Health risk assessment method, device, equipment, storage medium and program product

Abstract

The application discloses a health risk assessment method and device, equipment, a storage medium and a program product, and relates to the technical field of intelligent health, and comprises the following steps: collecting multi-source data of a monitoring object and extracting modal features corresponding to each data source from the multi-source data; performing weighted fusion of the modal features based on an attention mechanism to obtain feature representation input representing the comprehensive health state of the monitoring object, inputting the feature representation into a risk assessment model for assessment, and obtaining a health risk assessment result of the monitoring object in a current time period; the result represents whether the monitoring object has a health risk, and a risk probability, a risk type and a risk level when the health risk exists; the risk assessment model is constructed based on historical feature representation of a target group to which the monitoring object belongs, and is obtained by parameter fine-tuning of the monitoring object through an online learning mechanism. Through the attention mechanism, multi-source data is dynamically fused and personalized risk assessment is performed, so that real-time, accurate and personalized assessment of the health risk of the elderly is realized.

Description

Technical Field

[0001] This application relates to the field of intelligent health technology, and in particular to a health risk assessment method, device, equipment, storage medium, and program product. Background Technology

[0002] With the development of artificial intelligence technology in the field of health management, health risk assessment for the elderly has become an important issue. However, existing assessment methods generally suffer from drawbacks such as static assessment, coarse-grained early warning, and non-personalized services. First, traditional methods based on clinical consultations and physical examinations rely on periodic static clinical examinations, resulting in delayed assessments that depend primarily on user subjective statements, have limited data dimensions, and are difficult to capture dynamic health risks in a timely manner. Second, risk prediction models built on single clinical data suffer from problems such as one-sided data sources, long update cycles, and strong generalization of the population, making it difficult to achieve personalized real-time risk responses. In addition, while health monitoring based on wearable devices can continuously collect users' physiological data, it creates "data silos," limiting health risk assessments to superficial threshold alerts and lacking fusion analysis with other dimensions of data, leading to insufficient early warning accuracy. The root cause of these problems lies in the fact that existing health risk assessment methods have failed to effectively integrate and dynamically analyze multi-dimensional health-related data, resulting in delayed, inaccurate, and untargeted assessments.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide a health risk assessment method, device, equipment, storage medium, and program product, which aims to solve the technical problems of existing health risk assessment methods failing to effectively integrate and dynamically analyze multi-dimensional health-related data, resulting in assessments that are lagging, inaccurate, and lack specificity.

[0005] To achieve the above objectives, this application proposes a health risk assessment method, which includes: Collect multi-source data of the monitored object, and extract modal features corresponding to each data source from the multi-source data; Based on the attention mechanism, the modal features are weighted and fused to obtain a feature representation that characterizes the comprehensive health status of the monitored object; The feature representation is input into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored object in the current time period. The health risk assessment results characterize whether the monitored object has a health risk, and the probability, type and level of the risk when a health risk exists; the risk assessment model is constructed based on the historical characteristics of the target group to which the monitored object belongs, and is obtained by fine-tuning the parameters for the monitored object through an online learning mechanism.

[0006] In one embodiment, the multi-source data includes behavioral data, physiological data, and environmental data; the step of extracting modal features corresponding to each data source from the multi-source data includes: Physiological modal features characterizing the health status of the monitored subjects are extracted from the physiological data; the physiological modal features include at least heart rate variability, blood pressure rhythm, and sleep structure stages. The behavioral modal features of the monitored object are extracted from the behavioral data using a pre-trained behavior recognition algorithm; the behavioral modal features include at least the activity level change trend, behavior regularity index, and gait stability index. The environmental modal characteristics of the monitored object are calculated based on the environmental data; the environmental modal characteristics include at least the environmental comfort index and the degree of environmental variability.

[0007] In one embodiment, the step of weightedly fusing the modal features based on an attention mechanism to obtain a feature representation characterizing the comprehensive health status of the monitored object includes: Based on the attention mechanism, the attention weights of modal features of each data source are calculated; By weighting and fusing modal features from different data sources using the attention weights, a feature representation of the comprehensive health status of the monitored object is obtained.

[0008] In one embodiment, the feature is represented as a time-series feature, and the risk assessment model establishes a personalized baseline based on the historical data of the monitored object; the step of inputting the feature representation into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored object in the current time period includes: The feature representation is input into a pre-trained risk assessment model, and a spatiotemporal analysis of the feature representation is performed using a sliding time window. The feature change trend within the current time window is calculated to obtain the preliminary assessment result of the monitored object in the current time period corresponding to the current time window. If the preliminary assessment results indicate that the monitored subject has a health risk in the current time period, the risk probability and risk level of the health risk are determined based on the personalized baseline to obtain the health risk assessment result of the monitored subject in the current time period.

[0009] In one embodiment, after the step of inputting the feature representation into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored object in the current time period, the method further includes: If the health risk assessment results indicate that the monitored object has a health risk, early warning parameters are determined based on the health risk assessment results; the early warning parameters include early warning level and early warning method. Based on the aforementioned warning parameters, risk warnings are issued to the monitored objects. Based on the risk warning prompts, feedback data is obtained regarding the health risk assessment results; The risk assessment model is then fine-tuned and optimized using the feedback data.

[0010] In one embodiment, the multi-source data is heterogeneous data; before the step of extracting modal features corresponding to each data source from the multi-source data, the method further includes: The multi-source data is subjected to format conversion and standardization processing to obtain standardized data in a preset target format; The standardized data is then sequentially cleaned, denoised, and normalized to obtain normalized data. The normalized data is time-aligned according to the data acquisition time to form time-series data in the target format.

[0011] Furthermore, to achieve the above objectives, this application also proposes a health risk assessment device, which includes: The data processing module is used to collect multi-source data of the monitored object and extract modal features corresponding to each data source from the multi-source data; The feature fusion module is used to perform weighted fusion of the modal features based on an attention mechanism to obtain a feature representation that characterizes the comprehensive health status of the monitored object; The risk assessment module is used to input the feature representation into a pre-trained risk assessment model for evaluation, and obtain the health risk assessment result of the monitored object in the current time period. The health risk assessment results characterize whether the monitored object has a health risk, and the probability, type and level of the risk when a health risk exists; the risk assessment model is constructed based on the historical characteristics of the target group to which the monitored object belongs, and is obtained by fine-tuning the parameters for the monitored object through an online learning mechanism.

[0012] In addition, to achieve the above objectives, this application also proposes a health risk assessment device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the health risk assessment method as described above.

[0013] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and which, when executed by a processor, implements the steps of the health risk assessment method described above.

[0014] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the health risk assessment method described above.

[0015] One or more technical solutions proposed in this application have at least the following technical effects: By collecting multi-source data from monitored subjects and extracting their corresponding modal features, comprehensive health-related information from multiple sources can be utilized to provide a comprehensive data foundation for health assessment, thus solving the problem of one-sided or incomplete assessments from single data sources. Furthermore, by using an attention mechanism to weightedly fuse the extracted modal features, dynamic and adaptive weight allocation of features from different data sources can be achieved, effectively integrating the inherent correlations of multi-source data to obtain a feature representation of the monitored subjects' overall health status. This improves the generalization of the risk assessment model and the accuracy of the assessment results, achieving a precise characterization of an individual's overall health status. Based on this, the risk assessment model is used to evaluate the monitored subjects' health risk in the current time period, enabling real-time or near-real-time calculation of health risk for the current time period, transforming the assessment from static to dynamic and reducing assessment lag. Moreover, the risk assessment model is constructed based on the historical feature representation of the target group to which the monitored subjects belong, and is obtained through online learning mechanisms for parameter fine-tuning for the monitored subjects. This allows the risk assessment model to establish a baseline based on general group characteristics and to make personalized adaptations and adjustments based on the historical data of specific monitored subjects, thereby achieving dynamic, accurate, and personalized assessment of the health risks of monitored subjects. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the health risk assessment method of this application in Implementation Example 1. Figure 2 This is a flowchart illustrating Embodiment 2 of the health risk assessment method of this application; Figure 3 This is a flowchart illustrating Embodiment 3 of the health risk assessment method of this application; Figure 4 This is a schematic diagram of the module structure of the health risk assessment device according to an embodiment of this application; Figure 5 This is a schematic diagram of the device structure of the hardware operating environment involved in the health risk assessment method in the embodiments of this application.

[0019] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0021] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0022] The main solution of this application embodiment is: relying on the Internet of Things smart elderly care platform, by integrating multi-source heterogeneous data such as wearable devices, environmental sensors, behavior monitoring systems and medical databases, to achieve real-time, dynamic and quantitative assessment of the health risks of the elderly. It is applicable to health management scenarios of community elderly care, home-based elderly care and institutional elderly care, and provides decision support for preventing risks such as falls, chronic disease exacerbation and cognitive impairment.

[0023] It should be noted that the multi-source data mentioned in the various embodiments of this application, especially the data related to users (especially elderly people or senior citizens), were all obtained in a compliant manner with the authorization of the users, and data security was protected by means of encryption and desensitization.

[0024] Because existing technologies for assessing health risks for users, especially the elderly, suffer from one-sided and isolated data sources, they cannot effectively integrate and analyze isolated data, resulting in assessments that are delayed, inaccurate, and lack specificity.

[0025] Specifically, this is manifested in: Static assessment: The assessment behavior is discrete and periodic, which makes it impossible to achieve long-term continuous monitoring and risk warning, and there is a lag in the detection of acute events such as falls and sudden illnesses.

[0026] Coarse-grained early warning: Risk warnings are mostly based on a single indicator or simple rules, making it difficult to balance sensitivity and specificity. They cannot identify complex risks formed by the superposition of multiple weak signals and lack insight into the evolution trend of risks.

[0027] The service lacks personalization: the assessment model is highly general but lacks personalization, making it difficult to provide customized risk warnings and intervention recommendations for different elderly people based on their underlying diseases, lifestyle habits and genetic backgrounds.

[0028] This application provides a solution that fully utilizes IoT, big data and artificial intelligence technologies to provide a health risk assessment method based on multi-source data fusion. This method can break down data silos, achieve dynamic and continuous assessment of health risks for the elderly, and has accurate early warning capabilities.

[0029] Specifically, the system systematically integrates multi-source data, including clinical data, real-time physiological data, daily behavioral data, environmental perception data, and subjective report data, to construct a comprehensive digital profile of an individual's health. Through real-time data stream processing technology, it transforms traditional "static snapshot" assessments into dynamic, continuous assessments, enabling early detection and early warning of health risks. By utilizing machine learning, deep learning, and AI models, it mines complex nonlinear relationships from multi-dimensional, time-series data to identify potential health risk patterns, improving the accuracy and foresight of early warnings. Based on the dynamic assessment results, it generates personalized health improvement suggestions and risk intervention measures, and pushes early warning information in real time, forming a closed-loop health management service.

[0030] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or health risk assessment device capable of performing the above functions. The following description uses a health risk assessment device as an example to illustrate this embodiment and the subsequent embodiments.

[0031] This application provides a health risk assessment method applied to an IoT-based smart elderly care platform. Targeting elderly care scenarios such as institutional care, it constructs a closed-loop process of data collection, dynamic cleaning, feature extraction, risk assessment, and feedback optimization to achieve dynamic, accurate, and personalized assessment of the health risks of elderly individuals.

[0032] Specifically, refer to Figure 1 , Figure 1This is a flowchart illustrating the first embodiment of the health risk assessment method of this application.

[0033] In this embodiment, the health risk assessment method includes steps S10 to S30: Step S10: Collect multi-source data of the monitored object, and extract modal features corresponding to each data source from the multi-source data; Step S20: Based on the attention mechanism, the modal features are weighted and fused to obtain a feature representation of the comprehensive health status of the monitored object; Step S30: Input the feature representation into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored object in the current time period; The health risk assessment results characterize whether the monitored object has a health risk, and the probability, type and level of the risk when a health risk exists; the risk assessment model is constructed based on the historical characteristics of the target group to which the monitored object belongs, and is obtained by fine-tuning the parameters for the monitored object through an online learning mechanism.

[0034] Multi-source data is collected from the monitored subjects, and modal features corresponding to each data source are extracted from the collected multi-source data. It should be noted that the monitored subjects include, but are not limited to, elderly people, and the collected multi-source data includes data from different dimensions.

[0035] In one embodiment, the multi-source data is collected through various sensor nodes built into the IoT smart elderly care platform. The multi-source data can be multimodal data, with different data sources corresponding to different modes. Feature extraction is performed on the collected multi-source data to obtain the modal features under the corresponding modes of each data source.

[0036] Furthermore, based on the attention mechanism, the extracted modal features are weighted and fused to obtain a feature representation that characterizes the comprehensive health status of the monitored object.

[0037] In one embodiment, the feature representation used to characterize the overall health status of the monitored object is used to create a digital health profile of the monitored object.

[0038] The fused feature representations are input into a pre-trained risk assessment model for evaluation, yielding a health risk assessment result for the monitored object in the current time period. This health risk assessment result characterizes whether the monitored object faces a health risk, and if so, the risk parameters, including risk probability, risk type, and risk level.

[0039] Furthermore, the risk assessment model is constructed based on the historical characteristics of the target group to which the monitored object belongs, and is obtained by fine-tuning the parameters for the monitored object through an online learning mechanism.

[0040] In one embodiment, the risk assessment model is built based on machine learning, deep learning, or AI models. A sample dataset is constructed using the historical characteristics of the target group to which the monitored object belongs for model training, enabling the model to learn general risk patterns. When applied to a specific monitored object, the model's internal parameters or adaptation layers are quickly fine-tuned using a small-sample adaptation framework and the individual historical characteristics of the monitored object. This allows the model to rapidly capture the normal pattern baseline of the individual monitored object, thereby achieving deep personalization to adapt to individual health changes in the monitored object.

[0041] Subsequently, through an online learning mechanism, new monitoring data and user feedback (such as false alarm confirmations and actual events) are continuously used to incrementally fine-tune the personalized model parameters of the monitored objects, enabling them to adapt to the long-term evolution of the monitored objects' health status.

[0042] For health risk assessment results, if the monitored subject has health risks, and at least one of the risk probability and risk type is different, then the risk level is different.

[0043] In this embodiment, by collecting multi-source data of the monitored objects and extracting their corresponding modal features, multi-source health-related information can be comprehensively utilized to provide a comprehensive data foundation for health assessment, thereby solving the problem of one-sided or incomplete assessment from a single data source. Furthermore, by weighted fusion of the extracted modal features based on an attention mechanism, dynamic and adaptive weight allocation of features from different data sources can be achieved, effectively integrating the inherent correlations of multi-source data to obtain a feature representation of the monitored object's overall health status. This improves the generalization of the risk assessment model and the accuracy of the assessment results, achieving a precise characterization of an individual's overall health status. Based on this, the risk assessment model is used to evaluate the monitored object's health risk in the current time period, enabling real-time or near-real-time calculation of health risk within the current time period, transforming the assessment from static to dynamic and reducing assessment lag. Moreover, the risk assessment model is constructed based on the historical feature representation of the target group to which the monitored object belongs, and is obtained through online learning mechanisms for parameter fine-tuning for the monitored object. This allows the risk assessment model to establish a baseline based on general group characteristics and to perform personalized adaptation and adjustment based on the historical data of specific monitored objects, thereby achieving dynamic, accurate, and personalized assessment of the monitored object's health risk.

[0044] In one feasible implementation, the collected multi-source data includes behavioral data, physiological data, and environmental data. Physiological data can be collected through wearable devices and authorized historical clinical data from electronic health records. Environmental data is captured by environmental sensors. Behavioral data can be the interaction data between the monitored object and smart home devices, as well as behavioral data generated by its daily activities. The behavioral data of the monitored object can indirectly reflect its work-rest patterns, activity levels, and cognitive status.

[0045] The modal features extracted from multi-source data include at least physiological modal features, behavioral modal features, and environmental modal features. Based on this, step S10 may include steps S11-S13: Step S11: Extract physiological modal features from the physiological data to characterize the health status of the monitored object; the physiological modal features include at least heart rate variability, blood pressure rhythm, and sleep structure stages; Step S12: Extract the behavioral modal features of the monitored object from the behavioral data using a pre-trained behavior recognition algorithm; the behavioral modal features include at least the activity level change trend, behavior regularity index, and gait stability index. Step S13: Calculate the environmental modal characteristics of the monitored object based on the environmental data; the environmental modal characteristics include at least the environmental comfort index and the degree of environmental variability.

[0046] Physiological modal features for characterizing the health status of monitored subjects are extracted from physiological data from multiple sources. These physiological modal features include at least heart rate variability, blood pressure rhythm, and sleep structure stages.

[0047] Using a pre-trained behavior recognition algorithm, behavioral modal features of the monitored object are extracted from behavioral data from multiple sources. These behavioral modal features include at least the trend of activity level changes, behavioral regularity index, and gait stability index.

[0048] Environmental modal characteristics of the monitored object are calculated based on multi-source data. These environmental modal characteristics include at least the environmental comfort index and the degree of environmental variability.

[0049] In one embodiment, the physiological data of the monitored object includes at least heart rate, blood pressure, blood oxygen saturation, body temperature, and sleep data; the behavioral data of the monitored object includes activity trajectory, gait data, fall detection data, and daily activity frequency, which can be collected by positioning sensors, infrared sensors, millimeter-wave radar, and door magnetic sensors within the area accessible to the monitored object; the environmental data of the monitored object includes temperature, humidity, air quality, and light intensity of the area where the monitored object is located.

[0050] Modal features corresponding to each data source are extracted from multi-source data, including but not limited to calculating heart rate variability and gait stability indices. For example, the phase and amplitude features of the circadian rhythm of the monitored subject are extracted from continuous blood pressure data, the distribution entropy of the daytime activity hotspot area of ​​the monitored subject is extracted from indoor positioning trajectory, or acoustic and semantic features such as speech rate and response delay of the monitored subject are extracted from social voice data, thereby constructing modal features that characterize different aspects of the monitored subject's physiology, behavior, cognition, etc., and combining them into a multimodal feature vector.

[0051] In one embodiment, the collected multi-source data is heterogeneous data. Before feature extraction, the collected multi-source heterogeneous data can be preprocessed. Specifically, refer to... Figure 2 , Figure 2 Another flowchart illustrating the health risk assessment method provided in this application embodiment is shown below. Figure 2 As shown, after collecting multi-source data from the monitored objects, the collected multi-source data is preprocessed to obtain corresponding time-series data. Modal features corresponding to each data source are extracted from this time-series data. Then, based on an attention mechanism, the extracted modal features are weighted and fused to obtain a feature representation characterizing the comprehensive health status of the monitored objects; this feature representation is a time-series feature. Finally, the obtained feature representation is input into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored objects in the current time period.

[0052] Furthermore, the preprocessing of multi-source data may include steps S101 to S103: Step S101: Perform format conversion and standardization processing on the multi-source data to obtain standardized data in a preset target format; Step S102: The standardized data is sequentially cleaned, denoised, and normalized to obtain normalized data; Step S103: The normalized data is time-aligned according to the data acquisition time to form time-series data in the target format.

[0053] First, the collected multi-source heterogeneous data undergoes format conversion and standardization post-processing to obtain standardized data in the preset target format. This standardized data is then sequentially processed for denoising, denoising, and normalization to obtain normalized data. Finally, the normalized data is time-aligned according to the data acquisition time to form time-series data in the target format. Feature extraction is performed based on this time-series data to obtain modal features corresponding to each data source; each modal feature is a time-series feature.

[0054] As one embodiment, the fusion of modal features is based on an attention mechanism, calculating the weights of each modal feature, and then weighting the modal features based on the calculated weights to obtain the final feature representation. Therefore, step S20 may include steps S21-S22: Step S21: Based on the attention mechanism, calculate the attention weights of the modal features of each data source; Step S22: The modal features of different data sources are weighted and fused using the attention weights to obtain a feature representation of the comprehensive health status of the monitored object.

[0055] Based on the attention mechanism, the attention weights of the modal features corresponding to each data source are calculated. The modal features of different data sources are then weighted and fused using these attention weights to obtain a feature representation of the comprehensive health status of the monitored object.

[0056] Calculating the attention weights of each modality feature is to dynamically evaluate the importance of different data sources in the overall health status of the monitored object in the current event. Unlike simple averaging or fixed weight fusion, the attention mechanism enables the model to adaptively allocate attention based on the feature context of the current input.

[0057] In one implementation, multiple modal feature vectors extracted from physiological, behavioral, and environmental data can be concatenated or stacked to form a multimodal feature matrix. This matrix is ​​then input into an attention computation network, which calculates an attention score distribution, where each score corresponds to a specific modal feature (e.g., heart rate variability, gait stability index, environmental comfort index, etc.). This score is the attention weight.

[0058] As an example, the attention weights corresponding to features of the same modality may differ for different health risks. For instance, when detecting potential fall risk, behavioral modality features such as gait stability indicators and sudden drop in nighttime activity may be assigned higher weights; while when assessing the stable state of chronic diseases, physiological modality features such as heart rate variability and blood pressure diurnal rhythm may be more critical.

[0059] Modal features from the same data source can have multiple attention weights for different types of health risks. Based on the attention weights under the same risk type, different modal features are weighted and fused to obtain multiple corresponding feature representations, which are used to monitor different types of health risks.

[0060] The feature representation obtained through attention fusion not only includes all the original modal information of the multi-source data, but also emphasizes the features most relevant to the current health risk through attention weights, suppressing the interference of irrelevant or noisy features, thereby significantly improving the accuracy and interpretability of subsequent risk assessment.

[0061] In one feasible implementation, the features obtained through attention mechanism fusion are represented as time-series features. The risk assessment model establishes a personalized baseline based on the historical data of the monitored object. On this basis, step S30 may include steps S31-S32: Step S31: Input the feature representation into the pre-trained risk assessment model, perform spatiotemporal analysis on the feature representation using a sliding time window, calculate the feature change trend within the current time window, and obtain the preliminary assessment result of the monitored object in the current time period corresponding to the current time window. Step S32: If the preliminary assessment results indicate that the monitored object has a health risk in the current time period, the risk probability and risk level of the health risk are determined based on the personalized baseline to obtain the health risk assessment result of the monitored object in the current time period.

[0062] The feature representations characterizing the overall health status of the monitored subjects are input into a pre-trained risk assessment model. The model uses a sliding time window approach to analyze the input feature stream. For example, a 10-minute time window is set, and as new features are continuously input, the time window slides forward on the time axis. The model performs spatiotemporal analysis on all time-series features within the current window, calculating the changing trends of features relative to their recent history, such as trend slope, abrupt change points, and pattern anomalies. Through windowed spatiotemporal analysis, the model can capture short-term health fluctuations and potential risk precursors. Based on the analysis of feature changing trends within the current time window, a preliminary assessment result is obtained, which is mainly used to preliminarily determine whether the monitored subject has a health risk in the current time period.

[0063] Furthermore, if the preliminary assessment indicates that the monitored subjects face health risks in the current time period—that is, if the trend of characteristic changes is significantly abnormal compared to the personalized baseline—the model further determines the risk type, probability, and level based on the personalized baseline. Specifically, the model calculates the deviation of the current abnormal characteristics from the personalized baseline and, combined with prior knowledge of the risk type, quantifies the probability of the risk occurring, such as a 75% probability of a fall. Simultaneously, based on the level of risk probability, the severity of the abnormal characteristics, and their duration, the risk is categorized into different levels, such as low, medium, and high. The final judgment combining the risk probability and risk level constitutes a complete health risk assessment result for the monitored subjects in the current time period.

[0064] Among them, the features representing the overall health status of the monitored objects are represented as time-series features. This means that the features representing the overall health status extracted and fused from multi-source data are not isolated instantaneous snapshots, but a series of feature vectors that are continuously generated over time. These feature vectors have order and correlation in the time dimension and can reflect the physiological rhythms, behavioral habits and status change trends of the monitored objects, providing a basis for dynamic risk assessment.

[0065] For personalized baselines, during the model deployment or initialization phase for a specific monitored object, historical time-series characteristic data of the monitored object under normal and stable conditions over a period of time (e.g., several weeks) are collected. Based on the time-series characteristic data, the model learns and establishes a personalized health baseline that reflects the normal pattern of the monitored object. This health baseline can be a statistical distribution of characteristics, such as mean, variance, periodic pattern, or the feature space boundary under normal conditions, serving as an important reference for subsequent judgment of whether the real-time state of the monitored object deviates from normal and the degree of deviation.

[0066] In one embodiment, after obtaining the health risk assessment results, a tiered warning can be triggered based on the detected health risks. (See reference...) Figure 3 , Figure 3 This is another flowchart illustrating the health risk assessment method provided in the embodiments of this application, such as... Figure 3 As shown in step S40, after obtaining the health risk assessment result for the monitored object in the current time period, if the health risk assessment result indicates that the monitored object has a health risk, then a health risk warning is issued to the monitored object. Based on this, after step S30, step S40 is also included, and step S40 may further include steps S41 to S44: Step S41: If the health risk assessment result indicates that the monitored object has a health risk, determine the early warning parameters based on the health risk assessment result; the early warning parameters include the early warning level and the early warning method. Step S42: Based on the warning parameters, issue a risk warning to the monitored object; Step S43: Obtain feedback data on the health risk assessment results based on the risk warning prompt; Step S44: Fine-tune and optimize the risk assessment model using the feedback data.

[0067] When a health risk assessment indicates that a monitored individual faces a health risk, early warning parameters are determined based on the detailed content of the assessment results. These parameters include the early warning level and the early warning method, serving as metadata that triggers and guides early warning actions. They are determined according to the specific risk level (e.g., high, medium, low), risk type (e.g., fall, chronic disease exacerbation, cognitive impairment), the assessed risk probability, and the confidence level or urgency of the risk event. These parameters are the direct basis for implementing tiered early warning systems. If at least one of the risk probability, risk type, or risk level differs, the corresponding early warning parameters will differ, leading to the use of different early warning levels and / or methods for subsequent risk alerts.

[0068] After determining the warning parameters, risk warnings are issued to the monitored objects based on these parameters, achieving graded and personalized warnings. As one example, a graded warning mechanism is adopted. For instance, when the warning parameters indicate "low risk," only one warning log may be generated on the backend management platform; when the indication is medium risk, a notification may be pushed to the mobile application of caregivers or family members; when the indication is high risk and persists for a period of time (e.g., 5 minutes), a composite warning action is triggered, including platform push notifications, emergency notifications to family members, and even automatic initiation of linkage with the community emergency center. The selection of warning channels and methods is dynamically determined based on the warning parameters.

[0069] After issuing a risk warning, feedback data on the health risk assessment results is actively or passively acquired. This feedback data is crucial for optimizing the risk assessment model. Its sources are diverse and can include: user (or caregiver) confirmation of the warning information (e.g., clicking "received," "false alarm"), records of actual interventions taken after the warning (e.g., family members calling to inquire, community staff conducting home visits), records of subsequent verifiable health events (e.g., a fall confirmed by sensors, subsequent medical diagnosis), and subsequent physiological and behavioral data of the monitored subject for a period after the warning. This feedback data connects the predicted assessment results with reality, forming the truth labels for model optimization.

[0070] Finally, the acquired feedback data is used to fine-tune and optimize the risk assessment model, achieving incremental fine-tuning based on an online learning mechanism. For example, if the feedback data confirms a "false alarm," the feature representation that triggered the false alarm, along with the current risk label, can be used as new training samples to fine-tune the model parameters, reducing the probability of misjudging similar patterns in the future. Conversely, if the feedback data confirms a "missed report" or "effective warning," the corresponding samples can be used to enhance the model's ability to identify this type of health risk. Through a continuous feedback-optimization cycle, the risk assessment model can continuously adapt to short-term fluctuations and long-term changes in the health status of the monitored subjects, such as aging processes and chronic disease progression, and improve the accuracy of identifying individual-specific risk patterns, thereby ultimately achieving a continuous improvement in assessment effectiveness.

[0071] Understandably, risk assessment models and their personalized baselines can be dynamically optimized and continuously updated over time to adapt to changes in the health status of older adults.

[0072] In this embodiment, a fundamental improvement in health risk assessment for the elderly is achieved by constructing a complete technical closed loop of "multi-source data acquisition - dynamic cleaning and feature extraction - attention-based modal feature fusion - risk assessment based on sliding window and personalized baseline - early warning feedback optimization".

[0073] Specifically, by systematically integrating multi-source heterogeneous data from various sources such as wearable devices, environmental sensors, and behavioral monitoring systems, and using an attention mechanism for dynamic weighted fusion, a comprehensive and accurate digital profile of personal health is constructed. This effectively solves the problems of biased assessment and insufficient early warning accuracy caused by single data dimensions and data silos. By employing a sliding time window to analyze time-series characteristics in real time, and combining this with a risk assessment model to establish a personalized baseline for each monitored individual, the assessment is transformed from traditional static, periodic snapshots into continuous, dynamic, and ongoing monitoring. This significantly improves the timeliness of risk detection and can capture the progressive deterioration of the health status of the elderly and the precursor signals of acute events.

[0074] Furthermore, the feedback optimization closed loop of the hierarchical early warning and online learning mechanism enables the risk assessment model to be continuously fine-tuned and evolved based on early warning feedback and the latest multi-source data, thereby continuously enhancing the personalization and foresight of the assessment. Ultimately, this achieves a shift from extensive and lagging general assessment to precise, real-time, and personalized proactive health management, thereby improving the quality of care for the elderly.

[0075] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the health risk assessment method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0076] This application also provides a health risk assessment device, please refer to... Figure 4 The health risk assessment device includes: Data processing module 10 is used to collect multi-source data of the monitored object and extract modal features corresponding to each data source from the multi-source data; The feature fusion module 20 is used to perform weighted fusion of the modal features based on an attention mechanism to obtain a feature representation of the comprehensive health status of the monitored object; Risk assessment module 30 is used to input the feature representation into a pre-trained risk assessment model for evaluation, and obtain the health risk assessment result of the monitored object in the current time period; The health risk assessment results characterize whether the monitored object has a health risk, and the probability, type and level of the risk when a health risk exists; the risk assessment model is constructed based on the historical characteristics of the target group to which the monitored object belongs, and is obtained by fine-tuning the parameters for the monitored object through an online learning mechanism.

[0077] In one embodiment, the multi-source data includes behavioral data, physiological data, and environmental data; the data processing module 10 is further used for... Physiological modal features characterizing the health status of the monitored subjects are extracted from the physiological data; the physiological modal features include at least heart rate variability, blood pressure rhythm, and sleep structure stages. The behavioral modal features of the monitored object are extracted from the behavioral data using a pre-trained behavior recognition algorithm; the behavioral modal features include at least the activity level change trend, behavior regularity index, and gait stability index. The environmental modal characteristics of the monitored object are calculated based on the environmental data; the environmental modal characteristics include at least the environmental comfort index and the degree of environmental variability.

[0078] In one embodiment, the feature fusion module 20 is further configured to: Based on the attention mechanism, the attention weights of modal features of each data source are calculated; By weighting and fusing modal features from different data sources using the attention weights, a feature representation of the comprehensive health status of the monitored object is obtained.

[0079] In one embodiment, the feature is represented as a time-series feature, and the risk assessment model establishes a personalized baseline based on the historical data of the monitored object; the risk assessment module 30 is further configured to: The feature representation is input into a pre-trained risk assessment model, and a spatiotemporal analysis of the feature representation is performed using a sliding time window. The feature change trend within the current time window is calculated to obtain the preliminary assessment result of the monitored object in the current time period corresponding to the current time window. If the preliminary assessment results indicate that the monitored subject has a health risk in the current time period, the risk probability and risk level of the health risk are determined based on the personalized baseline to obtain the health risk assessment result of the monitored subject in the current time period.

[0080] In one embodiment, the health risk assessment device further includes an early warning optimization module, used for: If the health risk assessment results indicate that the monitored object has a health risk, early warning parameters are determined based on the health risk assessment results; the early warning parameters include early warning level and early warning method. Based on the aforementioned warning parameters, risk warnings are issued to the monitored objects. Based on the risk warning prompts, feedback data is obtained regarding the health risk assessment results; The risk assessment model is then fine-tuned and optimized using the feedback data.

[0081] In one embodiment, the multi-source data is heterogeneous data; the data processing module 10 is further configured to: The multi-source data is subjected to format conversion and standardization processing to obtain standardized data in a preset target format; The standardized data is then sequentially cleaned, denoised, and normalized to obtain normalized data. The normalized data is time-aligned according to the data acquisition time to form time-series data in the target format.

[0082] The health risk assessment device provided in this application, employing the health risk assessment method described in the above embodiments, addresses the technical problem that existing health risk assessment methods fail to effectively integrate and dynamically analyze multi-dimensional health-related data, resulting in assessments that are delayed, inaccurate, and lack specificity. Compared to the prior art, the beneficial effects of the health risk assessment device provided in this application are the same as those of the health risk assessment method provided in the above embodiments, and other technical features of the health risk assessment device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0083] This application provides a health risk assessment device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the health risk assessment method in Embodiment 1 above.

[0084] The following is for reference. Figure 5 The diagram illustrates a structural schematic suitable for implementing the health risk assessment device in the embodiments of this application. The health risk assessment device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The health risk assessment device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0085] like Figure 5As shown, the health risk assessment device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the health risk assessment device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the health risk assessment device to communicate wirelessly or wiredly with other devices to exchange data. While the figure shows health risk assessment devices with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0086] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0087] The health risk assessment device provided in this application, employing the health risk assessment method described in the above embodiments, addresses the technical problem that existing health risk assessment methods fail to effectively integrate and dynamically analyze multi-dimensional health-related data, resulting in assessments that are delayed, inaccurate, and lack specificity. Compared to existing technologies, the beneficial effects of the health risk assessment device provided in this application are the same as those of the health risk assessment method provided in the above embodiments, and other technical features of this health risk assessment device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0088] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0089] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0090] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to perform the health risk assessment method described in the above embodiments.

[0091] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0092] The aforementioned computer-readable storage medium may be included in the health risk assessment device; or it may exist independently and not incorporated into the health risk assessment device.

[0093] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the health risk assessment device, cause the health risk assessment device to: Collect multi-source data of the monitored object, and extract modal features corresponding to each data source from the multi-source data; Based on the attention mechanism, the modal features are weighted and fused to obtain a feature representation that characterizes the comprehensive health status of the monitored object; The feature representation is input into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored object in the current time period. The health risk assessment results characterize whether the monitored object has a health risk, and the probability, type and level of the risk when a health risk exists; the risk assessment model is constructed based on the historical characteristics of the target group to which the monitored object belongs, and is obtained by fine-tuning the parameters for the monitored object through an online learning mechanism.

[0094] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0095] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0096] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0097] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described health risk assessment method. This addresses the technical problem that existing health risk assessment methods fail to effectively integrate and dynamically analyze multi-dimensional health-related data, resulting in assessments that are delayed, inaccurate, and lack specificity. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the health risk assessment method provided in the above embodiments, and will not be elaborated upon here.

[0098] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the health risk assessment method described above.

[0099] The computer program product provided in this application addresses the technical problem that existing health risk assessment methods fail to effectively integrate and dynamically analyze multi-dimensional health-related data, resulting in assessments that are delayed, inaccurate, and lack specificity. Compared to existing technologies, the beneficial effects of the computer program product provided in this application are the same as those of the health risk assessment methods provided in the above embodiments, and will not be elaborated upon here.

[0100] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A health risk assessment method, characterized in that, The health risk assessment methods include: Collect multi-source data of the monitored object, and extract modal features corresponding to each data source from the multi-source data; Based on the attention mechanism, the modal features are weighted and fused to obtain a feature representation that characterizes the comprehensive health status of the monitored object; The feature representation is input into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored object in the current time period. The health risk assessment results characterize whether the monitored object has a health risk, and the probability, type and level of the risk when a health risk exists; the risk assessment model is constructed based on the historical characteristics of the target group to which the monitored object belongs, and is obtained by fine-tuning the parameters for the monitored object through an online learning mechanism.

2. The health risk assessment method as described in claim 1, characterized in that, The multi-source data includes behavioral data, physiological data, and environmental data; the step of extracting modal features corresponding to each data source from the multi-source data includes: Physiological modal features characterizing the health status of the monitored subjects are extracted from the physiological data; the physiological modal features include at least heart rate variability, blood pressure rhythm, and sleep structure stages. The behavioral modal features of the monitored object are extracted from the behavioral data using a pre-trained behavior recognition algorithm; the behavioral modal features include at least the activity level change trend, behavior regularity index, and gait stability index. The environmental modal characteristics of the monitored object are calculated based on the environmental data; the environmental modal characteristics include at least the environmental comfort index and the degree of environmental variability.

3. The health risk assessment method as described in claim 1, characterized in that, The step of weighted fusion of the modal features based on an attention mechanism to obtain a feature representation of the comprehensive health status of the monitored object includes: Based on the attention mechanism, the attention weights of modal features of each data source are calculated; By weighting and fusing modal features from different data sources using the attention weights, a feature representation of the comprehensive health status of the monitored object is obtained.

4. The health risk assessment method as described in claim 1, characterized in that, The features are represented as time-series features, and the risk assessment model establishes a personalized baseline based on the historical data of the monitored object. The step of inputting the feature representation into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored object in the current time period includes: The feature representation is input into a pre-trained risk assessment model, and a spatiotemporal analysis of the feature representation is performed using a sliding time window. The feature change trend within the current time window is calculated to obtain the preliminary assessment result of the monitored object in the current time period corresponding to the current time window. If the preliminary assessment results indicate that the monitored subject has a health risk in the current time period, the risk probability and risk level of the health risk are determined based on the personalized baseline to obtain the health risk assessment result of the monitored subject in the current time period.

5. The health risk assessment method according to any one of claims 1 to 4, characterized in that, After the step of inputting the feature representation into a pre-trained risk assessment model for evaluation to obtain the health risk assessment result of the monitored object in the current time period, the method further includes: If the health risk assessment results indicate that the monitored object has a health risk, early warning parameters are determined based on the health risk assessment results; the early warning parameters include early warning level and early warning method. Based on the aforementioned warning parameters, risk warnings are issued to the monitored objects. Based on the risk warning prompts, feedback data is obtained regarding the health risk assessment results; The risk assessment model is then fine-tuned and optimized using the feedback data.

6. The health risk assessment method according to any one of claims 1 to 4, characterized in that, The multi-source data is heterogeneous data; before the step of extracting modal features corresponding to each data source from the multi-source data, the method further includes: The multi-source data is subjected to format conversion and standardization processing to obtain standardized data in a preset target format; The standardized data is then sequentially cleaned, denoised, and normalized to obtain normalized data. The normalized data is time-aligned according to the data acquisition time to form time-series data in the target format.

7. A health risk assessment device, characterized in that, The health risk assessment device includes: The data processing module is used to collect multi-source data of the monitored object and extract modal features corresponding to each data source from the multi-source data; The feature fusion module is used to perform weighted fusion of the modal features based on an attention mechanism to obtain a feature representation that characterizes the comprehensive health status of the monitored object; The risk assessment module is used to input the feature representation into a pre-trained risk assessment model for evaluation, and obtain the health risk assessment result of the monitored object in the current time period. The health risk assessment results characterize whether the monitored object has a health risk, and the probability, type and level of the risk when a health risk exists; the risk assessment model is constructed based on the historical characteristics of the target group to which the monitored object belongs, and is obtained by fine-tuning the parameters for the monitored object through an online learning mechanism.

8. A health risk assessment device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the health risk assessment method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the health risk assessment method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the health risk assessment method as described in any one of claims 1 to 6.

Images