Disclosure of Invention
In order to solve the above problems, the present application provides a knowledge-graph-based medical records diagnosis and surgical ICD encoding method, comprising:
S10, establishing a knowledge graph of disease and operation ICD codes, wherein the knowledge graph of the ICD codes comprises diagnosis and treatment paths corresponding to each ICD code, and each diagnosis and treatment path comprises a plurality of diagnosis and treatment nodes;
S20, responding to the diagnosis and treatment information to be matched, and carrying out semantic analysis on the diagnosis and treatment information to extract keywords corresponding to the diagnosis and treatment information;
S30, determining a target node related to the diagnosis and treatment information based on the keywords;
S40, retrieving a knowledge graph of the ICD code, and matching the target node to the diagnosis and treatment nodes so as to determine at least one diagnosis and treatment path matched by the target node;
S50, outputting the at least one diagnosis and treatment path and the corresponding ICD codes.
Optionally, the step of establishing an ICD code knowledge graph, where the ICD code knowledge graph includes a diagnosis and treatment path corresponding to each ICD code, and each diagnosis and treatment path includes a plurality of diagnosis and treatment nodes includes:
Collecting medical data, and performing data cleaning on the medical data, wherein the medical data comprises disease information, diagnosis and treatment processes, medicine information, operation records and the like;
identifying named entities from the cleaned medical data, and giving identifiers of each named entity, wherein the named entities are used for representing semantic objects in the medical data;
And analyzing the association relation between the named entities, and integrating the identified named entities and the association relation into a preset knowledge frame to form the knowledge graph.
Optionally, the step of establishing an ICD code knowledge graph, where the ICD code knowledge graph includes a diagnosis and treatment path corresponding to each ICD code, and each diagnosis and treatment path includes a plurality of diagnosis and treatment nodes further includes:
carrying out knowledge verification on the knowledge graph, wherein the knowledge verification comprises expert verification and data cross verification;
and regularly collecting user and expert opinions, and updating knowledge of the knowledge graph.
Optionally, the step of performing semantic analysis on the diagnosis and treatment information to extract keywords corresponding to the diagnosis and treatment information in response to obtaining the diagnosis and treatment information to be matched includes:
Extracting target keywords according to the diagnosis and treatment information to be matched, and marking the part of speech of each target keyword;
And identifying the named entities in the diagnosis and treatment information based on the target keywords marked with parts of speech, and classifying and identifying the named entities.
Optionally, the step of determining the target node related to the diagnosis and treatment information based on the keyword includes:
determining the weight of each target keyword according to a preset strategy according to the importance degree of the target keyword in the diagnosis and treatment information;
determining semantic roles of each named entity in the diagnosis and treatment information based on the weight of the target keyword and the classification identification of the named entity so as to determine the association relationship between the named entities;
sequentially connecting entity nodes with association relations to construct a semantic association path;
And extracting diagnosis and treatment nodes in the semantic association path, and determining target nodes related to the diagnosis and treatment information.
Optionally, the step of retrieving the knowledge-graph of the ICD code, and matching the target node to the plurality of diagnosis and treat nodes to determine at least one diagnosis and treat path matched by the target node includes:
Performing query operation from the ICD encoded knowledge graph to search diagnosis and treatment nodes in the ICD encoded knowledge graph, so that the target nodes are matched with the diagnosis and treatment nodes one by one;
when the target node is successfully matched, searching along a relation chain between named entities based on the knowledge graph of the ICD code to call a diagnosis and treatment path connected with the target node;
And calculating the suitability of each called diagnosis and treatment path, and screening at least one diagnosis and treatment path with the suitability value exceeding a preset threshold, wherein the suitability is used for representing the matching degree of the diagnosis and treatment path and the target node.
Optionally, the step of performing a query operation from the knowledge graph of the ICD code to retrieve diagnosis and treat nodes in the knowledge graph of the ICD code, and performing matching calculation on the target nodes and the diagnosis and treat nodes one by one includes at least one of the following:
comparing entity names of the target node and the diagnosis and treatment node to perform name-based matching calculation;
Invoking the category of the target node, and comparing the category of the target node with the category of the diagnosis and treatment node based on the category of the target node so as to perform category-based matching calculation;
retrieving the attribute of the target node, and comparing the attribute of the target node with the attribute of the diagnosis and treatment node based on the attribute of the target node so as to perform attribute-based matching calculation;
And analyzing the semantic relation between the target nodes, and comparing the semantic relation of the target nodes with the semantics of the diagnosis and treatment nodes based on the semantic relation of the target nodes so as to perform semantic-based matching calculation.
Optionally, the step of calculating the suitability of each diagnosis and treatment path, and screening at least one diagnosis and treatment path with the suitability value exceeding a preset threshold value includes:
inquiring a medical database, and determining association frequencies among diagnosis and treatment nodes in the semantic association path;
The method comprises the steps of calling case data in historical diagnosis and treatment data, and analyzing diagnosis and treatment effective rate of the semantic association path in the historical diagnosis and treatment path;
and calculating the logic consistency of the current semantic association path and the called diagnosis and treatment path according to the association frequency and the diagnosis and treatment effective rate.
Optionally, the step of calculating the suitability of each diagnosis and treatment path, and screening at least one diagnosis and treatment path with the suitability value exceeding a preset threshold further includes:
extracting the disease, symptoms and treatment modes related to the diagnosis and treatment information by calling semantic analysis contents of the diagnosis and treatment information;
And matching diagnosis and treatment nodes in the semantic association path with the diagnosis and treatment information based on the extracted diseases, symptoms and treatment modes, and calculating the relativity of the diagnosis and treatment nodes in the semantic association path and the content of the diagnosis and treatment information.
Optionally, the step of calculating the suitability of each diagnosis and treatment path, and screening at least one diagnosis and treatment path with the suitability value exceeding a preset threshold further includes:
Determining evaluation indexes related to the diagnosis and treatment path, wherein the evaluation indexes comprise the relevance, the logic coherence degree, the treatment effect and the safety, and determining a corresponding weight coefficient for each evaluation index;
Performing score distribution on the evaluation index based on the diagnosis and treatment information so as to calculate a weighted total score value of the diagnosis and treatment path according to the weight coefficient of the evaluation index;
And normalizing the weighted total score value of the diagnosis and treatment path to determine the suitability of the diagnosis and treatment path.
The medical records diagnosis and operation ICD coding method based on the knowledge graph provided by the application is used for extracting keywords corresponding to diagnosis and treatment information by carrying out semantic analysis on the diagnosis and treatment information, determining target nodes related to the diagnosis and treatment information based on the keywords, calling a pre-established knowledge graph of ICD codes, matching the target nodes to the plurality of diagnosis and treatment nodes to determine at least one diagnosis and treatment path matched with the target nodes, and outputting the at least one diagnosis and treatment path and the ICD codes corresponding to the at least one diagnosis and treatment path. The diagnosis and treatment information can be subjected to semantic analysis, and the target node is matched to the plurality of diagnosis and treatment nodes based on the pre-established knowledge graph so as to determine at least one diagnosis and treatment path matched by the target node, so that accurate ICD codes can be accurately and rapidly matched for the diagnosis and treatment process of each patient, the code management cost is greatly reduced, and the user experience is improved.
Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the application. Rather, they are merely examples of apparatus and methods consistent with aspects of the application as detailed in the accompanying claims.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, the element(s) defined by the phrase "comprising one does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other like elements in different embodiments of the application having the same meaning as may be defined by the same meaning as they are explained in this particular embodiment or by further reference to the context of this particular embodiment.
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
First embodiment
The application provides a knowledge-based medical record diagnosis and operation ICD coding method, and FIG. 1 is a flow chart of a knowledge-based medical record diagnosis and operation ICD coding method according to an embodiment of the application.
As shown in fig. 1, in an embodiment, a method for diagnosing and operating an ICD encoding a medical record based on a knowledge graph includes:
And S10, establishing a knowledge graph of the ICD codes of the diseases and the operations, wherein the knowledge graph of the ICD codes comprises diagnosis and treatment paths corresponding to each ICD code, and each diagnosis and treatment path comprises a plurality of diagnosis and treatment nodes.
International disease classification (international Classification of diseases, ICD) is a system that classifies diseases according to rules and is represented by a coded method, based on certain characteristics of the disease. The Knowledge map (knowledgegraph), called Knowledge domain visualization or Knowledge domain mapping map in book condition report, is a series of various graphs showing Knowledge development process and structural relationship, and uses visualization technology to describe Knowledge resources and their carriers, and excavate, analyze, construct, draw and display Knowledge and their interrelationships. The method combines theory and method of subjects such as application mathematics, graphics, information visualization technology, information science and the like with methods such as metering introduction analysis, co-occurrence analysis and the like, and utilizes a visualized map to visually display the core structure, development history, leading edge field and overall knowledge architecture of ICD encoding subjects. In the application process of the ICD coding knowledge map, the complex medical ICD coding knowledge field can be displayed through data mining, information processing, knowledge metering and graphic drawing, the dynamic development rule of the medical knowledge field is revealed, and a practical and valuable reference is provided for ICD coding discipline research. Illustratively, the diagnosis and treatment path is a standardized diagnosis and treatment mode, and aims to standardize medical behaviors, improve medical quality and reduce medical cost. diagnosis and treatment path a standardized medical care plan with strict working sequence and accurate time requirements is formulated for the diagnosis, treatment, rehabilitation and care process of a single disease. The diagnosis and treatment path not only improves the medical quality and the management level, but also promotes the upgrading and informatization construction of the hospital information system. Clinical nodes refer to critical periods or key factors throughout the disease progression that are important in medicine, i.e., staged, vital nodes in the disease progression. Clinical nodes play a very important role in clinical practice and are the basis for the deep understanding and effective intervention of the disease progression process. A plurality of clinical diagnosis and treatment nodes are connected in series to form a determined diagnosis and treatment path.
And S20, carrying out semantic analysis on the diagnosis and treatment information in response to obtaining the diagnosis and treatment information to be matched so as to extract keywords corresponding to the diagnosis and treatment information.
Illustratively, the diagnosis and treatment information mainly comprises patient information clusters composed of various clinical and related contents such as personal basic information, registration information, treatment information, hospitalization doctor's advice information, expense information, image data and test results of patients. Such information plays a vital role in the medical procedure, not only for diagnosis and treatment, but also as a basis for medical institutions to provide high quality medical services. The semantic analysis (SEMANTIC ANALYSIS) refers to a process of performing deep understanding and analysis on natural language text of diagnosis and treatment information so as to identify semantic information such as entities, relations, emotions and the like in the text. The objective of semantic analysis is to convert natural language text into a form that can be understood and processed by a computer, including entity recognition, relationship extraction, emotion analysis, semantic role labeling, event extraction, semantic similarity calculation, and the like. In the process of carrying out semantic analysis on the diagnosis and treatment information, a rule-based method, a statistical learning-based method, a deep learning-based method and the like can be adopted.
And S30, determining a target node related to the diagnosis and treatment information based on the keywords.
And determining the semantic role of each identified entity based on the keywords extracted from the diagnosis and treatment information, so that the target node corresponding to the diagnosis and treatment information can be determined.
And S40, retrieving the knowledge graph of the ICD code, and matching the target node to the diagnosis and treatment nodes so as to determine at least one diagnosis and treatment path matched by the target node.
Establishing a knowledge graph containing the corresponding relation between the treatment scheme and the coding rule, taking the fixed knowledge of a doctor as the knowledge graph, and carrying out the diagnosis and treatment process of a patient with a disease according to a fixed diagnosis and treatment path. Illustratively, it is assumed that there are two modes of treatment for a disease, one taking a drug and one doing a surgery. Recording the fixed knowledge of doctors to the knowledge map, selecting the corresponding diagnosis and treatment path for the case taking the medicine, selecting the corresponding diagnosis and treatment path for the case doing the operation, and then selecting the corresponding diagnosis and treatment path for the case taking the medicine. The prior knowledge graph establishes the corresponding relation between the diagnosis and treatment path and ICD codes. After the patient comes in for diagnosis and treatment, the doctor determines diagnosis and treatment path nodes according to the records for diagnosing the patient, and automatically identifies at least one diagnosis and treatment path according to the diagnosis and treatment path nodes.
S50, outputting the at least one diagnosis and treatment path and the corresponding ICD codes.
Illustratively, the physician is well aware of the path of the diagnosis and treatment. In the search to determine ICD codes, nodes of the diagnosis and treatment path can be selected from any step. Can go upwards and can go downwards. For example, a CT is done today, and the source can be traced upward from the node record of the CT, and the treatment can also be performed from the CT to the subsequent treatment node.
The embodiment comprises the steps of carrying out semantic analysis on diagnosis and treatment information to extract keywords corresponding to the diagnosis and treatment information, determining target nodes related to the diagnosis and treatment information based on the keywords, retrieving a pre-established knowledge graph of ICD codes, matching the target nodes to the plurality of diagnosis and treatment nodes to determine at least one diagnosis and treatment path matched with the target nodes, and outputting the at least one diagnosis and treatment path and the ICD codes corresponding to the at least one diagnosis and treatment path. The ICD coding method can accurately and rapidly match the diagnosis and treatment process of each patient to the accurate ICD coding, greatly reduces the coding management cost and improves the user experience.
Optionally, the step of establishing an ICD code knowledge graph, where the ICD code knowledge graph includes a diagnosis and treatment path corresponding to each ICD code, and each diagnosis and treatment path includes a plurality of diagnosis and treatment nodes includes:
Collecting medical data, and performing data cleaning on the medical data, wherein the medical data comprises disease information, diagnosis and treatment processes, medicine information, operation records and the like;
identifying named entities from the cleaned medical data, and giving identifiers of each named entity, wherein the named entities are used for representing semantic objects in the medical data;
and analyzing the association relation between the named entities, and integrating the identified named entities and the association relation into a preset knowledge frame to construct the knowledge graph.
By way of example, relevant medical data may be collected first, including various medical ICD codes, disease information, medical procedures, drug information, and the like. Such data may originate from medical records, clinical guidelines, drug specifications, research papers, and the like. And cleaning the collected data, and processing incomplete, inaccurate or repeated data to ensure the quality and accuracy of the data. Key entities such as diseases, symptoms, medicines, examination items, etc. can be identified from the collected data and given unique identifiers. By further identifying relationships between entities, such as treatments for certain diseases, side effects of certain drugs, etc., and establishing a relationship between these entities. And finally, integrating the extracted entities and relations into a unified knowledge framework to form a structured knowledge graph.
Optionally, the step of establishing an ICD code knowledge graph, where the ICD code knowledge graph includes a diagnosis and treatment path corresponding to each ICD code, and each diagnosis and treatment path includes a plurality of diagnosis and treatment nodes further includes:
carrying out knowledge verification on the knowledge graph, wherein the knowledge verification comprises expert verification and data cross verification;
and regularly collecting user and expert opinions, and updating knowledge of the knowledge graph.
Illustratively, the constructed knowledge graph is verified to ensure its accuracy and integrity. This may include collaboration with an expert, data cross-validation, and the like. Since medical knowledge is continuously updated, knowledge maps need to be updated periodically to reflect the latest medical progress and research results. Alternatively, a suitable data storage means, such as a relational database, a graph database, etc., may be selected to store the constructed knowledge-graph. A structured, easy to query and analyze medical information resource may be provided by developing an interface or platform that allows a user to query and access information in a knowledge-graph.
Optionally, the step of performing semantic analysis on the diagnosis and treatment information to extract keywords corresponding to the diagnosis and treatment information in response to obtaining the diagnosis and treatment information to be matched includes:
Extracting target keywords according to the diagnosis and treatment information to be matched, and marking the part of speech of each target keyword;
And identifying the named entities in the diagnosis and treatment information based on the target keywords marked with parts of speech, and classifying and identifying the named entities.
Illustratively, the diagnosis and treatment information is preprocessed, including removal of irrelevant characters, conversion to a standard format (such as lowercase), word segmentation (segmentation of text into individual words or phrases). Assume that a diagnosis and treatment message of hypertension using aspirin is provided. After pretreatment, irrelevant characters (such as punctuation marks) are removed, converted into a standard format and segmented into terms such as "use", "aspirin", "treatment" and "hypertension".
Extracting keywords from the pre-processed text may be implemented by a variety of methods, including statistical-based methods (e.g., TF-IDF), semantic-based methods (e.g., part-of-speech tagging, named entity recognition), or deep learning-based methods (e.g., word embedding, neural network), for example. For example, the key word is extracted using the TF-IDF method. The extraction results may include "aspirin", "treatment", "hypertension".
Illustratively, the part of speech of each word, such as nouns, verbs, adjectives, etc., may be further identified and tagged to facilitate further semantic analysis. For example, "aspirin" is labeled "noun," treatment "is labeled" verb, "and" hypertension "is labeled" noun.
Illustratively, named entities in text, such as person names, place names, organization names, drug names, etc., are identified and categorized, which have important and wide-ranging uses in the medical field. For example, "aspirin" is identified as "drug name" and "hypertension" is identified as "disease name".
Optionally, the step of determining the target node related to the diagnosis and treatment information based on the keyword includes:
determining the weight of each target keyword according to a preset strategy according to the importance degree of the target keyword in the diagnosis and treatment information;
determining semantic roles of each named entity in the diagnosis and treatment information based on the weight of the target keyword and the classification identification of the named entity so as to determine the association relationship between the named entities;
sequentially connecting entity nodes with association relations to construct a semantic association path;
And extracting diagnosis and treatment nodes in the semantic association path, and determining target nodes related to the diagnosis and treatment information.
For example, phrases in text, such as a guest phrase, a bias phrase, etc., that typically contain keywords may be identified and extracted by a common algorithmic model. For example, the phrase "use aspirin" and the bias phrase "treat hypertension" in the extract of "treat hypertension with aspirin". The weight of the keywords may be calculated based on their importance in the text, e.g. using TF-IDF or a word embedding method based on deep learning. Weights for "aspirin", "treatment", "hypertension" are calculated, for example using TF-IDF, to determine their importance in sentences.
Illustratively, the extracted keywords are ranked or filtered to determine the most relevant keywords that will be used for subsequent diagnosis and treatment path matching. For example, according to the weight ranking, it is determined that "aspirin" and "hypertension" are the most important keywords for subsequent diagnosis and treatment path matching. Semantic roles of entities in sentences, such as agent, events, objects, etc., can be determined by semantic analysis algorithms, which facilitate understanding of relationships between entities. For example, it is determined that "aspirin" plays a role of "therapeutic means" in sentences, and "hypertension" plays a role of "disease".
By understanding the relation among the entities, the meaning of the diagnosis and treatment information can be more accurately understood, and a basis is provided for the matching of the subsequent medical ICD codes.
Illustratively, the association between entities is analyzed, which may involve direct association (e.g., a drug for treating a disease) or indirect association (e.g., association between symptoms and disease) between entities. For example, analyzing the relationship between "aspirin" and "hypertension" can determine that "aspirin" is a drug for treating "hypertension". Each entity may be given a semantic role-related weight based on the keyword weights extracted previously and the classification of named entities (e.g., disease name, symptom name, drug name, etc.). For example, in the diagnosis and treatment information "treating hypertension with aspirin," aspirin is a drug name, and "hypertension" is a disease name, and the weight is low.
Illustratively, each named entity is assigned a semantic role, such as "treatment," "disease," and the like. For example, "aspirin" may be assigned the role of "therapeutic" and "hypertension" may be assigned the role of "disease". And sequentially connecting entity nodes with direct or indirect association to form a diagnosis and treatment path with semantic association. For example, a myocardial patient diagnosis and treatment path including nodes of "cardiography" and "quick-acting heart-rescue bolus" means that during the treatment of myocardial patients, the "quick-acting heart-rescue bolus" is used for the early treatment and then the "cardiography" is used for the subsequent treatment. And finally, according to the entity relation analysis, determining the target node related to each diagnosis and treatment information. For example, in the diagnosis and treatment path of myocardial infarction patients, the target node may be a surgical treatment node of "cardiography" and a drug treatment node of "aspirin".
Optionally, the step of retrieving the knowledge-graph of the ICD code, and matching the target node to the plurality of diagnosis and treat nodes to determine at least one diagnosis and treat path matched by the target node includes:
Performing query operation from the ICD encoded knowledge graph to search diagnosis and treatment nodes in the ICD encoded knowledge graph, so that the target nodes are matched with the diagnosis and treatment nodes one by one;
when the target node is successfully matched, searching along a relation chain between named entities based on the knowledge graph of the ICD code to call a diagnosis and treatment path connected with the target node;
And calculating the suitability of each called diagnosis and treatment path, and screening at least one diagnosis and treatment path with the suitability value exceeding a preset threshold, wherein the suitability is used for representing the matching degree of the diagnosis and treatment path and the target node.
Illustratively, the relevant information may be retrieved from a stored medical ICD encoding knowledge-graph. This typically involves executing a query in a profile database. For example, the profile is queried to find diagnostic pathways associated with "hypertension" and "aspirin". Next, the extracted target nodes (e.g., "hypertension" and "aspirin") are matched to the nodes in the knowledge-graph. The matching may be based on the name, type, or other attribute of the entity. For example, matching "hypertension" nodes to "disease" categories in the profile, matching "aspirin" nodes to "drug" categories.
Once the matching nodes are found, the system searches the knowledge graph for a diagnosis and treatment path connecting the nodes. This may include searching along a chain of relationships between entities, such as from "disease" nodes to "treatment" nodes. For example, a diagnosis and treatment path from a "hypertension" node to a "medication" node is searched. After finding the diagnosis and treat path, the suitability of the found diagnosis and treat path can be evaluated, which may include checking the logical continuity of the path, the relevance of nodes on the path, and the like. For example, it is assessed whether the route of diagnosis from "hypertension" to "aspirin" is reasonable, i.e., whether "aspirin" is indeed an effective method of treating "hypertension".
And selecting the most suitable path according to the evaluation result. If there are multiple paths, the one that best matches the medical information is selected. The diagnosis and treatment path corresponding to the highest proper value exceeding the preset threshold can be selected through the calculation of the proper value, so that the system can search and select the diagnosis and treatment path most relevant to the diagnosis and treatment information to be matched from the medical ICD coding knowledge graph.
Optionally, the step of performing a query operation from the knowledge graph of the ICD code to retrieve diagnosis and treat nodes in the knowledge graph of the ICD code, and performing matching calculation on the target nodes and the diagnosis and treat nodes one by one includes:
And comparing entity names of the target node and the diagnosis and treatment node to perform name-based matching calculation.
Illustratively, in a name-based matching approach, if the target node is "hypertension" and there is a corresponding node in the knowledge-graph, the matching is performed by comparing the names. The final matching result is to determine that the "hypertension" node in the knowledge-graph is matched with the extracted target node.
Optionally, the step of performing a query operation from the knowledge graph of the ICD code to retrieve diagnosis and treat nodes in the knowledge graph of the ICD code, and performing matching calculation on the target nodes and the diagnosis and treat nodes one by one includes:
and calling the category of the target node, and comparing the category of the target node with the category of the diagnosis and treatment node based on the category of the target node so as to perform category-based matching calculation.
Illustratively, if the target node is "aspirin" and it is known to be a "drug", then all nodes of the "drug" type are found in the knowledge-graph. Eventually, all nodes of the "drug" type are found, and then it is determined which node is "aspirin".
Optionally, the step of performing a query operation from the knowledge graph of the ICD code to retrieve diagnosis and treat nodes in the knowledge graph of the ICD code, and performing matching calculation on the target nodes and the diagnosis and treat nodes one by one includes:
and searching the attribute of the target node, and comparing the attribute of the target node with the attribute of the diagnosis and treatment node based on the attribute of the target node so as to perform attribute-based matching calculation.
Illustratively, if the target node has specific attributes, such as the "chemical component" of "aspirin" or "manufacturer," the nodes having these attributes are found in the knowledge-graph. Eventually, nodes with matching properties are found, such as drug nodes containing "aspirin" chemicals.
Optionally, the step of performing a query operation from the knowledge graph of the ICD code to retrieve diagnosis and treat nodes in the knowledge graph of the ICD code, and performing matching calculation on the target nodes and the diagnosis and treat nodes one by one includes:
And analyzing the semantic relation between the target nodes, and comparing the semantic relation of the target nodes with the semantics of the diagnosis and treatment nodes based on the semantic relation of the target nodes so as to perform semantic-based matching calculation.
Illustratively, if the relationship between the target nodes is explicit, such as "aspirin" for "treating hypertension", then nodes with such semantic relationships are found in the knowledge-graph. Finally, aspirin is found as a node of the treatment method, and the association of aspirin with the hypertension node is determined. If certain rules or logic relations exist between the target node and the nodes in the knowledge graph, such as ' patients aged over 65 ' need regular physical examination ', matching can also be performed based on the rules.
Optionally, the step of calculating the suitability of each diagnosis and treatment path, and screening at least one diagnosis and treatment path with the suitability value exceeding a preset threshold value includes:
inquiring a medical database, and determining association frequencies among diagnosis and treatment nodes in the semantic association path;
The method comprises the steps of calling case data in historical diagnosis and treatment data, and analyzing diagnosis and treatment effective rate of the semantic association path in the historical diagnosis and treatment path;
and calculating the logic consistency of the current semantic association path and the called diagnosis and treatment path according to the association frequency and the diagnosis and treatment effective rate.
Illustratively, evaluating the suitability of the found path may include checking the logical continuity and node relevance of the path. It is checked whether the relationship between each node in the path is in accordance with medical wisdom and logic, and it is further determined whether each node in the path is related to the content in the medical information. Illustratively, a medical knowledge base or database, such as a medical document, clinical guideline, drug specification, etc., is queried to obtain authoritative medical information about each node in the path and the relationships between them. From the queried plurality of authoritative medical information, the frequency of occurrence of a certain treatment method (such as aspirin) in the treatment of a specific disease (such as hypertension) can be calculated, and the association frequency between diagnosis nodes can be determined. Further analyzing the therapeutic effect of a certain therapeutic method on a specific disease, indexes such as therapeutic response rate, survival rate and the like can be used. Further analysis of the correlation between disease and treatment method, the following expression may be used, using statistical methods such as correlation coefficients:
F =
wherein F is a first correlation coefficient, A is a disease index, B is an average disease index, C is a treatment method index, and D is an average treatment method index.
Optionally, the step of calculating the suitability of each diagnosis and treatment path, and screening at least one diagnosis and treatment path with the suitability value exceeding a preset threshold further includes:
extracting the disease, symptoms and treatment modes related to the diagnosis and treatment information by calling semantic analysis contents of the diagnosis and treatment information;
And matching diagnosis and treatment nodes in the semantic association path with the diagnosis and treatment information based on the extracted diseases, symptoms and treatment modes, and calculating the relativity of the diagnosis and treatment nodes in the semantic association path and the content of the diagnosis and treatment information.
Illustratively, the content of the diagnosis and treatment information is deeply analyzed, and specific diseases, symptoms, treatment methods and the like related to the diagnosis and treatment information are determined. For example, if the diagnosis and treatment information is "treating hypertension with aspirin", it is required to specify that "hypertension" is a disease and "aspirin" is a treatment method. And then matching the extracted target node with the content in the diagnosis and treatment information. For example, the "hypertension" is matched to the "disease" node, and the "aspirin" is matched to the "drug" node. It is also possible to compare whether the node attributes in the path coincide with the contents in the diagnosis and treat information, and analyze whether the relationship between the nodes coincides with the description in the diagnosis and treat information. In analyzing the correlation between the node attribute and the diagnosis and treatment information, the following expression may be used, and statistical methods such as a correlation coefficient may be used:
E =
wherein E is the second correlation coefficient, H is the node attribute, I is the average node attribute, J is the field content, and K is the average field content.
Optionally, the step of calculating the suitability of each diagnosis and treatment path, and screening at least one diagnosis and treatment path with the suitability value exceeding a preset threshold further includes:
Determining evaluation indexes related to the diagnosis and treatment path, wherein the evaluation indexes comprise the relevance, the logic coherence degree, the treatment effect and the safety, and determining a corresponding weight coefficient for each evaluation index;
Performing score distribution on the evaluation index based on the diagnosis and treatment information so as to calculate a weighted total score value of the diagnosis and treatment path according to the weight coefficient of the evaluation index;
And normalizing the weighted total score value of the diagnosis and treatment path to determine the suitability of the diagnosis and treatment path.
For example, a weight may be assigned according to the importance of each index. For example, the logical continuity weight is 0.3, the node relevance weight is 0.2, the treatment effect weight is 0.2, the safety weight is 0.15, and the cost effectiveness weight is 0.15. The score for each indicator may be multiplied by its weight and then added to give a total score according to the following expression:
total score = (logical degree of continuity average x logical degree of continuity weight) + (node degree of correlation average x node degree of correlation weight) + (treatment effect average x treatment effect weight) + (safety average x safety weight) + (cost benefit average x cost benefit weight).
For example, if logical continuity values and node relevance values have been calculated, a suitability value for the medical path may be calculated from these values. In determining the weights of the logical coherency and node coherency in the overall suitability assessment, for example, assume a logical coherency weight of 0.6 and a node coherency weight of 0.4. If the logical continuity score is 0.8, the node relevance score is 0.9. Multiplying the score by a corresponding weight, for example:
Logical continuity score = 0.8 x 0.6 = 0.48
Node relevance score = 0.9 x 0.4 = 0.36
Total score=0.48+0.36=0.84
Illustratively, the total score may be normalized to between 0 and 1 to facilitate comparison. For example, dividing the total score by the maximum possible total score (e.g., 1) results in a normalized suitability score. And finally, judging the suitability of the diagnosis and treatment path according to the suitability score. For example, a score of near 1 indicates that the path is very suitable, while a score of near 0 indicates that the path may not be very suitable.
By this method, a comprehensive fitness value can be calculated based on the logical coherence and node correlation to help evaluate and select the best diagnosis and treatment path.
Second embodiment
The application also provides a data processing system comprising a processor and a memory storing a computer program, which when run by the processor implements the method as described above.
Illustratively, the doctor gives the patient proper treatment according to his own diagnosis and treatment procedure during the treatment. And synchronously recording and collecting each diagnosis and treatment node of the diagnosis and treatment flow of the doctor in the diagnosis and treatment process. The patient with each disease entering the doctor diagnosis and treatment process determines the diagnosis and treatment process according to the fixed diagnosis and treatment path. One disease may have multiple treatment modes, one diagnosis and treatment process is drug treatment and one treatment process is operation treatment. In the data processing system of this embodiment, diagnosis and treatment knowledge fixed by a doctor is collected and sorted, and a knowledge graph including correspondence between treatment schemes and coding rules is established. The knowledge graph spectrum can be used as the data processing system to find the ICD encoding infrastructure.
Illustratively, after recording the fixed knowledge of the physician to the knowledge graph, a first diagnosis and treatment path corresponding to the medication may be selected according to the case of the medication, thereby corresponding to the first ICD code. According to the case of the surgical treatment, a second diagnosis and treatment path corresponding to the surgical treatment can be selected, so that the second ICD code is corresponding. According to the data processing system, the corresponding relation between the diagnosis and treatment path and the ICD code is established in the knowledge graph. After the patient enters the diagnosis and treatment process, the doctor diagnoses the patient and gives out each diagnosis and treatment node of the diagnosis and treatment path.
The data processing system of the embodiment can identify the case contents of doctors through the large model, automatically identify and match the corresponding ICD codes according to the case contents, and finally record the ICD codes in the case files. And the ICD codes corresponding to the diagnosis and treatment path can be identified by identifying and selecting the path on a computer according to the diagnosis and treatment path nodes, and finally recorded into the case file.
Illustratively, the physician may not determine the ICD code for each of the diagnostic pathways, but the physician is clearly aware of the diagnostic pathways. The user can select nodes of the diagnosis and treatment path from any step when searching the ICD code. Can go upwards and can go downwards. For example, a CT detection is performed today, and a trace-up search can be performed from a node of the CT detection, or a comparison search can be performed from a node of the CT detection to a subsequent treatment node.
For example, for patients with myocardial infarction, the heart stent may be relieved after diagnosis. At this time, the doctor uses the data processing system to query the knowledge graph as a user to determine the corresponding ICD code. The user can search for the stent in the system, for example, the three paths of the system all comprise the stent, different numbers of stents are placed for different patients, and different other diagnosis and treatment measures are needed to be carried out on the nodes before and after the diagnosis and treatment node where the stent is placed. Then doctor makes cardiography examination according to the diagnosis and treatment process path of own patient before and after making the support, for example, before placing two supports, and the final result searches an ICD code, for example, X+Y+Z, according to the path relation of the two diagnosis and treatment nodes. The ICD code X+Y+Z may be registered at this point in the patient's electronic case.
For example, if two stents are input for cardiac imaging and placement during the query process, and may also coincide with other diagnosis and treatment paths, the system may give multiple diagnosis and treatment paths and corresponding ICD codes, and the doctor may continue to match to the upstream or downstream nodes to find an ICD code that can be completely matched for registration.
Third embodiment
The application also provides a readable storage medium storing a computer program which, when executed by a processor, implements the steps of the method described above.
The medical records diagnosis and operation ICD coding method based on the knowledge graph provided by the application is used for extracting keywords corresponding to diagnosis and treatment information by carrying out semantic analysis on the diagnosis and treatment information, determining target nodes related to the diagnosis and treatment information based on the keywords, calling a pre-established knowledge graph of ICD codes, matching the target nodes to the plurality of diagnosis and treatment nodes to determine at least one diagnosis and treatment path matched with the target nodes, and outputting the at least one diagnosis and treatment path and the ICD codes corresponding to the at least one diagnosis and treatment path. The ICD coding method can accurately and rapidly match the diagnosis and treatment process of each patient to the accurate ICD coding, greatly reduces the coding management cost and improves the user experience.
In the present application, step numbers such as S10 and S20 are used for the purpose of more clearly and briefly describing the corresponding contents, and are not to constitute a substantial limitation on the sequence, and those skilled in the art may execute S20 first and then S10 in the specific implementation, which are all within the scope of the present application.
The embodiments of the system and the storage medium provided by the application may include all technical features of any one of the embodiments of the method, and the expansion and explanation contents of the description are basically the same as those of each embodiment of the method, and are not repeated herein.
Embodiments of the present application also provide a computer program product comprising computer program code which, when run on a computer, causes the computer to perform the method as in the various possible embodiments described above.
The embodiment of the application also provides a chip, which comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor is used for calling and running the computer program from the memory, so that the device provided with the chip executes the method in the various possible implementation manners.
It can be understood that the above scenario is merely an example, and does not constitute a limitation on the application scenario of the technical solution provided by the embodiment of the present application, and the technical solution of the present application may also be applied to other scenarios. For example, as one of ordinary skill in the art can know, with the evolution of the system architecture and the appearance of new service scenarios, the technical solution provided by the embodiment of the present application is also applicable to similar technical problems.
The foregoing embodiment numbers of the present application are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
The steps in the method of the embodiment of the application can be sequentially adjusted, combined and deleted according to actual needs.
The units in the device of the embodiment of the application can be combined, divided and deleted according to actual needs.
In the present application, the same or similar term concept, technical solution and/or application scenario description will be generally described in detail only when first appearing and then repeatedly appearing, and for brevity, the description will not be repeated generally, and in understanding the present application technical solution and the like, reference may be made to the previous related detailed description thereof for the same or similar term concept, technical solution and/or application scenario description and the like which are not described in detail later.
In the present application, the descriptions of the embodiments are emphasized, and the details or descriptions of the other embodiments may be referred to.
The technical features of the technical scheme of the application can be arbitrarily combined, and all possible combinations of the technical features in the above embodiment are not described for the sake of brevity, however, as long as there is no contradiction between the combinations of the technical features, the application shall be considered as the scope of the description of the application.
The foregoing description is only of the preferred embodiments of the present application, and is not intended to limit the scope of the application, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.