Energy metabolism disease health data processing method, device and equipment and storage medium

Abstract

This invention relates to the field of medical data processing, and more particularly to a method, apparatus, device, and storage medium for processing health data related to energy metabolism diseases. The method includes: acquiring multi-source heterogeneous health data related to a target individual; extracting different dimensional data features from the multi-source heterogeneous health data, and constructing a structured data capsule model recording the target individual's health status based on a preset hierarchical hybrid association architecture and the data features of each dimension, the structured data capsule model including at least two levels; responding to the detection of new events related to the target individual, and updating the data features of the new events to the structured data capsule model according to a preset update mechanism; and generating a dynamic attribute graph based on the updated structured data capsule model. By using multi-source heterogeneous health data, constructing a structured data capsule model, dynamically updating in response to new events, and generating a dynamic attribute graph, the method solves the problems of fragmented health data and lack of dynamic adjustment in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of electromedical data processing, and more particularly to a method, apparatus, device, and storage medium for processing health data of energy metabolism diseases. Background Technology

[0002] In the health management practice of energy metabolism diseases, existing data processing technologies generally rely on general medical storage systems or static data warehouse architectures, which at their core adopt a flat data organization model based on timestamps and numerical values. This type of technical solution results in a highly fragmented state of patient health data. Specifically, key data such as blood glucose monitoring records, diet logs, medication history, and complication diagnosis information are scattered across multiple independent sources, including hospital information systems, laboratory testing platforms, wearable devices, and third-party health applications. These data sources are not only physically isolated but also encompass heterogeneous formats such as text descriptions, structured tables, and medical images, lacking a unified logical framework for effective integration. Because the existing architecture does not construct a hierarchical relationship between health events, related data, and metabolic mechanisms, the system can only store raw numerical indicators (such as fasting blood glucose readings) and cannot embed medical semantic concepts (such as insulin resistance or β-cell dysfunction) and clinical guideline evidence, severely limiting data traceability capabilities. Clinicians struggle to quickly identify potential causal pathways between blood glucose fluctuations and specific dietary behaviors or medication interventions. Furthermore, the system lacks event-driven dynamic adjustment capabilities; when new health events occur (such as acute hypoglycemia or new diabetic complications), it cannot automatically trigger adaptive reconstruction of the data structure, requiring manual intervention to modify storage logic or recalculate data correlations. At the data application level, existing technologies employ a linear workflow model: passive multi-source data collection is directly stored in a flat database, followed by discrete data retrieval based on human experience. This model allows physicians to obtain only isolated numerical fragments during diagnosis, requiring significant time for manual screening and correlation inference. It fails to provide timely and coherent data support for clinical decision support, AI-assisted diagnosis, or multi-institutional collaborative research, significantly hindering the efficiency and accuracy of long-term dynamic management of energy metabolism diseases. Summary of the Invention

[0003] The main objective of this invention is to address the problems of fragmented health data management and lack of dynamic adjustment in the prior art.

[0004] The first aspect of this invention provides a method for processing health data related to energy metabolism diseases, comprising: Acquire multi-source heterogeneous health data related to the target individual; Extract different dimensions of data features from the multi-source heterogeneous health data, and construct a structured data capsule model that records the health status of the target individual based on a preset hierarchical hybrid association architecture and data features of each dimension. The structured data capsule model includes at least two levels, including an event structure layer for indexing health events and individuals and a data structure layer for aggregating associated data. In response to detecting a new event related to the target individual, the data features of the new event are updated to the structured data capsule model according to a preset update mechanism; Dynamic attribute graphs are generated based on the updated structured data capsule model.

[0005] Optionally, in a first implementation of the first aspect of the present invention, the event structure layer includes an identity feature layer and a health event layer, and the data structure layer includes a data element layer and an atomic data unit layer; wherein: The identity feature layer is used to bind the identity identifier and basic feature information of the target individual; The health event layer is used to record the event attributes of each health event, indexed by energy metabolism-related health events. The data element layer is used to classify and aggregate the data associated with each health event in the health event layer according to the medical logic dimension; The atomic data unit layer is used to store the specific data values ​​and status labels corresponding to various types of data in the data element layer.

[0006] Optionally, in a second implementation of the first aspect of the present invention, the hierarchical relationships between the levels in the structured data capsule model are a hybrid tree-and-network architecture, wherein: The structured data capsule model for recording the health status of the target individual is constructed based on a preset hierarchical hybrid association architecture and data features of various dimensions, including: An identity feature layer is constructed based on the identity features in the data features of each dimension; Event nodes are constructed based on event features from data features across various dimensions to obtain a health event layer; Data element nodes are constructed based on the indicator features in the data features of each dimension, thus obtaining the data element layer; Atomic data nodes are constructed based on the indicator value features in the data features of each dimension, resulting in an atomic data unit layer; The identity feature layer is used as the root node, and all event nodes in the health event layer are associated with it. The event nodes in the health event layer are associated with one or more data element nodes in the data element layer by using the association index; The data element nodes in the data element layer are associated with multiple event nodes in the health event layer through event association identifiers; By associating the data element nodes in the data element layer with one or more corresponding atomic data nodes in the atomic data unit layer, a structured data capsule model is obtained.

[0007] Optionally, in a third implementation of the first aspect of the present invention, the step of responding to the detection of a new event related to the target individual and updating the data features of the new event to the structured data capsule model according to a preset update mechanism includes: Extract data features from new events and determine whether the corresponding data features inside the structured data capsule model meet the triggering conditions in the preset update mechanism. The triggering conditions include at least one of the following: new event triggering conditions, data correction triggering conditions, and state change triggering conditions. If the conditions for triggering the new event are met, a new event node is instantiated in the structured data capsule model, and related data element nodes and related edges are instantiated according to the mapping rules in the preset update mechanism to obtain a new structured data capsule model. If the data correction triggering condition is met, the data value of the corresponding atomic data node in the atomic data unit layer is updated based on the data characteristics, and the correlation strength between the affected event node and the data element node is recalculated to obtain a new structured data capsule model. If the state change triggering condition is met, the priority level of the corresponding health event is calculated based on the value of the data feature, and its position in the health event layer is adjusted based on the priority level.

[0008] Optionally, in a fourth implementation of the first aspect of the present invention, the step of responding to the detection of a new event related to the target individual and updating the data features of the new event to the structured data capsule model according to a preset update mechanism further includes: In the process of instantiating related data element nodes and related edges according to the mapping rules in the preset update mechanism to obtain a new structured data capsule model, constraint verification is performed on any newly added event nodes and data element nodes in the new structured data capsule model based on a predefined set of medical logic constraints. If the constraint verification shows that the newly added event node and data element node conflict with any constraint in the medical logic constraint set, then the update operation for the new event will be blocked.

[0009] Optionally, in a fifth implementation of the first aspect of the present invention, after generating the dynamic attribute graph based on the updated structured data capsule model, the method further includes: The event nodes and / or data element nodes in the structured data capsule model are mapped to the standard medical ontology terminology system to establish a unified semantic index. Based on a knowledge graph in the field of energy metabolism diseases, semantic associations between event nodes and potential metabolic mechanism nodes are established through multi-hop reasoning and updated into a structured data capsule model.

[0010] Optionally, in a sixth implementation of the first aspect of the present invention, the step of establishing semantic associations between event nodes and potential metabolic mechanism nodes through multi-hop reasoning based on a knowledge graph of energy metabolism diseases, and updating it to the structured data capsule model, includes: Based on the set of connection paths between event nodes and metabolic mechanism nodes in the knowledge graph of energy metabolism diseases, a predefined association strength scoring function is used to calculate the event node... With metabolic mechanisms causal tightness

[0011] If the causal density is greater than or equal to the strong correlation threshold, then an event node is established in the structured data capsule model. With metabolic mechanisms Semantic associations.

[0012] A second aspect of the present invention provides a health data processing device for energy metabolism diseases, comprising: The data acquisition module is used to acquire multi-source heterogeneous health data related to the target individual; The model building module is used to extract different dimensions of data features from the multi-source heterogeneous health data, and to build a structured data capsule model that records the health status of the target individual based on a preset hierarchical hybrid association architecture and data features of each dimension. The structured data capsule model includes at least two levels, including an event structure layer for indexing health events and individuals and a data structure layer for aggregating associated data. A dynamic update module is used to respond to the detection of new events related to the target individual and update the data features of the new events to the structured data capsule model according to a preset update mechanism; The data output module is used to generate dynamic attribute graphs based on the updated structured data capsule model.

[0013] A third aspect of the present invention provides an electronic device comprising: a memory and at least one processor, wherein the memory stores a request, and the memory and the at least one processor are interconnected via a circuit; the at least one processor invokes the request in the memory to cause the electronic device to perform the steps of the above-described energy metabolism disease health data processing method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium storing a request that, when executed on a computer, causes the computer to perform the steps of the above-described energy metabolism disease health data processing method.

[0015] In the technical solution of this invention, multi-source heterogeneous health data related to a target individual is acquired; different dimensions of data features are extracted from the multi-source heterogeneous health data; and a structured data capsule model recording the health status of the target individual is constructed based on a preset hierarchical hybrid association architecture and the data features of each dimension. The structured data capsule model includes at least two levels; in response to the detection of new events related to the target individual, the data features of the new events are updated to the structured data capsule model according to a preset update mechanism; and a dynamic attribute graph is generated based on the updated structured data capsule model. By using multi-source heterogeneous health data, constructing a structured data capsule model, dynamically updating in response to new events, and generating a dynamic attribute graph, the problems of fragmented health data and lack of dynamic adjustment in existing technologies are solved. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the first embodiment of the health data processing method for energy metabolism diseases in this invention; Figure 2 This is a schematic diagram of a second embodiment of the health data processing method for energy metabolism diseases in this invention. Figure 3 This is a schematic diagram of one embodiment of the health data processing device for energy metabolism diseases in this invention. Figure 4 This is a schematic diagram of one embodiment of the electronic device in this invention. Detailed Implementation

[0017] The technical solution of this invention solves the problems of data fragmentation and lack of dynamic adjustment in the prior art by acquiring multi-source heterogeneous health data, constructing a structured data capsule model, dynamically updating in response to new events and generating a dynamic attribute graph. It has the advantages of realizing unified integration and dynamic updating of health data, and improving the efficiency and accuracy of clinical decision-making.

[0018] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar elements and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] In traditional energy metabolism disease health data management, multi-source heterogeneous health data is scattered across multiple heterogeneous platforms such as hospital HIS systems, laboratory LIS systems, blood glucose meters, and various health apps, resulting in fragmented data with formats including text, tables, and images. Because current technologies lack a hierarchical relationship architecture of "health event - related data - metabolic mechanism," the system only records raw values ​​without linking them to medical semantics and clinical guidelines, reducing data traceability. Furthermore, existing technologies lack dynamic update mechanisms, failing to automatically trigger adaptive adjustments to the data structure when new health events occur, thus affecting the timeliness and completeness of data management and failing to meet the core needs of long-term dynamic management of energy metabolism diseases.

[0020] For example, when managing the health data of a patient with type 2 diabetes, blood glucose monitoring data comes from independent blood glucose meters, dietary records are stored in third-party health applications, medication history is kept in the hospital's HIS system, and complication diagnosis information is scattered across the LIS system. When the patient experiences an acute hypoglycemic event, the system cannot automatically correlate the previous high carbohydrate intake record with the insulin injection dose. Doctors must manually retrieve data from multiple heterogeneous platforms and perform empirical correlation analysis, leading to a prolonged clinical intervention decision-making process and making it easy to miss key causal relationships due to data fragmentation.

[0021] If these issues are not addressed, the fragmentation of health data will persist, and insufficient data traceability will reduce the reliability of clinical decision support systems, thereby affecting the accuracy of individualized treatment plans for patients. Furthermore, the lack of a dynamic update mechanism will make it difficult for the system to adapt to the constantly changing health events in long-term disease management, ultimately hindering the implementation of AI-driven multi-center data sharing and advanced analytics applications.

[0022] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1The first embodiment of the energy metabolism disease health data processing method in this invention includes: 101. Acquire multi-source heterogeneous health data related to the target individual; In this embodiment, multi-source heterogeneous health data refers to health-related data originating from different systems, devices, or platforms, and whose data formats, structures, and semantics may differ. For example, this data can come from hospital information systems (HIS), laboratory information systems (LIS), electronic medical records (EMR), wearable devices, health management apps, genetic testing reports, etc., and covers various types such as physiological indicators, lifestyle habits, medication records, diagnostic results, and imaging data.

[0023] Specifically, this multi-source heterogeneous health data can be entered manually, such as by medical staff or patients manually inputting data such as blood glucose monitoring results, dietary records, and medication information into the system.

[0024] The system also allows for batch uploading of data files (such as CSV and Excel files) exported from different medical devices or health applications to the data processing system via a simple file import function. This data may exist in multiple formats and originate from different information systems, such as hospital electronic medical record systems, independent laboratory testing systems, and smart wearable devices used by users.

[0025] 102. Extract data features of different dimensions from multi-source heterogeneous health data, and construct a structured data capsule model that records the health status of target individuals based on a pre-defined hierarchical hybrid association architecture and data features of each dimension. The structured data capsule model includes at least two levels, including an event structure layer for indexing health events and individuals and a data structure layer for aggregating associated data. Specifically, after obtaining the original multi-source heterogeneous health data through step 101, the system first determines whether the data formats of the multi-source heterogeneous health data are the same or whether they are currently parseable data formats. If not, the system calls a preset protocol conversion mechanism to convert the health data of each data format. Then, the converted health data is preliminarily processed to identify meaningful information fragments, i.e., data features.

[0026] For example, key information in the data can be manually identified and labeled, such as manually extracting "hyperglycemia event" or "insulin dosage" from a text description. These extracted data features are then used to construct a structured data model. This model is designed to contain at least two levels: one level is designated as the event structure layer, which indexes health events and individual information; the other level is designated as the data structure layer, which aggregates the specific data associated with these events. For example, a simple two-level structure can be predefined, with a top-level list of events and a bottom-level list of raw data corresponding to each event, and the association between events and data can be manually established.

[0027] In another embodiment, when performing data format conversion, the source of the collected health data can be identified, and the corresponding data protocol can be matched based on the correspondence between the source and the preset format and data system. The health data can then be converted using the conversion relationship between the data protocol and the conversion protocol of the target data format to obtain health data in the target data format.

[0028] Furthermore, text recognition technology can be used to identify data from various health data files, then reorganize the data according to the target data format structure, and output the health data corresponding to the target data format.

[0029] 103. In response to the detection of new events related to the target individual, update the data features of the new events into the structured data capsule model according to the preset update mechanism; In this embodiment, by using the target individual's identity characteristics as a monitoring index, changes in health data across various data source systems are monitored in real time. These changes can be monitored from a time dimension, or from dimensions such as tags, values, or even from the size of data packets. The "new event" refers to an event that occurs or is newly detected during the target individual's health management process and is related to the individual's health status. Examples include a new abnormal blood glucose record, a new medication adjustment, or a new diagnosis of a complication.

[0030] If changes are detected in these dimensions, a new event is identified, and the corresponding health data is collected. It should be noted that when collecting health data for a new event, either the complete data or the changed portion of the data can be collected.

[0031] Specifically, when the system detects any new health event related to the target individual, such as a new medical examination report or a new medication adjustment, the method adjusts the existing structured data capsule model according to a set of pre-defined update rules. For example, when new blood glucose measurement data is recorded, the data can be manually added to the data structure layer, and the corresponding blood glucose monitoring event in the event structure layer can be manually updated. This update mechanism ensures that the model reflects the target individual's latest health status.

[0032] 104. Generate dynamic attribute graphs based on the updated structured data capsule model.

[0033] In this embodiment, the dynamic attribute graph refers to a graph-structured data representation generated based on the updated structured data capsule model. Specifically, based on the events and data in the model, drawing software identifies the nodes and relationships within them to create a simple graph, where nodes represent events or data, and edges represent the relationships between them.

[0034] For example, one node in the graph might represent a "hyperglycemia event," and another node might represent an "insulin dosage event." They are connected by edges, representing the effect of medication on blood glucose. Simultaneously, these event nodes are connected to specific data nodes, such as "blood glucose level" and "insulin dosage." When the model is updated, the attribute graph is also refreshed in real time; for example, new hyperglycemia event nodes and their associated data are added to the graph, or the association strength of existing events changes.

[0035] In this embodiment, a structured data capsule model is constructed to achieve a unified structured organization of multi-source heterogeneous health data. Furthermore, the ability to respond to new events and dynamically update the structured data capsule model according to a preset update mechanism is introduced, enabling automatic adaptive adjustments, instantiation of new event nodes, and updating of associated data. This contrasts sharply with the static model in existing technologies, which requires manual adjustment of the database structure or recalculation of correlations, significantly improving the efficiency and timeliness of data management and meeting the core needs of long-term dynamic management of energy metabolism diseases. It also solves the problems of fragmented health data and lack of dynamic adjustment in existing technologies.

[0036] Please see Figure 2 A second embodiment of the energy metabolism disease health data processing method of the present invention includes: 201. Construct middleware for multi-protocol adaptation based on the data format and target data format of a multi-data source system; In this step, the middleware built is actually a protocol conversion interface that adapts to multiple systems. Its main function is to convert the health data collected from various systems into a unified data format.

[0037] Specifically, this middleware is built by decoupling the data access, transformation, and distribution logic to achieve dual-mode data output. The implementation steps are as follows: Step 1: Construct a multi-source heterogeneous data ingestion layer. Specifically, first construct a data access gateway based on the protocol adapter pattern, and define the input data set Din={ddevice,dsystem,dmanual}; Then, design a protocol parsing function P(x) to map heterogeneous data from different sources to the system's internal standard data object Ostd: ; Furthermore, non-blocking I / O and message queue buffering mechanisms are adopted to ensure that the access latency Tingest≤2s under high concurrency requests, and the data access success rate Rsuccess≥98% based on the ACK confirmation mechanism.

[0038] Step 2: Construct a full-graph serialization output layer for independent applications. In practical applications, full-scale atlas serialization engines can be designed for independent scenarios such as clinical research. A four-level structured data capsule model G is defined, where the four-level structure is completely mapped to the target format file through the output function Eindep(G, fmt).

[0039] In the model, through structured mapping By utilizing JSON-LD's Context mechanism to preserve semantic tags, the interpretability of exported data in the SPSS / R environment is ensured.

[0040] Furthermore, the model also includes the definition of a structured integrity index. η is ≥ 99%, where V and E are the sets of nodes and edges, respectively. A recursive traversal algorithm is used to ensure that η ≥ 99%.

[0041] In another implementation, in addition to the aforementioned transformation, the middleware also includes setting up a Data Projection & Push Service. Specifically, for the core energy metabolism management system, a context-aware data projection mechanism is designed to extract only key decision features and avoid redundant data transmission. A projection operator Πcore is defined to filter only L2 event summaries and L3 aggregated data, while filtering L1 sensitive information and L4 atomic details. , where Agg() is an aggregation function (such as taking the mean or extreme value).

[0042] Furthermore, Dcore is encapsulated as a standard FHIR Observation and Condition resource package, and actively pushed via RESTful API (Push Mode). No ETL operation is required in the core system, the interface handshake success rate is 100%, and the end-to-end synchronization latency Tsync ≤ 3s.

[0043] In another embodiment, the middleware further includes: Attribute-based dynamic access control (ABAC & Privacy Masking). Based on the Attribute-Based Access Control (ABAC) model, fine-grained privacy protection is implemented at the interface layer.

[0044] Access control and data masking logic: Define access policy Policy(Userrole, TagL1), and output the masked data view Vsafe. ; If Check() returns False, access is denied; if it passes, differential privacy or masking is applied to sensitive fields (such as patient name and ID) depending on the scenario.

[0045] Security and compliance: The permission verification logic is embedded in the API gateway layer to ensure that all outbound traffic is verified, with a verification coverage of 100%, meeting compliance requirements such as HIPAA / GDPR.

[0046] 202. Extracting different dimensions of data features from multi-source heterogeneous health data based on middleware; Suppose user A is an individual with an energy metabolism disorder (such as type 2 diabetes). In traditional health management models, user A's blood glucose monitoring data might be stored in a blood glucose meter app, dietary records in another health app, medication information in a hospital's electronic medical record system, and physical examination reports in paper or PDF format. This data is scattered and in various formats, making unified management and correlation analysis difficult.

[0047] In this embodiment, the system first acquires multi-source heterogeneous health data related to user A. Specifically, the system can obtain daily blood glucose levels from user A's blood glucose meter app, dietary information from a diet record app, medication records and diagnostic information from the hospital's HIS system, and the latest medical examination report from the medical examination center via a data interface. This collected data may include various formats such as numerical values, text, and images.

[0048] Next, the system extracts different dimensions of data features from these multi-source heterogeneous health data. For example, it extracts features such as "blood glucose level" and "measurement time" from blood glucose data; features such as "food type" and "intake" from dietary records; features such as "drug name," "dosage," and "dosage time" from medication records; and features such as "glycated hemoglobin (HbA1c)" and "blood lipid level" from physical examination reports. Subsequently, based on a pre-defined hierarchical hybrid association architecture and these extracted data features, the system constructs a structured data capsule model recording user A's health status. This model is designed to include an event structure layer and a data structure layer. For example, the event structure layer may contain nodes such as "hyperglycemia event," "insulin medication event," and "abnormal physical examination event," while the data structure layer aggregates specific data related to these events, such as "blood glucose level 12.5 mmol / L," "insulin dosage 10 units," and "HbA1c 8.0%." Event nodes in the event structure layer and data nodes in the data structure layer are connected through preset association rules to form an organic whole. For example, the "high blood sugar event" node can be associated with specific "blood sugar value" data nodes and "dietary record" data nodes.

[0049] 203. Construct a structured data capsule model to record the health status of target individuals based on a pre-defined hierarchical hybrid association architecture and data features of various dimensions; In this embodiment, the hierarchical hybrid association architecture refers to a data organization structure characterized by combining hierarchical (such as tree-like) and non-hierarchical (such as network-like) association methods. In this architecture, data elements are organized into different levels, while complex many-to-many relationships exist within or between levels, in order to more flexibly and comprehensively reflect the inherent connections between data.

[0050] Specifically, the structured data capsule model includes at least two levels: an event structure layer for indexing health events and individuals, and a data structure layer for aggregating related data; the event structure layer includes an identity feature layer and a health event layer, and the data structure layer includes a data element layer and an atomic data unit layer; wherein: The identity feature layer is used to bind the identity identifier and basic feature information of the target individual; wherein, the identity feature layer L1 serves as the unique identifier layer of the data capsule, binding the patient's core identity and basic features, including a unique capsule ID (encoded using "patient ID + timestamp"), basic information (age, gender, BMI), basic metabolic features (such as insulin sensitivity typing, carbohydrate tolerance), and data permission tags (such as "available for clinical research" and "shared in core systems"); once the data in this layer is entered, it only supports permission updates and does not change dynamically with events.

[0051] The health event layer is used to record the event attributes of each health event, indexed by energy metabolism-related health events. Among them, the health event layer L2 serves as the core driving layer of the data capsule, indexed by energy metabolism-related health events, including basic event attributes and association rules. Event types are classified according to medical logic as follows: diagnosis and treatment events (such as "diabetes diagnosis" and "complication screening"), monitoring events (such as "blood glucose monitoring" and "blood lipid testing"), intervention events (such as "dietary adjustment" and "medication execution"), and abnormal events (such as "blood glucose spike" and "hypoglycemic episode"). Each event is bound to attributes such as "event ID, event type, occurrence timestamp, event level (P1-P3 level), and associated data element index", forming the core framework of the data organization.

[0052] The data element layer is used to classify and aggregate the data associated with each health event in the health event layer according to the medical logic dimension. Among them, the data element layer L3 serves as the event-related data aggregation layer, classifying and aggregating event-related data according to the "medical logic dimension". Each data element is bound to a unique data element ID and an event association identifier. The classification dimensions include: physiological indicator data elements (such as "blood glucose indicator group" including fasting / 2-hour postprandial blood glucose), behavioral intervention data elements (such as "dietary data group" including carbohydrate / fat intake), diagnosis and treatment data elements (such as "medication data group" including drug name / dosage), and multi-omics data elements (such as "gene data group" including TCF7L2 gene mutation). The association between data elements and events adopts a "many-to-many" rule, supporting a single event to be associated with multiple data elements and a single data element to be associated with multiple related events (such as the "blood glucose spike" event being associated with "blood glucose indicator", "diet", and "medication" data elements).

[0053] The atomic data unit layer is used to store the specific data values ​​and status labels corresponding to various types of data in the data element layer. Among them, the atomic data unit layer L4 serves as the smallest data unit of the data capsule, storing specific original data and status labels. Each atomic unit is bound to a unique atomic ID, data element association identifier, data value, collection time, and data quality label (such as "valid", "outlier", "missing"). For example, the data element "blood glucose index group" is associated with atomic units such as "fasting blood glucose -12.5mmol / L-2025-12-05-valid" and "2h postprandial blood glucose -16.3mmol / L-2025-12-05-valid".

[0054] Furthermore, the above four layers adopt a hybrid "tree + network" association architecture. The L1 layer is the root node that associates all L2 layer events; L2 layer events point to L3 layer data elements through "association data element index" (tree association); L3 layer data elements associate multiple L2 layer events through "event association identifier" (network association); the L3 layer and L4 layer are in a "one-to-many" tree association to ensure the integrity and efficiency of data traceability.

[0055] It is understood that the event types in the health event layer include at least one of diagnosis and treatment events, monitoring events, intervention events, and abnormal events; the data classification dimensions of the data element layer include physiological indicator data elements, behavioral intervention data elements, diagnosis and treatment data elements, and / or multi-omics data elements.

[0056] Specifically, in constructing the structured data capsule model, the event structure layer is first clearly divided into an identity feature layer and a health event layer. The identity feature layer is responsible for establishing the unique identity and basic characteristic information of the target individual, providing a clear individual context for the attribution of all subsequent health data. Based on this, the health event layer focuses on energy metabolism-related health events, indexing and recording the attributes of each individual's health event to ensure the integrity and traceability of the events. Simultaneously, the data structure layer is refined into a data element layer and an atomic data unit layer. The data element layer, based on a pre-defined medical logic dimension, classifies and aggregates the raw data associated with each health event in the health event layer, abstracting the scattered raw data into data elements with business significance, facilitating high-level analysis and understanding. The atomic data unit layer, as the lowest-level data storage, accurately preserves the specific data values ​​and status labels corresponding to various types of data in the data element layer, ensuring the originality and verifiability of the data. Through this hierarchical division, the structured data capsule model can more effectively integrate multi-source heterogeneous health data in a structured manner. When new health data is acquired, the system can first identify the target individual through the identity feature layer, then associate it with specific energy metabolism-related events through the health event layer, then classify and aggregate it according to medical logic dimensions through the data element layer, and finally store the raw data in the atomic data unit layer.

[0057] In another embodiment, the structured data capsule model employs a hybrid tree-and-network relational architecture between its various levels, wherein: The structured data capsule model for recording the health status of the target individual is constructed based on a preset hierarchical hybrid association architecture and data features of various dimensions, including: An identity feature layer is constructed based on the identity features in the data features of each dimension; Event nodes are constructed based on event features from data features across various dimensions to obtain a health event layer; Data element nodes are constructed based on the indicator features in the data features of each dimension, thus obtaining the data element layer; Atomic data nodes are constructed based on the indicator value features in the data features of each dimension, resulting in an atomic data unit layer; The identity feature layer is used as the root node, and all event nodes in the health event layer are associated with it. The event nodes in the health event layer are associated with one or more data element nodes in the data element layer by using the association index; The data element nodes in the data element layer are associated with multiple event nodes in the health event layer through event association identifiers; By associating the data element nodes in the data element layer with one or more corresponding atomic data nodes in the atomic data unit layer, a structured data capsule model is obtained.

[0058] The hybrid tree-and-network architecture is a data organization method that combines hierarchical structures and complex many-to-many relationships. Tree structures are typically used to represent explicit parent-child or containment relationships, such as the attribution relationship between an individual and an event; while network structures allow for more flexible, non-hierarchical multi-directional associations between nodes, for example, a data metric may be associated with multiple health events. This architecture can be implemented using graph databases (such as Neo4j), where nodes represent entities at different levels and edges represent the relationships between them; or it can be simulated using multi-table associations and views in relational databases, establishing connections using foreign keys and intermediate association tables. The identity feature layer is the starting point of the entire data capsule model, used to uniquely identify and describe the target individual. During construction, identity features extracted from multi-source heterogeneous health data (such as patient ID, name, gender, age, etc.) are used to create nodes in the identity feature layer. These nodes carry the individual's basic information and can be standardized to ensure uniqueness. The health event layer is used to record energy metabolism-related health events occurring in the target individual. During construction, event features (such as diagnostic results, medical records, medication records, and examination reports) are identified and extracted from multi-source heterogeneous health data. Each event feature corresponds to a specific health event and is instantiated as an event node in the health event layer. These nodes may contain attributes such as event type, occurrence time, and event description. The data element layer is used to classify and aggregate data associated with health events. During construction, indicator features extracted from multi-source heterogeneous health data (such as blood glucose levels, blood pressure levels, and various indicators in blood lipid reports) are used to construct data element nodes in the data element layer. Each data element node represents a type of indicator with medical logical significance and may contain meta-information such as indicator name, unit, and normal range. The atomic data unit layer is the lowest level for storing specific, raw data values ​​and status labels. During construction, indicator value features (i.e., specific numerical or textual results) extracted from multi-source heterogeneous health data are used to construct atomic data nodes in the atomic data unit layer. Each atomic data node stores a specific measurement value or observation result and may contain status information such as data collection time and data source. Using the identity feature layer as the root node, and associating it with all event nodes in the health event layer, aims to establish a direct hierarchical relationship between individuals and all health events, ensuring that all events belong to a specific individual. This can be achieved by creating "own" or "occur" type edges between identity feature layer nodes and health event layer nodes. By using association indexes to link event nodes in the health event layer to one or more data element nodes in the data element layer, the aim is to clarify which specific medical indicators or data elements a health event involves. This can be achieved by storing a list of data element node IDs in the health event layer nodes or creating "contains" type edges.Associating data element nodes in the data element layer with multiple event nodes in the health event layer through event association identifiers aims to allow a single data element to be referenced by multiple health events, forming a network of relationships. This can be achieved by storing a list of health event node IDs in the data element layer nodes or by creating reverse "referenced" edges. Associating data element nodes in the data element layer with one or more corresponding atomic data nodes in the atomic data unit layer aims to connect abstract medical indicators with specific measurements or observations. This can be achieved by including a reference to its parent data element layer node in the atomic data unit layer node.

[0059] Specifically, firstly, the identity feature layer, acting as the root node, establishes a direct hierarchical association with all event nodes in the health event layer, ensuring that all health events can be traced back to a specific target individual. Secondly, event nodes in the health event layer point to one or more data element nodes in the data element layer through association indexes, clarifying the specific data indicators involved in each health event. Simultaneously, data element nodes in the data element layer are back-associated with multiple event nodes in the health event layer through event association identifiers, thereby achieving data element reuse and many-to-many associations across different health events. Finally, data element nodes in the data element layer are associated with corresponding atomic data nodes in the atomic data unit layer, connecting abstract indicators with specific raw data values. This refined construction process enables multi-source heterogeneous health data to be organized and managed efficiently and accurately, forming a model with both a clear hierarchical structure and complex network association capabilities. This approach solves the problem of effectively establishing many-to-many associations between events and data in complex health data, providing a solid foundation for subsequent data querying, analysis, and updates, and significantly improving the flexibility and expressive power of the data model.

[0060] As a specific implementation method, when constructing a structured data capsule model, health data of the target individual can first be obtained from multiple sources such as electronic medical record systems, wearable device data, and laboratory test reports. For example, for a diabetic patient, their identity characteristics (such as patient ID "P001" and name "Zhang San") are used to construct a node in the identity characteristic layer. Next, event characteristics extracted from the medical record, such as "type 2 diabetes diagnosis," "insulin medication record," and "regular blood glucose monitoring," are instantiated as multiple event nodes in the health event layer. Simultaneously, indicator characteristics extracted from laboratory reports, such as "fasting blood glucose," "2-hour postprandial blood glucose," and "glycated hemoglobin," are constructed as data element nodes in the data element layer. Specific blood glucose measurements (such as "fasting blood glucose: 7.5 mmol / L" and "2-hour postprandial blood glucose: 12.3 mmol / L") are constructed as atomic data nodes in the atomic data unit layer. In terms of hierarchical association, the "Zhang San" node in the identity characteristic layer serves as the root node, directly associated with all event nodes in the health event layer, such as the "type 2 diabetes diagnosis" event node. In the health event layer, the "Type 2 Diabetes Diagnosis" event node points to data element nodes such as "Fasting Blood Glucose," "2-Hour Postprandial Blood Glucose," and "HbA1c" in the data element layer via association indexes, indicating that these indicators are related to diabetes diagnosis. Conversely, the "Fasting Blood Glucose" data element node in the data element layer can be associated with multiple event nodes in the health event layer, such as "Type 2 Diabetes Diagnosis" and "Regular Blood Glucose Monitoring," through event association identifiers, reflecting its importance in different events. Finally, the "Fasting Blood Glucose" data element node in the data element layer is associated with specific atomic data nodes such as "7.5 mmol / L" in the atomic data unit layer, thus connecting abstract indicators with specific values.

[0061] 204. In response to the detection of new events related to the target individual, extract the data features of the new events and determine whether the corresponding data features inside the structured data capsule model meet the triggering conditions in the preset update mechanism; 205. Update the structured data capsule model based on the judgment results; In this embodiment, the triggering conditions include at least one of the following: a new event triggering condition, a data correction triggering condition, and a state change triggering condition.

[0062] In this step, if the new event triggering condition is met, a new event node is instantiated in the structured data capsule model, and related data element nodes and associated edges are instantiated according to the mapping rules in the preset update mechanism to obtain a new structured data capsule model; if the data correction triggering condition is met, the data values ​​of the corresponding atomic data nodes in the atomic data unit layer are updated based on the data features, and the association strength between the affected event nodes and data element nodes is recalculated to obtain a new structured data capsule model; if the state change triggering condition is met, the priority level of the corresponding health event is calculated based on the value of the data features, and its position in the health event layer is adjusted based on the priority level.

[0063] Understandably, extracting data features from new events refers to identifying and separating key information from newly received health data. This information describes the type of event, its occurrence time, relevant numerical values, and the identities of the individuals involved. For example, for a new laboratory test report, the test name, result value, test date, and patient ID can be extracted. This extraction process can be implemented using various techniques. For instance, predefined parsing rules can be used to extract fields from structured data; or, for unstructured or semi-structured data, Natural Language Processing (NLP) techniques can be used for entity recognition and relation extraction.

[0064] The system determines whether the data features within the structured data capsule model meet the trigger conditions in a preset update mechanism. These trigger conditions include at least one of three: new event trigger conditions, data correction trigger conditions, and state change trigger conditions. This means the system evaluates whether the data features of a new event meet the initiation conditions for a certain update operation based on a pre-defined set of rules. The preset update mechanism is a set of strategies and rules for managing the dynamic changes of the data capsule model. Trigger conditions are the criteria for initiating specific update operations. A new event trigger condition occurs when a completely independent health event not previously recorded in the model is detected, such as the first diagnosis of a disease or a new medical examination. A data correction trigger condition occurs when information that corrects or supplements existing data in the model is detected, such as the correction or update of a historical test result. A state change trigger condition occurs when a change in the key attributes (such as severity or priority) of a health event is detected, such as a patient's condition worsening or improving.

[0065] If the conditions for triggering a new event are met, a new event node is instantiated in the structured data capsule model. Based on the mapping rules in the preset update mechanism, associated data element nodes and edges are also instantiated to obtain a new structured data capsule model. This means that when the system identifies a new health event, a new data entity is created in the structured data capsule model. Instantiating a new event node means creating a node representing the new event in the health event layer and assigning it corresponding attributes. The mapping rules define how to correspond the data characteristics of the new event to the hierarchical structure of the data capsule model. For example, mapping a certain examination result to a specific data element node type and connecting these data element nodes to the newly created event node through associated edges.

[0066] If the data correction triggering condition is met, the data values ​​of the corresponding atomic data nodes in the atomic data unit layer are updated based on the data characteristics, and the correlation strength between the affected event nodes and data element nodes is recalculated to obtain a new structured data capsule model. This means that when the system identifies a correction to existing data, it directly modifies the specific data values ​​stored in the corresponding atomic data nodes in the atomic data unit layer. For example, if a blood glucose value is corrected, the value of that atomic data node is directly updated. Simultaneously, to reflect the impact of data changes on the relationship between events and data elements, the system recalculates the correlation strength between the affected event nodes and data element nodes to ensure the accuracy of the relationships in the model. Correlation strength can reflect the relevance, importance, or timeliness between data and events.

[0067] If the aforementioned state change triggering condition is met, the priority level of the corresponding health event is calculated based on the value of the data feature, and its position in the health event layer is adjusted based on the priority level. This means that when the system identifies a change in the state of a health event, it reassesses the urgency or importance of the health event based on the value of the new data feature (e.g., the degree of abnormality of a certain indicator), thereby calculating a new priority level. For example, the priority level of a previously stable chronic disease event might be increased if a key indicator suddenly worsens. Subsequently, the system adjusts the logical or physical position of the event node in the health event layer based on this new priority level, for example, placing it in a more prominent position or prioritizing its return during data queries to reflect its higher level of attention.

[0068] This application's solution continuously monitors new events related to the target individual and intelligently determines the type of new event (addition, correction, or status change) according to a preset update mechanism. It then adopts targeted update strategies to dynamically adjust the structured data capsule model. When a new health event occurs, the system can promptly instantiate new event nodes and related data nodes in the model, expanding the model's coverage. When existing data needs correction, the system can accurately update the data values ​​at the atomic data unit layer and reassess the correlation strength between events and data elements, ensuring data accuracy. When the priority of a health event changes, the system can recalculate its priority level and adjust its position in the model to reflect its importance. This dynamic update mechanism enables the structured data capsule model to reflect the target individual's latest health status in real time, avoiding the problems of lagging or inaccurate model data.

[0069] In another embodiment, the correlation strength between the recalculated event nodes and data element nodes is calculated using a multi-factor weighted attenuation formula, which is: ; Where Δt is the normalized time difference between the occurrence time of the health event and the data collection time, simulating the time decay effect; For ontology semantic similarity calculated based on energy metabolism knowledge graph; To score data quality; α, β, and γ are adjustment coefficients, and .

[0070] In practical applications, the update mechanism in steps 204 and 205 refers to establishing a dynamic graph evolution mechanism of "event triggering - topology reconstruction - weight calculation," treating the data capsule as a dynamic attribute graph G=(V, E), where V is the set of nodes at each level and E is the set of edges connecting the levels. By monitoring changes in the state vector, the adaptive evolution of the data capsule structure is achieved.

[0071] This mechanism includes mathematical modeling and state monitoring of multidimensional triggering conditions: Specifically, the system deploys a state monitoring daemon, defines the current state vector of the data capsule as St, and the update trigger function as F(St)→ΔS, including the following three trigger models: Topology Growth Triggered by New Events: When a new event entity enew is detected in the input stream, if enew... If VL2 (the set of L2 layer nodes) is reached, the graph growth operation is triggered. The system automatically calculates the set of L3 layer data element nodes that need to be instantiated based on a preset mapping rule set R. and the corresponding edge set Enew. The topology update logic is represented as: .

[0072] Weight recalculation is triggered by data correction. When the value Val(aij) of the L4 layer atomic unit aij is supplemented or corrected, not only are the node attributes updated, but the event node e of its corresponding L2 layer is also triggered; and the data element node dj of the L3 layer is also affected. Association Strength Recalculation. The correlation strength W(ei,dj) is calculated using a multi-factor weighted attenuation formula: Where Δt is the normalized time difference between the event occurrence time and the data acquisition time, simulating the time decay effect; SimKG is the ontology semantic similarity (0-1 interval) calculated based on the energy metabolism knowledge graph; Q(dj) is the data quality score (based on completeness and confidence). α, β, γ are adjustment coefficients, and .

[0073] Level Transition Trigger: When the data value Val(α) crosses the preset threshold interval [θmin, θmax], the priority level P(e) of the L2 layer event node is triggered to transition. The level adjustment function is defined as follows:

[0074] Level transitions will simultaneously trigger storage policy change signals.

[0075] In this embodiment, although a dynamic evolution of the model is proposed by responding to the detection of new events related to the target individual and updating the data features of the new events into the structured data capsule model according to a preset update mechanism, if the newly added data is not effectively validated during the instantiation of new event nodes and associated data element nodes, it may lead to the introduction of data that contradicts or is inconsistent with the existing medical logic into the model, thereby affecting the accuracy and reliability of the model.

[0076] In response to the detection of a new event related to the target individual, and the updating of the data features of the new event to the structured data capsule model according to a preset update mechanism, this embodiment further proposes: In the process of instantiating related data element nodes and related edges according to the mapping rules in the preset update mechanism to obtain a new structured data capsule model, constraint verification is performed on any newly added event nodes and data element nodes in the new structured data capsule model based on a predefined set of medical logic constraints. If the constraint verification shows that the newly added event node and data element node conflict with any constraint in the medical logic constraint set, then the update operation for the new event will be blocked.

[0077] The predefined set of medical logical constraints refers to a series of rules, conditions, or patterns summarized from large amounts of medical data by medical experts, domain knowledge engineers, or through machine learning methods before or during system operation. These constraints aim to ensure the rationality, consistency, and validity of health data in medicine. For example, it can include provisions regarding disease diagnostic criteria, drug contraindications, ranges of physiological indicators, and the order of events. Implementation methods can include, but are not limited to: storing them in the form of logical expressions parsable by the rule engine, such as SWRL (Semantic WebRule Language) rules; defining relationships and attribute constraints between concepts in the form of ontology; or encoding them as production rules (IF-THEN rules) in expert systems. The constraint verification refers to the process of automatically checking new event data (including newly instantiated event nodes and data element nodes) before the system receives new event data and attempts to integrate it into the structured data capsule model. This process aims to evaluate whether the new data conforms to the provisions in the predefined set of medical logical constraints. The verification process can be implemented in several ways: using a rule engine to match and reason with the new data and the rule set; using a semantic inference machine to check whether the new data violates the constraints defined in the ontology; or using a custom verification function to check the validity of specific data fields. Blocking the update operation for the new event means that when constraint verification finds a conflict between the new event node or data element node and any constraint in the medical logic constraint set, the system will stop the current new event update process and will not write the conflicting data into the structured data capsule model. This mechanism ensures that only verified data that conforms to medical logic can be included in the model, thereby maintaining the integrity and accuracy of the model. The blocking operation can manifest as: rolling back the database transaction; issuing a warning notification to the user or administrator; recording a conflict log for subsequent manual intervention or rule adjustments; or directly refusing to receive the new event data.

[0078] In response to the detection of new events related to a target individual and the updating of data features from these new events into the structured data capsule model according to a pre-defined update mechanism, this application introduces a rigorous verification mechanism to ensure the medical rationality and consistency of the model data. Specifically, when the system instantiates new event nodes and associated data element nodes according to the mapping rules in the pre-defined update mechanism and prepares to integrate them into the structured data capsule model, the update is not completed directly. In this crucial step, the system performs a comprehensive constraint verification on these newly added event nodes and data element nodes based on a predefined set of medical logical constraints. This verification process checks whether the new data conforms to known medical knowledge, clinical guidelines, or data consistency rules. For example, it verifies the rationality of a new diagnosis, the effective range of new test results, or whether there are logical conflicts between different events. Only when all newly added event nodes and data element nodes pass the constraint verification—that is, when they do not conflict with any constraints in the medical logical constraint set—will the system allow the update operation to continue, thereby obtaining a new structured data capsule model. Conversely, if any newly added event node or data element node is found to conflict with any constraint in the medical logic constraint set during the verification process, the system will immediately block the update operation for the new event. This mechanism effectively prevents data that does not conform to medical logic or is contradictory from being written into the model, thereby ensuring the data quality and medical reliability of the structured data capsule model during its dynamic evolution, enabling it to more accurately reflect the true health status of the target individual.

[0079] In practical applications, the "copy-on-write" approach is combined with an incremental update strategy to ensure low system latency (Latency≤3s): hierarchical incremental update: only the subgraph GsubCG involved in the change is locked for write operations, keeping the L1 level and irrelevant branches in a read-only state, reducing the performance overhead caused by full lock.

[0080] Logical constraint verification: A medical logical constraint verification machine is introduced, defining a constraint set C = {C1, C2, ..., Cn}. For any newly added association (e, d), all Boolean constraints must be satisfied:

[0081] For example, if Type(e) = "hypoglycemia" and Type(d) = "high carbohydrate intake", based on the knowledge graph reasoning rule ck: if it is judged as a logical conflict, then Validity = False, the system will automatically block the update and throw an exception, with a verification pass rate of ≥98%.

[0082] 206. Calculate the access popularity score of event nodes in the structured data capsule model using a predefined popularity scoring function; In one embodiment, the formula for calculating the popularity rating function is: ,in, For indicator functions, This is the time decay factor; The calculation formula for the storage medium selection strategy is as follows: ,in, The threshold for popularity scoring is Redis, which is a memory cache, and HDFS is a distributed file system.

[0083] 207. Select the target storage medium based on the access popularity score and storage medium selection strategy, and schedule event nodes with different popularity and their associated data element nodes and atomic data nodes to the target storage medium for storage. Among them, the data element nodes associated with event nodes with high popularity scores are stored using a dual-write mechanism of memory caching and persistent storage. In this embodiment, the popularity scoring function is an algorithm used to quantify the access frequency or importance of event nodes in a structured data capsule model. The concept involves assigning a numerical value to an event node based on its usage pattern, time sensitivity, or business importance to reflect its "popularity." For example, the function can be calculated based on the number of queries to an event node within a specific time window, or by assigning different weights based on the interval between the event node's occurrence time and the current time. The access popularity score of an event node is a numerical value calculated by this function, representing the activity level of a specific event node. This score can be a continuous value or a discrete level classification. The storage medium selection strategy is a set of rules that determine the data storage location based on the access popularity score of the event node, aiming to optimize the utilization of storage resources and improve data access efficiency. Its implementation can include setting multiple popularity thresholds to map data of different popularity ranges to different storage media, or dynamically allocating storage media based on their cost, performance, capacity, and other characteristics, combined with data popularity. The target storage medium is a physical or logical storage device determined according to the storage medium selection strategy for storing data with specific popularity. It can be different types such as memory, solid-state drives, hard disk drives, or distributed file systems. Scheduling event nodes of varying popularity, along with their associated data element nodes and atomic data nodes, to the target storage medium refers to migrating or copying this data from its current storage location to the selected target storage medium. This can be a background task or a real-time data stream processing process. The dual-write mechanism of memory caching and persistent storage is a data storage strategy designed to ensure that high-frequency data receives fast access while also guaranteeing data persistence and reliability. Its implementation typically involves simultaneously writing data to a high-speed memory cache and a low-speed but persistent storage. When reading data, it is first retrieved from the memory cache; if not present in the cache, it is read from persistent storage and loaded into the cache.

[0084] Building upon the aforementioned structured data capsule model construction and updates, this application introduces a storage optimization mechanism based on data popularity to address the storage and access efficiency challenges arising from model growth. The system first calculates an access popularity score for each event node in the structured data capsule model using a predefined popularity scoring function. This score reflects the activity level or importance of the event node. Subsequently, based on these calculated access popularity scores and a preset storage medium selection strategy, the system intelligently selects the most suitable target storage medium for event nodes with different access popularity and their associated data element nodes and atomic data nodes. For example, for event nodes with high access popularity, the system uses a dual-write mechanism of memory caching and persistent storage for their associated data element nodes to ensure fast data retrieval while maintaining data reliability. For event nodes with lower access popularity, they may be scheduled to use lower-cost persistent storage media. In this way, this application achieves hierarchical storage management of data in the structured data capsule model, enabling faster response times for frequently accessed data and optimizing the overall utilization efficiency of storage resources. This mechanism, combined with the aforementioned structured data capsule model construction and update process, forms a dynamic and efficient data management system that effectively solves the performance bottleneck of large-scale health data storage and access.

[0085] In one feasible embodiment, after constructing the structured data capsule model of the target individual, the system can periodically or upon triggering specific events evaluate the popularity of event nodes in the model. For example, a popularity scoring function can be defined, which comprehensively considers the number of times an event node has been accessed in the past week, the time of its most recent access, and the importance weight of the event type. If an event node has been accessed more than 100 times in the past week, and its most recent access was within 24 hours, and its event type is marked as "high-risk warning," its popularity score will be significantly increased. Based on this popularity score, the system can adopt a hierarchical storage strategy. For example, a popularity threshold can be set; event nodes exceeding this threshold, along with their associated data element nodes and atomic data nodes, are considered "hot data." This hot data is scheduled to be cached in a high-performance in-memory database and simultaneously written to a distributed file system for persistent storage, forming a dual-write mechanism. When new data is written or updated, both the in-memory database and the distributed file system are updated simultaneously. When reading this hot data, the system will preferentially retrieve it from the in-memory database to achieve millisecond-level response. Event nodes and their associated data below this threshold are considered "cold data" and are stored directly in the distributed file system to save memory resources.

[0086] 208. Generate dynamic attribute graphs based on the updated structured data capsule model; 209. Map the event nodes and / or data element nodes in the structured data capsule model to the standard medical ontology terminology system to establish a unified semantic index; In this step, the event nodes and / or data element nodes in the structured data capsule model are vectorized into text to obtain corresponding semantic vectors; the cosine similarity between the text vectors and the corresponding term vectors in the standard medical term vector library is calculated; if the cosine similarity value is greater than or equal to a preset acceptance threshold, the corresponding event nodes and / or data element nodes are mapped to the standard terms corresponding to the cosine similarity, and a unified semantic index is recommended.

[0087] Specifically, text vectorization is the process of converting textual information into numerical vectors that capture the semantic information of the text. For example, methods based on word embeddings (such as Word2Vec, GloVe, and FastText) can be used to map words into a low-dimensional dense vector space, and then the vector representation of the entire text can be obtained through averaging, weighted averaging, etc.; alternatively, methods based on pre-trained language models (such as BERT and RoBERTa) can be used to directly generate high-quality text vectors using their powerful semantic understanding capabilities. Semantic vectors are the result of text vectorization; they represent the semantic meaning of the text in a multi-dimensional space. Their characteristic is that semantically similar texts are closer together in the vector space, while semantically dissimilar texts are farther apart. A standard medical terminology vector library is a pre-built database that contains standard medical ontology terms (such as SNOMEDCT, ICD-10, LOINC, etc.) and their corresponding semantic vectors. These terminology vectors can be obtained by training on a large corpus of medical texts or by directly constructing them using the structural information of existing medical ontologies. Cosine similarity is a metric that measures the cosine of the angle between two non-zero vectors. It's commonly used to assess text similarity, with values ​​ranging from -1 to 1. A value closer to 1 indicates a more consistent direction between the two vectors, meaning greater semantic similarity. A preset acceptance threshold is a numerical standard used to determine if the semantic similarity is high enough for mapping. This threshold can be set according to the specific application scenario and the required mapping accuracy; for example, it could be set to 0.8 or 0.9. Only when the calculated cosine similarity value reaches or exceeds this threshold is a valid semantic association considered to exist between nodes in the structured data capsule model and standard medical terms. Mapping to the standard term corresponding to cosine similarity means establishing an association between event nodes and / or data element nodes in the structured data capsule model and the standard medical term vector library that is semantically most similar and meets the threshold condition. This mapping can be a direct ID association or storing a reference to the standard term. Establishing a unified semantic index refers to providing a standardized and interoperable semantic identifier for event nodes and / or data element nodes with non-standardized or internal representations in the structured data capsule model through the above mapping process. This index enables health data from different sources and with different description methods to be understood, queried, and analyzed within a unified semantic framework.

[0088] This application's solution performs text vectorization on event nodes and / or data element nodes in a structured data capsule model, converting the text descriptions of these nodes into high-dimensional semantic vectors. This allows textual information, which was previously difficult to compare directly, to be quantified and calculated numerically. Subsequently, these semantic vectors are compared with terminology vectors in a pre-built standard medical terminology vector library using cosine similarity calculation. Cosine similarity effectively measures the directional consistency between two vectors in the semantic space, reflecting the degree of semantic association between them. By setting a preset acceptance threshold, the system can determine whether the calculated similarity is sufficiently high. If the similarity value is greater than or equal to the threshold, a reliable semantic correspondence is considered to exist between the node in the structured data capsule model and the corresponding standard medical terminology, and the node is then mapped to that standard term. Through this series of steps, this application's solution systematically aligns the internally represented health data with the external standard medical ontology terminology system, thereby establishing a unified semantic index. This not only overcomes the obstacles caused by inconsistent data descriptions but also lays a solid foundation for subsequent knowledge reasoning and data integration based on standard medical ontology, significantly improving the standardization and interoperability of health data processing.

[0089] The following example illustrates this. Suppose there is an event node in the structured data capsule model with the text description "hyperglycemia attack". First, the system uses a pre-trained word embedding model (e.g., a FastText model trained on a medical corpus) to convert the phrase "hyperglycemia attack" into a semantic vector. Simultaneously, the system maintains a standard medical terminology vector library, which includes standard terms such as "hyperglycemia", "diabetes", and "acute hyperglycemia" and their corresponding semantic vectors. Next, the system calculates the cosine similarity between the semantic vector of "hyperglycemia attack" and all term vectors in the standard medical terminology vector library. For example, the calculation result might show a cosine similarity of 0.92 with "hyperglycemia" and 0.75 with "diabetes". If the preset acceptance threshold is set to 0.90, then since 0.92 is greater than 0.90, the system will map the event node "hyperglycemia attack" to the standard term "hyperglycemia" and establish a corresponding semantic index. In this way, even if there are subtle differences in the original data description, it can be accurately normalized to standard medical concepts, ensuring the consistency of data semantics.

[0090] For example, a multi-source term mapping mechanism is established to map L2 / L3 layer nodes to the standard medical ontology space O. The mapping function M(x)→tstd is defined, and a semantic vector matching algorithm based on the Bert-based BioBERT pre-trained model is used. In the L2 event mapping, the L2 event description text is vectorized into Vevent, and its cosine similarity with the ICD-11 disease code and SNOMEDCT event terminology vector library Vstd is calculated: .

[0091] Set the acceptance threshold δ=0.95 (corresponding to an association accuracy of ≥95%) and select standard terms that satisfy max(Sim)≥δ (e.g., map “blood sugar spike” to the SNOMEDCT concept ID “Hyperglycemia due to insulin resistance”).

[0092] In the L3 layer data element mapping, a strategy combining exact matching and fuzzy matching is adopted to map L3 layer data element attributes to the LOINC laboratory terminology set (e.g., "blood glucose index group" is mapped to LOINC Code "2345-7"), thus constructing a unified semantic index.

[0093] 210. Based on the knowledge graph of energy metabolism diseases, establish semantic associations between event nodes and potential metabolic mechanism nodes through multi-hop reasoning, and update them in the structured data capsule model; In this step, based on the set of connected paths between event nodes and metabolic mechanism nodes in the knowledge graph of energy metabolism diseases, a predefined association strength scoring function is used to calculate the causal closeness between event nodes and metabolic mechanisms; if the causal density is greater than or equal to the strong association threshold, then a semantic association between event nodes and metabolic mechanisms is established in the structured data capsule model.

[0094] Wherein, the correlation strength scoring function

[0095]

[0096] in, node To metabolic mechanisms The set of all connected paths, For path Mid-single jump relationship The weight, Metabolic mechanism nodes in the knowledge graph of energy metabolism diseases The confidence level of the prior probability.

[0097] In the knowledge graph of energy metabolism diseases, the set of connected paths between event nodes and metabolic mechanism nodes refers to the set of all possible relational paths between a specific event node and a potential metabolic mechanism node within a pre-constructed knowledge graph containing energy metabolism-related entities (such as diseases, symptoms, test indicators, genes, proteins, metabolic pathways, etc.) and their interrelationships (such as "cause," "influence," "participate in," "is the mechanism of..."). These paths encode complex, multi-layered association information between events and mechanisms and form the basis for calculating causal tightness. This set of connected paths can be obtained by searching all paths from a specified event node to a metabolic mechanism node in the knowledge graph using graph traversal algorithms, such as depth-first search (DFS) or breadth-first search (BFS); or, these paths can be efficiently retrieved and extracted using specialized graph database query languages, such as Cypher or Gremlin.

[0098] A predefined association strength scoring function is a mathematical model or algorithm used to quantify the strength or reliability of the causal relationship between event nodes and metabolic mechanisms. This function comprehensively considers multiple factors, including the length of the connected path, the type of relationship on the path, the importance of nodes, and the confidence level of each element on the path. Its role is to transform complex knowledge graph path information into a comparable numerical value, thereby determining the strength of the association. For example, the function can be designed to penalize path length (the longer the path, the lower the association strength), assign higher weights to specific types of causal relationships, and combine the confidence levels of all nodes and edges on the path through product or weighted summation. Furthermore, this function can also be a machine learning-based model, trained on a labeled causal relationship dataset to learn how to predict causal tightness from path features.

[0099] Causal tightness is a numerical value calculated using the aforementioned association strength scoring function. It directly reflects the strength or reliability of the causal relationship between event nodes and metabolic mechanism nodes. This value serves as the basis for determining whether to establish semantic associations in the structured data capsule model.

[0100] A strong association threshold is a preset value used to define the minimum acceptable level of causal closeness. This threshold serves as a screening criterion, ensuring that only sufficiently strong, clinically or biologically significant associations are established, thereby avoiding the introduction of weak associations or noisy information and guaranteeing the quality and accuracy of the structured data capsule model. This threshold can be set by domain experts based on clinical experience, biological principles, or existing research evidence; alternatively, it can be determined through data-driven methods, such as performance evaluation on validation datasets, to select the optimal threshold that maximizes predictive accuracy or F1 score.

[0101] Establishing semantic associations between event nodes and metabolic mechanisms in a structured data capsule model refers to creating new edges or relationships between event nodes and metabolic mechanism nodes within the data structure of the structured data capsule model to explicitly represent their validated semantic associations. This can be manifested by adding a reference or link to the relevant metabolic mechanism node in the event node's data structure, with attributes such as association strength; or by creating a new type of relationship (e.g., "potential mechanism is", "causally related to") to connect event nodes and metabolic mechanism nodes, thereby enriching the model's semantic information.

[0102] This application's solution utilizes rich semantic information from a knowledge graph in the field of energy metabolism diseases and combines it with a predefined association strength scoring function to quantitatively evaluate the connectivity paths between event nodes and metabolic mechanism nodes, thereby calculating the causal tightness. Subsequently, the calculated causal tightness is compared with a preset strong association threshold. Only when the causal tightness reaches or exceeds this threshold is the semantic association between the event node and the metabolic mechanism explicitly established in the structured data capsule model. This process ensures that the established associations are rigorously screened and highly reliable, avoiding the introduction of weak or uncertain information into the health data model. In this way, the structured data capsule model can delve deeper from apparent health events to the underlying biological mechanisms, providing a deeper insight into understanding the health status of target individuals.

[0103] As a specific implementation method, assume there is a "hyperglycemia event" node in the structured data capsule model and a "metabolic mechanism" node in the energy metabolism disease domain knowledge graph. The system first searches the energy metabolism disease domain knowledge graph for all connected paths from "hyperglycemia event" to "insulin resistance." For example, it might find a path: "hyperglycemia" → "leads to" → "increased insulin secretion" → "long-term effects" → "insulin receptor downregulation" → "manifests as" → "insulin resistance." Simultaneously, it might find another path: "hyperglycemia" → "related to..." → "obesity" → "leads to" → "abnormal adipokines" → "induces" → "insulin resistance." Next, a predefined association strength scoring function is used to calculate the causal strength of these paths. This function can weight the paths based on factors such as path length, relationship type (e.g., "leads to" has a higher weight than "related to..."), and the confidence level of the nodes on the path. Assume the calculated causal strength of the first path is 0.85, and the causal strength of the second path is 0.70. If the preset strong association threshold is 0.75, then since 0.85 is greater than 0.75, the system will establish a semantic association between the "hyperglycemia event" node and the "insulin resistance" node in the structured data capsule model, for example, by adding an edge marked "potential mechanism is". However, for paths with a causal tightness of 0.70, since they do not reach the strong association threshold, no corresponding semantic association will be established.

[0104] In another embodiment, after step 210, the method further includes: Context-Aware Tagging: A dynamic rule-based inference engine is constructed, introducing L1 layer basic features as the context vector CL1, and conditionally semantically annotating L3 layer data elements DL3. The tag generation function Tag(DL3|CL1) is defined as follows: .

[0105] For example, define the rule: if CL1.BMI≥28 (obese) and DL3.Type="Carbohydrate_Intake", then automatically trigger the injection of the "high-risk" label. This label has dynamic listening characteristics (Observer Pattern), and once CL1 in the L1 layer is updated, the inference engine recalculates the label state in real time.

[0106] Semantic Subgraph Retrieval: This technology overcomes the limitations of traditional keyword matching by employing "Query Expansion." When a user inputs a query term q (e.g., "insulin resistance"), the system first retrieves its neighboring node set N(q) from the knowledge graph, constructing an expanded query set Qexp={q}∪N(q). The retrieval result set R is defined as follows: .

[0107] The graph traversal algorithm returns a complete subgraph containing L2 events, L3 data elements, and L4 atomic data in one go. Experiments show that the combined recall and precision of this mechanism are ≥92%, and the retrieval efficiency is improved by 60% compared to the traditional inverted index due to the reduction of multiple IO queries.

[0108] In another embodiment, after establishing semantic associations between event nodes and potential metabolic mechanism nodes through multi-hop reasoning based on the knowledge graph of energy metabolism diseases, and updating it in the structured data capsule model, the method further includes: In response to a push request from a user on the monitoring application interface, key decision feature data is projected and extracted from the structured data capsule model, packaged into a standard medical data exchange resource package, and pushed to the monitoring application with a notification.

[0109] Specifically, responding to push requests from users on the monitoring application interface refers to the system receiving data push instructions issued by users through the user interface of the monitoring application. Its function is to trigger subsequent data extraction, packaging, and push processes, ensuring that data transmission is performed on demand. For example, users can click the "Get Latest Report" button in the monitoring application, or the system can automatically trigger pushes at specific times according to user-preset timed push rules. Projecting and extracting key decision-making feature data from the structured data capsule model refers to filtering and extracting a subset of data that is significant for user decision-making from the complex structured data capsule model. Its function is to refine information, avoid presenting users with too much raw data, and improve information readability and decision-making efficiency. This can be achieved by predefining a set of data extraction rules, such as extracting only the most recently occurring health events, high-risk metabolic mechanism associations, and key physiological indicator summary data; or dynamically selecting the data type and granularity to be extracted based on the user's personalized configuration in the monitoring application. Packaging into a standard medical data exchange resource package refers to formatting and packaging the extracted key decision-making feature data according to preset medical data exchange standards. Its function is to ensure that data can interoperate, be correctly parsed and understood between different medical information systems or applications, and improve the efficiency and accuracy of data exchange. For example, the HL7 FHIR (Fast Healthcare Interoperability Resources) standard can be used to encapsulate data into resource packages in JSON or XML format, or the DICOM (Digital Imaging and Communications in Medicine) standard can be used to encapsulate specific types of medical image or report data. Pushing to the monitoring application and providing notifications refers to sending the encapsulated standard medical data exchange resource package to the user's monitoring application and providing corresponding notifications or reminders on the application interface. Its function is to promptly deliver processed health data to the user and ensure that the user is aware of the arrival of new data. This can be achieved by pushing data to the monitoring application in real time via the WebSocket protocol and displaying a notification on the application interface; or by sending notifications via push notification services on mobile devices (such as APNs or FCM), allowing the user to access the application and view the data after clicking the notification.

[0110] Understandably, when the system receives a push request from the user on the monitoring application interface, this request acts as a trigger signal, initiating the data extraction process. The system intelligently "projects" and extracts key decision-making feature data from the structured data capsule model. This "projection extraction" is not a simple data copying process, but rather, based on preset rules or the user's personalized needs, it filters out the key information that best reflects the target individual's health status, risk assessment, and potential metabolic mechanisms from the multi-level structured data capsule model, thus avoiding redundancy and complexity of the original data. Subsequently, to ensure that this key information can flow seamlessly and be correctly interpreted between different medical information systems or monitoring applications, the extracted data is packaged into a standard medical data exchange resource package. This standardized packaging not only ensures data interoperability but also lays the foundation for subsequent integration and analysis. Finally, the packaged resource package is pushed to the user's monitoring application along with prompts, ensuring that the user can obtain highly refined and standardized health data in a timely and convenient manner, thereby assisting the user in making more informed health management decisions. In this way, this application transforms in-depth health data analysis results into standardized information that users can directly utilize, greatly improving the efficiency and security of users acquiring and understanding complex health data.

[0111] In this embodiment, the step of projecting and extracting key decision feature data from the structured data capsule model includes: Extract event summary information from the health event layer of the structured data capsule model; Aggregated data, after aggregation calculation, is extracted from the data element layer of the structured data capsule model; Sensitive identity information in the identity feature layer and detailed atomic data in the atomic data unit layer of the structured data capsule model are filtered to form the key decision feature data.

[0112] Extracting event summary information from the health event layer of the structured data capsule model refers to filtering out refined information that summarizes the core content, key time points, or main impacts of an event from the health event layer. The health event layer is a crucial component of the structured data capsule model, used to record the event attributes of each health event, indexed by energy metabolism-related health events. For example, for a "hospitalization for hyperglycemia" event, the summary information could be "admitted for treatment due to hyperglycemia"; for a "medication adjustment" event, the summary information could be "insulin dosage adjustment." This extraction can be achieved based on predefined rules, keyword matching, or natural language processing techniques. Extracting aggregated data after aggregation calculation from the data element layer of the structured data capsule model refers to obtaining data that has undergone statistical, summary, or trend analysis processing from the data element layer. The data element layer is used to classify and aggregate data associated with each health event in the health event layer according to medical logic dimensions. For example, it can extract the patient's average glycated hemoglobin (HbA1c) value over the past three months, blood glucose fluctuation range, and the increasing or decreasing trends of specific indicators. These aggregated data are typically more valuable for decision-making than the raw atomic data, providing a macro-level view of health status. Filtering sensitive identity information in the identity feature layer and detailed atomic data in the atomic data unit layer of the structured data capsule model refers to anonymizing, pseudonyming, or directly removing personally identifiable information (such as name, ID number, precise date of birth, etc.) contained in the identity feature layer to protect patient privacy. The identity feature layer is used to bind the identity identifier and basic characteristic information of the target individual. Simultaneously, raw data points in the atomic data unit layer that are too detailed, not directly used for decision-making, or may indirectly leak privacy (such as each blood glucose measurement value, detailed medication records, etc.) are filtered or omitted, retaining only high-level information or aggregated results useful for decision-making. The atomic data unit layer is used to store the specific data values ​​and status labels corresponding to various types of data in the data element layer. The resulting key decision feature data refers to a refined, secure, and highly valuable set of data generated after the above extraction and filtering processes. This data is designed to provide surveillance applications with the necessary and sufficient information to quickly assess the health status of target individuals, identify potential risks, or assist in developing intervention measures, while avoiding information redundancy and privacy breaches.

[0113] Specifically, firstly, event summary information is extracted from the health event layer, enabling monitoring applications to quickly understand the key health events experienced by a target individual and grasp the overall picture of their health journey. Secondly, aggregated data, after aggregation calculations, is extracted from the data element layer. This provides statistically significant indicators, such as averages and trends, processed with professional medical logic, avoiding the need for monitoring applications to directly process large amounts of raw data and improving decision-making efficiency. Simultaneously, to ensure data security and privacy protection, this solution rigorously filters sensitive identity information in the identity feature layer and detailed atomic data in the atomic data unit layer. This filtering mechanism ensures that the data pushed to the monitoring application meets both decision-making needs and privacy protection standards. Through the synergistic effect of the above steps, this solution transforms the rich but complex raw data in the structured data capsule model into refined, secure, and easily understandable key decision feature data, effectively solving the problems of information overload, privacy leakage, and low decision-making efficiency that monitoring applications may face when receiving and processing raw data.

[0114] By implementing the methods provided above, compared with existing technologies, this invention achieves significant improvements in data organization efficiency, dynamic adaptability, semantic understanding ability, and system engineering performance through a collaborative technology system of "multi-level structural modeling + event-driven update + semantic association + interface adaptation," specifically including: By constructing a four-level structured model consisting of an identity feature layer, a health event layer, a data element layer, and an atomic data unit layer, a unified hierarchical organization of multi-source heterogeneous health data is achieved, avoiding the data fragmentation problem caused by the traditional flat storage of "timestamp + numerical value". At the same time, through a combination of tree and network association mechanisms, a stable mapping relationship is formed between health events and related data.

[0115] By introducing an event-driven mechanism, when a new health event, data correction, or state change is detected, topology expansion, relationship reconstruction, and event level adjustment are automatically triggered without manual intervention in the database structure. At the same time, incremental updates and local locking strategies are adopted to update only the relevant subgraphs, reducing system overhead.

[0116] By constructing a many-to-many relationship between health events and data elements, and combining it with a correlation strength calculation mechanism based on time difference, semantic similarity, and data quality, the system can dynamically identify key influencing factors; at the same time, through the hierarchical path of "event-data element-atomic data", the system can achieve full-link traceability.

[0117] By introducing medical semantic association rules, health events and data elements are mapped to a standard medical terminology system, and semantic associations between health events and metabolic mechanisms are established based on knowledge graphs, transforming raw numerical data into information with medical semantics; at the same time, personalized semantic enhancement is achieved through a context-aware tagging mechanism.

[0118] Through standardized interface adaptation design, the system enables unified access and standardized output of multi-source heterogeneous data, and supports two modes: full-scale atlas output and simplified projection output. This allows the system to be used independently in clinical research or as a data input module for core systems.

[0119] By adopting a tiered storage strategy based on access frequency, a dual-write mechanism of caching and distributed storage is used for high-frequency events, while persistent storage is used for low-frequency data, thereby reducing storage costs while ensuring query performance.

[0120] By employing an attribute-based access control (ABAC) mechanism, permissions are verified for different user roles, and differential privacy or masking is performed in conjunction with data tags to achieve fine-grained data security control. All data access is verified at the interface layer, effectively preventing the leakage of sensitive information and meeting medical data compliance requirements.

[0121] By constructing an event-driven and semantically enhanced data processing system, data can be transformed from "passive storage" to "active service," directly supporting clinical decision analysis, artificial intelligence model training, and multi-center research. In the long-term management of energy metabolism diseases, continuous dynamic tracking and risk warning can be achieved, thereby improving the efficiency of diagnosis and treatment and the accuracy of decision-making.

[0122] The above describes the health data processing method for energy metabolism diseases in embodiments of the present invention. The following describes the health data processing apparatus for energy metabolism diseases in embodiments of the present invention. Please refer to [link / reference]. Figure 3 One embodiment of the energy metabolism disease health data processing device of the present invention includes: Data acquisition module 310 is used to acquire multi-source heterogeneous health data related to the target individual; The model building module 320 is used to extract different dimensions of data features from the multi-source heterogeneous health data, and to build a structured data capsule model that records the health status of the target individual based on a preset hierarchical hybrid association architecture and data features of each dimension. The structured data capsule model includes at least two levels, including an event structure layer for indexing health events and individuals and a data structure layer for aggregating associated data. The dynamic update module 330 is used to respond to the detection of a new event related to the target individual and update the data features of the new event to the structured data capsule model according to a preset update mechanism. Data output module 340 is used to generate dynamic attribute graphs based on the updated structured data capsule model.

[0123] In this embodiment, the event structure layer includes an identity feature layer and a health event layer, and the data structure layer includes a data element layer and an atomic data unit layer; wherein: The identity feature layer is used to bind the identity identifier and basic feature information of the target individual; The health event layer is used to record the event attributes of each health event, indexed by energy metabolism-related health events. The data element layer is used to classify and aggregate the data associated with each health event in the health event layer according to the medical logic dimension; The atomic data unit layer is used to store the specific data values ​​and status labels corresponding to various types of data in the data element layer.

[0124] In this embodiment, the relationships between the levels in the structured data capsule model are a hybrid tree-and-network architecture, wherein: The model building module 320 is specifically used for: An identity feature layer is constructed based on the identity features in the data features of each dimension; Event nodes are constructed based on event features from data features across various dimensions to obtain a health event layer; Data element nodes are constructed based on the indicator features in the data features of each dimension, thus obtaining the data element layer; Atomic data nodes are constructed based on the indicator value features in the data features of each dimension, resulting in an atomic data unit layer; The identity feature layer is used as the root node, and all event nodes in the health event layer are associated with it. The event nodes in the health event layer are associated with one or more data element nodes in the data element layer by using the association index; The data element nodes in the data element layer are associated with multiple event nodes in the health event layer through event association identifiers; By associating the data element nodes in the data element layer with one or more corresponding atomic data nodes in the atomic data unit layer, a structured data capsule model is obtained.

[0125] In this embodiment, the dynamic update module 330 is specifically used for: Extract data features from new events and determine whether the corresponding data features inside the structured data capsule model meet the triggering conditions in the preset update mechanism. The triggering conditions include at least one of the following: new event triggering conditions, data correction triggering conditions, and state change triggering conditions. If the conditions for triggering the new event are met, a new event node is instantiated in the structured data capsule model, and related data element nodes and related edges are instantiated according to the mapping rules in the preset update mechanism to obtain a new structured data capsule model. If the data correction triggering condition is met, the data value of the corresponding atomic data node in the atomic data unit layer is updated based on the data characteristics, and the correlation strength between the affected event node and the data element node is recalculated to obtain a new structured data capsule model. If the state change triggering condition is met, the priority level of the corresponding health event is calculated based on the value of the data feature, and its position in the health event layer is adjusted based on the priority level.

[0126] In this embodiment, the dynamic update module 330 is further configured to: In the process of instantiating related data element nodes and related edges according to the mapping rules in the preset update mechanism to obtain a new structured data capsule model, constraint verification is performed on any newly added event nodes and data element nodes in the new structured data capsule model based on a predefined set of medical logic constraints. If the constraint verification shows that the newly added event node and data element node conflict with any constraint in the medical logic constraint set, then the update operation for the new event will be blocked.

[0127] In this embodiment, the dynamic update module 330 is further configured to: The event nodes and / or data element nodes in the structured data capsule model are mapped to the standard medical ontology terminology system to establish a unified semantic index. Based on a knowledge graph in the field of energy metabolism diseases, semantic associations between event nodes and potential metabolic mechanism nodes are established through multi-hop reasoning and updated into a structured data capsule model.

[0128] In this embodiment, the dynamic update module 330 is specifically used for: Based on the set of connected paths between event nodes and metabolic mechanism nodes in the knowledge graph of energy metabolism diseases, a predefined association strength scoring function is used to calculate the causal tightness between event nodes and metabolic mechanisms. If the semantic association between event nodes and metabolic mechanisms is established in the causal density structured data capsule model.

[0129] In this embodiment of the invention, the energy metabolism disease health data processing device operates the aforementioned energy metabolism disease health data processing method to acquire multi-source heterogeneous health data related to a target individual; extracts different dimensional data features from the multi-source heterogeneous health data, and constructs a structured data capsule model recording the health status of the target individual based on a preset hierarchical hybrid association architecture and the data features of each dimension. The structured data capsule model includes at least two levels; in response to the detection of new events related to the target individual, the data features of the new events are updated to the structured data capsule model according to a preset update mechanism; and a dynamic attribute graph is generated based on the updated structured data capsule model. By using multi-source heterogeneous health data, constructing a structured data capsule model, dynamically updating in response to new events, and generating a dynamic attribute graph, the problems of overly fragmented health data and lack of dynamic adjustment in the prior art are solved.

[0130] above Figure 3 The energy metabolism disease health data processing device in the embodiments of the present invention will be described in detail from the perspective of modular functional entities. The electronic device in the embodiments of the present invention will be described in detail from the perspective of hardware processing.

[0131] Figure 4 This is a schematic diagram of the structure of an electronic device 400 provided in an embodiment of the present invention. The electronic device 400 can vary significantly due to differences in configuration or performance. It may include one or more central processing units (CPUs) 410 (e.g., one or more processors) and a memory 420, and one or more storage media 430 (e.g., one or more mass storage devices) for storing application programs 433 or data 432. The memory 420 and storage media 430 can be temporary or persistent storage. The program stored in the storage media 430 may include one or more modules (not shown in the diagram), each module including a series of request operations on the electronic device 400. Furthermore, the processor 410 may be configured to communicate with the storage media 430 and execute the series of request operations on the storage media 430 on the electronic device 400 to implement the steps of the above-described energy metabolism disease health data processing method.

[0132] Electronic device 400 may also include one or more power supplies 440, one or more wired or wireless network interfaces 450, one or more input / output interfaces 460, and / or one or more operating systems 431, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 4The illustrated electronic device structure does not constitute a limitation on the electronic device provided in this application. It may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0133] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores a request that, when the request is executed on a computer, causes the computer to perform the steps of the energy metabolism disease health data processing method.

[0134] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0135] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several requests to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0136] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for processing health data related to energy metabolism diseases, characterized in that, The method includes: Acquire multi-source heterogeneous health data related to the target individual; Extract different dimensions of data features from the multi-source heterogeneous health data, and construct a structured data capsule model that records the health status of the target individual based on a preset hierarchical hybrid association architecture and data features of each dimension. The structured data capsule model includes at least two levels, including an event structure layer for indexing health events and individuals and a data structure layer for aggregating associated data. In response to detecting a new event related to the target individual, the data features of the new event are updated to the structured data capsule model according to a preset update mechanism; Dynamic attribute graphs are generated based on the updated structured data capsule model.

2. The method for processing health data on energy metabolism diseases according to claim 1, characterized in that, The event structure layer includes an identity feature layer and a health event layer; the data structure layer includes a data element layer and an atomic data unit layer; wherein: The identity feature layer is used to bind the identity identifier and basic feature information of the target individual; The health event layer is used to record the event attributes of each health event, indexed by energy metabolism-related health events. The data element layer is used to classify and aggregate the data associated with each health event in the health event layer according to the medical logic dimension; The atomic data unit layer is used to store the specific data values ​​and status labels corresponding to various types of data in the data element layer.

3. The method for processing health data on energy metabolism diseases according to claim 2, characterized in that, The structured data capsule model uses a hybrid tree-and-network relational architecture between its various levels, where: The structured data capsule model for recording the health status of the target individual is constructed based on a preset hierarchical hybrid association architecture and data features of various dimensions, including: An identity feature layer is constructed based on the identity features in the data features of each dimension; Event nodes are constructed based on event features from data features across various dimensions to obtain a health event layer; Data element nodes are constructed based on the indicator features in the data features of each dimension, thus obtaining the data element layer; Atomic data nodes are constructed based on the indicator value features in the data features of each dimension, resulting in an atomic data unit layer; The identity feature layer is used as the root node, and all event nodes in the health event layer are associated with it. The event nodes in the health event layer are associated with one or more data element nodes in the data element layer by using the association index; The data element nodes in the data element layer are associated with multiple event nodes in the health event layer through event association identifiers; By associating the data element nodes in the data element layer with one or more corresponding atomic data nodes in the atomic data unit layer, a structured data capsule model is obtained.

4. The method for processing health data on energy metabolism diseases according to claim 1, characterized in that, The response to detecting a new event related to the target individual, and updating the data features of the new event to the structured data capsule model according to a preset update mechanism, includes: Extract data features from new events and determine whether the corresponding data features inside the structured data capsule model meet the triggering conditions in the preset update mechanism. The triggering conditions include at least one of the following: new event triggering conditions, data correction triggering conditions, and state change triggering conditions. If the conditions for triggering the new event are met, a new event node is instantiated in the structured data capsule model, and related data element nodes and related edges are instantiated according to the mapping rules in the preset update mechanism to obtain a new structured data capsule model. If the data correction triggering condition is met, the data value of the corresponding atomic data node in the atomic data unit layer is updated based on the data characteristics, and the correlation strength between the affected event node and the data element node is recalculated to obtain a new structured data capsule model. If the state change triggering condition is met, the priority level of the corresponding health event is calculated based on the value of the data feature, and its position in the health event layer is adjusted based on the priority level.

5. The method for processing health data on energy metabolism diseases according to claim 4, characterized in that, The step of responding to the detection of a new event related to the target individual and updating the data features of the new event to the structured data capsule model according to a preset update mechanism further includes: In the process of instantiating related data element nodes and related edges according to the mapping rules in the preset update mechanism to obtain a new structured data capsule model, constraint verification is performed on any newly added event nodes and data element nodes in the new structured data capsule model based on a predefined set of medical logic constraints. If the constraint verification shows that the newly added event node and data element node conflict with any constraint in the medical logic constraint set, then the update operation for the new event will be blocked.

6. The method for processing health data on energy metabolism diseases according to any one of claims 1-5, characterized in that, After generating the dynamic attribute graph based on the updated structured data capsule model, the method further includes: The event nodes and / or data element nodes in the structured data capsule model are mapped to the standard medical ontology terminology system to establish a unified semantic index. Based on a knowledge graph in the field of energy metabolism diseases, semantic associations between event nodes and potential metabolic mechanism nodes are established through multi-hop reasoning and updated into a structured data capsule model.

7. The method for processing health data on energy metabolism diseases according to claim 6, characterized in that, The knowledge graph based on the energy metabolism disease domain establishes semantic associations between event nodes and potential metabolic mechanism nodes through multi-hop reasoning, and updates them to the structured data capsule model, including: Based on the set of connection paths between event nodes and metabolic mechanism nodes in the knowledge graph of energy metabolism diseases, a predefined association strength scoring function is used to calculate the event node... With metabolic mechanisms causal tightness If the causal density is greater than or equal to the strong correlation threshold, then an event node is established in the structured data capsule model. With metabolic mechanisms Semantic associations.

8. A health data processing device for energy metabolism diseases, characterized in that, The device includes: The data acquisition module is used to acquire multi-source heterogeneous health data related to the target individual; The model building module is used to extract different dimensions of data features from the multi-source heterogeneous health data, and to build a structured data capsule model that records the health status of the target individual based on a preset hierarchical hybrid association architecture and data features of each dimension. The structured data capsule model includes at least two levels, including an event structure layer for indexing health events and individuals and a data structure layer for aggregating associated data. A dynamic update module is used to respond to the detection of new events related to the target individual and update the data features of the new events to the structured data capsule model according to a preset update mechanism; The data output module is used to generate dynamic attribute graphs based on the updated structured data capsule model.

9. An electronic device, characterized in that, The electronic device includes a memory and at least one processor, the memory storing instructions; the at least one processor invokes the instructions in the memory to cause the electronic device to perform the steps of the energy metabolism disease health data processing method as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the various steps of the energy metabolism disease health data processing method as described in any one of claims 1-7.

Images