Clinical research medical inspection method and system based on artificial intelligence

Abstract

The invention provides a clinical research medical supervision method and system based on artificial intelligence, and relates to the technical field of medical supervision. The method comprises the following steps: constructing a medical supervision ontology knowledge base by obtaining related medical knowledge texts, and mapping medical supervision data sample data into an ontology framework based on the knowledge base to form a medical supervision knowledge graph. Then, knowledge extraction is carried out on the medical supervision application data to obtain review elements, the review elements are mapped into a knowledge graph to obtain related ontology concept nodes, and a review element graph is formed; then, reasoning and judging the review element graph based on the medical supervision knowledge graph and a preset rule to form a preliminary review result; and finally, performing quantitative scoring on the preliminary review result, and performing review decision based on the score and the preliminary result to obtain a final review decision result. According to the method, the medical inspection efficiency and the inspection quality are improved.

Description

Technical Field

[0001] The present application relates to the field of medical monitoring technology, and in particular to a clinical research medical monitoring method and system based on artificial intelligence. Background Art

[0002] Clinical trials are an indispensable and critical link in the development of new drugs and medical devices. Their purpose is to comprehensively evaluate the safety, effectiveness and quality controllability of candidate drugs or medical devices. In order to ensure the authenticity, completeness, standardization and reliability of clinical trial data, strict medical monitoring must be carried out. Medical monitoring aims to review the original data and documents from the clinical trial site, identify the doubts and risks therein, raise medical questions in a timely manner, and ensure the high-quality completion of clinical trials.

[0003] Traditional medical monitoring is mainly carried out manually, with professional medical monitors carefully reviewing and comparing test protocols, case report forms, medical records and other documents to identify possible data anomalies, protocol violations, and ethical violations. However, with the rapid growth in the number of new drug developments and the continuous expansion of clinical trials, medical monitoring work faces many challenges such as large amounts of data, high professionalism, and heavy workload.

[0004] Under current technical conditions, medical monitors need to invest a lot of time and energy to systematically review a large amount of trial site documents and data, relying on their own professional knowledge and experience to discover hidden doubts and risks. This way of working is inefficient, easily affected by human factors, and difficult to ensure the consistency and stability of monitoring quality. This seriously affects the efficiency and quality of medical monitoring. Summary of the invention

[0005] The present application provides an artificial intelligence-based clinical research medical monitoring method and system to improve the efficiency of medical monitoring and the quality of review.

[0006] In a first aspect, the present application provides a clinical research medical monitoring method based on artificial intelligence, the method comprising: Obtain medical knowledge texts related to medical monitoring, and build a medical monitoring ontology knowledge base based on the medical knowledge texts; Obtain sample data of medical monitoring materials, and map the sample data of medical monitoring materials to the ontology framework based on the ontology knowledge base to form a semantic network containing concept nodes, attribute edges, and relationship edges, and obtain a medical monitoring knowledge graph; Obtain medical monitoring application materials, and perform knowledge extraction on the medical monitoring application materials to obtain review elements; Map the review elements to the medical monitoring knowledge graph, obtain the review element related ontology concept nodes, and form a review element graph based on the related ontology concept nodes; Based on the medical monitoring knowledge graph and the preset medical monitoring rules, the review factor graph is inferred and judged to form a preliminary review result; Quantitatively score the preliminary review results to obtain a quantitative review score; A review decision is made based on the review quantitative score and the preliminary review results to obtain a review decision result.

[0007] In the above technical solution, by obtaining medical knowledge texts related to medical monitoring and constructing a medical monitoring ontology knowledge base based on the medical knowledge texts, a solid knowledge foundation is laid for the subsequent construction of a medical monitoring knowledge graph. On this basis, by obtaining sample data of medical monitoring materials and mapping the sample data of medical monitoring materials to the ontology framework based on the ontology knowledge base, a semantic network containing concept nodes, attribute edges, and relationship edges is formed, thereby obtaining a medical monitoring knowledge graph with rich content and complete structure. This method of converting unstructured medical knowledge into structured and semantic knowledge representation greatly improves the processing efficiency and standardization of massive medical monitoring data.

[0008] Furthermore, after obtaining the medical monitoring application materials, this application uses knowledge extraction technology to intelligently extract the review elements therein, and maps the review elements to the medical monitoring knowledge graph to obtain the ontology concept nodes related to the review elements, and on this basis forms a refined review element graph. This process makes full use of the semantic association information of the knowledge graph, quickly locks in the core review points of the application materials, and greatly reduces the burden of manual review.

[0009] On this basis, this application makes inferences and judgments on the review factor graph based on the medical monitoring knowledge graph and preset medical monitoring rules, automatically identifies possible problems in the application materials, and forms preliminary review results. This intelligent reasoning method based on domain knowledge and rules significantly improves the sensitivity and accuracy of medical monitoring and effectively prevents data quality risks in clinical trials. At the same time, this application also introduces a quantitative scoring mechanism to objectively evaluate the preliminary review results and obtain a quantitative review score. This scoring method establishes a unified monitoring quality assessment standard and improves the consistency and comparability of monitoring results.

[0010] In a second aspect of the present application, a clinical research medical monitoring system based on artificial intelligence is provided, the system comprising: An ontology knowledge base construction module is used to obtain medical knowledge texts related to medical monitoring and to construct a medical monitoring ontology knowledge base based on the medical knowledge texts; The medical monitoring knowledge graph construction module is used to obtain sample data of medical monitoring materials and map the sample data of medical monitoring materials to the ontology framework based on the ontology knowledge base to form a semantic network containing concept nodes, attribute edges, and relationship edges to obtain a medical monitoring knowledge graph; The review element acquisition module is used to obtain the medical monitoring application materials and perform knowledge extraction on the medical monitoring application materials to obtain the review elements; The review factor graph acquisition module is used to map the review factors into the medical monitoring knowledge graph, obtain the review factor related ontology concept nodes, and form the review factor graph according to the related ontology concept nodes; The preliminary review result acquisition module is used to infer and judge the review factor graph based on the medical monitoring knowledge graph and the preset medical monitoring rules to form a preliminary review result; A quantitative scoring module is used to quantify the scores of preliminary review results and obtain the review quantitative scores; The review decision module is used to make review decisions based on the review quantitative scores and preliminary review results to obtain review decision results.

[0011] In a third aspect of the present application, a computer storage medium is provided, wherein the computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the above method steps.

[0012] In the fourth aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the above method.

[0013] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This application establishes a structured and semantic medical monitoring knowledge representation system by acquiring medical knowledge texts related to medical monitoring and constructing a medical monitoring ontology knowledge base based on the medical knowledge texts. The ontology knowledge base organizes the core concepts, relationships, and attributes in the field of medical monitoring in a formalized way, laying a solid knowledge foundation for subsequent intelligent review. By mapping the sample data of medical monitoring materials into the ontology framework, a semantic network containing concept nodes, attribute edges, and relationship edges is formed, and a medical monitoring knowledge graph with rich content and complete structure is further obtained. The knowledge graph not only covers various regulatory requirements for medical monitoring, but also reflects the empirical knowledge contained in the case data, so that the review process can make full use of relevant information of previous cases and improve the comprehensiveness and accuracy of the review.

[0014] 2. This application uses artificial intelligence technologies such as natural language processing and ontology construction to automatically and efficiently mine the intrinsic knowledge of medical knowledge texts, and systematically define concepts and their relationships, thereby constructing a medical monitoring ontology knowledge base, forming a comprehensive, detailed, scientific and standardized regulatory knowledge organization system.

[0015] 3. This application is based on the ontology knowledge base. Through a series of methods such as data preprocessing, concept node mapping, attribute edge mapping, and relationship edge extraction, it can efficiently, accurately, and comprehensively map unstructured medical monitoring material sample data into a standardized ontology framework, thereby forming a medical monitoring knowledge graph covering concept connotations, attribute characteristics, and relationship semantics. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of a process flow of a clinical research medical monitoring method based on artificial intelligence provided in an embodiment of the present application; Figure 2 An architecture diagram of a clinical research medical monitoring system based on artificial intelligence provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in this application. DETAILED DESCRIPTION

[0017] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0018] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0019] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0020] In order to facilitate understanding of the method and system provided by the embodiments of the present application, before introducing the embodiments of the present application, the background of the embodiments of the present application is first introduced.

[0021] As an indispensable and key link in scientific research activities, medical monitoring shoulders the important mission of ensuring compliance of the research process, protecting the rights and interests of subjects, and promoting scientific research integrity. For a long time, medical monitoring has been mainly carried out manually. Reviewers need to invest a lot of time and energy to carefully review the application materials, deeply analyze the compliance of the research plan, comprehensively assess potential ethical risks, and make prudent review decisions on this basis. Although this traditional manual review model has played an important role, it has gradually exposed some problems that need to be improved. First, the efficiency of manual review is low and it is difficult to cope with the growing demand for review. With the vigorous development of scientific research activities, the number of applications for medical monitoring continues to rise, and manual processing methods can no longer adapt to this explosive growth trend. Reviewers need to invest a lot of time and energy to review each application document one by one. This high-intensity workload not only affects the efficiency of the review, but may also lead to a decline in the quality of the review. Secondly, manual review is difficult to avoid the problems of subjectivity and inconsistent standards. The reviewer's knowledge background, experience level, values ​​and other factors may affect the review decision, resulting in a certain degree of subjectivity in the review results. Different reviewers may have different judgments on the same case, and it is difficult to achieve complete unification of the review standards. Such subjectivity and inconsistency in standards may affect the fairness and standardization of the review and cause doubts and dissatisfaction among applicants.

[0022] After the background introduction of the above content, those skilled in the art can understand the problems existing in the prior art. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0023] Based on the above background technology, please refer to Figure 1 , Figure 1 A flowchart of a clinical research medical monitoring method based on artificial intelligence provided in an embodiment of the present application. The system can be implemented by a computer program or can be run as an independent tool application. Specifically, in an embodiment of the present application, the method can be applied on a server, but can also be applied to electronic devices such as servers. A clinical research medical monitoring method based on artificial intelligence includes the following steps: S101, obtaining medical knowledge texts related to medical monitoring, and constructing a medical monitoring ontology knowledge base based on the medical knowledge texts; Specifically, by systematically collecting and organizing authoritative text materials such as various laws and regulations, departmental regulations, industry norms, and guiding principles in the field of medical monitoring, a comprehensive and rich corpus of medical knowledge texts is formed. These medical knowledge texts cover the requirements and standards of all aspects of medical monitoring and are an important basis for conducting review work. On the basis of obtaining medical knowledge texts, ontology construction technology is used to conduct in-depth analysis and knowledge extraction of medical knowledge texts, identify the core concepts, relationships, and attributes in the field of medical monitoring, and organize them into a structured ontology knowledge base in a formalized way.

[0024] Based on the above embodiment, as an optional embodiment, a medical monitoring ontology knowledge base is constructed according to medical knowledge text, including: S201, uses natural language processing technology to extract core concepts related to medical monitoring from medical knowledge texts; Specifically, the medical knowledge text needs to be preprocessed first, including sentence segmentation, word segmentation, part-of-speech tagging, named entity recognition, etc., to convert the unstructured text into a structured sequence of semantic units, providing a basis for concept extraction. This preprocessing step is usually implemented using mature natural language processing toolkits or platforms, such as Jieba, THULAC, StanfordCoreNLP, etc.

[0025] After preprocessing, concept extraction mainly adopts rule-based, statistical and deep learning-based methods. The rule-based method manually defines a series of templates and rules based on the grammatical and semantic features of the concept in the sentence to match and extract concepts. Common templates include "concept + attributive clause", "concept + of + noun", "adjective + concept", etc. The advantage of this method is that the accuracy is high, but the recall rate is low, and it relies heavily on the knowledge of domain experts. The statistical method automatically discovers concepts in the text by measuring statistical indicators such as the co-occurrence frequency of concepts in documents and vocabulary composition. Common statistical indicators include TF-IDF, Mutual Information, and X2 statistics.

[0026] S202, based on the ontology core concepts and the medical knowledge text, determining the semantic relationship between the ontology core concepts, and performing ontology relationship definition according to the semantic relationship between the ontology core concepts, to obtain the relationship definition between each ontology core concept; Specifically, there are two main strategies for defining concept relationships: one is to use the ontology description language to directly formally define the relationship between concepts; the other is to automatically mine the relationship between concepts from the text, then abstract it into ontology relationships and formally define them.

[0027] For the first strategy, domain experts and knowledge engineers work together to make intuitive judgments about the relationships between concepts by referring to the context of concepts in medical knowledge texts, and then formalize them into machine-readable relationship definitions using ontology description languages ​​such as RDF and OWL. For example, it can be defined that there is a "protection" relationship between "informed consent" and "subject rights", that is, "informed consent protects subject rights"; and then expressed as (informed consent, protection, subject rights) using RDF triples (subject, predicate, object). Although this manual definition method is time-consuming and labor-intensive, the accuracy and consistency of the relationship definition are high, and it is more in line with the expert's understanding logic.

[0028] For the second strategy, natural language processing technology, especially semantic dependency analysis and open knowledge extraction technology, is first used to automatically extract semantic dependency relationships between concepts from medical knowledge texts. For example, the semantic relationship of "protection" can be identified from the sentence "When conducting human trials, the privacy and personal information of subjects must be protected", and then the two triples of (human trials, protection, privacy of subjects) and (human trials, protection, personal information) can be extracted. After obtaining the preliminary relationship extraction results, experts will screen, verify and abstract them, remove redundant and erroneous relationships, merge semantically similar relationships into the same type, and define them as formalized relationships using ontology language. For example, the relationships of "protection", "maintenance" and "guarantee" are uniformly abstracted as the "protection" relationship.

[0029] S203, defining the attributes of each ontology core concept based on the relationship definition between each ontology core concept, to obtain the attribute definition of each ontology core concept; Specifically, the definition of concept attributes mainly adopts two approaches: one is to combine the relationship between concepts in the ontology, extract and generalize the intrinsic attributes of the concept from the relationship objects; the other is to refer to the contextual information of the concept in the medical knowledge text and manually define the external attributes that reflect the characteristics of the concept.

[0030] The first idea is to make full use of the rich conceptual relationships that have been defined, and abstract and extract conceptual attributes from the semantics of conceptual relationships. The advantage of this approach is that the attribute definition is highly consistent with the relationship definition, and no conflicts or errors are introduced. At the same time, by revealing the attributes through relationships, the inherent connection between relationships and attributes is reflected, and the semantic consistency of the ontology is enhanced. For example, from the relationship "the ethics committee consists of medical professionals", the attribute "members = medical professionals" of the "ethics committee" can be extracted. For another example, from the relationship "informed consent must include a description of the purpose of the research", the attribute "essential elements = description of the purpose of the research" of "informed consent" can be extracted. These attributes extracted from conceptual relationships form an organic part of the connotation of the concept.

[0031] The second idea is for domain experts to directly refer to medical knowledge texts and manually define external attributes that reflect concept characteristics. Although this attribute definition method based on background knowledge is highly subjective, it can cover concept characteristics that are not covered by relationships and is a necessary supplement to attribute extraction based on relationships. By analyzing the semantic context of concepts in medical knowledge texts, experts can often identify some key attributes that are easy to understand and grasp. For example, by reading relevant clauses, it is not difficult to find that "subjects" have attributes such as "voluntary participation", "right to know", "right to privacy", "right to withdraw midway", etc. For another example, "research projects" have attributes such as "name", "purpose", "start and end time", "participants", "funding source", etc. These manually defined attributes, based on expert experience and domain common sense, highlight the significant characteristics of concepts and facilitate people to quickly recognize concepts.

[0032] In practice, these two attribute definition methods are usually used interchangeably and in conjunction with each other. Taking the concept of "informed consent" as an example, first of all, we can extract attributes such as "protected object = subject's rights" and "necessary procedures = subject's signature" from the relationships such as "informed consent protects the rights and interests of subjects" and "informed consent requires the signature of subjects". Then, experts refer to the original text of the regulations to define external attributes such as "notification content", "notification method", and "signature format". These two types of attributes complement each other and together constitute a comprehensive and detailed "informed consent" attribute spectrum, making the originally abstract concept more three-dimensional, vivid, perceptible and knowable. S204, constructing a medical monitoring ontology knowledge base based on the relationship definitions of each ontology core concept and the attribute definitions of each ontology core concept.

[0033] Specifically, the construction of the medical monitoring ontology knowledge base mainly includes two aspects: one is to formalize the definitions of concepts, relationships, and attributes into an ontology knowledge base based on the ontology description language and use ontology development tools; the other is to manually review and evaluate the constructed ontology knowledge base and make necessary revisions, supplements, and extensions to the ontology.

[0034] For the first aspect, ontology description languages ​​such as OWL (Web Ontology Language) are mainly used, and with the help of ontology development tools such as Protégé and TopBraid Composer, the definitions of concepts, relationships, and attributes are transformed into formal ontology knowledge base representations. Specifically, the ontology header information is first defined, and the ontology namespace, version, and metadata such as the introduction of external ontologies are declared; then the class (Class) hierarchy is defined, and the core concepts are organized into a tree-like or graph-like type hierarchy according to the hierarchical relationship of concepts; then the relationship (Property) between types is defined, and according to the definition of conceptual relationships, logical properties such as SubProperty, InverseOf, and TransitiveProperty are used to characterize the semantic characteristics of the relationship; finally, the instance (Individual) of the class is defined and the attribute values ​​of the instance are described. According to the definition of conceptual attributes, the type instance is associated with the attribute value domain and data type using ObjectProperty, DatatypeProperty, etc. For example, define the "subject" class and use SubClassOf to define it as a subclass of the "research participant" class; then define ObjectProperties such as "right to know", "right to privacy", and "benefit compensation" to associate "subject" with other related classes; finally define instances such as "Zhang San" and "Li Si" to describe their properties such as right to know, right to privacy, and benefit compensation. Through layer-by-layer definition, concepts, relationships, and attributes are organically organized into a complete knowledge system, forming an ontology knowledge base with a clear structure and rich semantics.

[0035] S102, obtaining sample data of medical monitoring materials, and mapping the sample data of medical monitoring materials to the ontology framework based on the ontology knowledge base, forming a semantic network including concept nodes, attribute edges, and relationship edges, and obtaining a medical monitoring knowledge graph; Specifically, we first collect sample data of medical monitoring materials from historical review cases, typical samples and other channels, including various types of unstructured text data such as applications, informed consent forms, research plans, and review opinions. These sample data contain rich review practice experience and case knowledge, and are an important data source for building knowledge graphs. After obtaining the sample data, it is necessary to preprocess it, such as format conversion, text cleaning, word segmentation, part-of-speech tagging, etc., to convert the unstructured text into a form that can be understood and processed by computers. Next, based on the constructed medical monitoring ontology knowledge base, the preprocessed sample data is mapped to the ontology framework. This process is mainly achieved through natural language processing and knowledge extraction technology. First, the named entity recognition (NER) method is used to identify entities corresponding to ontology concepts from the sample data, such as "subjects", "research purpose", "informed consent", etc. Then, the semantic relationships between entities are extracted from the sample data using relationship extraction technology, such as rule-based methods or machine learning methods, such as "research project-involving-subjects", "researcher-obtaining-informed consent", etc. Through entity recognition and relationship extraction, the key information in the sample data is transformed into a form compatible with the ontology knowledge base.

[0036] After the entities and relationships are extracted, the extracted entities are mapped to the corresponding concept nodes in the ontology knowledge base, and the extracted relationships are mapped to the attribute edges or relationship edges between the concept nodes, thus forming a semantic network closely connected to the ontology framework. For example, the "subject" entity identified in the sample data is mapped to the "subject" concept node in the ontology knowledge base, and the "research project-involved-subject" relationship extracted from the sample data is mapped to the relationship edge between the "research project" and "subject" concept nodes. Through this mapping process, the scattered sample data are integrated into a unified semantic framework, forming a highly connected and semantically rich medical monitoring knowledge graph.

[0037] The construction process of the medical monitoring knowledge graph usually adopts a semi-automatic approach, combining manual intervention and computer automatic processing. First, experts define mapping rules and strategies to guide the mapping process of sample data to the ontology framework. Then, natural language processing and machine learning technologies, such as conditional random fields (CRF), support vector machines (SVM), deep learning and other methods, are used to automatically complete tasks such as entity recognition, relationship extraction and mapping. On the basis of automatic processing, experts are also required to perform manual verification and correction to ensure the quality and accuracy of the knowledge graph. If necessary, experts can also expand and optimize the knowledge graph, add new concept nodes, attribute edges or relationship edges to make it more complete and comprehensive.

[0038] Based on the above embodiment, as an optional embodiment, the sample data of medical monitoring materials is mapped to the ontology framework based on the ontology knowledge base to form a semantic network including concept nodes, attribute edges, and relationship edges, and obtain a medical monitoring knowledge graph, including: S301, preprocessing the sample data of the medical monitoring information to obtain medical monitoring data; Specifically, the preprocessing of sample data of medical monitoring materials mainly includes the following steps: data cleaning, data conversion, data integration, data reduction, etc.

[0039] S302, performing data mapping on the medical monitoring data based on the ontology knowledge base, mapping entities in the medical monitoring data to corresponding ontology concept nodes in the ontology knowledge base, and mapping attributes in the medical monitoring data to attribute edges of the ontology concept nodes; Specifically, the process of data mapping of medical monitoring data based on the ontology knowledge base mainly includes two aspects: one is to map the entities in the review data to the corresponding concept nodes in the ontology knowledge base; the other is to map the attributes in the medical monitoring data to the attribute edges of the concept nodes.

[0040] S303, extracting relationships from the medical monitoring data based on the ontology knowledge base, extracting semantic relationships between named entities in the medical monitoring data, and mapping the semantic relationships to relationship edges between ontology concept nodes; Specifically, relation extraction of medical monitoring data based on the ontology knowledge base mainly includes two steps: one is to extract the semantic relationship between named entities from the data text; the other is to map the extracted semantic relationship to the relation edge of the concept node in the ontology knowledge base.

[0041] For the first step, semantic relationship extraction, natural language processing technology is mainly used to identify named entities from the unstructured text of the review data and determine the types of semantic relationships between entities. Specifically, named entity recognition models, such as conditional random fields and recurrent neural networks, are first used to identify multiple types of named entities such as people, institutions, documents, and drugs from the text; then relationship extraction models, such as convolutional neural networks and attention mechanisms, are used to determine whether there is a semantic association between the identified entity pairs and predict the specific type of association, such as "author-work", "sponsor-sponsored project", "researcher-informed consent", etc. Most of the current mainstream relationship extraction models use the remote supervised learning paradigm, automatically generate large-scale training corpus through the ontology knowledge base, and learn the semantic mapping between entity pairs and relationship types at the sentence level. Some work also integrates pre-trained language models such as BERT to improve the model's understanding of semantic relationships through pre-training. In addition, rule template matching, bootstrapping and other methods are also used to mine structured relationship patterns between entities from syntactic analysis trees. Through these technologies, structured triple knowledge such as "drug A - indication - disease B" and "case C - inclusion criteria - condition D" are identified from the review data text, and the correlation between the review elements is preliminarily mined.

[0042] For the second step, i.e., relationship-edge mapping, the main task is to semantically match the semantic relationship extracted from the data with the relationship edges defined in the ontology knowledge base to form a unified graphical representation. Specifically, based on dictionary matching, vector space semantic similarity calculation and other methods, it is first determined whether the semantics of the extracted semantic relationship and the ontology relationship edge are equivalent, that is, whether the head entity and the tail entity in the triple correspond to the head concept and the tail concept connected by the relationship edge, and whether the relationship type corresponds to the definition of the relationship edge; for semantically equivalent relationships, they are converted into corresponding relationship edge instances; for relationships that cannot be directly matched, the superior-subordinate relationship of the existing concept nodes in the ontology is used to infer a relationship path that spans at most one intermediate concept, and the relationship edges in the path are combined to generate a new composite relationship edge. For example, the semantic relationship of "patient A-inclusion-trial plan B" is identified and mapped to the "subject-participation-clinical trial" relationship edge defined in the ontology, and an instance of the relationship edge is directly generated; for another example, the semantic relationship of "researcher C-record-adverse reaction D" is identified, and no directly corresponding relationship edge can be found in the ontology, but the composite association of "researcher-record-adverse reaction" can be inferred based on the two existing relationship edges of "researcher-conduct-research" and "research-involved-adverse event", and a new composite relationship edge is generated accordingly. Through the above mapping and reasoning, the semantic associations extracted from the data are uniformly mapped to the relationship framework of the ontology knowledge base, so that it inherits the structured semantic information such as the type definition, value range, and constraint rules of the relationship edge.

[0043] S304, construct a medical monitoring knowledge graph based on ontology concept nodes, attribute edges and relationship edges.

[0044] Specifically, constructing a medical monitoring knowledge graph based on ontology concept nodes, attribute edges, and relationship edges mainly includes three steps: defining the graph Schema, storing graph data, and optimizing graph queries.

[0045] The first step is to define the Schema mode of the knowledge graph, which is mainly to sort out the knowledge organization framework of the graph, including meta-information such as concept types, relationship types, attribute types and their semantic constraints, so as to inherit and expand the semantic system of the ontology knowledge base. Schema definition generally uses ontology description languages ​​such as RDF and OWL, and standardizes the semantic types of concept nodes and relationship edges through axiom constraints such as domain, value range, and cardinality. For example, define core concept classes such as "researcher" and "subject", and map the necessary attributes of their ontology definitions such as "name" and "institution" to data attributes of the graph; define core relationship classes such as "participation" and "signature", and clarify the semantic restrictions of their head and tail concept classes. In addition, the Schema definition should also consider physical storage optimization, adopt attribute embedding for frequently co-occurring attributes, and implement shard storage for concepts with a large number of instances. A well-designed graph Schema should strike a balance between semantic completeness and storage efficiency. Secondly, the specific data such as entities, attributes, and relationships of the graph are stored. The physical storage is mainly carried out by mainstream graph databases such as Neo4j and JanusGraph, which store semantic network data mapped by ontology in a node-edge-attribute diagrammatic structure. When storing, the database tables of concept nodes, relationship edges, and attributes are first created according to the Schema definition, and then the ETL tool is used to convert the previous heterogeneous review data into standard graph data formats such as RDF and CSV, and then imported into the graph database through the transaction mechanism through graph query languages ​​such as Cypher and Gremlin. The diagrammatic physical storage allows nodes to be seamlessly linked directly through relationship edges, avoiding the iterative foreign key query of traditional relational libraries. For example, to find "subjects who participated in trial X and had adverse reaction Y", the library only needs to quickly locate the "trial" and "adverse event" nodes through pattern matching, and then track the relationship edges of their first-degree neighbors to efficiently filter out the target "subject" node. In fact, the graph traversal algorithm of the library can convert complex joins of multiple tables into single-table pattern matching, greatly improving the retrieval efficiency of related data.

[0046] S103, obtaining medical monitoring application materials, and performing knowledge extraction on the medical monitoring application materials to obtain review elements; Specifically, first obtain the complete information of the medical monitoring application, including the application form, research plan, informed consent form, researcher qualification certificate and other parts. These materials are usually submitted in the form of paper or electronic documents, covering all aspects of the research project, and are the basic materials for conducting medical monitoring. After obtaining the application materials, it is necessary to perform necessary preprocessing on them, such as format conversion, text extraction, image recognition, etc., to convert the unstructured application materials into a form that can be understood and processed by computers. Next, knowledge extraction is performed on the preprocessed application materials, with the goal of identifying the review elements closely related to medical monitoring. This process is mainly achieved using natural language processing and text mining technology. First, the named entity recognition (NER) method is used to identify key entities from the application materials, such as "research purpose", "subject selection criteria", "informed consent process", "data confidentiality measures", etc. Since the review elements usually involve professional terms and concepts in a specific field, it is necessary to use pre-built domain dictionaries and rule bases to improve the accuracy and recall rate of entity recognition.

[0047] On the basis of entity recognition, we further use technologies such as keyword extraction and phrase mining to extract keywords and phrases related to review elements from the application materials, such as "risk", "subject rights", "privacy protection", etc. These keywords and phrases can usually reflect the focus and risks of the application project in medical monitoring, and provide important clues for subsequent review and judgment.

[0048] Based on the above embodiment, as an optional embodiment, knowledge extraction is performed on the medical monitoring application materials to obtain review elements, including: S401, structurally processing the medical monitoring application materials to obtain structured application materials, and performing text analysis on the structured application materials to obtain key information elements; Specifically, knowledge extraction of medical monitoring application materials to obtain review elements mainly includes two steps: structuring the application materials and text parsing of the structured application materials to extract key information elements.

[0049] For the first step, i.e., structural processing, the main purpose is to convert the original application documents in PDF, Word and other formats into structured review information, providing standardized input for subsequent text analysis. Specifically, firstly, document analysis tools such as Apache PDFBox, Apache POI, etc. are used to convert the PDF and Word documents of the original review materials into structured HTML or XML formats, retaining important document structure information such as text content, title hierarchy, table, list, etc.; then the converted structured documents are parsed to extract their title tree, paragraph text, table cells, list items and other structured elements, and based on mechanisms such as XPath or CSS selectors, a unique positioning expression is generated for each structural element; finally, the extracted structured information is stored in a document database such as MongoDB, and a structured index of "document ID-location path-text content" is established to provide fast random access for subsequent analysis and extraction. The key to this step is to retain as much semi-structured information as possible that is crucial to semantic understanding, such as the name of the entry item in the review form, the chapter title of the research plan, etc., on the basis of removing the noise of the original document, so that the subsequent analysis can focus on areas with higher information density. In addition, attention should be paid to establishing a unified structural index for multiple application documents of the same review project (such as research plans, informed consent forms, medical monitoring application forms, etc.) to provide support for the associated extraction of multi-source information. Practice has shown that the use of document structured preprocessing can increase the speed of subsequent text parsing by 1-2 orders of magnitude.

[0050] For the second step, i.e. the extraction of key information elements, natural language processing and machine learning technologies are mainly used to perform semantic analysis and information extraction on the structured review materials, and identify the key information elements required for medical monitoring decisions. Specifically, the structured text is first preprocessed by NLP, including sentence segmentation, word segmentation, part-of-speech tagging, named entity recognition, syntactic dependency analysis, etc., to build basic features at the vocabulary and syntactic levels for semantic analysis; then a text parsing engine is developed for review elements, mainly using rule-based, template-based, and knowledge-based methods to extract review elements from the three levels of vocabulary, syntax, and semantics. At the lexical level, dictionary matching, regular expressions and other methods are mainly used to quickly identify keywords and phrases related to review elements, such as "informed consent", "research purpose", "subject rights", etc., and preliminarily locate sentences containing review elements. At the syntactic level, syntactic analysis, dependency analysis and other technologies are mainly used to extract descriptive modifiers of review elements from relevant sentences according to syntactic rules such as verb-object structure and attributive structure, such as "written consent" and "clear research purpose". At the semantic level, the extracted vocabulary and syntactic units are mapped to the structured representation of the review elements using the predefined review element ontology and extraction rules. For example, the sentence "This project ensures that the subjects are fully aware of the research purpose, process and risks by signing a written informed consent form" is identified, and based on the extraction template of "research project-through-written informed consent-guarantee-subjects-know-research purpose / process / risk", a structured representation of "informed consent method: written; informed content: research purpose, research process, research risk" is formed. The above three levels complement each other. The vocabulary level realizes the coarse-grained identification and positioning of key information, the syntactic level realizes the structured organization of modifying and limiting phrases, and the semantic level realizes the ontology mapping of structured semantic components to review elements, which together constitute a progressive extraction from unstructured to structured.

[0051] S402, performing entity naming based on key information elements and structured application materials to obtain key entities corresponding to the review elements; Specifically, entity naming is performed based on key information elements and structured application materials to obtain key entities corresponding to the review elements. The naming method based on dictionaries and rules is mainly adopted, supplemented by a machine learning model trained with a small amount of manually annotated data. First, for the core entity types in the medical monitoring knowledge graph, such as "research project", "research institution", "research subject", "informed consent", "research risk", "conflict of interest", etc., a named entity dictionary covering their synonyms, variants, and aliases is constructed, and key entities belonging to these types are quickly identified through string matching. For example, keywords such as "clinical trial", "randomized control", and "research project" are mapped to "research project" entities; "patient", "subject", and "volunteer" are mapped to "research subject" entities; "informed consent form" and "informed document" are mapped to "informed consent" entities. Although dictionary matching is simple and efficient, it is easily interfered by cross-context ambiguity and needs to be supplemented by context rules for result filtering. Therefore, based on dictionary matching, a series of entity naming rules are manually defined for specific combination patterns of review elements, and context features are used to improve the accuracy of naming. For example, it is stipulated that only "subjects" in contexts such as "the subjects of this study are healthy volunteers over the age of 18" can be named as "research subjects", excluding other situations such as "subject rights" and "subject informed consent". For another example, sentences where "researcher" and "conflict of interest" entities co-occur are marked as "conflict of interest" elements, excluding general declarative sentences where "researcher" appears alone. On the basis of dictionaries and rules, machine learning sequence annotation models such as conditional random fields (CRF) and recurrent neural networks (RNN) are further introduced. By manually annotating a small amount of training corpus, the distribution law of review elements in the context is learned, and the generalization and robustness of naming recognition are further improved.

[0052] S403, performing named entity recognition on the structured application materials to extract key entities corresponding to the review elements; Specifically, named entity recognition is performed on structured application materials to extract key entities corresponding to review elements, mainly using multi-strategy recognition methods such as dictionary matching, rule filtering and sequence annotation. First, for common entity types in the field of medical monitoring, such as "research design", "statistical analysis" and "qualification certificate", a domain dictionary covering their terms, aliases and abbreviations is constructed. Then, these dictionaries are used to quickly match structured texts and preliminarily identify candidate entities. Next, a series of filtering rules are manually defined to eliminate ambiguous naming and irrelevant entities using contextual features such as part of speech and syntax to improve recognition accuracy. For example, it is stipulated that "subjects" in "subjects of this study signed informed consent" only represent research subjects, excluding interference items such as "subject rights"; for another example, sentences where researchers and conflict of interest entities co-occur are annotated as conflict of interest elements, excluding general statements where researchers appear alone. On the basis of dictionary matching and rule filtering, machine learning sequence annotation models such as CRF and BiLSTM are further introduced. By manually annotating a small amount of training corpus, the distribution pattern of review entities in context is learned to overcome the limitations of insufficient dictionary coverage and weak rule generalization ability. Specifically, syntactic features such as part of speech and dependency analysis are extracted, and automatically combined into high-dimensional feature vectors that describe the entity context through feature templates. Combined with manually annotated corpus, discriminant models such as CRF are used to learn feature weights to form a review entity recognizer with stronger domain adaptability. In addition, active learning and human-computer interaction are used to optimize the manual annotation process, obtain the most representative training samples at the lowest annotation cost, and further improve model performance.

[0053] S404, performing type definition on the key entities to obtain defined key entities, and performing relationship extraction based on the defined key entities to determine the semantic associations between the defined key entities; Specifically, the key entities are defined by type definition to obtain the defined key entities, and the relationship extraction is performed based on the defined key entities to determine the semantic associations between the defined key entities. The main method is based on knowledge base and rules, supplemented by a machine learning model trained with a small amount of labeled data. First, refer to the knowledge graph in the field of medical monitoring to build an entity type system, such as dividing entities into major categories such as research projects, research institutions, research objects, and risks, and further subdividing them into subcategories such as cross-sectional studies and drug clinical trials. Then, using the upper and lower relationships and attribute characteristics of similar entities in the knowledge base, a series of type mapping rules are manually defined, such as the research design entity in "This study adopts a randomized controlled trial design" is mapped to the "randomized controlled trial" type, and the institution entity in "The research team comes from Peking University First Hospital" is defined as the "research hospital" type. At the same time, the type information of existing entities in the knowledge base is quickly aligned using methods such as dictionary matching. In addition, for the tail and new entities that are insufficiently covered by the knowledge base, active learning is used to guide experts to label the types of a small number of seed entities. Then, through remote supervision techniques such as bootstrpping, type reference patterns in large-scale unstructured texts are automatically mined, the annotation scale is expanded, and machine learning classification models such as CRF and CNN are trained to automatically classify new entities.

[0054] S405, defining attributes based on the defined key entities to obtain the attributes of each defined key entity; Specifically, attribute definition is performed based on the defined key entities to obtain the attributes of each defined key entity. The main method is based on knowledge base and context, supplemented by active learning to introduce manual annotation. First, refer to the medical monitoring knowledge ontology, enumerate the common attributes of key entities, and construct an attribute dictionary. For example, the attribute dictionary of the "researcher" entity contains name, unit, title, etc. Then, the dictionary is used to automatically match and identify the attribute mentions of each entity, and filter irrelevant attributes based on contextual features such as syntax and semantics to eliminate cross-context attribute reference ambiguity. For example, it is stipulated that the "professor" attribute value of "Wang Ming" in "Professor Wang Ming of Peking University School of Pharmacy" is his "School of Pharmacy", excluding redundant attributes of other part-time units. At the same time, attribute values ​​are automatically extracted from the corresponding paragraphs of the application materials, such as extracting "2022-010" from "This trial has passed the review of the ethics committee, and the approval number is 2022-010" as the "approval number" attribute value of the "medical monitoring" entity. For long-tail attribute values ​​in unstructured expressions, regular expressions, dependency analysis and other methods are used to extract the grammatical form of attribute values. For new and sparse attribute types, active learning methods guide experts to annotate them, learn the semantic similarity between attribute mentions and attribute types through distributed representation, and automatically mine training corpora for annotating new attributes.

[0055] S406, organizing the defined key entities, the attributes of each defined key entity and the semantic associations between the defined key entities into a structured knowledge representation to obtain review elements.

[0056] Specifically, in the process of knowledge extraction of medical monitoring application materials, it is a critical step to organize the defined key entities, the attributes of each defined key entity and the semantic associations between the defined key entities into a structured knowledge representation to obtain the review elements. This is because the review elements are essentially structured and standardized knowledge representations extracted from the medical monitoring application materials, which contain the core information elements of the application materials. Only by systematically organizing and representing the key entities, attributes and semantic associations can the content of the application materials be fully and accurately described, and a clear and complete information basis can be provided for subsequent review decisions. Specifically, in the previous steps, the medical monitoring application materials have been structured to obtain the key information elements, and the entities have been named based on the key information elements to obtain the key entities corresponding to the review elements. Then, the key entities are defined in type and the type of each key entity is determined. On the basis of the type definition, the relationship extraction of the defined key entities is further performed to clarify the semantic associations between the defined key entities, that is, the relationships between them. At the same time, the attributes of the defined key entities are defined to determine the attributes of each key entity. On this basis, the defined key entities, the attributes of each defined key entity and the semantic associations between the defined key entities are organized into a structured knowledge representation, which can be specifically carried out in the following ways: first, each defined key entity is taken as the basic unit of knowledge representation, and each key entity is represented by a standardized term or identifier; second, for each key entity, its attributes are associated with it in the form of attribute name and attribute value, where the attribute name represents the type or name of the attribute, and the attribute value gives the specific value of the entity on the attribute; third, for the semantic association between key entities, two key entities are connected by a relationship name to represent the relationship between them, and the relationship consists of three elements: the starting entity, the relationship name and the ending entity; finally, the key entities, attributes and semantic associations are organically organized in the form of triples (entity-attribute-attribute value, or entity-relationship-entity) to form a complete structured knowledge representation, namely, the review element.

[0057] S104, mapping the review elements to the medical monitoring knowledge graph, obtaining the review element related ontology concept nodes, and forming a review element graph according to the related ontology concept nodes; Specifically, the extracted review elements are first mapped one by one with the ontology concepts in the medical monitoring knowledge graph. This process is mainly achieved using semantic matching and similarity calculation techniques. For each review element, search for semantically similar or related ontology concept nodes in the knowledge graph, and calculate the similarity or correlation between them. The similarity calculation can consider multiple factors such as the literal similarity, semantic similarity, and hierarchical relationship between the review element and the ontology concept to comprehensively evaluate their matching degree. Usually, ontology concept nodes whose similarity or correlation exceeds a certain threshold are considered to be related to the review element and are selected as the mapping result.

[0058] After completing the mapping of review elements to ontology concept nodes, a subgraph closely related to the review elements is formed in the medical monitoring knowledge graph based on these related ontology concept nodes, namely the review element graph. Specifically, with the mapped related ontology concept nodes as the center, their adjacent nodes, attribute edges and relationship edges are extracted in the knowledge graph to form a connected and compact semantic subnetwork. This subnetwork encompasses concepts, attributes and relationships directly or indirectly related to the review elements, reflecting the semantic context and association environment of the review elements in the knowledge graph.

[0059] Based on the above embodiment, as an optional embodiment, the review elements are mapped to the medical monitoring knowledge graph to obtain the review element related ontology concept nodes, and the review element graph is formed according to the related ontology concept nodes, including: S501, mapping the review elements to the ontology concept nodes in the medical monitoring knowledge graph to obtain the ontology concept nodes related to the review elements; Specifically, the review elements are mapped to the ontology concept nodes in the medical monitoring knowledge graph to obtain the ontology concept nodes related to the review elements, mainly using methods based on semantic similarity and graph random walk. First, the ontology concept nodes in the medical monitoring knowledge graph are traversed, and the semantic similarity between the review elements and the concept nodes is calculated using pre-trained word vectors such as BERT and Word2Vec. Considering the diversity of concept expressions in the field of medical ethics, character-level features such as glyph similarity are incorporated into the calculation of semantic similarity to alleviate the OOV problem. Then, the high-order semantic representation of ontology concepts is learned using methods such as topic models and knowledge distillation, and the semantic association matrix between concepts is constructed, so that the thematic semantics behind the vocabulary can be captured when mapping the review elements. For example, for the review element "the project team obtains a list of research subjects from the village committee", even if its apparent vocabulary is far from the concept of "informed consent", the deep association between the two in terms of "privacy protection" can be explored by analyzing the corresponding themes of the vocabulary. After calculating the semantic similarity matrix, breakpoint filtering, TOP-K and other strategies are used to select the top K ontology concept nodes most relevant to the review elements. At the same time, considering that it is difficult to reveal indirect associations by relying solely on semantic similarity, the graph random walk method is further used to explore the multi-hop propagation paths of related concepts and capture the high-order associations of review elements in the ontology network. For example, for the element "there is a conflict of interest between the research team and the participating enterprises", through multi-hop association analysis, it can be found that there is an implicit violation relationship between it and the concept of "the researcher's integrity obligation". Finally, the ontology concept nodes directly or indirectly related to the review elements are summarized to obtain the set of ontology concept nodes related to the review elements.

[0060] S502, expanding the relevant nodes in the medical monitoring knowledge graph according to the review factor related ontology concept nodes to obtain the review factor expansion nodes; Specifically, the related nodes in the medical monitoring knowledge graph are expanded according to the review factor related ontology concept nodes to obtain the review factor extension nodes, mainly using the method based on graph neural network and cross-graph mapping. First, the review factor related ontology concept nodes are taken as the central nodes in the medical monitoring knowledge graph, and the low-dimensional representation of the nodes in the graph is learned by using the graph neural network method, so that the semantically related nodes in the embedding space are close in the vector space. Considering the heterogeneity of entity-relationship-entity triples, heterogeneous graph neural networks such as R-GCN can be used to distinguish the semantic contributions of different relationship types to node embedding. At the same time, the mutual influence between concept nodes in the modeling graph is modeled by using graph attention networks such as GAT, and the aggregate node neighbor information is adjusted by attention weight to highlight the semantic contribution of strongly related nodes. After learning the node embedding, the nearest neighbor search method based on Euclidean distance, cosine similarity, etc. is used to obtain the top N extension nodes most related to the central node. Furthermore, in order to expand the coverage of knowledge, a node expansion method across knowledge graphs is introduced. By manually mapping seed nodes and learning the joint representation of cross-graph nodes, we construct a semantic mapping between the medical monitoring knowledge graph and the knowledge graphs in the fields of medicine and law, and expand the extension scope of related concept nodes to multiple external knowledge graphs. For example, by associating the legal field graph, the related concepts of "informed consent" can be expanded to related concepts from a legal perspective such as "contracting ability" and "contract validity". Finally, the extended nodes of the internal and external graphs of the comprehensive ontology are deduplicated and integrated to obtain the extended node set of the review elements.

[0061] S503, extracting a connected review factor subgraph from the medical monitoring knowledge graph based on the review factor extension node and the review factor-related ontology concept node; Specifically, a connected subgraph of review elements is extracted from the medical monitoring knowledge graph based on the review element extension nodes and the review element related ontology concept nodes, mainly using the method based on graph community discovery and minimum spanning tree. First, the review element related ontology concept nodes and the extension node set are merged as the candidate node set. Then, a graph community discovery algorithm is run on the medical monitoring knowledge graph, such as the Louvain algorithm based on modularity optimization, which aims to divide the graph into several node communities with tight internal and loose external nodes. Next, each graph community is traversed to select the Top-K communities with the largest number of candidate nodes. Considering that relying only on local communities may miss key paths, the minimum spanning tree algorithm is further used to extract the shortest connected subgraph of candidate nodes in the knowledge graph, aiming to connect as many candidate nodes as possible with as few edges as possible. At this time, it is necessary to balance the information content and noise of the subgraph. The thresholds of the number of nodes and edges can be set to appropriately constrain the size of the subgraph. Finally, the graph community is integrated with the minimum spanning tree, redundant nodes and edges are pruned, and one-hop neighbors are expanded using local search to improve information completeness, and finally the connected subgraph of the review element is obtained. It is worth mentioning that for complex scenarios involving multiple review elements, connected subgraphs of each element can be constructed in parallel, and alignment and fusion between subgraphs can be achieved through iterative optimization of graph matching and other technologies, thereby constructing an association network covering multiple review elements and internal semantic connectivity.

[0062] S504, performing redundant pruning on the review factor subgraph to obtain a review factor graph.

[0063] Specifically, the review factor subgraph is redundantly pruned to obtain the review factor graph, mainly using methods based on betweenness centrality and structural holes. First, the betweenness centrality of each node in the subgraph is counted, that is, the frequency of the node as the intermediary of the shortest path between other nodes, aiming to characterize the importance of information dissemination of the node in the graph structure. At the same time, the structural hole theory is used to measure the sparsity of the neighborhood where the node is located, that is, the difficulty of connecting different communities connected by the node on other paths, aiming to characterize the information control ability of the node in the graph structure. Next, the subgraph nodes are jointly scored by combining the betweenness centrality and structural hole indicators to balance the information dissemination and control ability of the nodes. Considering the dependence of association analysis on core nodes, a threshold is set to retain the top-K% high-scoring nodes as backbone nodes. On this basis, an optimization method similar to PCSF is adopted to minimize the total edge weight of the subgraph under the constraint that the backbone nodes remain unchanged, aiming to extract the most dense and concise node connection subset. Drawing on the idea of ​​the minimum spanning tree, the local topological structure is regenerated from the backbone node set, and parameters are set to balance the edge scale and path length, trying to connect the most key nodes with the least edges. Through iterative optimization, we finally get a concise, compact, and semantically rich review factor graph. In the case of multiple review factors, we further apply frequent subgraph mining to each factor graph to extract the co-occurring concept combination pattern and reveal the general rules of multi-factor association.

[0064] S105, based on the medical monitoring knowledge graph and the preset medical monitoring rules, the review factor graph is inferred and judged to form a preliminary review result; Specifically, we first need to express the preset medical monitoring rules in a form compatible with the knowledge graph, such as Horn clauses, predicate logic, etc. These review rules are derived from laws and regulations, ethical guidelines, expert consensus, etc. related to medical monitoring, covering the requirements and standards of various aspects of medical monitoring. For example, "all human research must obtain informed consent from the subjects" can be expressed as a Horn clause: "human research (X) ∧ involving (X, subjects) → informed consent (X, subjects)". By formalizing the review rules, they can be directly matched and interoperated with the ontology concepts and relationship edges in the knowledge graph, providing a basis for reasoning and judgment based on the knowledge graph. Next, the review factor graph is fused with the medical monitoring knowledge graph to form a more complete and related knowledge network. Specifically, the concept nodes, attribute edges, and relationship edges in the review factor graph are mapped to the corresponding ontology elements in the knowledge graph, so that the two graphs are aligned and connected at the semantic level. This fusion process can be completed automatically based on technologies such as semantic similarity matching and ontology mapping, and it also requires necessary manual inspection and confirmation by experts. The fused knowledge graph contains a seamless combination of specific information of the application project and knowledge in the field of medical monitoring, providing a comprehensive and accurate knowledge basis for reasoning and judgment.

[0065] On the basis of knowledge graph fusion, tools such as ontology reasoning engine or rule engine are used to make inference and judgment on the fused knowledge graph. The reasoning process mainly includes two steps: rule matching and conflict resolution. First, the reasoning engine traverses the knowledge graph and searches for matching graph fragments according to the preset medical monitoring rules. For example, for the rule "Human research must obtain informed consent", the reasoning engine searches for nodes marked as "Human research" in the graph and checks whether it has an "involved" relationship edge with the "subject" node, and whether there is a corresponding "informed consent" node and relationship edge. The successfully matched rules indicate that the application project meets the ethical requirements in the corresponding aspects, while the unmatched rules indicate potential ethical risks or deficiencies.

[0066] After completing the rule matching, the inference engine further performs conflict detection and resolution on the matching results. Since there may be overlaps, conflicts or exceptions between medical monitoring rules, simple rule matching may lead to contradictory or wrong judgment results. Therefore, it is necessary to use mechanisms such as priority setting and exception handling to comprehensively analyze and adjust the rule matching results to eliminate conflicts and inconsistencies. For example, if the "informed consent" rule is successfully matched, but the exception rule of "subject's lack of understanding ability" is triggered at the same time, it is necessary to further determine whether the subject meets the exception conditions and adjust the review results accordingly.

[0067] After rule matching and conflict resolution, the inference engine finally forms a preliminary review result, including compliance assessment, risk warnings and improvement suggestions for the application project in various medical monitoring aspects. This preliminary review result is presented in a structured and organized form, clearly pointing out the medical monitoring focus, compliance evidence and content that needs to be supplemented and improved for the application project, providing a comprehensive and reliable reference for expert review.

[0068] S106, quantify and score the preliminary review results to obtain a quantitative review score; Specifically, we first need to formulate a set of scientific and standardized medical monitoring scoring standards and scoring rules. The scoring standards start from the various key dimensions of medical monitoring, such as informed consent, risk control, and subject protection, and set a number of evaluation indicators and weight coefficients for each dimension. For example, the "informed consent" dimension can include indicators such as "completeness of informed consent" and "comprehensibility of informed consent", and each indicator is assigned a different weight value according to its importance. The scoring rules clarify the scoring level, score range and judgment conditions of each indicator. For example, the "completeness of informed consent" indicator can be set as "complete (5 points)", "basically complete (3 points)", and "incomplete (0 points)" three levels, and the judgment criteria for each level are listed in detail. The formulation of scoring standards and rules needs to fully absorb the opinions of stakeholders such as medical monitoring experts, legal professionals, and the public to ensure their scientificity, rationality and wide acceptance.

[0069] After clarifying the scoring criteria and rules, the preliminary review results will be matched and evaluated item by item to obtain the original score of each indicator. Specifically, for each medical monitoring dimension and indicator involved in the review results, the corresponding scoring level is determined according to the degree to which it complies with the scoring rules, and the corresponding score is assigned. This matching and evaluation process can be completed manually by experts or automated through technologies such as natural language processing and pattern matching. Regardless of the method adopted, it is necessary to base it on the preliminary review results, strictly follow the scoring criteria and rules, and try to avoid subjectivity or arbitrariness. After evaluating all indicators, a set of original scores is obtained, which reflects the basic situation of the application project in various medical monitoring aspects.

[0070] In order to obtain a quantitative review score for the overall application project, it is also necessary to perform weighted synthesis based on the original score. Specifically, according to the weight coefficient of each indicator, the original score is weighted to obtain the weighted score of the indicator. Then, the weighted scores of all indicators are added together to obtain the comprehensive score of the application project in the entire medical monitoring dimension. This weighted synthesis process usually uses a linear weighted method, that is, multiplying the weight of each indicator by the original score and then summing them. However, in practice, other weighting strategies can also be used as needed, such as geometric mean, logistic regression, etc., to better reflect the interaction and combination effects between different indicators.

[0071] For example, suppose the "informed consent" dimension includes three indicators, with weights of 0.5, 0.3, and 0.2, respectively, and the original scores are 4 points, 3 points, and 5 points after evaluation. The weighted scores of these three indicators are 2 points, 0.9 points, and 1 point, respectively, totaling 3.9 points, which is the score of the application project in the "informed consent" dimension. If the total weight of the "informed consent" dimension is 0.4, and the weights of the other four medical monitoring dimensions are 0.3, 0.1, 0.1, and 0.1, respectively, and the scores are 4.2, 3.5, 4.0, and 3.8 after weighted synthesis, the overall review quantitative score of the application project is 3.99 points (i.e. 0.43.9+0.34.2+0.13.5+0.14.0+0.1*3.8). This comprehensive score can intuitively reflect the ethical compliance level of the application project and provide an important quantitative reference for the formation of the final review results.

[0072] Based on the above embodiment, as an optional embodiment, a review decision is made based on the review quantitative score and the preliminary review result to obtain a review decision result, including: S601, fusion analysis is performed on the review quantitative score and the preliminary review result to obtain the review quantitative score; Specifically, the review quantitative score and the preliminary review result are fused and analyzed to obtain the fused review quantitative score, mainly using the method based on DS evidence theory and hierarchical analysis method. First, the qualitative description in the preliminary review result is mapped to the corresponding quantitative sub-score. By constructing a corresponding rule base of description statements and sub-scores, the key sentences in the review results are automatically extracted and mapped to the corresponding sub-scores using natural language processing technology, and the review results are structured and quantified. At the same time, the DS evidence theory is used to model the credibility of the sub-scores of each dimension, and the uncertainty of each score is characterized by setting the basic probability confidence of the sub-score problem. On this basis, the DS evidence synthesis rule is used to integrate the quantitative scores and confidence of each dimension to form the initial value of the fusion score. Considering the different importance of each scoring dimension, the hierarchical analysis method is further used to determine the weight of each dimension. The importance of the dimension is relatively measured by comparing the judgment matrix between the two dimensions, and the weight is calculated by the eigenvalue method to balance the influence of each dimension. Finally, the initial value of the fusion score and the dimension weight are combined to form a weighted sum, and the fusion score that comprehensively considers the quantitative score and qualitative analysis is obtained, which is fed back to the review quantitative score. In addition, in order to enhance the interpretability of the fusion analysis, methods such as fuzzy comprehensive evaluation are used to compare the score differences before and after the fusion, trace the key factors behind the differences, and automatically generate text explanations to intuitively present the impact of qualitative analysis on the score.

[0073] S602, making a review decision based on a preset scoring threshold and a review quantitative score to obtain a review decision result.

[0074] Specifically, based on the preset scoring threshold and the review quantitative score, the review decision is made to obtain the review decision result, mainly using the method based on decision tree and fuzzy comprehensive evaluation. First, according to the historical review data and expert experience, the threshold is set for the integrated review quantitative score, and the score range is divided into several discrete intervals. At the same time, a clear decision label is set for each score interval to form a mapping rule base from quantitative score to review decision. Next, the decision tree algorithm is used to train and optimize the mapping rule base. By constructing a decision tree with the score interval as a non-leaf node and the decision label as a leaf node, and splitting and pruning nodes based on information entropy, Gini index and other criteria, a concise and efficient decision rule is formed. On this basis, the quantitative score of the project to be reviewed is input into the decision tree, and the corresponding review decision result is automatically inferred. Considering that there may be uncertainty in the decision when the score is at the critical point of the interval, the fuzzy comprehensive evaluation method is further used for auxiliary judgment. The degree of proximity between the score and the critical point is measured by the membership function, and the membership threshold is set. The critical point samples are marked as the category that needs to be reviewed, prompting manual intervention. In addition, for projects that require rectification and reexamination, a method similar to functional magnetic resonance imaging is used to compare the state matrices before and after, quantitatively evaluate the degree of project optimization, and provide a basis for adjusting the review strategy.

[0075] S107, making a review decision based on the review quantitative score and the preliminary review result to obtain a review decision result.

[0076] Specifically, it is necessary to first set scientific and reasonable decision thresholds and judgment rules based on the quantitative review scores. Usually, corresponding score ranges can be set for different scoring levels (such as excellent, good, qualified, unqualified, etc.), and the types of review results corresponding to each range can be clarified. For example, it can be stipulated that scores above 90 points are "excellent", corresponding to direct pass; 80-90 points are "good", corresponding to pass after modification; 60-80 points are "qualified", corresponding to conditional pass; and below 60 points are "unqualified", corresponding to failure. When determining the threshold and judgment rules, it is necessary to comprehensively consider factors such as the legal and regulatory requirements of medical supervision, the characteristics of the subject area, and the social ethical environment, and widely absorb expert opinions and practical experience to ensure its scientificity and operability.

[0077] Based on the above embodiment, as an optional embodiment, the preliminary review result is quantitatively scored to obtain the review quantitative score, including: S701, according to the requirements and standards of medical monitoring, determine the quantitative dimensions and quantitative indicators of the quantitative scoring and set scoring rules for each quantitative indicator; Specifically, according to the requirements and standards of medical monitoring, the quantitative dimensions and quantitative indicators of quantitative scoring are determined, and scoring rules are set for each quantitative indicator. The methods based on hierarchical analysis and Delphi method are mainly adopted. First, the general requirements of medical monitoring are systematically sorted out through literature review, and several core elements are summarized from the perspectives of scientificity, ethics, and compliance. Experts in the field are invited to supplement and revise, and finally the top-level quantitative dimensions are determined. Under each top-level dimension, the specific evaluation points are further refined, and multiple rounds of expert opinions are organized through the Delphi method to converge to form several operational and assessable quantitative indicators. Based on the expert judgment, machine learning methods are used to optimize the indicator settings through cluster analysis, association analysis and other technologies, and the hierarchical analysis method is used to determine the indicator weights to balance the importance of each factor. Next, scoring rules are formulated for each quantitative indicator. The idea of ​​fuzzy comprehensive evaluation is adopted, and the corresponding membership level is set based on the degree of compliance of the indicator with the corresponding review points, and matched to the percentage score range. At the same time, additional bonus and deduction items are set from the aspects of the completeness and authenticity of the review materials to deal with special circumstances. On this basis, we further formed a scorecard template, clarified the scoring method of each indicator, the list of required materials and the key points of review, and finally established a structured quantitative scoring system. For dynamically adjusted scoring rules, we updated the scorecard and related guidelines in a timely manner to ensure the synchronous iteration of the system.

[0078] S702, assigning weights to the quantitative dimensions and quantitative indicators to obtain a weight ratio table; Specifically, weights are assigned to quantitative dimensions and quantitative indicators to obtain a weight ratio table, mainly using methods based on hierarchical analysis and machine learning. First, based on the "dimension-indicator-rule" framework, a three-level hierarchical weight model is constructed using the hierarchical analysis method. Through pairwise comparison, the relative importance of each top-level dimension to the review decision is judged from a macro level, and a judgment matrix is ​​constructed and consistency test is performed. On the basis of the determination of the weight at the dimension level, the relative importance of each indicator within each dimension is further compared, and the weight at the indicator level is determined by a similar method. The top-level dimension weights and the corresponding indicator layer weights are comprehensively analyzed and normalized, and finally a top-down hierarchical weight ratio scheme is formed. Considering that the hierarchical analysis method mainly relies on expert experience and has a certain degree of subjectivity, machine learning methods are further introduced to optimize it. A number of representative historical review projects are selected, their quantitative scoring data are extracted, and the projects are labeled according to the final review results, distinguishing different categories such as pass, rectification, and rejection. On this basis, common metrics such as information gain and Gini index are used to analyze the correlation between the quantitative scores of different indicators and the review results through algorithms such as classification decision trees and logistic regression, and the weight values ​​given by experts are calibrated accordingly. The weights obtained by comprehensive qualitative analysis and quantitative calculation are used to obtain a dynamically integrated weight ratio table through weighted averaging and other methods. Considering the dynamic nature of review needs, new review data is regularly collected to continuously optimize the weight ratio and continuously improve the scientific nature of weight setting. In addition, in response to the dynamic adjustment of weights, the scoring system is updated synchronously to achieve automatic changes in score calculations and ensure the synchronization of scoring rules and weight ratios.

[0079] S703, setting scoring rules and weight ratio tables based on quantitative indicators, quantifying the scores of the preliminary review results, and obtaining the review quantitative scores.

[0080] Specifically, based on the scoring rules and weight ratio table set by the quantitative indicators, the preliminary review results are quantitatively scored to obtain the review quantitative score, mainly using a hierarchical scoring method based on weighted summation. First, the preliminary review results of the review object to be reviewed are matched with the scoring rule table, and the degree of compliance with the corresponding requirements of each indicator is judged item by item, and the score is given with reference to the rule table. For elements that are difficult to judge subjectively, the qualitative and quantitative combination methods such as fuzzy comprehensive judgment can be combined with expert opinions to give points as appropriate. On the basis that each indicator has a clear score, refer to the weight ratio table and use the weighted summation method to calculate the weighted average of the scores of each indicator in each top-level dimension as the comprehensive score of the dimension. Further, the same method is used to sum the scores of each top-level dimension according to the weight to obtain the preliminary quantitative score of the review object. Taking into account the characteristics of different review matters, appropriate fine-tuning can be made on the basis of common weights and combined with specific circumstances. For example, for projects involving human trials, the weight of informed consent can be increased; for big data application research, more emphasis should be placed on data security and privacy protection. For the adjusted personalized weights, the same hierarchical weighted calculation method is used to obtain the final quantitative score that matches the specific review object. During the calculation process, it is necessary to convert the percentage score into a decimal representation in the range of 0-1, and retain 2-3 decimal places in the result to ensure accuracy. In addition, a constant term can be set appropriately to avoid the extreme situation where the overall score is low due to a single indicator scoring zero.

[0081] On the other hand, the present application also provides a clinical research medical monitoring system based on artificial intelligence, such as Figure 2 The system comprises: The ontology knowledge base construction module 1 is used to obtain medical knowledge texts related to medical monitoring and construct a medical monitoring ontology knowledge base based on the medical knowledge texts; The medical monitoring knowledge graph construction module 2 is used to obtain sample data of medical monitoring materials, and map the sample data of medical monitoring materials to the ontology framework based on the ontology knowledge base to form a semantic network including concept nodes, attribute edges, and relationship edges to obtain a medical monitoring knowledge graph; The review element acquisition module 3 is used to obtain the medical monitoring application materials and perform knowledge extraction on the medical monitoring application materials to obtain the review elements; The review factor graph acquisition module 4 is used to map the review factor to the medical monitoring knowledge graph, obtain the review factor related ontology concept nodes, and form a review factor graph according to the related ontology concept nodes; A preliminary review result acquisition module 5 is used to infer and judge the review factor graph based on the medical monitoring knowledge graph and the preset medical monitoring rules to form a preliminary review result; A quantitative scoring module 6 is used to quantitatively score the preliminary review results to obtain a review quantitative score; The review decision module 7 is used to make a review decision based on the review quantitative score and the preliminary review result to obtain a review decision result.

[0082] Please refer to Figure 3 The application also discloses an electronic device. Figure 3 The electronic device 300 may include: at least one processor 301 , at least one network interface 304 , a user interface 303 , a memory 305 , and at least one communication bus 302 .

[0083] The communication bus 302 is used to realize the connection and communication between these components.

[0084] The user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0085] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0086] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field~Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 301 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301, and it can be implemented separately through a chip.

[0087] Among them, the memory 305 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read~Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may optionally be at least one storage system located away from the aforementioned processor 301. Refer to Figure 3 , the memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program for a clinical research medical monitoring method based on artificial intelligence.

[0088] exist Figure 3 In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 301 can be used to call the application program storing the road assessment method in the memory 305. When executed by one or more processors 301, the electronic device 300 executes one or more methods in the above-mentioned embodiments. It should be noted that for the aforementioned method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited to the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application. In the above embodiments, the description of each embodiment has its own emphasis. For the parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0089] In the several implementations provided in this application, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of the system or unit can be electrical or other forms.

[0090] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0091] The present application also provides a computer storage medium that can store multiple instructions, which are suitable for being loaded and executed by a processor as described above. Figure 1 The road assessment method of the embodiment shown in the figure can be found in the specific implementation process. Figure 1 The specific description of the illustrated embodiment will not be repeated here.

[0092] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0093] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, 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. The computer software product is stored in a memory, including a number of instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0094] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0095] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A clinical research medical monitoring method based on artificial intelligence, characterized in that the method include: Acquire medical knowledge texts related to medical monitoring, and construct a medical monitoring ontology knowledge base based on the medical knowledge texts; Acquire sample data of medical monitoring materials, and map the sample data of medical monitoring materials to an ontology framework based on the ontology knowledge base to form a semantic network including concept nodes, attribute edges, and relationship edges, and obtain a medical monitoring knowledge graph; Obtaining medical monitoring application materials, and performing knowledge extraction on the medical monitoring application materials to obtain review elements; Mapping the review elements to the medical monitoring knowledge graph to obtain the review element related ontology concept nodes, and forming a review element graph according to the related ontology concept nodes; Based on the medical monitoring knowledge graph and the preset medical monitoring rules, the review factor graph is inferred and judged to form a preliminary review result; Quantitatively scoring the preliminary review results to obtain a quantitative review score; A review decision is made based on the review quantitative score and the preliminary review result to obtain a review decision result.

2. The method according to claim 1, characterized in that The step of constructing a medical monitoring ontology knowledge base according to the medical knowledge text includes: Using natural language processing technology to extract core concepts of ontology related to medical monitoring from the medical knowledge text; Based on the ontology core concepts and the medical knowledge text, the semantic relationship between the ontology core concepts is determined, and the ontology relationship definition is performed according to the semantic relationship between the ontology core concepts to obtain the relationship definition between each ontology core concept; Based on the relationship definition between each ontology core concept, define the attributes of each ontology core concept to obtain the attribute definition of each ontology core concept; The medical monitoring ontology knowledge base is constructed based on the relationship definitions of each ontology core concept and the attribute definitions of each ontology core concept.

3. The method according to claim 1, characterized in that The method maps the sample data of the medical monitoring materials to the ontology framework based on the ontology knowledge base to form a semantic network including concept nodes, attribute edges, and relationship edges, and obtains a medical monitoring knowledge graph, including: Preprocessing the sample data of the medical monitoring information to obtain medical monitoring data; Based on the ontology knowledge base, the medical monitoring data is data mapped, entities in the medical monitoring data are mapped to corresponding ontology concept nodes in the ontology knowledge base, and attributes in the medical monitoring data are mapped to attribute edges of the ontology concept nodes; Performing relationship extraction on the medical monitoring data based on the ontology knowledge base, extracting semantic relationships between named entities in the medical monitoring data, and mapping the semantic relationships into relationship edges between the ontology concept nodes; A medical monitoring knowledge graph is constructed based on the ontology concept nodes, the attribute edges and the relationship edges.

4. The method according to claim 1, characterized in that The knowledge extraction of the medical monitoring application materials is performed to obtain review elements, including: Structuring the medical monitoring application materials to obtain structured application materials, and performing text parsing on the structured application materials to obtain key information elements; Perform entity naming based on the key information elements and the structured application materials to obtain key entities corresponding to the review elements; Perform named entity recognition on structured application materials to extract key entities corresponding to the review elements; Performing type definition on the key entities to obtain defined key entities, and performing relationship extraction based on the defined key entities to determine semantic associations between the defined key entities; Defining attributes based on the defined key entities to obtain the attributes of each of the defined key entities; The defined key entities, the attributes of each of the defined key entities and the semantic associations between the defined key entities are organized into a structured knowledge representation to obtain the review elements.

5. The method according to claim 1, characterized in that The step of mapping the review elements to the medical monitoring knowledge graph to obtain the review element related ontology concept nodes, and forming a review element graph according to the related ontology concept nodes includes: Mapping the review elements to ontology concept nodes in the medical monitoring knowledge graph to obtain ontology concept nodes related to the review elements; Expanding relevant nodes in the medical monitoring knowledge graph according to the ontology concept nodes related to the review factors to obtain review factor expansion nodes; Extracting a connected review factor subgraph from the medical monitoring knowledge graph based on the review factor extension node and the review factor-related ontology concept node; The review factor graph is obtained by performing redundant pruning on the review factor sub-graph.

6. The method according to claim 1, characterized in that The review decision is made based on the review quantitative score and the preliminary review result to obtain the review decision result, including: The review quantitative score and the preliminary review results are integrated and analyzed to obtain the review quantitative score; A review decision is made based on a preset scoring threshold and the review quantitative score to obtain the review decision result.

7. The method according to claim 1, characterized in that The quantitative scoring of the preliminary review results to obtain the review quantitative score includes: According to the requirements and standards of medical monitoring, determine the quantitative dimensions and quantitative indicators of the quantitative score and set scoring rules for each of the quantitative indicators; Allocating weights to the quantitative dimensions and the quantitative indicators to obtain a weight ratio table; The scoring rules and the weight ratio table are set based on the quantitative indicators to quantify the preliminary review results and obtain the review quantitative score.

8. A clinical research medical monitoring system based on artificial intelligence, characterized in that: include: An ontology knowledge base construction module is used to obtain medical knowledge texts related to medical monitoring and construct a medical monitoring ontology knowledge base based on the medical knowledge texts; A medical monitoring knowledge graph construction module is used to obtain sample data of medical monitoring materials, and map the sample data of medical monitoring materials to an ontology framework based on the ontology knowledge base to form a semantic network including concept nodes, attribute edges, and relationship edges, thereby obtaining a medical monitoring knowledge graph; A review element acquisition module is used to obtain medical monitoring application materials and perform knowledge extraction on the medical monitoring application materials to obtain review elements; A review factor graph acquisition module, used to map the review factor to the medical monitoring knowledge graph, obtain the review factor related ontology concept nodes, and form a review factor graph according to the related ontology concept nodes; A preliminary review result acquisition module is used to infer and judge the review factor graph based on the medical monitoring knowledge graph and preset medical monitoring rules to form a preliminary review result; A quantitative scoring module is used to quantitatively score the preliminary review results to obtain a review quantitative score; The review decision module is used to make a review decision based on the review quantitative score and the preliminary review result to obtain a review decision result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes a processor, a memory and a transceiver, the memory is used to store instructions, the transceiver is used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method according to any one of claims 1 to 7.

Images