Special disease queue data capturing method and system based on intelligent medical knowledge graph

Abstract

The invention discloses a special disease queue data capturing method and system based on an intelligent medical knowledge graph, and relates to the technical field of medical information, and the method comprises the following steps: S1, constructing a special disease intelligent medical knowledge graph which comprises a bidirectional mapping relation between standard terms of a single disease category and clinical actual corpora, clinical text data is accumulated in a mode of combining manual annotation and machine learning, and a domain exclusive knowledge base containing symptoms, diagnosis and examination indexes is formed. According to the special disease queue data capturing method and system provided by the invention, by constructing the special disease intelligent medical knowledge graph, bidirectional mapping of single disease specification terms and clinical actual corpora is realized, and the problem of insufficient semantic understanding when non-standardized clinical corpora are processed by a traditional method is effectively solved; the entity information in the unstructured medical data can be accurately extracted by utilizing a natural language processing model and an inference engine.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a method and system for capturing disease-specific queue data based on intelligent medical knowledge graphs. Background Art

[0002] In the field of medical informationization and clinical research, the efficient collection and structured processing of disease-specific cohort data are key links in promoting the development of precision medicine. The large amount of unstructured medical data accumulated by hospitals, such as medical records, external hospital examination reports, imaging descriptions, etc., contains rich disease characteristics and diagnosis and treatment information, but the diversity of their formats and the complexity of their language expressions make it difficult for traditional technologies to be directly converted into structured data that can be used for scientific research analysis. When constructing a disease-specific cohort database for specific diseases such as tumors, cardiovascular diseases, and pituitary tumors, it is necessary to accurately capture key information such as symptoms, diagnoses, and examination indicators from massive non-standardized clinical corpus to support research such as drug intervention effect evaluation and disease risk prediction. However, existing technologies face many challenges in processing such data.

[0003] Traditional data capture methods primarily rely on techniques such as keyword matching and regular expression search. These methods have significant limitations when dealing with personalized physician descriptions, colloquial patient expressions, and complex medical terminology. On the one hand, their semantic understanding of non-standard clinical descriptions is insufficient, making it difficult to identify the inherent connections between diverse expressions such as "chest tightness" and "chest tightness," "rapid heartbeat" and "palpitations," resulting in incomplete data capture or misjudgment. On the other hand, they lack in-depth analysis of the contextual logic of medical terminology and are unable to infer implicit information from causal relationships between disease characteristics, affecting the integrity and accuracy of the data. Furthermore, existing knowledge graph update mechanisms rely on manual maintenance, making it difficult to incorporate newly emerging clinical corpora in real time. Their low level of automation makes them unable to adapt to the dynamic changes in medical knowledge, resulting in a system with limited processing capabilities for long-tail data. To address this issue, we propose a method and system for capturing disease-specific cohort data based on an intelligent medical knowledge graph. Summary of the Invention

[0004] To solve the above technical problems, a method and system for capturing disease-specific cohort data based on an intelligent medical knowledge graph are provided. This technical solution solves the problems of the above-mentioned methods relying on keyword matching, insufficient semantic understanding when processing non-standardized clinical corpus, difficulty in identifying diverse expression associations, and prone to errors in data capture; lack of term context analysis, inability to infer implicit information; reliance on manual labor for knowledge graph updates, low automation, difficulty in adapting to dynamic changes, and weak long-tail data processing capabilities.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is: The method for capturing disease-specific cohort data based on intelligent medical knowledge graph includes the following steps: S1. Build a disease-specific intelligent medical knowledge graph, which includes a bidirectional mapping relationship between standardized terminology for a single disease and actual clinical data. By combining manual annotation with machine learning, clinical text data is accumulated to form a domain-specific knowledge base containing symptoms, diagnoses, and examination indicators. S2: Input unstructured medical data into a pre-trained natural language processing model and extract entity information from the text through semantic analysis, including symptom descriptions, examination results, and diagnosis conclusions; S3. Match the extracted entities with the mapping relationships in the knowledge graph, use the disease feature association logic in the graph to infer implicit information, and generate structured intermediate results; S4. Update the knowledge graph through a dynamic self-optimization mechanism: When the model detects uncovered clinical descriptions, a workflow is triggered to collect new corpus annotated by doctors, which is then cleaned and formatted before being added to the knowledge graph. The model is then retrained to adapt to the updated graph. S5. Fill the matched structured data into the target queue database according to the disease-specific database paradigm to complete data capture.

[0006] Preferably, the construction of the disease-specific intelligent medical knowledge graph in step S1 specifically includes: Extract clinical data from unstructured medical records on the hospital intranet, external hospital examination reports, and image descriptions, and form an initial corpus through manual annotation; Establishing a mapping relationship between standard terms and clinical corpus, the mapping relationship includes synonym mapping, context association and semantic expansion rules to achieve the conversion of non-standard descriptions into standard terms; Use graph databases to store mapping relationships and disease feature associations, including the causal relationship between symptoms and diagnoses, and the association between examination indicators and disease stages; By continuously annotating new corpora and combining clustering algorithms to identify similar expressions, the coverage of the knowledge graph can be dynamically expanded; the knowledge graph is versioned to record the source of the corpus, the annotator, and the timestamp of each update.

[0007] Preferably, the logical inference in step S3 specifically includes: Based on the association of disease features in the knowledge graph, logical reasoning is performed on implicit symptoms that are not clearly described; Analyze the evolution of symptoms using time series data to infer the stage of disease development; Use the probabilistic model to calculate the matching degree between entities and disease characteristics, and set the threshold to filter high-confidence results; Conflict detection is performed on contradictory data, triggering a manual review process to correct mismatches.

[0008] Preferably, the dynamic self-optimization mechanism of step S4 specifically includes: Design a user interface for doctors to annotate unidentified clinical descriptions or correct incorrect matching results; The labeled data is automatically transferred to the backend through the workflow, and new mapping relationships are generated after deduplication and standardization. Trigger the incremental training process and fine-tune the model parameters only for the newly added corpus; Regularly evaluate the coverage and model accuracy of the knowledge graph and optimize the iteration strategy.

[0009] Preferably, the method further includes the following steps of image data processing: Extract features from medical images and generate structured description text; The image feature description and text data are input into the semantic processing module to perform multimodal data fusion analysis; Establish association rules between image features and disease diagnosis in the knowledge graph, including the mapping relationship between image description and diagnostic criteria; support doctors to mark image description errors through a workflow mechanism and update the image feature mapping relationship in the graph.

[0010] A disease-specific cohort data capture system based on an intelligent medical knowledge graph is used to implement the disease-specific cohort data capture method based on an intelligent medical knowledge graph, including: The knowledge graph construction module is used to integrate the standardized terminology of a single disease with actual clinical data, establish a bidirectional mapping relationship, and support dynamic updates; The semantic processing module is used to include pre-trained natural language processing models and inference engines for parsing unstructured text and extracting entities; The matching and reasoning module is used to compare the extracted entities with the knowledge graph and generate structured intermediate results based on the disease feature association rules; The dynamic update module is used to collect new corpus annotated by doctors through a workflow mechanism, automatically update the knowledge graph and trigger model retraining; The data warehousing module is used to store structured data in a specialized disease cohort database according to preset field formats, supports the joint processing of multimodal data, and integrates a privacy protection mechanism.

[0011] Preferably, the semantic processing module includes: The multimodal data integration unit supports the combined input of text, OCR-recognized image reports, and image feature descriptions; the entity recognition submodule uses a domain-adapted pre-trained language model to perform named entity recognition and distinguish symptoms, diagnoses, medications, and examination indicators; The context understanding submodule uses the attention mechanism to analyze the semantic associations between entities; The output interface converts the parsing results into a structured data format for subsequent module calls.

[0012] Preferably, the dynamic update module specifically includes: The annotation management unit is used to provide visual tools for doctors to circle text segments, add comments or select preset tags; The data cleaning unit is used to remove noise data using regular expressions and rule engines; The version control unit is used to record the update history of the knowledge graph and support version rollback; The model fine-tuning unit is used to perform incremental training tasks using a distributed training framework.

[0013] Preferably, the data storage module includes: Data verification unit, which checks the integrity of structured data based on the rules in the knowledge graph; Data conversion unit, mapping intermediate results to fixed fields according to the disease-specific database paradigm; Batch import interface, supporting scheduled tasks or real-time streaming writing; Data visualization unit generates statistical charts to show the distribution characteristics of queue data.

[0014] Preferably, the privacy protection measures of the privacy protection mechanism include: The entire process is deployed on the hospital intranet, and data transmission is prohibited; Encrypt and desensitize the patient's name and ID number to generate an anonymous identifier; Data access rights are managed in a hierarchical manner, and only authorized personnel can operate original data and knowledge graphs; Audit logs record all data operations, including query, modification, and export records.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The method and system for capturing data from specialized disease cohorts proposed in the present invention achieve a two-way mapping between standardized terminology for a single disease and actual clinical corpus by constructing an intelligent medical knowledge graph for specialized diseases, effectively solving the problem of insufficient semantic understanding of traditional methods when processing non-standardized clinical corpus. By utilizing natural language processing models and inference engines, it can accurately extract entity information from unstructured medical data and infer implicit information through the disease feature association logic in the knowledge graph, thereby improving the accuracy and completeness of data capture. The dynamic self-optimization mechanism ensures real-time updating of the knowledge graph, adapts to the dynamic changes of medical knowledge, enhances the system's processing capabilities for long-tail data, supports the joint processing of multimodal data, and integrates a privacy protection mechanism, which not only meets the needs of clinical research but also ensures the security of patient data. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the method of the present invention; Figure 2It is a system module framework diagram of the present invention. DETAILED DESCRIPTION

[0017] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0018] Reference Figure 1 As shown in FIG, the method for capturing disease-specific cohort data based on the intelligent medical knowledge graph includes the following steps: S1. Construct a disease-specific intelligent medical knowledge graph. This knowledge graph includes a bidirectional mapping relationship between standardized terminology for a single disease and actual clinical data. By combining manual annotation with machine learning, clinical text data is accumulated to form a domain-specific knowledge base encompassing symptoms, diagnoses, and examination indicators. The bidirectional mapping design takes into account both the standard terminology system and the diversity of actual clinical expressions. S2. Input unstructured medical data into a pre-trained natural language processing model. Through semantic analysis, the model extracts entity information from the text, including symptom descriptions, examination results, and diagnosis conclusions. The pre-trained model uses prior knowledge from a large-scale medical corpus to improve entity recognition accuracy. S3. Match the extracted entities with the mapping relationships in the knowledge graph, use the disease feature association logic in the graph to infer implicit information, generate structured intermediate results, and quantify the association probability between the entity and the disease through a logical reasoning formula, where the logical reasoning formula is: Where P(d|e) is the association probability of disease d given entity e, e is the input entity, d is the target disease, r∈R is the relationship type defined in the knowledge graph, and w r is the weight of relation r, f(e,r,d) is the indicator function, which takes the value of 1 when entity e is associated with disease d through relation r, otherwise it is 0, and e′ is all possible other entities in the knowledge graph; S4. Update the knowledge graph through a dynamic self-optimization mechanism: When the model detects uncovered clinical descriptions, a workflow is triggered to collect new data annotated by doctors. After cleaning and formatting, it is added to the knowledge graph, and the model is retrained to adapt to the updated graph. This closed-loop optimization ensures that the system continues to adapt to the evolution of clinical practice. S5. Fill the matched structured data into the target cohort database according to the disease-specific database paradigm to complete data capture and store them in a standardized manner to support subsequent scientific research analysis and clinical decision-making.

[0019] The construction of the disease-specific intelligent medical knowledge graph described in step S1 specifically includes: extracting actual clinical corpus from the unstructured medical records, external hospital examination reports and image descriptions on the hospital intranet, forming an initial corpus through manual annotation, and collecting multi-source data to ensure the comprehensiveness of knowledge coverage; establishing a mapping relationship between standard terms and clinical corpus, and the mapping relationship includes: synonym mapping, context association and semantic extension rules to realize the conversion of non-standard descriptions to standard terms, and the rule engine solves the ambiguity problem of clinical expression; using a graph database to store mapping relationships and disease feature associations, including the causal relationship between symptoms and diagnosis, and the association between examination indicators and disease stages. The graph structure supports efficient multi-hop relationship queries; by continuously annotating new corpus and combining clustering algorithms to identify similar expressions, the coverage of the knowledge graph is dynamically expanded, and incremental expansion reduces manual maintenance costs; the knowledge graph is versioned and managed, and the source of the corpus, the annotation personnel and the timestamp of each update are recorded. The version traceability function meets the compliance requirements of medical data.

[0020] The logical inference in step S3 specifically includes: based on the disease feature association in the knowledge graph, logical reasoning is performed on implicit symptoms that are not clearly described, and potential symptom discovery enhances data integrity; combining time series data to analyze the evolution of symptoms, infer the stage of disease development, and smooth the observed values ​​to identify trends. The time series model formula is: S t =αS t-1 +(1-α)O t Where S t is the smoothed symptom evolution state at the current time step t, S t-1 is the smoothed symptom evolution state of the previous time step t-1, O t is the original observation value at the current time step t, and α is the smoothing coefficient, which is used to control the weight ratio of the historical state and the current observation; The probabilistic model is used to calculate the matching degree between entities and disease characteristics, and thresholds are set to filter high-confidence results. Threshold filtering reduces interference from false positives. Conflict detection is performed on contradictory data, triggering a manual review process to correct incorrect matches. The human-computer collaborative mechanism ensures data quality.

[0021] The dynamic self-optimization mechanism of step S4 specifically includes: designing a user interface for doctors to annotate unidentified clinical descriptions or correct erroneous matching results, and visual tools to improve the efficiency of doctor participation; automatically transmitting the annotated data to the background through the workflow, generating new mapping relationships after deduplication and standardization, and the automated process shortens the knowledge update cycle; triggering the incremental training process, only fine-tuning the model parameters for the newly added corpus, and the incremental learning strategy saves computing resources; regularly evaluating the coverage and model accuracy of the knowledge graph, optimizing the iteration strategy, and using evaluation indicators to drive continuous improvement of the system.

[0022] It also includes image data processing steps: feature extraction of medical images, generation of structured descriptive text, and multimodal fusion to expand data dimensions; inputting image feature descriptions and text data into the semantic processing module for multimodal data fusion analysis, and cross-modal association to improve diagnostic accuracy; establishing association rules between image features and disease diagnosis in the knowledge graph, including the mapping relationship between image descriptions and diagnostic standards, and integrating image knowledge to improve the disease feature system; supporting doctors to mark image description errors through a workflow mechanism, updating the image feature mapping relationship in the graph, and closed-loop optimization to improve image analysis accuracy.

[0023] Reference Figure 2 As shown, a disease-specific cohort data capture system based on an intelligent medical knowledge graph is used to implement the disease-specific cohort data capture method based on an intelligent medical knowledge graph, including: a knowledge graph construction module for integrating the standardized terminology of a single disease with actual clinical corpus, establishing a two-way mapping relationship, and supporting dynamic updates, and a modular design for easy functional expansion; a semantic processing module for including a pre-trained natural language processing model and an inference engine for parsing unstructured text and extracting entities, and integrated processing to improve data parsing efficiency; a matching and inference module for comparing the extracted entities with the knowledge graph, and generating structured intermediate results based on disease feature association rules, and rule-driven inference to ensure logical rigor; a dynamic update module for collecting new corpus annotated by doctors through a workflow mechanism, automatically updating the knowledge graph and triggering model retraining, and adaptively responding to the evolution of clinical terminology; a data warehousing module for storing structured data in a disease-specific cohort database according to a preset field format, supporting the joint processing of multimodal data, and integrating a privacy protection mechanism, and standardized storage to meet scientific research data governance requirements.

[0024] The semantic processing module includes: a multimodal data integration unit that supports the joint input of text, image reports after OCR recognition, and image feature descriptions, and a unified interface to simplify the data processing process; an entity recognition submodule that uses a domain-adapted pre-trained language model to perform named entity recognition, distinguish symptoms, diagnoses, drugs, and examination indicators, and fine-tune the domain to improve the recognition rate of professional terms; a context understanding submodule that uses an attention mechanism to analyze the semantic associations between entities and capture key contextual relationships. The attention weight formula is: Where, α ij is the attention weight of entity i to entity j, e ij is the original correlation score of entity i to entity j, ∑ k exp(e ik ) is the normalization of the association scores of entity i with all other entities k.

[0025] The output interface converts the parsing results into a structured data format for subsequent modules to call. The format standardization reduces system coupling.

[0026] The dynamic update module specifically includes: a labeling management unit for providing visual tools for doctors to circle text fragments, add annotations or select preset tags, and the interactive design conforms to clinical work habits; a data cleaning unit for using regular expressions and rule engines to remove noise data, and the rule library maintains a data quality baseline; a version control unit for recording the update history of the knowledge graph and supporting version rollback, and version management to deal with the risk of accidental data contamination; a model fine-tuning unit for using a distributed training framework to perform incremental training tasks, and distributed computing accelerates the model iteration speed.

[0027] The data warehousing module includes: a data verification unit, which detects the integrity of structured data based on the rules in the knowledge graph, and rule verification prevents data missing; a data conversion unit, which maps intermediate results to fixed fields according to the paradigm of the disease-specific database, and the paradigm conversion ensures the consistency of the data structure; a batch import interface, which supports scheduled tasks or real-time streaming writing, and flexible access methods adapt to the informationization levels of different hospitals; a data visualization unit, which generates statistical charts to display the distribution characteristics of queue data, and visual analysis assists scientific research insights.

[0028] The privacy protection measures of the privacy protection mechanism include: the entire process is deployed on the hospital intranet, data transmission is prohibited, and physical isolation meets medical data security regulations; the patient's name and ID number are encrypted and desensitized to generate anonymous identifiers, and the encryption algorithm uses the SHA-256 hash function to ensure irreversibility; data access rights are managed in a hierarchical manner, and only authorized personnel can operate original data and knowledge graphs, and the authority control matrix prevents data leakage; audit logs record all data operation behaviors, including query, modification and export records, and operation tracing supports compliance review.

[0029] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for capturing disease-specific cohort data based on intelligent medical knowledge graph, characterized by: The following steps are involved: S1. Build a disease-specific intelligent medical knowledge graph, which includes a bidirectional mapping relationship between standardized terminology for a single disease and actual clinical data. By combining manual annotation with machine learning, clinical text data is accumulated to form a domain-specific knowledge base containing symptoms, diagnoses, and examination indicators. S2: Input unstructured medical data into a pre-trained natural language processing model and extract entity information from the text through semantic analysis, including symptom descriptions, examination results, and diagnosis conclusions; S3. Match the extracted entities with the mapping relationships in the knowledge graph, use the disease feature association logic in the graph to infer implicit information, and generate structured intermediate results; S4. Update the knowledge graph through a dynamic self-optimization mechanism: When the model detects uncovered clinical descriptions, a workflow is triggered to collect new corpus annotated by doctors, which is then cleaned and formatted before being added to the knowledge graph. The model is then retrained to adapt to the updated graph. S5. Fill the matched structured data into the target queue database according to the disease-specific database paradigm to complete data capture.

2. The method for capturing disease-specific queue data based on intelligent medical knowledge graph according to claim 1 is characterized in that: The construction of the disease-specific intelligent medical knowledge graph in step S1 specifically includes: Extract clinical data from unstructured medical records on the hospital intranet, external hospital examination reports, and image descriptions, and form an initial corpus through manual annotation; Establishing a mapping relationship between standard terms and clinical corpus, the mapping relationship includes synonym mapping, context association and semantic expansion rules to achieve the conversion of non-standard descriptions into standard terms; Use graph databases to store mapping relationships and disease feature associations, including the causal relationship between symptoms and diagnoses, and the association between examination indicators and disease stages; By continuously annotating new corpora and combining clustering algorithms to identify similar expressions, the coverage of the knowledge graph can be dynamically expanded. Perform version management on the knowledge graph, and record the source of the corpus, the annotator, and the timestamp of each update.

3. The method for capturing disease-specific queue data based on intelligent medical knowledge graph according to claim 1 is characterized in that: The logical inference in step S3 specifically includes: Based on the association of disease features in the knowledge graph, logical reasoning is performed on implicit symptoms that are not clearly described; Analyze the evolution of symptoms using time series data to infer the stage of disease development; Use the probabilistic model to calculate the matching degree between entities and disease characteristics, and set the threshold to filter high-confidence results; Conflict detection is performed on contradictory data, triggering a manual review process to correct mismatches.

4. The method for capturing disease-specific queue data based on intelligent medical knowledge graph according to claim 1 is characterized in that: The dynamic self-optimization mechanism of step S4 specifically includes: Design a user interface for doctors to annotate unidentified clinical descriptions or correct incorrect matching results; The labeled data is automatically transferred to the backend through the workflow, and new mapping relationships are generated after deduplication and standardization. Trigger the incremental training process and fine-tune the model parameters only for the newly added corpus; Regularly evaluate the coverage and model accuracy of the knowledge graph and optimize the iteration strategy.

5. The method for capturing disease-specific queue data based on intelligent medical knowledge graph according to claim 1 is characterized in that: It also includes image data processing steps: Extract features from medical images and generate structured description text; The image feature description and text data are input into the semantic processing module to perform multimodal data fusion analysis; Establish association rules between image features and disease diagnosis in the knowledge graph, including the mapping relationship between image descriptions and diagnostic criteria; The workflow mechanism supports doctors in marking image description errors and updating the image feature mapping relationship in the atlas.

6. The disease-specific queue data capture system based on intelligent medical knowledge graph is characterized by: A method for capturing disease-specific cohort data based on an intelligent medical knowledge graph as described in any one of claims 1 to 5, comprising: The knowledge graph construction module is used to integrate the standardized terminology of a single disease with actual clinical data, establish a bidirectional mapping relationship, and support dynamic updates; The semantic processing module is used to include pre-trained natural language processing models and inference engines for parsing unstructured text and extracting entities; The matching and reasoning module is used to compare the extracted entities with the knowledge graph and generate structured intermediate results based on the disease feature association rules; The dynamic update module is used to collect new corpus annotated by doctors through a workflow mechanism, automatically update the knowledge graph and trigger model retraining; The data warehousing module is used to store structured data in a specialized disease cohort database according to preset field formats, supports the joint processing of multimodal data, and integrates a privacy protection mechanism.

7. The disease-specific queue data capture system based on intelligent medical knowledge graph according to claim 6 is characterized in that: The semantic processing module includes: Multimodal data integration unit, supporting the joint input of text, OCR-recognized image reports, and image feature descriptions; The entity recognition submodule uses a domain-adapted pre-trained language model to perform named entity recognition and distinguish symptoms, diagnoses, medications, and examination indicators; The context understanding submodule uses the attention mechanism to analyze the semantic associations between entities; The output interface converts the parsing results into a structured data format for subsequent module calls.

8. The disease-specific queue data capture system based on intelligent medical knowledge graph according to claim 6 is characterized in that: The dynamic update module specifically includes: The annotation management unit is used to provide visual tools for doctors to circle text segments, add comments or select preset tags; The data cleaning unit is used to remove noise data using regular expressions and rule engines; The version control unit is used to record the update history of the knowledge graph and support version rollback; The model fine-tuning unit is used to perform incremental training tasks using a distributed training framework.

9. The disease-specific queue data capture system based on intelligent medical knowledge graph according to claim 6 is characterized in that: The data storage module includes: Data verification unit, which checks the integrity of structured data based on the rules in the knowledge graph; Data conversion unit, mapping intermediate results to fixed fields according to the disease-specific database paradigm; Batch import interface, supporting scheduled tasks or real-time streaming writing; Data visualization unit generates statistical charts to show the distribution characteristics of queue data.

10. The disease-specific queue data capture system based on intelligent medical knowledge graph according to claim 6 is characterized in that: The privacy protection measures of the privacy protection mechanism include: The entire process is deployed on the hospital intranet, and data transmission is prohibited; Encrypt and desensitize the patient's name and ID number to generate an anonymous identifier; Data access rights are managed in a hierarchical manner, and only authorized personnel can operate original data and knowledge graphs; Audit logs record all data operations, including query, modification, and export records.

Images