Electronic medical record quality evaluation method and system

Abstract

The invention relates to the technical field of electronic medical record quality evaluation through knowledge engineering and artificial intelligence, and discloses an electronic medical record quality evaluation method and system. The evaluation method comprises the following steps: S1, constructing a western medicine diagnosis and treatment standard metadata ontology according to a DC metadata standard, an industry guidance specification, international disease classification and expert knowledge, and supplementing ontology knowledge to form a knowledge organization strategy; s2, aiming at the western medicine diagnosis and treatment standard data set, organizing through the knowledge organization strategy, and constructing a western medicine diagnosis and treatment standard knowledge base; s3, performing quality evaluation on the electronic medical record according to the western medicine diagnosis and treatment standard knowledge base; the quality evaluation comprises form normalization evaluation, content semantic correctness evaluation and logic consistency evaluation. Fine-grained, intelligent and scientific evaluation can be carried out on the electronic medical records, and hospitals are assisted in improving the medical supervision level.

Description

Technical Field

[0001] The present invention relates to the technical field of evaluating the quality of electronic medical records using knowledge engineering and artificial intelligence, and in particular to a method and system for evaluating the quality of electronic medical records. Background Art

[0002] The electronic medical record text consists of several prescribed parts. It is the cognitive result of the doctor's processing of the patient's source information. The doctor's subjective thoughts are contained in these written contents. As a core component of the modern medical system, it carries multiple functions such as patient health information, clinical decision support, medical service coordination and medical quality improvement. Therefore, it is imperative to conduct quality assessment of electronic medical records.

[0003] The guiding data set of Western medical diagnosis and treatment standards is a mixed language expression of knowledge from various disciplines of Western medicine, containing a large number of terms, numerical values, and especially a large amount of descriptive text. Due to the large differences in the expression methods of different disciplines, although this knowledge is scientific and rigorous in logic, it is obviously lacking in machine usability. Lacking systematic knowledge that can be used by machines, the existing evaluation methods of electronic medical records basically rely on manual labor, which is time-consuming, labor-intensive, high-cost, and low-efficiency.

[0004] There are some results in existing research on using neural network models to complete electronic medical record evaluation. However, the technical principle of the neural network model is the "machine learning" method, and the accuracy of its evaluation depends on the scale and quality of the training data. It is not suitable for application scenarios with key problems such as multi-disciplinary Western medicine, uneven quality of training data sources, and insufficient medical knowledge of the annotators themselves. The quality evaluation results of electronic medical records predicted by the neural network model are not accurate.

[0005] Therefore, there is an urgent need for an electronic medical record quality assessment method and system that can perform scientific and accurate quality assessment of electronic medical records. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides an electronic medical record quality assessment method, comprising the following steps: S1: Based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge, construct the metadata ontology of Western medicine diagnosis and treatment standards and supplement the ontology knowledge to form a knowledge organization strategy; S2: Organizing the Western medicine diagnosis and treatment standard data set through the knowledge organization strategy to build a Western medicine diagnosis and treatment standard knowledge base; S3: Perform quality assessment on electronic medical records based on the Western medicine diagnosis and treatment standard knowledge base; the quality assessment includes formal normativeness assessment, content semantic correctness assessment, and logical consistency assessment.

[0007] As an implementation method of the present application, in S1, based on industry guidance specifications, international disease classification and expert knowledge, a knowledge organization strategy is formed based on DC metadata, including: S11: Based on the retrieval point specification of DC metadata and in accordance with industry guidance specifications, a metadata description framework for Western medicine diagnosis and treatment standards is constructed; the metadata description framework for Western medicine diagnosis and treatment standards includes description elements for organizing Western medicine diagnosis and treatment standard document units and concept elements and relationship elements for organizing Western medicine diagnosis and treatment standard knowledge units; the description elements include source, disease name, overview, and department classification; the concept elements include etiology, clinical manifestations, complications, examinations, diagnosis, and treatment; the relationship elements include metadata element relationships; S12: Based on the metadata description framework of western medicine diagnosis and treatment standards, abstract the classes and semantic relationships in the metadata of western medicine diagnosis and treatment standards; take each class as a knowledge unit; S13: Based on the correspondence between department classification and disease names in the International Classification of Diseases and expert knowledge, department concept trees and disease word bags are constructed; S14: Using the disease name as a connection and the semantic relationship as a relationship, the department concept tree is associated with the classes in the Western medicine diagnosis and treatment standard metadata to form a knowledge organization strategy.

[0008] As an implementation method of the present application, in S1, the classes in the metadata ontology of the Western medicine diagnosis and treatment standards include department classes and disease classes; the subclasses of the disease class include: cause, clinical manifestations, examination, diagnosis, and treatment.

[0009] As an implementation method of the present application, in S1, the semantic relationship includes belonging to, causing, phenomenon expression, concurrent, having inspection, having diagnosis and having treatment.

[0010] As an implementation method of the present application, the formal normative evaluation is: performing text parsing on the electronic medical record, and performing formal normative evaluation on the parsed electronic medical record according to the formal requirements of the electronic medical record writing project, including the following steps: S3A1: Use regular expressions to parse the text of electronic medical records and extract the predefined fields in the electronic medical records and the filled content of each field; S3A2: Based on the case writing standards, a standard template for format review is prepared; S3A3: Evaluate the completeness of the predefined fields in the electronic medical record and the formality of the content filled in each field based on the format review standard template.

[0011] As an implementation method of the present application, the content semantic correctness evaluation includes the following steps: S3B1: for the western medicine diagnosis and treatment standard knowledge base, extract knowledge elements and corresponding knowledge element entities through named entity recognition; For the electronic medical records to be evaluated, substructure elements and corresponding substructure element entities are extracted through named entity recognition; S3B2: Through the pre-trained language model, the knowledge meta-entities and sub-structure element entities are vectorized to obtain the knowledge meta-entity vector set and the sub-structure element entity vector set respectively; S3B3: Determine the knowledge element similarity calculation method based on the category of the substructure element entity vector set: If the class of the substructure element entity vector set is the cause, the common sense causal feature model is used to calculate the knowledge element similarity; If the class of the substructure element entity vector set is part, the similarity of knowledge element is calculated using the upper and lower concept tree; If the class of the substructure element entity vector set is not the cause and location, the cosine similarity is used to calculate the knowledge element similarity; S3B4: Based on each knowledge element entity vector set and the substructure element entity vector set, the selected similarity calculation method is used to calculate and obtain an initial knowledge element similarity matrix; S3B5: For the initial knowledge element similarity matrix, only the maximum similarity value of each column is retained, and the remaining similarity values ​​are assigned to 0, thereby obtaining a knowledge element similarity matrix; S3B6: Generate a knowledge unit similarity matrix corresponding to the knowledge unit according to the knowledge element similarity matrix; S3B7: Based on the knowledge unit similarity matrix, weighted calculation is performed using the weights corresponding to each knowledge element entity to obtain the comprehensive similarity of each knowledge unit; S3B8: The comprehensive similarity of each knowledge unit is averaged and the calculation result is compared with the preset global matching threshold to determine the content semantic correctness evaluation result.

[0012] As an implementation method of the present application, the logic consistency evaluation includes the following steps: S3C1: extracting knowledge elements and corresponding knowledge element entities through named entity recognition for the western medicine diagnosis and treatment standard knowledge base; For the electronic medical records to be evaluated, substructure elements and corresponding substructure element entities are extracted through named entity recognition; S3C2: constructing logic consistency evaluation rules based on triples in knowledge meta-entities; the triples are: disease, with examination, examination; S3C3: According to the construction logic consistency evaluation rules, the disease in the substructure element entity is located, and it is determined whether the fine-grained entity corresponding to the inspection entity in the substructure element entity is consistent with the fine-grained entity in the construction logic consistency evaluation rules to obtain the logic consistency evaluation result.

[0013] Another aspect of the present application further provides an electronic medical record quality assessment system, which executes any of the electronic medical record quality assessment methods described above, and the system includes a knowledge organization strategy generation module, a knowledge base generation module and a quality assessment module: The knowledge organization strategy generation module builds the metadata ontology of Western medicine diagnosis and treatment standards and supplements the ontology knowledge to form a knowledge organization strategy based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge; A knowledge base generation module organizes the Western medicine diagnosis and treatment standard data set through the knowledge organization strategy to construct a Western medicine diagnosis and treatment standard knowledge base; The quality assessment module is used to perform quality assessment on electronic medical records based on the Western medicine diagnosis and treatment standard knowledge base; the quality assessment includes formal normativeness assessment, content semantic correctness assessment and logical consistency assessment.

[0014] The embodiments of the present invention have the following technical effects: This application proposes an electronic medical record quality assessment method, which constructs a metadata ontology of Western medicine diagnosis and treatment standards and supplements ontology knowledge based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge to form a knowledge organization strategy; organizes the Western medicine diagnosis and treatment standard data set through the knowledge organization strategy to construct a Western medicine diagnosis and treatment standard knowledge base. This solves the problem of the computability of the Western medicine diagnosis and treatment standard knowledge base in the machine, converts the Western medicine diagnosis and treatment standard data set into systematic knowledge that can be used by the machine, and provides a basis for subsequent electronic medical record quality assessment.

[0015] Then, the quality of the electronic medical record is evaluated based on the Western medicine diagnosis and treatment standard knowledge base. In view of the rational use of systematic domain knowledge, the evaluation results of the present invention do not depend on the scale and quality of the model training data. It also realizes the intelligent automation of the quality evaluation of electronic medical records, greatly improves the processing efficiency and standardization of medical records, reduces labor intensity, reduces the reliance on manual operation during the evaluation of electronic medical records, reduces costs and improves efficiency.

[0016] The quality assessment of this application includes formal normativeness assessment, content semantic correctness assessment, and logical consistency assessment. It conducts a comprehensive assessment of the electronic medical records from three aspects: form, content, and logic, thereby improving the comprehensiveness of the assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 is a flow chart of an electronic medical record quality assessment method provided by an embodiment of the present invention; Figure 2 is a schematic diagram of the metadata ontology of Western medicine diagnosis and treatment standards provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of a department concept tree provided by an embodiment of the present invention; Figure 4 is a schematic diagram of an electronic medical record provided by an embodiment of the present invention; Figure 5 It is a diagram showing the instantiation of the metadata ontology of Western medicine diagnosis and treatment standards in the form of diseases provided by an embodiment of the present invention; Figure 6 It is a schematic diagram of upper and lower concepts using the respiratory system composition as an example provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0020] The knowledge organization methods in the existing Western medical diagnosis and treatment standards were created at an early time, and their purpose was to guide doctors' diagnosis and treatment behaviors rather than for machine use. Therefore, when this type of general metadata standard was established, it did not focus on reflecting the content of knowledge units, semantic features, etc., resulting in great obstacles to the automated use of machines, especially the basic inability to support the computability of knowledge. It cannot be directly applied to the intelligent automated quality assessment of electronic medical records. The accuracy of predicting quality assessment results through neural network models depends on training. The uneven quality of training data sources and the lack of medical knowledge of the annotators themselves result in poor quality of training data, which reduces the accuracy of model prediction. In addition, the multidisciplinary characteristics of Western medicine increase the difficulty of model learning.

[0021] Based on the above problems, the present invention provides a method for evaluating the quality of electronic medical records. Figure 1 As shown, the evaluation method comprises the following steps: S1: Based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge, construct the metadata ontology of Western medicine diagnosis and treatment standards and supplement the ontology knowledge to form a knowledge organization strategy; Reconstruct the DC metadata standard, form the disease resource metadata standard, abstract the metadata elements in the disease resource metadata standard to construct the disease resource metadata knowledge ontology and supplement the ontology with knowledge, and form a knowledge organization strategy that takes into account both traditional document units and semantic-centered knowledge units.

[0022] As an implementation method of the present application, S1 specifically includes: S11: Based on the DC metadata access point specification and industry guidance specifications, a metadata description framework for Western medicine diagnosis and treatment standards is established; The important retrieval points of the reused DC metadata content description part are the description elements (document units) of the Western medical diagnosis and treatment standards; in comparison with the specifications, the conceptual elements (knowledge units) and semantic elements (knowledge units) in the Western medical diagnosis and treatment standards are abstracted, and the three elements together constitute the core elements. Combined with the core elements, the metadata DIS of the Western medical diagnosis and treatment standards is compiled to form the metadata description framework of the Western medical diagnosis and treatment standards, realizing the transformation of the knowledge management unit from the document unit to the knowledge unit with semantic core.

[0023] The metadata description framework of Western medicine diagnosis and treatment standards includes description elements for organizing Western medicine diagnosis and treatment standard document units and concept elements and relationship elements for organizing Western medicine diagnosis and treatment standard knowledge units; the description elements include source, disease name, overview, and department classification; the concept elements include etiology, clinical manifestations, complications, examinations, diagnosis, and treatment; the relationship elements include metadata element relationships; S12: Based on the metadata description framework of western medicine diagnosis and treatment standards, abstract the classes and semantic relationships in the metadata of western medicine diagnosis and treatment standards; take each class as a knowledge unit; As an implementation method of the present application, in S1, the classes in the metadata ontology of the Western medicine diagnosis and treatment standards include department classes and disease classes; the subclasses of the disease class include: cause, clinical manifestations, examination, diagnosis, and treatment.

[0024] As an implementation method of the present application, in S1, the semantic relationship includes belonging to, causing, phenomenon expression, concurrent, having inspection, having diagnosis and having treatment.

[0025] S13: Based on the correspondence between department classification and disease names in the International Classification of Diseases and expert knowledge, department concept trees and disease word bags are constructed; The International Classification of Diseases is a standard classification method for coding diseases, injuries and poisonings that the World Health Organization (WHO) requires all member countries to adopt in health statistics. It is a unified, internationally accepted disease classification method that encodes diseases through letters and numbers to facilitate computer processing and data analysis. Expert knowledge refers to the professional knowledge of experts in the medical field in terms of understanding of diseases, diagnostic criteria and treatment methods. Based on the correspondence between department classification and disease names in the International Classification of Diseases and expert knowledge, the department concept tree and disease word bag are constructed to identify and classify disease information in electronic medical records, thereby supporting an objective and comprehensive evaluation of the quality of electronic medical records.

[0026] S14: Using the disease name as a connection and the semantic relationship as a relationship, the department concept tree is associated with the classes in the Western medicine diagnosis and treatment standard metadata to form a knowledge organization strategy.

[0027] Construct metadata ontology of western medicine diagnosis and treatment standards and expand ontology knowledge: abstract classes and inter-class relationships from DIS to complete the knowledge of metadata scheme; further hierarchical analysis of classes is performed to form concept trees, special word bags, etc. to supplement ontology knowledge, improve the knowledge system of the present invention, and form a knowledge base of western medicine diagnosis and treatment standards.

[0028] For example, DC metadata (ISO 15836), also known as "Dublin Core metadata", is the most widely used and influential standardized metadata in the current library community. The Dublin Core metadata set mainly includes 15 basic elements, which comprehensively describe resources from the perspectives of resource content description, intellectual property rights, and external attributes. DC implementers can choose from traditional metadata elements based on the specific requirements of the application, and no element is mandatory. In order to more fully comply with the best practices of machine-processable metadata, it is expected that implementers can (and are encouraged by DCMI) use more semantically precise terms. This provides support for the self-compiled field metadata of the present invention.

[0029] The present invention extracts important access points of the DC metadata content description part and reuses them as description elements, as shown in Table 1, which briefly describes and classifies disease resources from a literature perspective.

[0030] Table 1 Important access points in the DC content description

[0031] In order to realize the evolution of the basic unit of knowledge management from the document unit to the knowledge unit with semantics as the core, the concepts and relationships of disease resources of different disciplines in the data set are abstracted from the perspective of knowledge units in accordance with the Basic Specifications for Medical Record Writing and Management (2022 Edition) (hereinafter referred to as the "Specifications"), as the conceptual elements and relationship elements of the knowledge organization strategy of the present invention. The present invention compiled the metadata standard DIS, with a total of 11 metadata elements, namely, description elements: source, disease name, overview, department classification; concept elements: cause, clinical manifestations, complications, examination, diagnosis, treatment; relationship elements: metadata element relationship. The complete information is shown in Table 2.

[0032] Table 2 Metadata description framework for Western medicine diagnosis and treatment standards

[0033] The use of qualifiers can further refine the meaning of clinical manifestations and metadata element relationships. The specific qualifiers are shown in Table 3.

[0034] Table 3 Qualifiers of metadata elements

[0035] Ontology is the knowledge of metadata scheme, with rich semantic and logical reasoning functions, providing tools for direct knowledge manipulation for this system. The "disease" class is abstracted from DIS, and "source", "disease name" and "overview" are classified as data attributes of the disease class; considering that the essence of complications is "disease", no additional "complication" class is constructed. Department classification, concept elements, and relationship elements correspond to the classes and object attributes in the metadata ontology of Western medicine diagnosis and treatment standards. The ontology attribute definition description and semantic relationship description are shown in Table 4 and Table 5 respectively. For details of the ontology, see Figure 2 .

[0036]

[0037] In order to support the machine to obtain the relevant departments, disease entities and the relationship between the two in the electronic medical record, and solve the problem of mixed language expression of knowledge of various disciplines of Chinese and Western medicine in the data set. This system constructs a department concept tree (such as Figure 3) and disease word bags. Examples of disease word bags are as follows: Internal Medicine - Respiratory Medicine [airway burns, idiopathic interstitial pneumonia, bronchial and pulmonary tumor-like lesions, mixed acid-base imbalance, ..., pleural effusion after liver transplantation, bronchitis, health care-associated pneumonia, severe pneumonia, ossifying tracheobronchial disease, plastic bronchitis, nontuberculous mycobacterial lung disease, nocardiosis, hyperventilation syndrome, pulmonary allergic granulomatous vasculitis, tracheoesophageal fistula]; Ophthalmology - Ophthalmology [solar retinal burns, hypotrichosis, post-streptococcal glomerulonephritis, ..., acute retinal necrosis syndrome, benign lymphoepithelial lesions, solitary plasmacytoma of the orbit, posterior retinal membrane, eye burns, intracranial venous sinus thrombosis-related eye diseases].

[0038] S2: The Western medicine diagnosis and treatment standard data set is organized by the knowledge organization strategy to construct a Western medicine diagnosis and treatment standard knowledge base; the constructed Western medicine diagnosis and treatment standard knowledge base is computable and covers multiple disciplines.

[0039] Western medicine diagnosis and treatment standard data sets refer to information sets that summarize, integrate and standardize disease, symptom, examination, treatment and other related data in accordance with certain norms and standards in the field of Western medicine to meet the needs of clinical diagnosis and treatment, teaching, scientific research, etc. These data sets usually contain a large number of medical terms, definitions, classifications, diagnostic criteria, treatment plans, etc., which provide doctors with reliable references and help improve the accuracy and efficiency of diagnosis and treatment. However, they all exist in the form of books, papers, etc. in non-data formats.

[0040] This application proposes an electronic medical record quality assessment method, which constructs a metadata ontology of Western medicine diagnosis and treatment standards and supplements ontology knowledge based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge to form a knowledge organization strategy; organizes the Western medicine diagnosis and treatment standard data set through the knowledge organization strategy to construct a Western medicine diagnosis and treatment standard knowledge base. This solves the problem of the computability of the Western medicine diagnosis and treatment standard knowledge base in the machine, converts the Western medicine diagnosis and treatment standard data set into systematic knowledge that can be used by the machine, and provides a basis for subsequent electronic medical record quality assessment.

[0041] For example, this application compares GB / T 36344-2018 Information Technology Data Quality Evaluation Indicators (hereinafter referred to as "Indicators") and Basic Specifications for Medical Record Writing and Management (2022 Edition) (hereinafter referred to as "Specifications") to establish an electronic medical record quality evaluation indicator system and interpretation based on the Western medicine diagnosis and treatment standard knowledge base in this system. Different from previous studies, this application solves the problem of electronic medical record quality evaluation based on natural language at the knowledge level.

[0042] S3: Perform quality assessment on electronic medical records based on the Western medicine diagnosis and treatment standard knowledge base; the quality assessment includes formal normativeness assessment, content semantic correctness assessment, and logical consistency assessment.

[0043] With reference to GB / T 36344-2018 Information Technology Data Quality Evaluation Indicators and Basic Specifications for Medical Record Writing and Management (2022 Edition), an electronic medical record quality evaluation indicator system is established. The indicator system consists of three parts: formal standardization, content correctness, and logical consistency of electronic medical records. Within the indicator system, the Western medicine diagnosis and treatment standard knowledge base is used to conduct a knowledge-level quality assessment of electronic medical records. The quality assessment includes formal standardization assessment, content semantic correctness assessment, and logical consistency assessment. It is possible to conduct a fine-grained, intelligent, and scientific assessment of electronic medical records, assisting hospitals in improving the level of medical supervision.

[0044] According to the Western medicine diagnosis and treatment standard knowledge base, the quality of electronic medical records is evaluated. In view of the rational use of systematic domain knowledge, the evaluation results of the present invention do not depend on the scale and quality of model training data. It also realizes the intelligent automation of electronic medical record quality evaluation, greatly improves the processing efficiency and standardization of medical records, reduces labor intensity, reduces the reliance on manual operation during electronic medical record evaluation, reduces costs and improves efficiency.

[0045] The quality assessment of this application includes formal normative assessment, content semantic correctness assessment and logical consistency assessment, which comprehensively evaluates the electronic medical records from the three aspects of form, content and logic, thus improving the comprehensiveness of the assessment. Among them, formal normativeness refers to the complete value range and complete format of the written items in the electronic medical record text to be evaluated; content semantic correctness refers to the fact that the written items of the electronic medical record text contain the structural elements required by the "Specifications" and the structural elements can be mapped to the knowledge elements of several knowledge units in the Western medicine diagnosis and treatment standard knowledge base; logical consistency is reflected in the internal writing items of the electronic medical record, that is, the writing items in the electronic medical record need to satisfy the causal logic characteristic relationship in the knowledge base.

[0046] As an implementation method of the present application, in order to realize the standardization inspection of electronic medical records, the standardization degree of medical records and the problem list of non-standard medical records are recorded. The formal standardization evaluation is: text parsing of electronic medical records, and performing formal standardization evaluation on the parsed electronic medical records according to the formal requirements of the electronic medical record writing items, including the following steps: S3A1: Use regular expressions to parse the text of electronic medical records and extract the predefined fields in the electronic medical records and the filled content of each field; Exemplarily, regular expressions may be used to perform text parsing to extract predefined fields in the electronic medical record and the filled-in content of each field.

[0047] S3A2: Based on the case writing standards, a standard template for format review is prepared; Exemplarily, the formal requirements for each writing item of the electronic medical record are obtained from the "Specifications" for the electronic medical record, which are used as basic parameters of the format reviewer to generate a format review standard template to obtain a framework representation of the electronic medical record.

[0048] S3A3: Evaluate the completeness of the predefined fields in the electronic medical record and the formality of the content filled in each field based on the format review standard template.

[0049] For example, check whether the predefined fields in the template are filled in. Verify whether the field content meets the formal specifications of the filled content, for example: the number of words in the chief complaint (chief complaint field) does not exceed 20 and the content is not empty. In the history of present illness (history of present illness field), quotes need to be added for drug names, diagnosis names, surgery names, etc. Record the number of compliant fields, the total number of fields, and the list of non-compliant issues in the medical record.

[0050] The framework representation of the electronic medical record is as follows: <?xml version="1.0" encoding="UTF-8"?> <emr> <field name="病区"> <restrict> Fill in the ward name, cannot be empty< / restrict> < / field> <field name="姓名"> <restrict> Patient name, cannot be empty< / restrict> < / field> <field name="性别"> <restrict> Must be "Male" or "Female", cannot be empty< / restrict> < / field> <field name="年龄"> <restrict> Enter the integer age, cannot be empty< / restrict> < / field> <field name="病案号"> <restrict> Unique identifier, cannot be empty< / restrict> < / field> <field name="民族"> <restrict> Fill in ethnic information, cannot be empty< / restrict> < / field> <field name="婚姻"> <restrict> Fill in the marital status, cannot be left blank< / restrict> < / field> <field name="工作单位"> <restrict> Fill in the work unit, cannot be left blank< / restrict> < / field> <field name="职业"> <restrict> Fill in occupation information, cannot be empty< / restrict> < / field> <field name="现住址"> <restrict> Fill in the detailed address, cannot be empty< / restrict> < / field> <field name="籍贯"> <restrict> Fill in the place of origin information, cannot be empty< / restrict> < / field> <field name="入院日期"> <restrict> The date format is YYYY-MM-DD and cannot be empty< / restrict> < / field> <field name="病历记录日期"> <restrict> The date format is YYYY-MM-DD and cannot be empty< / restrict> < / field> <field name="病情陈述者"> <restrict> Fill in the medical statement, cannot be left blank< / restrict> < / field> <field name="与患者的关系"> <restrict> Fill in the relationship with the patient, cannot be empty< / restrict> < / field> <field name="联系电话"> <restrict> Please fill in your phone number. It cannot be left blank.< / restrict> < / field> <field name="主诉"> <restrict> Describe the main complaint, cannot be empty, no more than 20 words< / restrict> < / field> <field name="现病史"> <restrict> Describe the current medical history in detail. This field cannot be empty. The drug name, diagnosis name, and surgery name must be enclosed in quotation marks ("")< / restrict> < / field> <field name="既往史"> <restrict> Fill in the medical history, cannot be left blank< / restrict> < / field> <field name="个人史"> <restrict> Fill in personal history, cannot be empty< / restrict> < / field> <field name="婚育史"> <restrict> Fill in the marriage and childbearing history, cannot be empty< / restrict> < / field> <field name="家族史"> <restrict> Fill in family history, cannot be left blank< / restrict> < / field> <field name="体格检查"> <restrict> Fill in the physical examination results, cannot be empty< / restrict> < / field> <field name="专科情况"> <restrict> Fill in the specialty information, cannot be left blank< / restrict> < / field> <field name="辅助检查"> <restrict> Fill in the auxiliary inspection results, cannot be empty< / restrict> < / field> <field name="初步诊断"> <restrict> Fill in the preliminary diagnosis, cannot be left blank< / restrict> < / field> <field name="上级医师签名"> <restrict> Fill in the signature of the superior physician, cannot be left blank< / restrict> < / field> <field name="医师签名"> <restrict> Fill in the doctor's signature, cannot be left blank< / restrict> < / field> < / emr> As an implementation method of the present application, the content semantic correctness evaluation includes the following steps: S3B1: for the western medicine diagnosis and treatment standard knowledge base, extract knowledge elements and corresponding knowledge element entities through named entity recognition; For the electronic medical records to be evaluated, substructure elements and corresponding substructure element entities are extracted through named entity recognition; Specifically, according to the requirements of the "Specification" for electronic medical record writing specifications and the various classes in the ontology, fine-grained classes are determined. Specifically, fine-grained classes include fine-grained entities, which are 13 categories. The first-level entities are Disease: disease name, Reason: cause, Symptom: symptom, Check: inspection method, Drug: drug name, Operation: operation name; in order to refine the description of symptoms and inspections, and improve the evaluation accuracy of content correctness inspection and logical consistency inspection, the Symptom secondary entity is defined: Part: part, Character: property, Degree: degree, Factor: influencing factor, Time: time law; Check secondary entity is defined: Material: inspection material, Value: inspection index value.

[0051] For the knowledge base of western medicine diagnosis and treatment standards, named entity recognition is performed based on fine-grained entities to obtain knowledge elements and corresponding knowledge element entities.

[0052] Similarly, for the electronic medical records to be evaluated, named entity recognition is performed based on the fine-grained entities to obtain substructure elements and corresponding substructure element entities.

[0053] For example, the Western medicine diagnosis and treatment standard knowledge base contains N knowledge units, each of which Contains m types of knowledge elements. Each type of knowledge element may be empty, indicating that there is no entity of this category in this knowledge unit; each type of knowledge element may also have more than one. The i-th type of knowledge element in the n-th knowledge unit is represented as , where p represents the number of entities in the i-th type of knowledge element, It is the pth knowledge element entity of the i-th knowledge element of the n-th knowledge unit. 1≤n≤N, 0≤i≤m and 0≤p≤M. M is the number of classes of the above-mentioned fine-grained entity 13. In this application, N is the number of classes (i.e., the number of knowledge units), a total of 7, namely, the classes: department class and disease class, and the subclasses of the disease class include: cause, clinical manifestation, examination, diagnosis, and treatment.

[0054] Similarly, the electronic medical records to be evaluated include Class substructure element. Each substructure element may be empty, indicating that the class has no entity substructure elements and substructure element entities in the writing project, which can be represented as follows: , where W j is the jth substructure element, w jq is the qth substructure element entity in the jth substructure element, where 0≤j≤m and 0≤q≤M.

[0055] The writing items of electronic medical records usually include: general items, chief complaints, current medical history, past medical history, personal history, marital history, menstrual and reproductive history, family history, physical examination, auxiliary examinations, and preliminary diagnosis.

[0056] Formal normative assessment is conducted for all the above items.

[0057] The evaluation of semantic correctness and logical consistency of the content was conducted on “chief complaint, history of present illness, auxiliary examinations and preliminary diagnosis”.

[0058] Figure 4 The red boxes in the middle are different writing items in the electronic medical record, and the orange boxes are sub-structure element entities, which belong to the "symptoms" sub-structure element.

[0059] Figure 5 Taking the disease "acute tracheobronchitis" as an example, the instantiation form of the "Western medicine diagnosis and treatment standard metadata ontology" is given. The red box is the knowledge unit, and the dotted box is the complete display of the "clinical manifestation" knowledge unit. The orange box is the "symptom" knowledge meta-entity belonging to the "symptom" knowledge meta. The green box is the "property" knowledge meta-entity belonging to the "property" knowledge meta.

[0060] S3B2: Through the pre-trained language model, the knowledge meta-entities and sub-structure element entities are vectorized to obtain the knowledge meta-entity vector set and the sub-structure element entity vector set respectively; Before matching, knowledge elements and substructure elements need to be extracted and vectorized.

[0061] Knowledge base: From each knowledge unit Extract various knowledge meta-entities and construct knowledge meta-entity sets .

[0062] Writing project: Extract entities of various sub-structure elements from the writing project W and construct a sub-structure element entity set .

[0063] Convert the extracted entities into vector representations to facilitate subsequent similarity calculations. Use the following method to vectorize the entities: Embedding model: Utilize a pre-trained language model (BERT) to map each entity into a high-dimensional vector space.

[0064] Vectorized representation: For each knowledge entity and substructure element entities , simplify the hidden layer output of the language model and generate the corresponding vector representation (the vector representation of the jth entity of the i-th knowledge unit of the n-th knowledge unit) and (The k-th entity of the j-th type of substructure element). 0≤j≤q; 0≤k≤q.

[0065] After entity vectorization, each knowledge unit, each knowledge element entity and each sub-structure element entity of the writing project are converted into a vector set that can be used for quantitative calculation. and .

[0066] S3B3: Determine the knowledge element similarity calculation method based on the category of the substructure element entity vector set: If the class of the fine-grained entity of the substructure element entity vector set is the cause, the common sense causal feature model is used to calculate the knowledge element similarity; If the class of the fine-grained entity of the substructure element entity vector set is a part, the similarity of the knowledge element is calculated using the upper and lower concept tree; If the class of the fine-grained entity of the substructure element entity vector set is not the cause and the part, the cosine similarity is used to calculate the knowledge element similarity; The substructure elements are compared with the knowledge elements of the same category to preliminarily generate the initial knowledge element similarity matrix. Because the "cause" category has causal characteristics, the "part" category has superior and inferior characteristics, and the other categories meet the equivalent characteristics, the present invention adopts different similarity calculation methods for different categories to improve the accuracy of the evaluation results.

[0067] Vectorized representation for each knowledge unit Each type of knowledge element is represented by vector And write each type of substructure element in the project vector representation , calculate the i-th category The similarity matrix , where i=j.

[0068] No matter which similarity calculation method you choose, you must first handle the case of missing entities: if or is empty (i.e., there is no entity in the knowledge element or substructure element), the similarity matrix of this category is not calculated, and Set to empty.

[0069] In subsequent calculations, the similarity calculation of this category is skipped to avoid affecting the final comprehensive similarity.

[0070] S3B4: Based on each knowledge element entity vector set and the substructure element entity vector set, the selected similarity calculation method is used to calculate and obtain an initial knowledge element similarity matrix; (1) If the class of the substructure element entity vector set is not the cause and location, the class with the same characteristics is calculated using cosine similarity The initial knowledge element similarity matrix , where each element Representing knowledge entities Text substructure element entity The similarity between: ; (2) If the class of the substructure element entity vector set is part, the similarity of knowledge elements is calculated using the hierarchical concept tree.

[0071] For classes with hyponymy, use the hyponymy concept tree to derive the i-th class The initial knowledge element similarity matrix , where each element Representing knowledge entities Text substructure element entity The similarity between: ; The genus-species relationship, the whole-part relationship, and the inclusion relationship in the traditional thesaurus can all be classified as a hierarchical relationship. Specifically, the relationship between biological species, product types, the relationship between the body's system and organs, the relationship between subject branches or professional fields, administrative agencies and social groups, etc., all belong to a hierarchical relationship. Therefore, the hierarchical concept tree in the present invention is a body system-organ composition concept tree, and the construction of the concept tree requires the support of medical books from different departments.

[0072] Exemplarily, when constructing a hierarchical concept tree, the application obtains concepts from medical books of different departments and determines the hierarchical relationship between concepts, which is achieved through manual annotation. Then, according to the determined hierarchical relationship, these concepts are organized into a tree structure. Each concept node is connected to its direct superordinate concept and direct subordinate concept to form a complete hierarchical concept tree. Figure 6 Take the respiratory system as an example: the two red boxes in the example represent two branches, and the same box becomes the "same branch of the hierarchical concept tree". According to the constructed hierarchical concept tree, the semantic depth method is selected to calculate the similarity of knowledge elements.

[0073] (3) If the class of the substructure element entity vector set is the cause, the common sense causal rule base is used to calculate the knowledge element similarity.

[0074] For example, event A leads to event B, which is a causal rule. The causal rule library in this application is formulated using similar causal rules. In this application, common sense causal relationships widely accepted in the medical field are first collected through expert interviews, literature review, etc.; then the big model technology is used to automatically extract and expand causal relationships from a large amount of text data, and then the construction of the common sense causal rule library is automatically completed.

[0075] The knowledge element similarity is calculated using the common sense causal rule base: First, the etiology knowledge meta-entities in the knowledge base knowledge unit are organized into standard etiology word bags. Specifically, taking the disease "acute tracheobronchitis" as an example, there is a cause in the Western medicine diagnosis and treatment standard (cold air, leading to acute tracheobronchitis), and cold air is the cause in the Western medicine diagnosis and treatment standard.

[0076] Then, call the GPT-4 API to expand the common sense causal relationship: Use the GPT-4 API to expand the relevant common sense causal relationship by inputting the cause knowledge meta-entity in the word bag, so as to build a relatively complete common sense causal rule library. Specifically, let the big model give the synonyms of "cold air" and get "cool wind, cold air, get cold, freeze, blow cold wind, get cold", and expand and improve the causal rule to ([cold air, cool wind, cold air, get cold, freeze, blow cold wind, get cold], leading to acute tracheobronchitis).

[0077] Finally, we traverse the common sense causal rules one by one to get the i-th category The initial knowledge element similarity matrix , where each element Representing knowledge entities Text substructure element entity The similarity between: ; Specifically, the knowledge element in the knowledge base is cold air, and the substructure element in the electronic medical record is catching a cold. The two are in the same common sense causal rule, and the similarity is 1.

[0078] Finally, assuming that the knowledge element of the i-th category contains p entities and the j-th structural element of the writing project contains q entities, then the similarity matrix of the i-th category is It is a p×q matrix, that is, the number of rows is the number of knowledge element entities p, and the number of columns is the number of substructure element entities q.

[0079] S3B5: For the initial knowledge element similarity matrix, only the maximum similarity value of each column is retained, and the remaining similarity values ​​are assigned to 0, thereby obtaining a knowledge element similarity matrix; For non-empty similarity matrices of various knowledge elements , retaining the maximum similarity value of each column to ensure that each substructure element entity finds the closest knowledge meta-entity in the category.

[0080] Find the maximum value in each column: Find the maximum value in the kth column of , record the position, and set other values ​​in the column to 0.

[0081] Constructing the updated knowledge element similarity matrix : Generate a matrix that retains only the maximum value in each column. For example, if for: ; If the maximum value of column 1 is , the maximum value of column 2 is , the maximum value of column 3 is , then update for: ; S3B6: Generate a knowledge unit similarity matrix corresponding to the knowledge unit according to the knowledge element similarity matrix; Vectorize knowledge units The knowledge element similarity matrix of all non-empty categories The knowledge unit similarity matrix composed of corresponding knowledge units , forming a block diagonal matrix: ; If some categories are empty, the corresponding blocks will be omitted in the overall similarity matrix. This block diagonal structure shows the matching of knowledge units with each non-empty category of the written items and provides the basis for the subsequent comprehensive similarity calculation.

[0082] S3B7: Based on the knowledge unit similarity matrix, weighted calculation is performed using the weights corresponding to each knowledge element entity to obtain the comprehensive similarity of each knowledge unit; Calculate the weighted average similarity for each class for The similarity matrix of each class Assign weight , set the weight of the category corresponding to the first-level entity higher than the weight of the category corresponding to the second-level entity. In medical texts, the first-level entities are core to understanding the entire text content and scenario because they directly describe the patient's condition, treatment methods, etc. In contrast, second-level entities are more used to supplement the specific details of these main concepts. Therefore, the weight of the first-level entity is set higher to ensure that when matching, the system can pay more attention to these core information and avoid being diluted by too much detailed information, thereby affecting the overall judgment. While ensuring that the weight of the category corresponding to the first-level entity is higher than the weight of the category corresponding to the second-level entity, the weights corresponding to each first-level entity category and the weights corresponding to each second-level entity category are added, and the weighted average similarity is calculated: ; in, ; It means selecting the largest value in the kth column of the matrix. or If it is empty, the similarity calculation of this type will be skipped and the total weight will be adjusted to ensure the accuracy of the calculation.

[0083] Computational Knowledge Unit The comprehensive similarity Add the weighted average similarities of all non-empty categories to get the knowledge unit The comprehensive similarity of: .

[0084] S3B8: The comprehensive similarity of each knowledge unit is averaged and the calculation result is compared with the preset global matching threshold to determine the content semantic correctness evaluation result.

[0085] The comprehensive similarity of each knowledge unit can be displayed in a list form, such as: .

[0086] Set the global matching threshold , the comprehensive similarity of each knowledge unit is averaged to obtain the average comprehensive similarity Avg Overall Similarity. If Avg Overall Similarity ≥ , then it is considered that the written project matches the knowledge base successfully, that is, the quality assessment result of the electronic medical record is good; otherwise, it is considered that it does not match and the assessment result is poor.

[0087] By counting the historical matching results, we can determine the matching accuracy under different global matching thresholds. Based on these statistical results, we can select an optimal global matching threshold that can balance the accuracy. The analysis of historical data can help us understand the distribution of matching successes and failures under a specific global matching threshold, so as to set the global matching threshold more reasonably.

[0088] While obtaining the quality assessment results, it is also possible to find knowledge units with not particularly good results based on the comprehensive similarity values ​​of each knowledge unit, providing guidance for the improvement direction of subsequent medical record writing.

[0089] As an implementation method of the present application, the logic consistency evaluation includes the following steps: S3C1: extracting knowledge elements and corresponding knowledge element entities through named entity recognition for the western medicine diagnosis and treatment standard knowledge base; For the electronic medical records to be evaluated, substructure elements and corresponding substructure element entities are extracted through named entity recognition; The specific method is the same as step S3B1. The knowledge element entity and substructure element entity obtained in S3B1 can also be directly called to extract parts from them, and obtain the substructure element entities of the inspection method, inspection material and inspection index value in the auxiliary inspection writing item; and obtain the substructure element entity of the disease name in the preliminary diagnosis writing item.

[0090] S3C2: construct logical consistency evaluation rules based on triples in knowledge meta-entities; the triples are: disease, with examination, examination, and the record format is (disease, with examination, examination); S3C3: According to the construction logic consistency evaluation rules, the disease in the substructure element entity is located, and it is determined whether the fine-grained entity corresponding to the inspection entity in the substructure element entity is consistent with the fine-grained entity in the construction logic consistency evaluation rules to obtain the logic consistency evaluation result.

[0091] Logical consistency is reflected in the internal writing items of the electronic medical record, that is, the evaluation of the writing items in the electronic medical record satisfies the causal logic characteristics in the knowledge base. Regarding the intelligent quality assessment of electronic medical records, the "Specifications" propose that when conducting quality assessments of electronic medical records, it is necessary to consider "auxiliary examinations should record the main examinations and their results related to the current disease before admission", with special emphasis on "related to the current disease", involving logical tests between the two writing items "examination" and "preliminary diagnosis". This part returns a logical consistency assessment report: prompt "Examination items and disease Meet logical consistency" or "check items and disease Does not satisfy logical consistency. It is an entity in the substructure element "Check: Check method", It is an entity in the substructure element - "Disease:Disease name".

[0092] by As input, search the Western medicine diagnosis and treatment standard knowledge base to obtain In the inspection knowledge unit of the knowledge base, call the knowledge meta-entity and substructure element entity obtained by S3B1 again, extract some of them, and obtain the inspection method, inspection material and inspection index value knowledge meta-entity in the inspection knowledge unit , , .Establish - There is inspection -> : Logical consistency assessment standard. For example: {Bacterial infection}-with examination->{Routine blood test}: {White blood cell count}{≥10,000 / uL}, indicating that the disease is bacterial infection, white blood cell count ≥10,000 / uL, and a routine blood test is logical, which can indicate "Examination item {Routine blood test} and disease {Bacterial infection} meet logical consistency".

[0093] The logical consistency evaluation standard is used to evaluate whether the electronic medical record writing items "auxiliary examination" and "preliminary diagnosis" meet the causal characteristics in the knowledge base. If they meet, it will prompt "examination item {} and disease {} meet logical consistency". If not, it will prompt "examination item {} and disease {} do not meet logical consistency".

[0094] Another aspect of the present application further provides an electronic medical record quality assessment system, which executes any of the electronic medical record quality assessment methods described above, and the system includes a knowledge organization strategy generation module, a knowledge base generation module and a quality assessment module: The knowledge organization strategy generation module compiles the metadata ontology of Western medicine diagnosis and treatment standards and supplements the ontology knowledge based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge to form a knowledge organization strategy; A knowledge base generation module organizes the Western medicine diagnosis and treatment standard data set through the knowledge organization strategy to construct a Western medicine diagnosis and treatment standard knowledge base; The quality assessment module is used to perform quality assessment on electronic medical records based on the Western medicine diagnosis and treatment standard knowledge base; the quality assessment includes formal normativeness assessment, content semantic correctness assessment and logical consistency assessment.

[0095] It should be noted that the terms used in the present invention are only for describing specific embodiments, rather than limiting the scope of the present application. As shown in the present specification, unless the context clearly indicates an exception, the words "one", "a", "a kind of" and / or "the" do not specifically refer to the singular, but may also include the plural. The terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of more restrictions, the elements defined by the sentence "include one..." do not exclude the presence of other identical elements in the process, method or device including the elements.

[0096] It should also be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", etc. should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the quality of electronic medical records, characterized in that: The evaluation method comprises the following steps: S1: Based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge, construct the metadata ontology of Western medicine diagnosis and treatment standards and supplement the ontology knowledge to form a knowledge organization strategy; S2: Organizing the Western medicine diagnosis and treatment standard data set through the knowledge organization strategy to build a Western medicine diagnosis and treatment standard knowledge base; S3: Perform quality assessment on electronic medical records based on the Western medicine diagnosis and treatment standard knowledge base; the quality assessment includes formal normativeness assessment, content semantic correctness assessment, and logical consistency assessment.

2. The electronic medical record quality assessment method according to claim 1, characterized in that: In S1, based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge, a metadata ontology of Western medicine diagnosis and treatment standards is constructed and supplemented with knowledge to form a knowledge organization strategy, including: S11: Based on the retrieval point specification of DC metadata and in accordance with industry guidance specifications, a metadata description framework for Western medicine diagnosis and treatment standards is constructed; the metadata description framework for Western medicine diagnosis and treatment standards includes description elements for organizing Western medicine diagnosis and treatment standard document units and concept elements and relationship elements for organizing Western medicine diagnosis and treatment standard knowledge units; the description elements include source, disease name, overview, and department classification; the concept elements include etiology, clinical manifestations, complications, examinations, diagnosis, and treatment; the relationship elements include metadata element relationships; S12: Based on the metadata description framework of western medicine diagnosis and treatment standards, abstract the classes and semantic relationships in the metadata of western medicine diagnosis and treatment standards; take each class as a knowledge unit; S13: Based on the correspondence between department classification and disease names in the International Classification of Diseases and expert knowledge, department concept trees and disease word bags are constructed; S14: Using the disease name as a connection and the semantic relationship as a relationship, the department concept tree is associated with the classes in the Western medicine diagnosis and treatment standard metadata to form a knowledge organization strategy.

3. The electronic medical record quality assessment method according to claim 2, characterized in that: In S1, the classes in the metadata ontology of Western medicine diagnosis and treatment standards include department class and disease class; the subclasses of the disease class include: cause, clinical manifestation, examination, diagnosis, and treatment.

4. The electronic medical record quality assessment method according to claim 2, characterized in that: In S1, the semantic relations include belonging to, causing, expressing phenomenon, concurrent, having examination, having diagnosis and having treatment.

5. The electronic medical record quality assessment method according to claim 1, characterized in that: The formal normative evaluation is: performing text analysis on the electronic medical record, and performing formal normative evaluation on the analyzed electronic medical record according to the formal requirements of the electronic medical record writing project, including the following steps: S3A1: Use regular expressions to parse the text of electronic medical records and extract the predefined fields in the electronic medical records and the filled content of each field; S3A2: Based on the case writing standards, a standard template for format review is prepared; S3A3: Evaluate the completeness of the predefined fields in the electronic medical record and the formality of the content filled in each field based on the format review standard template.

6. The electronic medical record quality assessment method according to claim 1, characterized in that: The content semantic correctness evaluation includes the following steps: S3B1: for the western medicine diagnosis and treatment standard knowledge base, extract knowledge elements and corresponding knowledge element entities through named entity recognition; For the electronic medical records to be evaluated, substructure elements and corresponding substructure element entities are extracted through named entity recognition; S3B2: Through the pre-trained language model, the knowledge meta-entities and sub-structure element entities are vectorized to obtain the knowledge meta-entity vector set and the sub-structure element entity vector set respectively; S3B3: Determine the knowledge element similarity calculation method based on the category of the substructure element entity vector set: If the class of the substructure element entity vector set is the cause, the common sense causal feature model is used to calculate the knowledge element similarity; If the class of the substructure element entity vector set is part, the similarity of knowledge element is calculated using the upper and lower concept tree; If the class of the substructure element entity vector set is not the cause and location, the cosine similarity is used to calculate the knowledge element similarity; S3B4: Based on each knowledge element entity vector set and the substructure element entity vector set, the selected similarity calculation method is used to calculate and obtain an initial knowledge element similarity matrix; S3B5: For the initial knowledge element similarity matrix, only the maximum similarity value of each column is retained, and the remaining similarity values ​​are assigned to 0, thereby obtaining a knowledge element similarity matrix; S3B6: Generate a knowledge unit similarity matrix corresponding to the knowledge unit according to the knowledge element similarity matrix; S3B7: Based on the knowledge unit similarity matrix, weighted calculation is performed using the weights corresponding to each knowledge element entity to obtain the comprehensive similarity of each knowledge unit; S3B8: The comprehensive similarity of each knowledge unit is averaged and the calculation result is compared with the preset global matching threshold to determine the content semantic correctness evaluation result.

7. The electronic medical record quality assessment method according to claim 1, characterized in that: The logical consistency assessment comprises the following steps: S3C1: extracting knowledge elements and corresponding knowledge element entities through named entity recognition for the western medicine diagnosis and treatment standard knowledge base; For the electronic medical records to be evaluated, substructure elements and corresponding substructure element entities are extracted through named entity recognition; S3C2: constructing logic consistency evaluation rules based on triples in knowledge meta-entities; the triples are: disease, with examination, examination; S3C3: According to the construction logic consistency evaluation rules, the disease in the substructure element entity is located, and it is determined whether the fine-grained entity corresponding to the inspection entity in the substructure element entity is consistent with the fine-grained entity in the construction logic consistency evaluation rules to obtain the logic consistency evaluation result.

8. An electronic medical record quality assessment system, executing an electronic medical record quality assessment method according to any one of claims 1 to 7, characterized in that: The system includes a knowledge organization strategy generation module, a knowledge base generation module and a quality assessment module: The knowledge organization strategy generation module builds the metadata ontology of Western medicine diagnosis and treatment standards and supplements the ontology knowledge to form a knowledge organization strategy based on DC metadata standards, industry guidance specifications, international disease classification and expert knowledge; A knowledge base generation module organizes the Western medicine diagnosis and treatment standard data set through the knowledge organization strategy to construct a Western medicine diagnosis and treatment standard knowledge base; The quality assessment module is used to perform quality assessment on electronic medical records based on the Western medicine diagnosis and treatment standard knowledge base; the quality assessment includes formal normativeness assessment, content semantic correctness assessment and logical consistency assessment.

Images