynamically managing personal health records in combination with virtual technology. Background Art

[0002] With the rapid development of the medical and health field, personal health records, as a comprehensive record of individual health information, have become an important part of health management. Traditional personal health records are mainly statically stored and only updated and maintained manually. This method has the following defects: first, it is impossible to realize real-time dynamic update of the records, and it is difficult to adapt to the rapid changes in the user's health status; second, it lacks the ability to perceive user scenarios and behaviors, resulting in delayed or biased updates of the archive content; third, the data presentation method is single, and it is difficult to provide users with intuitive health trend analysis and early warning information. Some existing technologies attempt to use sensors, wearable devices, etc. to collect health data, but most technologies are limited to the collection and simple analysis of a single data point, and cannot combine health records with real-time user scenarios for comprehensive analysis. In addition, the current health record management system generally lacks a flexible trigger mechanism, cannot dynamically adjust the monitoring focus and strategy according to different scenarios, and is difficult to achieve accurate management of multi-dimensional health indicators. Summary of the invention

[0003] The present application provides a method and device for dynamic management of personal health records combined with virtual technology, which is used to solve the technical problem that it is difficult to dynamically synchronize user health status and scene perception data in existing personal health record management, resulting in delayed health record updates and inability to accurately reflect the user's real-time health status.

[0004] In view of the above problems, the present application provides a method and device for dynamic management of personal health records combined with virtual technology.

[0005] The first aspect of the present application provides a method for dynamic management of personal health records in combination with virtual technology, the method comprising: after obtaining the user's authorization permission, reading the user's personal health record, and encrypting the personal health record and storing it offline in a local storage space; calling a local identification channel, performing file analysis on the personal health record based on the local identification channel, and establishing an indicator focus based on the analysis result; activating a wearable device, performing scene joint perception of the user based on the wearable device, establishing a scene perception result, controlling the wearable device to collect user indicators based on the indicator focus, and establishing a time series indicator data set; configuring a minimum granularity window, extracting data features of the time series indicator data set based on the minimum granularity window, establishing a granular feature extraction result, clustering the granular feature extraction results by feature trend, and establishing a feature trend clustering result; configuring an indicator trigger channel through the personal health record and the scene perception result, triggering and identifying the granular feature extraction result and the feature trend clustering result with the indicator trigger channel, and establishing an indicator trigger identification result; synchronously updating virtual data according to the personal health record and the indicator trigger identification result, and visually presenting it using virtual data.

[0006] The second aspect of the present application provides a personal health record dynamic management device combined with virtual technology, the device comprising: a personal health record reading module, the personal health record reading module is used to read the user's personal health record after obtaining the user's authorization, and encrypt the personal health record and store it offline in a local storage space; a file parsing module, the file parsing module is used to call a local identification channel, perform file parsing on the personal health record based on the local identification channel, and establish an indicator focus according to the parsing result; a perception collection module, the perception collection module is used to activate a wearable device, perform scene joint perception of the user based on the wearable device, establish a scene perception result, and control the wearable device to collect user indicators based on the indicator focus. Establish a time series indicator data set; a feature trend clustering module, the feature trend clustering module is used to configure a minimum granularity window, extract data features of the time series indicator data set based on the minimum granularity window, establish a granular feature extraction result, perform feature trend clustering on the granular feature extraction result, and establish a feature trend clustering result; a trigger identification module, the trigger identification module is used to configure an indicator trigger channel through personal health records and scene perception results, perform trigger identification of granular feature extraction results and feature trend clustering results through the indicator trigger channel, and establish an indicator trigger identification result; a visualization presentation module, the visualization presentation module is used to synchronously update virtual data according to personal health records and indicator trigger identification results, and use virtual data for visualization presentation.

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

[0008] After obtaining the user's authorization, the user's personal health file is read, and the personal health file is encrypted and stored offline in the local storage space; the local identification channel is called, and the personal health file is analyzed based on the local identification channel, and the indicator focus is established according to the analysis result; the wearable device is activated, and the user's scene joint perception is performed based on the wearable device, and the scene perception result is established. The wearable device is controlled to collect user indicators based on the indicator focus, and a time series indicator data set is established; the minimum granularity window is configured, and data features are extracted from the time series indicator data set based on the minimum granularity window, and a granular feature extraction result is established. The granular feature extraction result is clustered with feature trends, and a feature trend clustering result is established; the indicator trigger channel is configured through the personal health file and the scene perception result, and the granular feature extraction result and the feature trend clustering result are triggered and identified with the indicator trigger channel, and the indicator trigger identification result is established; the virtual data is updated synchronously according to the personal health file and the indicator trigger identification result, and the virtual data is visualized. The technical effect of combining scene perception to accurately collect and analyze health indicators and improve the intelligence and personalization level of health management is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] 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 creative work.

[0010] Figure 1 A flowchart of a method for dynamic management of personal health records in combination with virtual technology provided in an embodiment of the present application.

[0011] Figure 2 A schematic diagram of the structure of a personal health record dynamic management device combined with virtual technology provided in an embodiment of the present application.

[0012] Explanation of the accompanying drawings: personal health record reading module 1, record parsing module 2, perception collection module 3, feature trend clustering module 4, trigger recognition module 5, visualization presentation module 6. DETAILED DESCRIPTION

[0013] This application provides a method and device for dynamic management of personal health records combined with virtual technology, which is used to solve the technical problem that it is difficult to dynamically synchronize user health status and scene perception data in existing personal health record management, resulting in delayed health record updates and inability to accurately reflect the user's real-time health status.

[0014] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0015] Embodiment 1, as Figure 1 As shown, the present application provides a method for dynamic management of personal health records in combination with virtual technology, the method comprising:

[0016] After obtaining the user's authorization, the user's personal health record is read, and the personal health record is encrypted and stored offline in the local storage space.

[0017] Specifically, after the user explicitly authorizes and agrees, the user's personal health records will be read, including the user's basic health information, medical history, medical records and other data; in order to ensure the privacy and security of user data, these health records will be encrypted, and sensitive information will be converted into a ciphertext form that cannot be directly read through encryption algorithms (such as symmetric encryption algorithms, hash algorithms, etc.). The encrypted files will not be uploaded to the online server, but will be saved in the local storage space of the device, thereby avoiding the risk of data leakage caused by network transmission or cloud storage. This offline storage method not only ensures the security of user data, but also facilitates the rapid call of health record data in subsequent analysis.

[0018] The local identification channel is called, the personal health record is analyzed based on the local identification channel, and the indicator focus is established according to the analysis result.

[0019] Specifically, by calling the local identification channel in the device, the personal health records stored locally are parsed. The parsing process includes extracting key data from the health records, such as the user's medical history, test results, and health indicators, and classifying, screening, and structuring these data. Based on the information obtained from the analysis, the indicators that need to be focused on in the user's health status, such as blood pressure, blood sugar, or weight changes, are automatically identified, and a list, namely the indicator focus, is generated. This list is used to mark and track the user's key health data, providing reference and direction for subsequent data collection and analysis. The entire process is completed locally to ensure the privacy and security of user data.

[0020] For the local recognition channel, its core is to achieve accurate extraction of indicators that need to be focused on in health records through model training and reasoning. Taking the decision tree as an example, first, prepare the training data, including input data and output data. The input data is the sample personal health record, which contains basic information (such as age, gender), past medical history (such as diabetes, hypertension, etc.), medical records (such as medication, hospitalization history) and various health indicators (such as blood pressure, blood sugar, heart rate, etc.); the output data is the health indicators that need to be focused on (such as blood pressure, blood sugar, etc.), which are marked by medically defined threshold rules (such as blood pressure>140 is marked as needing attention). The input and output data are stored in a structured table format, and each record represents a sample health record. Data preprocessing is a key step in building an identification channel. In this step, the health record data is split into independent features and formatted, continuous variables (such as blood pressure and blood sugar) are normalized using the maximum and minimum methods, categorical variables (such as gender) are converted into numerical form (such as male = 0, female = 1), and missing data are filled or interpolated. In addition, health indicators need to be annotated for abnormalities, and the data needs to be labeled according to medical standards to mark which indicators belong to the category that needs attention, forming the target value of the training data set; then, with the decision tree algorithm as the core, the goal is to train an identification channel with classification function that can predict the health indicators that need to be focused on based on the input health record characteristics. The training process starts from the root node, selects features (such as blood pressure) for splitting, finds the best split point (such as whether blood pressure is higher than 140) based on the principle of maximizing information gain or Gini index, recursively splits data, and builds a multi-layer structure of decision tree until the preset stop condition is met (such as too few node samples or the depth of the tree reaches the limit); after the recognition channel training is completed, the decision tree is optimized, and the risk of overfitting is reduced through post-pruning. At the same time, the priority distribution of features is adjusted according to the importance of specific indicators in the health record (such as blood pressure has a higher weight for hypertensive patients); and then the recognition channel performance (such as accuracy, recall rate, F1 score) is evaluated through cross-validation to ensure that the recognition channel can accurately predict the indicators that need to be paid attention to and has good generalization ability. In practical applications, the local recognition channel uses the learned knowledge to infer the real-time input health records. Each time a new health record is received, the recognition channel parses the health indicators in it and identifies the indicators that need to be focused on according to the learned mapping relationship. For example, for a file containing blood pressure of 150 and blood sugar of 130, the blood pressure will be marked as an indicator requiring attention based on the rule of blood pressure>140, and the blood sugar will be marked as an indicator requiring attention based on the rule of previous history of diabetes and blood sugar>120, thereby outputting a list, namely indicators of attention, such as blood pressure and blood sugar.

[0021] Activate wearable devices, perform joint scene perception of users based on wearable devices, establish scene perception results, control wearable devices to collect user indicators based on indicator attention, and establish a time series indicator data set;

[0022] Specifically, by activating the user's wearable device (such as a smart watch, health bracelet or portable sensor), the device enters the working state, and uses its built-in sensors (such as temperature sensors, humidity sensors, light intensity sensors, etc.) to collect environmental information around the user. The wearable device obtains environmental data through scene joint perception technology, such as the user's current temperature, humidity, light intensity, noise level, etc., to form a complete scene perception result. This result is intended to reflect the specific characteristics of the user's current environment and provide support for subsequent health monitoring; on the basis of obtaining the scene perception result, according to the previously established indicator focus, the wearable device is controlled to collect the user's health indicator data in a targeted manner. For example, if the indicators of concern include heart rate and body temperature, the wearable device will give priority to activating the relevant sensors, collect these health data on demand, and record their time series changes (such as the fluctuation of heart rate or body temperature for several consecutive minutes); the collected data is integrated into a time series indicator data set for subsequent analysis and processing. This method ensures that the device's collection work is more accurate and efficient, and at the same time provides richer background support for health monitoring in combination with scene information.

[0023] Configure the minimum granularity window, extract data features of the time series indicator data set based on the minimum granularity window, establish granular feature extraction results, perform feature trend clustering on the granular feature extraction results, and establish feature trend clustering results.

[0024] Specifically, first, a minimum granularity window is configured to divide and process continuous data segments in the time series indicator data set. The size of the minimum granularity window is dynamically adjusted according to the change rate and collection frequency of the specific indicator. For example, for heart rate data, the window may be set to one minute. Through the minimum granularity window, refined features can be extracted from the time series indicator data set. These features constitute the granular feature extraction results, which contain the key statistical characteristics of the health indicators in each window; then, the granular feature extraction results are clustered according to feature trends, aiming to discover patterns and laws in the changing trends of health indicators. In the clustering process, by setting multiple initial clustering centers, the granular features are classified into corresponding cluster clusters according to similar values. However, there may be intermediate obstacles in the clustering process (such as short-term fluctuations in feature values ​​or discontinuities in data distribution). Such obstacles can be handled by a breakthrough mechanism, which analyzes the similarity of adjacent features and the breakthrough conditions of cluster clusters, such as judging whether two cluster clusters should be merged based on a specific threshold. Finally, after completing the clustering of all granular features, the feature trend clustering results are established. These results reflect the overall change trend of health indicators and their potential patterns, providing a basis for subsequent indicator analysis and abnormal monitoring.

[0025] In a possible implementation, the step of performing feature trend clustering on the granular feature extraction result to establish the feature trend clustering result further includes:

[0026] The granular feature extraction results are sorted in time series, and a predetermined number of cluster centers are evenly distributed based on the time series sorting; with the cluster center as the search center, a time series adjacent similarity search is performed, and a breakthrough coefficient mapped to the cluster size is established; search clustering is performed based on the time series adjacent similarity value and the breakthrough coefficient, and feature trend clustering is completed according to the search clustering results.

[0027] Specifically, the granular feature results extracted from the time series indicator data set are sorted in chronological order. Each granular feature segment corresponds to a time segment. The purpose of sorting is to ensure that the clustering analysis can reflect the dynamic change trend of the feature over time. For example, for heart rate data, the granular feature can be the average heart rate in each minute. After sorting, these minute-level data are arranged in chronological order; in order to initialize the clustering process, a set of initial cluster centers are evenly distributed according to the range of granular feature data sorted by time series. The number of cluster centers can be pre-set according to the data scale and analysis requirements. For example, for a data set containing 100 time series features, 10 initial cluster centers can be set, which are distributed within the range of time series features. The initial values ​​of these cluster centers can be equidistant from the data. Percentiles (such as every 10% percentile value) can also be statistical feature points of data (such as mean, median, etc.); in the clustering process, each cluster center is used as a search starting point to expand the search to adjacent data points. The core of adjacent similarity search is to calculate the similarity value between the current feature and the cluster center. For example, the distance between two features is measured by Euclidean distance, cosine similarity, etc. For the similarity of each data point in the search process, if the similarity value between the point and the cluster center is within the set threshold range, it will be included in the current cluster cluster. At the same time, the cluster center is updated in real time to the average value or centroid of the current cluster to ensure that subsequent searches more accurately reflect the distribution of the cluster; clustering process In the cluster search process, for each feature point that is not included in the cluster, it is determined in turn whether it satisfies the adjacent similarity value with a certain cluster center that is less than or equal to the set similarity threshold, and the calculated breakthrough coefficient is less than or equal to the preset breakthrough threshold; if all these conditions are met, the feature point is classified into the corresponding cluster, and the cluster center of the cluster is updated. If the conditions are not met, a new cluster center is generated and a new cluster is reinitialized. This process is iterated continuously until all feature points are classified into a cluster or the stopping condition is met (such as reaching the maximum number of cluster centers). Through the above process, the trend clustering of the granular feature extraction results is completed to form the final clustering results, namely the feature trend clustering results. These results include the cluster center and range of each cluster, the time span and trend description of each cluster. The clustering results can be used to analyze the change pattern of health indicators, for example, to identify the abnormal fluctuation trend of heart rate in a certain period of time or the significant increase of blood sugar in a certain stage, which provides an important basis for health monitoring and decision-making.

[0028] In a possible implementation, the searching and clustering based on the time-series adjacent similarity values ​​and the breakthrough coefficients of the clustering clusters further includes:

[0029] Based on the size of the clustering cluster, the breakthrough coefficient is matched, and the matching result is established. The breakthrough analysis is performed based on the matching result and the time-series adjacent similarity value; if the breakthrough analysis is a breakthrough result, the corresponding granular feature extraction result is clustered into the clustering cluster, and the clustering search is continued; if the breakthrough analysis is a non-breakthrough result, the clustering is stopped, and a new instruction is generated; based on the new instruction, a cluster center is added to the non-clustered granular feature extraction result, and a clustering search is performed until the clustering of all granular feature extraction results is completed.

[0030] Specifically, during the clustering process, the size of each cluster will change dynamically, indicating the number of data points in the cluster. In order to deal with obstacles in the time series feature data (such as short-term anomalies or discontinuous data), a breakthrough coefficient is calculated for each cluster. This breakthrough coefficient is calculated by comparing the distance between the current feature value and the cluster center and the number of data points in the current cluster. Through the breakthrough coefficient, it can be evaluated whether the obstacle point reasonably belongs to the cluster. Subsequently, the calculated breakthrough coefficient is compared with the preset matching threshold to establish a matching result. If the breakthrough coefficient is less than or equal to the preset breakthrough threshold, it means that the obstacle point is close to the cluster. If the distance between clusters is small, it can be classified into the cluster. If the breakthrough coefficient is greater than this threshold, it means that the correlation between the obstacle point and the cluster is weak and cannot be reasonably attributed. After that, the breakthrough analysis is performed based on the determined matching relationship temporal adjacent similarity value. The purpose of the breakthrough analysis is to combine the matching result and the temporal adjacent similarity value to determine whether the obstacle point can cross the obstacle and continue to cluster. If the matching result meets the breakthrough condition and the temporal adjacent similarity value is within the allowable range, the obstacle point can be classified into the current cluster. If any condition is not met, the obstacle point cannot be associated with the cluster. When the breakthrough analysis is a breakthrough result, that is, the breakthrough condition is met and the temporal adjacent similarity value is within the allowable range, the obstacle point can be classified into the current cluster. Within the allowed range, the obstacle point (i.e., the current granular feature extraction result) is classified into the corresponding clustering cluster, and the center value of the clustering cluster is updated to recalculate the center position by incorporating the data in the cluster after the obstacle point, and then continue to perform clustering search, starting from the updated clustering cluster center, and analyze the next time series feature point. In this way, the clustering cluster can continue to expand across obstacles, handle short-term fluctuating data characteristics, improve the continuity of clustering, and avoid interrupting the clustering process due to minor anomalies; when the breakthrough analysis is a non-breakthrough result, that is, any condition is not met, the expansion of the current clustering cluster is stopped, and the cluster is considered to be completed. At this time, a new instruction is generated for initialization. In the process of initialization, the current obstacle point will be selected as the new cluster center, and a new cluster will be initialized to include the obstacle point. In this way, the significant abnormal or discontinuous feature points in the data are processed separately to prevent the analysis results from being distorted due to incorrect clustering. Then, according to the newly added instructions, the cluster search is re-executed, starting from the new cluster center, and the new cluster is expanded forward in the time series order. The breakthrough coefficient matching and breakthrough analysis steps are repeated until all granular feature points are classified. When all granular feature extraction results have been classified into a certain cluster cluster, the clustering results of the granular features are completed, and the feature trend clustering results are output. This mechanism of dynamically adjusting the cluster center, combined with the flexible processing of the breakthrough coefficient, can take into account the continuity of the time series data and the independence of the abnormal points, ensuring that the final clustering results are both accurate and globally meaningful.

[0031] The indicator trigger channel is configured through personal health records and scene perception results, and the indicator trigger channel is used to trigger the recognition of granular feature extraction results and feature trend clustering results to establish the indicator trigger recognition results.

[0032] Specifically, according to the user's personal health record and the scene perception results obtained through wearable devices, an indicator trigger channel is jointly configured. The function of this trigger channel is to dynamically associate personal health information with current scene information, so as to accurately determine the health indicators that need further analysis. The configuration process of the trigger channel includes extracting the user's key health indicators (such as blood pressure, heart rate, etc.) from the health record, combining with scene perception data (such as ambient temperature, exercise status, etc.), and setting trigger conditions. For example, in high temperature scenes, heart rate and body temperature are monitored first; when the indicator trigger channel is started, the granular feature extraction results and feature trend clustering results are analyzed based on the configured trigger conditions. Screening and analysis are performed to determine whether there are features that exceed the normal range or have large fluctuations (such as a continuous increase in heart rate) in the granular feature extraction results; if an indicator meets the trigger condition (such as heart rate trend clustering shows that the abnormal duration exceeds the threshold), the indicator is marked as requiring attention and an indicator trigger identification result is generated; finally, the output indicator trigger identification result includes a list of triggered health indicators and their status descriptions. For example, an abnormal increase in heart rate lasting 10 minutes may pose a risk. These results provide a clear direction and basis for subsequent health analysis and early warning, and can also guide wearable devices to adjust their collection strategies and conduct more intensive monitoring of key indicators.

[0033] In a possible implementation, configuring an indicator trigger channel through personal health records and scene perception results, and using the indicator trigger channel to trigger identification of granular feature extraction results and feature trend clustering results, also includes:

[0034] Generate N static threshold ranges based on the personal health record, and use the N static threshold ranges to build a static trigger mechanism; perform scene adaptation of the N static threshold ranges based on the personal health record and the scene perception results, establish M dynamic expansion thresholds, and establish a dynamic trigger mechanism with the M dynamic expansion thresholds, where M is less than or equal to N; configure the indicator trigger channel with the static trigger mechanism and the dynamic trigger mechanism to complete trigger identification.

[0035] Specifically, the user's basic health information and key indicators, such as blood pressure, blood sugar, heart rate, etc., are extracted from the personal health record. The static threshold range of each indicator is defined according to medical standards and user historical data, for example: blood pressure 90~140mmHg, blood sugar 70~120mg / dL, heart rate 60~100 times / minute. The normal range of each indicator is marked as safe, and the value exceeding this range is defined as an abnormal area, forming a total of N static threshold ranges. Each static threshold range corresponds to an indicator, which is used to reflect the user's health baseline of the indicator under normal circumstances; then, according to each static threshold range, a trigger rule is set for each health indicator. For example, if the blood pressure is higher than 140mmHg or lower than 90mmHg, abnormal blood pressure is triggered, and if the heart rate is higher than 100 times / minute, a fast heart rate is triggered. These rules constitute a static trigger mechanism for basic health monitoring without scene dependence; then, combined with the scene perception results collected by the wearable device, for example, whether the user is in motion (running, walking), ambient temperature, humidity, light For some specific indicators, some static threshold ranges will be selected for expansion according to the scene perception results to generate dynamic expansion thresholds. For example, in the state of exercise, the normal range of heart rate is expanded to 80-150 beats / minute, and in a high temperature environment, the normal range of body temperature is adapted to 36.5-37.8℃. During the scene adaptation process, only the indicators directly related to the scene are expanded. For example, heart rate and body temperature will be expanded in the state of exercise, but blood sugar may maintain the static threshold. After the scene adaptation, only part of the static threshold range is expanded to generate dynamic expansion thresholds, a total of M dynamic thresholds (M≤N), and indicators not affected by the scene continue to maintain the original static thresholds. For example, if the user is in a sports scene, heart rate and body temperature generate dynamic expansion thresholds, while blood pressure and blood sugar retain static thresholds. Afterwards, according to these dynamic expansion thresholds, M new trigger rules are defined, that is, dynamic trigger mechanisms. For example, if the user is in a fast running scene and the heart rate is higher than 150 beats / minute, the exercise heart rate is too fast, and if the user is in a high temperature scene and the body temperature is higher than 37.8℃, abnormal body temperature is triggered. The dynamic trigger mechanism can flexibly adjust the judgment criteria of indicators in different scenarios and identify potential abnormalities more accurately. Then, the static trigger mechanism and the dynamic trigger mechanism are encapsulated to configure the indicator trigger channel. The trigger channel gives priority to the dynamic trigger mechanism according to the actual scenario. Only when there is no scenario perception data or the indicator does not generate a dynamic threshold, it returns to the static trigger mechanism. For indicators that have generated dynamic expansion thresholds (M), the trigger channel is based on the dynamic trigger mechanism. For indicators that have not generated dynamic expansion thresholds (NM), the trigger channel is based on the static trigger mechanism. The granular feature extraction results and feature trend clustering results are matched in real time through the trigger channel. If the feature value of an indicator exceeds the dynamic expansion threshold, the abnormal warning of the indicator is triggered. If the dynamic expansion threshold is not triggered and the feature value exceeds the static threshold, the basic abnormality is triggered through the static trigger mechanism. The indicator trigger channel finally outputs the indicator trigger recognition result containing the abnormal indicator and its status. For example, abnormal heart rate: the heart rate reaches 160 times / minute in the exercise state, which exceeds the dynamic threshold range (80-150 times / minute). Through the above process, the indicator trigger channel can combine static and dynamic trigger mechanisms to achieve accurate monitoring and abnormal identification of user health indicators. Among them, the dynamic expansion threshold mechanism enhances the scenario adaptability of indicator triggering, while the static threshold mechanism provides a global basic guarantee to ensure the efficient operation of the trigger channel.

[0036] In a possible implementation, the triggering identification of the granular feature extraction result and the feature trend clustering result by using the indicator trigger channel also includes:

[0037] Establish an association rule analysis layer, and couple the association rule analysis layer to the indicator trigger channel; when the indicator trigger channel is used for trigger identification, synchronously call the association rule analysis layer to perform time series correlation analysis of multiple indicators to generate association analysis results; use the association analysis results to establish a joint trigger, and compensate the indicator trigger identification results through the joint trigger.

[0038] Specifically, the association rule analysis layer is used to discover potential correlations between health indicators, especially multi-indicator association patterns in time series data, which are constructed through historical monitoring data and association rule mining algorithms (such as the Apriori algorithm or the FP-Growth algorithm); taking the Apriori algorithm as an example, key health indicators (such as blood pressure, heart rate, blood sugar, etc.) are first extracted from historical monitoring data and divided into multiple transaction data sets according to time windows. Each transaction represents a combination of abnormal indicators within a time segment. For example, a transaction may contain abnormal heart rate and elevated blood sugar; then, parameters such as minimum support, minimum confidence, and lift are set to define the minimum requirements for rule mining, where the minimum support is obtained by comparing the number of transactions containing the combination with the total number of transactions in multiple transaction data sets. The minimum confidence is calculated by calculating the ratio of the number of transactions containing the condition and result combination to the number of transactions containing only the condition combination, and the lift is calculated by calculating the ratio of the confidence and support of the result combination; the algorithm starts from a single item set and generates candidate item sets layer by layer, for example, from abnormal heart rate to the combination of abnormal heart rate and elevated blood sugar, and filters out frequent item sets by calculating support; then association rules are generated from the frequent item sets, and the confidence and lift of each rule are calculated, and only high-quality rules that meet the threshold are retained. For example, if the heart rate is abnormal, the rule of elevated blood sugar may have a confidence of 67% and a lift of 1.5; on this basis, the filtered high-quality rules are stored as a standardized rule base, and the association rule analysis layer is called when the indicator trigger channel is started by loading the rule base. When the indicator trigger channel identifies a single abnormal indicator, the analysis layer will automatically search for related rules based on the rule base, calculate the correlation of multiple indicators in combination with the current time series data, measure whether the changes of indicators in the same time window are related, and output the association analysis results, including indicators with strong correlation, corresponding rule matching, and possible joint trigger conditions; then, through the association analysis results, define the joint trigger conditions, for example, abnormal heart rate + elevated blood sugar → joint trigger: hypertension risk; if the single indicator trigger identification result does not cover all potential risks, the joint trigger will compensate for these unidentified abnormalities, for example, the single indicator trigger only detects abnormal heart rate, but does not detect changes in blood sugar and blood pressure. Through association analysis, the joint trigger result of hypertension risk is compensated, and it is recommended to monitor blood pressure, etc., to make up for the shortcomings of single indicator analysis. Through the above process, the introduction of the association rule analysis layer and the construction of the joint trigger mechanism have greatly improved the accuracy and comprehensiveness of the indicator trigger channel, which can more effectively identify potential health risks and provide precise guidance for health management.

[0039] In a possible implementation, establishing an indicator to trigger the recognition result further includes:

[0040] A cumulative anomaly model is established, and based on the cumulative anomaly model, cumulative identification of trigger identification is performed based on a multi-stage trigger mechanism to establish cumulative identification results; the indicator trigger identification results are compensated through the cumulative identification results, and user early warning management is performed based on the compensated indicator trigger identification results.

[0041] Specifically, the cumulative anomaly model is established by recording the abnormal conditions of each health indicator in multiple time periods, comprehensively calculating the abnormal count, abnormal duration and severity, and generating a cumulative anomaly value for each indicator. The model is calculated using the following formula: ;in, is the cumulative outlier value, N is the total number of time periods in the time window, is the number of exceptions triggered within time period t, is the total duration of the abnormality in time period t, is the anomaly weight (reflecting severity) within time period t; in the cumulative identification stage, the model is applied to short-term and long-term trigger identification. Short-term identification is based on a shorter time window (such as 5 minutes). When the cumulative anomaly value exceeds the preset short-term threshold, a short-term warning is triggered. Long-term identification is based on a longer time window (such as 1 hour). When multiple short-term cumulative anomalies continue to exceed the standard, a long-term warning is triggered. By dynamically monitoring the accumulation of abnormal values ​​of indicators, the model can accurately identify short-term and long-term potential risks; when the cumulative identification result is generated in the trigger identification process, the cumulative result is compared with the single trigger result, and the compensation mechanism is activated. If the cumulative identification result shows that the abnormal risk is high, but the single trigger does not detect the abnormality, the modification If the trigger result is abnormal, the abnormal priority is adjusted to a higher level. On the contrary, if the cumulative result shows that the risk is low, the trigger recognition priority is reduced or invalid triggers are cleared. This compensation mechanism can make up for the shortcomings of a single trigger and improve the comprehensiveness and accuracy of recognition. The compensated trigger recognition results are used to generate user warning management information, and the warning level is divided according to the severity of the result. For example, a mild warning prompts the user to pay attention and recommends self-monitoring. A moderate warning recommends taking measures such as rest or adjusting behavior. A severe warning directly prompts the user to seek medical treatment immediately and may notify the medical service provider. At the same time, the warning information is pushed to the user through a visual interface, mobile phone or wearable device to ensure that the user can take timely countermeasures. In the scenario of severe warning, abnormal logs will also be recorded for subsequent medical analysis to ensure the comprehensiveness and effectiveness of user health monitoring.

[0042] The virtual data is updated synchronously based on the personal health records and indicator trigger identification results, and is visualized using the virtual data.

[0043] Specifically, virtual data is updated in real time based on the user's personal health record and the latest indicator trigger identification results. A comprehensive and real-time health status mapping is formed by dynamically integrating basic information in the health record (such as past medical history, basic health parameters) with current indicator trigger data (such as the type, degree and duration of abnormal indicators). During the update process, key indicators in the personal health record are retrieved, such as blood pressure, heart rate, blood sugar, etc., and compared with the abnormal indicators detected in the trigger results. According to the changes in the trigger results, the relevant parameters of the virtual data are dynamically adjusted. For example, when the trigger result shows that the duration of the abnormal heart rate exceeds a certain threshold, the virtual data will record the time span and severity of the abnormality. The updated virtual data is presented as an intuitive graphical health status display through visualization tools, such as a real-time health trend chart, a risk heat map or a dynamic change curve of abnormal indicators. The visualization content can display key information of the user's current health status, such as the fluctuation trend of the heart rate over a period of time and the duration of abnormal blood sugar. At the same time, the visualization interface can highlight indicators that require special attention from the user and provide simple suggestions, such as reducing exercise intensity or measuring blood pressure regularly. This mechanism of synchronous update and visual presentation not only improves the user's intuitive understanding of their own health status, but also helps users quickly determine whether their health status requires further intervention, thereby achieving more efficient health management and early warning response.

[0044] In one possible implementation, the method further includes:

[0045] An indicator warning is created based on the identification result triggered by the indicator, a warning feedback is generated, and a verification window is simultaneously established; if the real-time indicator in the verification window fails to meet the calibration threshold of the warning feedback, a virtual hazard fitting is established based on the real-time indicator; and the virtual hazard fitting is visualized.

[0046] Specifically, according to the indicator trigger identification results, the corresponding indicator warning is first created. The creation of the warning includes extracting abnormal indicators (such as blood pressure, heart rate, etc.) and their abnormal types (such as too high or too low), duration and severity from the trigger results; based on this information, the warning feedback content is generated to clearly inform the user of the status of the abnormal indicators, possible risks and recommended response measures. For example, if the heart rate continues to rise abnormally (heart rate greater than 150bpm, lasting 10 minutes), it is recommended to rest and monitor. At the same time, a verification window is synchronously established to verify the accuracy of the warning feedback in real time. The time length and parameter range of the verification window are determined by the characteristics of the triggering indicator. For example, the heart rate may be set to a 10-minute window and the blood sugar may be set to 30 minutes; within the verification window, the data of the relevant indicators are collected in real time and compared with the calibration thresholds in the warning feedback to determine whether the indicators are gradually returning to the normal range. ; If within the verification window, the indicators collected in real time fail to meet the recovery threshold calibrated in the early warning feedback, the possible risk scenarios will be deduced through the virtual hazard fitting mechanism. The virtual hazard fitting is based on the current status, historical trends and possible impact relationships of the abnormal indicators. For example, an abnormal increase in heart rate may lead to the risk of cardiovascular disease, or persistent abnormal blood sugar may be associated with the possibility of acute complications of diabetes; the generated virtual hazard fitting results are presented in a visual way to intuitively show the potential risks to the user, for example, the continuous trend of abnormal heart rate is shown through a dynamic curve chart, the high-risk areas of the body are indicated through a heat map, or the hazard level and recommended health management measures are displayed in a graphical way; users can quickly understand the source and severity of health risks through this visual interface. This process enhances the accuracy and action guidance of the early warning and provides efficient support for user health management.

[0047] Embodiment 2 is based on the same inventive concept as the method for dynamic management of personal health records in combination with virtual technology in the above embodiment. Figure 2 As shown, the present application provides a personal health record dynamic management device combined with virtual technology, wherein the device includes:

[0048] A personal health record reading module 1 is used to read the user's personal health record after obtaining the user's authorization, and encrypt the personal health record and store it offline in the local storage space; a file analysis module 2 is used to call the local identification channel, perform file analysis on the personal health record based on the local identification channel, and establish indicator focus according to the analysis result; a perception acquisition module 3 is used to activate the wearable device, perform scene joint perception of the user based on the wearable device, establish scene perception results, control the wearable device to collect user indicators based on the indicator focus, and establish a time series indicator data set; a feature trend clustering module 4, The feature trend clustering module 4 is used to configure the minimum granularity window, extract data features of the time series indicator data set based on the minimum granularity window, establish granular feature extraction results, perform feature trend clustering on the granular feature extraction results, and establish feature trend clustering results; the trigger identification module 5, the trigger identification module 5 is used to configure the indicator trigger channel through the personal health record and the scene perception result, use the indicator trigger channel to trigger and identify the granular feature extraction results and the feature trend clustering results, and establish the indicator trigger identification result; the visualization presentation module 6, the visualization presentation module 6 is used to synchronously update the virtual data according to the personal health record and the indicator trigger identification result, and use the virtual data for visualization presentation.

[0049] Furthermore, the characteristic trend clustering module 4 also includes:

[0050] The granular feature extraction results are sorted in time series, and a predetermined number of cluster centers are evenly distributed based on the time series sorting; with the cluster center as the search center, a time series adjacent similarity search is performed, and a breakthrough coefficient mapped to the cluster size is established; search clustering is performed based on the time series adjacent similarity value and the breakthrough coefficient, and feature trend clustering is completed according to the search clustering results.

[0051] Furthermore, the characteristic trend clustering module 4 also includes:

[0052] Based on the size of the clustering cluster, the breakthrough coefficient is matched, and the matching result is established. The breakthrough analysis is performed based on the matching result and the time-series adjacent similarity value; if the breakthrough analysis is a breakthrough result, the corresponding granular feature extraction result is clustered into the clustering cluster, and the clustering search is continued; if the breakthrough analysis is a non-breakthrough result, the clustering is stopped, and a new instruction is generated; based on the new instruction, a cluster center is added to the non-clustered granular feature extraction result, and a clustering search is performed until the clustering of all granular feature extraction results is completed.

[0053] Furthermore, the trigger identification module 5 also includes:

[0054] Generate N static threshold ranges based on the personal health record, and use the N static threshold ranges to build a static trigger mechanism; perform scene adaptation of the N static threshold ranges based on the personal health record and the scene perception results, establish M dynamic expansion thresholds, and establish a dynamic trigger mechanism with the M dynamic expansion thresholds, where M is less than or equal to N; configure the indicator trigger channel with the static trigger mechanism and the dynamic trigger mechanism to complete trigger identification.

[0055] Furthermore, the trigger identification module 5 also includes:

[0056] Establish an association rule analysis layer, and couple the association rule analysis layer to the indicator trigger channel; when the indicator trigger channel is used for trigger identification, synchronously call the association rule analysis layer to perform time series correlation analysis of multiple indicators to generate association analysis results; use the association analysis results to establish a joint trigger, and compensate the indicator trigger identification results through the joint trigger.

[0057] Furthermore, the trigger identification module 5 also includes:

[0058] A cumulative anomaly model is established, and based on the cumulative anomaly model, cumulative identification of trigger identification is performed based on a multi-stage trigger mechanism to establish cumulative identification results; the indicator trigger identification results are compensated through the cumulative identification results, and user early warning management is performed based on the compensated indicator trigger identification results.

[0059] Furthermore, the visual presentation module 6 also includes:

[0060] An indicator warning is created based on the identification result triggered by the indicator, a warning feedback is generated, and a verification window is simultaneously established; if the real-time indicator in the verification window fails to meet the calibration threshold of the warning feedback, a virtual hazard fitting is established based on the real-time indicator; and the virtual hazard fitting is visualized.

[0061] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0062] The above description is only a preferred 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 should be included in the protection scope of the present application.

[0063] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for dynamic management of personal health records combined with virtual technology, characterized in that: The method comprises: After obtaining the user's authorization, read the user's personal health record, encrypt the personal health record and store it offline in the local storage space; Calling a local identification channel, performing file analysis on the personal health file based on the local identification channel, and establishing indicator focus according to the analysis result; Activate wearable devices, perform joint scene perception of users based on wearable devices, establish scene perception results, control wearable devices to collect user indicators based on indicator attention, and establish a time series indicator data set; Configure the minimum granularity window, extract data features from the time series indicator data set based on the minimum granularity window, establish granularity feature extraction results, perform feature trend clustering on the granularity feature extraction results, and establish feature trend clustering results; Configure the indicator trigger channel through personal health records and scene perception results, use the indicator trigger channel to trigger the recognition of granular feature extraction results and feature trend clustering results, and establish the indicator trigger recognition results; The virtual data is updated synchronously based on the personal health records and indicator trigger identification results, and is visualized using the virtual data.

2. The method for dynamic management of personal health records combined with virtual technology as claimed in claim 1, characterized in that: The step of clustering the granularity feature extraction results into feature trends and establishing feature trend clustering results includes: Sorting the granular feature extraction results in time series, and evenly distributing a predetermined number of cluster centers based on the time series sorting; Taking the cluster center as the search center, perform temporal adjacent similarity search and establish the breakthrough coefficient mapped with the cluster size; Search clustering is performed based on time series adjacent similarity values ​​and breakthrough coefficients, and feature trend clustering is completed based on the search clustering results.

3. The method for dynamic management of personal health records in combination with virtual technology as claimed in claim 2, characterized in that: The searching and clustering based on the time-series adjacent similarity values ​​and the breakthrough coefficients of the clustering clusters includes: Based on the size of clusters, the breakthrough coefficients are matched, matching results are established, and breakthrough analysis is performed based on the matching results and time series adjacent similarity values; If the breakthrough analysis is a breakthrough result, the corresponding granularity feature extraction results are clustered into clusters, and the cluster search is continued; If the breakthrough analysis shows no breakthrough result, the clustering is stopped and a new instruction is generated; Based on the newly added instructions, a cluster center is added to the unclustered granular feature extraction results, and a cluster search is performed until the clustering of all granular feature extraction results is completed.

4. The method for dynamic management of personal health records in combination with virtual technology as claimed in claim 1, characterized in that: The configuration of the indicator trigger channel through the personal health profile and the scene perception result, and the trigger identification of the granular feature extraction result and the feature trend clustering result by the indicator trigger channel, include: Generate N static threshold ranges based on personal health records, and use the N static threshold ranges to build a static trigger mechanism; Based on the personal health record and the scene perception result, scene adaptation of N static threshold ranges is performed, M dynamic expansion thresholds are established, and a dynamic trigger mechanism is established with the M dynamic expansion thresholds, where M is less than or equal to N; The indicator trigger channel is configured with the static trigger mechanism and the dynamic trigger mechanism to complete trigger identification.

5. The method for dynamic management of personal health records in combination with virtual technology as claimed in claim 1, characterized in that: The method further comprises: Creating an indicator warning based on the indicator trigger identification result, generating warning feedback, and simultaneously establishing a verification window; If the real-time indicators in the verification window fail to meet the calibration threshold of the early warning feedback, a virtual hazard fit is established based on the real-time indicators; Visualize the virtual hazard fit.

6. The method for dynamic management of personal health records in combination with virtual technology as claimed in claim 1, characterized in that: The triggering identification of granular feature extraction results and feature trend clustering results using the indicator trigger channel includes: Establish an association rule analysis layer and couple the association rule analysis layer to the indicator trigger channel; When trigger identification is performed using the indicator trigger channel, the association rule analysis layer is synchronously called to perform time series correlation analysis of multiple indicators to generate association analysis results; The results of correlation analysis are used to establish joint triggering, and the indicator trigger identification results are compensated through joint triggering.

7. The method for dynamic management of personal health records in combination with virtual technology as claimed in claim 1, characterized in that: The establishing of the indicator triggering identification result includes: Establish a cumulative anomaly model, perform cumulative identification of trigger identification based on a multi-stage trigger mechanism based on the cumulative anomaly model, and establish a cumulative identification result; The indicator-triggered recognition results are compensated by accumulating the recognition results, and user early warning management is performed based on the compensated indicator-triggered recognition results.

8. A personal health record dynamic management device combined with virtual technology, characterized in that: The personal health record dynamic management device combined with virtual technology is used to execute the personal health record dynamic management method combined with virtual technology according to any one of claims 1 to 7, comprising: Personal health record reading module: after obtaining the user's authorization, read the user's personal health record, encrypt the personal health record and store it offline in the local storage space; File parsing module: calling the local identification channel, parsing the personal health file based on the local identification channel, and establishing indicator focus according to the parsing result; Perception and collection module: activates wearable devices, performs joint perception of user scenarios based on wearable devices, establishes scenario perception results, controls wearable devices to collect user metrics based on metrics attention, and establishes a time series metric data set; Feature trend clustering module: configure the minimum granularity window, extract data features from the time series indicator data set based on the minimum granularity window, establish granular feature extraction results, perform feature trend clustering on the granular feature extraction results, and establish feature trend clustering results; Trigger identification module: configure the indicator trigger channel through personal health records and scene perception results, use the indicator trigger channel to trigger the identification of granular feature extraction results and feature trend clustering results, and establish the indicator trigger identification results; Visualization presentation module: Synchronously update virtual data based on personal health records and indicator-triggered identification results, and use virtual data for visualization.

Images