Construction method and device of diagnosis and treatment event knowledge graph and electronic equipment thereof

Abstract

The invention discloses a diagnosis and treatment event knowledge graph construction method and device and electronic equipment thereof, and relates to the field of artificial intelligence, the field of medical text understanding or other related fields, and the method comprises the steps: obtaining a to-be-processed diagnosis and treatment text, employing an event ontology model to extract an event type in the diagnosis and treatment text, extracting event key elements of diagnosis and treatment events included in the event types by adopting an element extraction model to obtain a diagnosis and treatment event type set and a diagnosis and treatment event element set, generating a diagnosis and treatment event sequence based on the diagnosis and treatment event type set and the diagnosis and treatment event element set, and fusing the diagnosis and treatment event sequence to obtain a diagnosis and treatment event sequence; and obtaining a diagnosis and treatment event atlas associated with different event types. According to the method and the device, the technical problems that a lot of human resources are needed and the accuracy is poor when the knowledge graph is constructed in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, the field of medical text understanding or other related fields. Specifically, it relates to a method and device for constructing a knowledge graph of diagnosis and treatment events and an electronic device thereof. Background Art

[0002] In recent years, with the wide application of big data in basic medicine, clinical medicine and public health fields, how to obtain knowledge from the big data generated in the clinical diagnosis and treatment process and construct a knowledge graph has become one of the research hotspots and cores in the field of natural language understanding. The continuous development of medical technology has made the methods of medical diagnosis for patients by medical institutions more efficient and rich, and the obtained patient condition information is also more accurate, complete and huge. At present, the vast majority of knowledge graphs study entity-oriented knowledge graphs, but the entity information is one-sided without specific context, lacking deep semantic information. As an even higher-level semantic unit, an event expresses the objective facts of the interaction of specific people, things and events at a specific time and specific place. Therefore, in order to represent various factual information more clearly and accurately, based on the characteristics of such semantic units as events, constructing an event knowledge graph has become a new research direction in the current knowledge graph field. Combining medicine with artificial intelligence technology to obtain events from the whole process of patient treatment and construct relevant event knowledge graphs to facilitate doctors' daily work and scientific research work has become an irresistible trend. At present, the related research on event knowledge graphs mainly focuses on the automatic acquisition of event knowledge, including event extraction and event relationship extraction, which are also the key technologies for constructing event knowledge graphs.

[0003] In the related art, a large amount of human resources are required in the process of constructing the indication pictures of medical texts, and the accuracy is relatively low.

[0004] To solve the above problems, no effective solution has been proposed yet. Summary of the Invention

[0005] Embodiments of the present invention provide a method and device for constructing a knowledge graph of diagnosis and treatment events and an electronic device thereof, so as to at least solve the technical problems in the related art that a large amount of human resources are required and the accuracy is poor when constructing a knowledge graph.

[0006] According to one aspect of the embodiments of the present invention, a method for constructing a medical treatment event knowledge graph is provided, including: obtaining medical treatment texts to be processed; extracting event types in the medical treatment texts by using an event ontology model, and extracting event key elements of medical treatment events included under the event types by using an element extraction model to obtain a medical treatment event type set and a medical treatment event element set; generating a medical treatment event sequence based on the medical treatment event type set and the medical treatment event element set; and fusing the medical treatment event sequence to obtain a medical treatment event graph associating different event types.

[0007] Optionally, after obtaining the medical treatment texts to be processed, it further includes: parsing the medical treatment texts; extracting punctuation marks, characters, and stop words in the medical treatment texts based on a preset standard term library; and performing text cleaning on the medical treatment texts to complete preprocessing of the medical treatment texts, where the text cleaning includes: deleting punctuation marks, deleting characters, and deleting stop words.

[0008] Optionally, after completing the preprocessing of the medical treatment texts, performing structured processing on the preprocessed medical treatment texts, including: extracting medical treatment original words in the medical treatment texts based on the preset standard term library; analyzing the similarity between the medical treatment original words in the medical treatment texts and each standard diagnosis word in a preset medical classification library; taking all the medical treatment original words with a similarity greater than a preset similarity threshold as synonyms of the corresponding standard diagnosis words; and mapping all the medical treatment original words defined as synonyms to the corresponding standard diagnosis words to complete the structured processing of the medical treatment texts.

[0009] Optionally, the event ontology model is pre-constructed. When constructing the event ontology model, it includes: constructing an entity model based on pre-classified entity objects, where the types of the entity objects include at least one of the following: patient, disease, consulting room, drug, symptom, examination, test, gene, surgery, body part, time; and constructing the event ontology model according to the entity model and real-time patient medical treatment information, where each type of event in the event ontology model is defined by using multi-tuple elements.

[0010] Optionally, the event types include at least one of the following: diagnosis event, medical treatment event, test event, examination event, surgery event, medication event, death event, infection event, hyperplasia event, myogenesis event, and rehabilitation event, and the event key elements include at least one of the following: data source, medical treatment object, diagnosis result, medical treatment operation, and medical treatment start time and end time.

[0011] Optionally, the step of generating a sequence of medical treatment events based on the set of medical treatment event types and the set of medical treatment event elements includes: using each medical treatment object in the set of medical treatment event elements as a graph node to obtain medical treatment object nodes; extracting the medical treatment event types associated with each medical treatment object node and the medical treatment events corresponding to each medical treatment event type to obtain all the medical treatment events associated with the medical treatment object nodes; determining the association relationships of the medical treatment object nodes based on all the medical treatment events of the medical treatment object nodes; establishing different case connection nodes of the medical treatment objects based on the association relationships; using the chronological order as the sequence axis, and generating a sequence of medical treatment events associated with the medical treatment objects based on the case connection nodes, where the sequence of medical treatment events includes: treatment time, treatment object, event type, and at least one of the event key elements.

[0012] Optionally, the step of fusing the sequence of medical treatment events to obtain a medical treatment event graph associated with different event types includes: fusing the sequence of medical treatment events to obtain common features between different event types; establishing an event change relationship between medical treatment events based on the common features between different event types to obtain the medical treatment event graph.

[0013] According to another aspect of the embodiments of the present invention, there is also provided a device for constructing a medical treatment event knowledge graph, including: an acquisition unit for acquiring medical treatment texts to be processed; an extraction unit for extracting event types in the medical treatment texts using an event ontology model and extracting event key elements of medical treatment events included under the event types using an element extraction model to obtain a set of medical treatment event types and a set of medical treatment event elements; a generation unit for generating a sequence of medical treatment events based on the set of medical treatment event types and the set of medical treatment event elements; and a fusion unit for fusing the sequence of medical treatment events to obtain a medical treatment event graph associated with different event types.

[0014] Optionally, the device for constructing the medical treatment event knowledge graph further includes: a parsing sub-module for parsing the medical treatment texts after acquiring the medical treatment texts to be processed; a first extraction sub-module for extracting punctuation marks, characters, and stop words in the medical treatment texts based on a preset standard term library; and a text cleaning sub-module for performing text cleaning on the medical treatment texts to complete preprocessing of the medical treatment texts, where the text cleaning includes: deleting punctuation marks, deleting characters, and deleting stop words.

[0015] Optionally, the construction device of the diagnosis and treatment event knowledge graph further includes: a second extraction sub-module, configured to extract the original diagnosis and treatment words in the diagnosis and treatment text based on the preset standard term library after completing the preprocessing of the diagnosis and treatment text; an analysis sub-module, configured to analyze the similarity between the original diagnosis and treatment words in the diagnosis and treatment text and each standard diagnosis word in the preset medical classification library; a definition sub-module, configured to use all the original diagnosis and treatment words with a similarity greater than the preset similarity threshold as synonyms of the corresponding standard diagnosis words; a mapping sub-module, configured to map all the original diagnosis and treatment words defined as synonyms to the corresponding standard diagnosis words, and complete the structuring process of the diagnosis and treatment text.

[0016] Optionally, the construction device of the diagnosis and treatment event knowledge graph further includes: a first construction module, configured to construct an entity model based on pre-classified entity objects, where the types of the entity objects include at least one of the following: patient, disease, consulting room, drug, symptom, examination, test, gene, surgery, body part, time; a second construction module, configured to construct the event ontology model according to the entity model and the real-time patient medical treatment information, where each type of event in the event ontology model is defined by a multi-tuple element.

[0017] Optionally, the event types mentioned in the construction device of the diagnosis and treatment event knowledge graph are at least one of the following: diagnosis event, medical treatment event, test event, examination event, surgery event, medication event, death event, infection event, hyperplasia event, myogenesis event, and rehabilitation event, and the key event elements include at least one of the following: data source, diagnosis and treatment object, diagnosis result, diagnosis and treatment operation, and diagnosis and treatment start time, end time.

[0018] Optionally, the generating unit includes: a first determination module, configured to use each diagnosis and treatment object in the diagnosis and treatment event element set as a graph node to obtain a diagnosis and treatment object node; an extraction module, configured to extract the diagnosis and treatment event types associated with each diagnosis and treatment object node and the diagnosis and treatment events corresponding to each diagnosis and treatment event type to obtain all the diagnosis and treatment events associated with the diagnosis and treatment object node; a second determination module, configured to determine the association relationship of the diagnosis and treatment object node based on all the diagnosis and treatment events of the diagnosis and treatment object node; a building module, configured to build different case connection nodes of the diagnosis and treatment object based on the association relationship; a generating module, configured to use the time sequence as the sequence axis and generate a diagnosis and treatment event sequence associated with the diagnosis and treatment object based on the case connection nodes, where the diagnosis and treatment event sequence includes: diagnosis and treatment time, diagnosis and treatment object, event type, and at least one of the key event elements.

[0019] Optionally, the fusion unit includes: a fusion module configured to fuse the sequence of diagnosis and treatment events to obtain common features between different event types; and an establishment module configured to establish an event change relationship between diagnosis and treatment events based on the common features between different event types, thereby obtaining the diagnosis and treatment event graph.

[0020] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for constructing a diagnosis and treatment event knowledge graph according to any one of the above.

[0021] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, which includes one or more processors and a memory. The memory is configured to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method for constructing a diagnosis and treatment event knowledge graph according to any one of the above.

[0022] In the present disclosure, when constructing an event knowledge graph, it is possible to first obtain the diagnosis and treatment text to be processed, extract the event types in the diagnosis and treatment text by using an event ontology model, and extract the event key elements of the diagnosis and treatment events included under the event types by using an element extraction model, so as to obtain a diagnosis and treatment event type set and a diagnosis and treatment event element set. Based on the diagnosis and treatment event type set and the diagnosis and treatment event element set, a sequence of diagnosis and treatment events is generated, and the sequence of diagnosis and treatment events is fused to obtain a diagnosis and treatment event graph that associates different event types.

[0023] In the present disclosure, when constructing an event knowledge graph, by automatically extracting the event types in the diagnosis and treatment text and the event key elements of the diagnosis and treatment events, and then generating a sequence of diagnosis and treatment events based on the diagnosis and treatment event type set and the diagnosis and treatment event element set, and fusing the sequence of diagnosis and treatment events, it is possible to obtain a diagnosis and treatment event graph that associates different event types. In the present disclosure, when constructing an event knowledge graph, it is possible to accurately construct a corresponding knowledge graph for each type of medical event, construct an event knowledge graph of the patient's diagnosis and treatment process, and more effectively improve the efficiency of medical research work, thereby solving the technical problems in the related art that a large amount of human resources are required and the accuracy is poor when constructing a knowledge graph.

[0024] In the present disclosure, by automatically extracting the event types in the diagnosis and treatment text and the event key elements of the diagnosis and treatment events, and then through the characteristics of the entities and the dependency relationships between them, it is closer to the human cognitive structure, constructs an application for the clinical diagnosis and treatment process, and can more quickly understand and analyze the patient's condition. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings described herein are used to provide a further understanding of the present invention and form a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0026] Figure 1 is a flowchart of an optional method for constructing a knowledge graph of diagnosis and treatment events according to an embodiment of the present invention;

[0027] Figure 2 is a flowchart framework diagram of an optional method for mining electronic medical record data based on the real world according to an embodiment of the present invention;

[0028] Figure 3 is an entity model relationship diagram constructed based on data and the clinical experience of hospital doctors according to an embodiment of the present invention;

[0029] Figure 4 is an event ontology model diagram constructed based on the entity model and the medical treatment situation of actual patients according to an embodiment of the present invention;

[0030] Figure 5 is a schematic diagram of optional data collection and collation according to an embodiment of the present invention;

[0031] Figure 6 is a schematic diagram of optional key element extraction according to an embodiment of the present invention;

[0032] Figure 7 is a schematic diagram of an optional device for constructing a knowledge graph of diagnosis and treatment events according to an embodiment of the present invention;

[0033] Figure 8 is a hardware structure block diagram of an electronic device (or mobile device) for a method of constructing a knowledge graph of diagnosis and treatment events according to an embodiment of the present invention. Detailed implementation manners

[0034] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0035] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0036] For the convenience of those skilled in the art to understand the present invention, the following explains some terms or nouns involved in each embodiment of the present invention:

[0037] International Classification of Diseases, 10th Revision, abbreviated as ICD-10, is a pre-set medical classification library compiled and released by the World Health Organization for classifying and coding various health-related events.

[0038] Medical Subject Headings, abbreviated as MeSH, is a controlled vocabulary used for indexing and retrieving biomedical literature.

[0039] Systematized Nomenclature of Medicine-Clinical Terms, abbreviated as SNOMED CT, is a clinical terminology system and a set of internationally unified standard clinical terminology systems.

[0040] International Classification of Functioning, Disability and Health, abbreviated as ICF, provides a unified and standard classification of the states of all functions and disabilities related to human health.

[0041] It should be noted that the method and device for constructing the knowledge graph of diagnosis and treatment events in the present disclosure can be used in the field of medical language understanding when constructing a knowledge graph for diagnosis and treatment events, and can also be used in any field other than the field of medical language understanding when constructing a knowledge graph for diagnosis and treatment events. The application field of the method and device for constructing the knowledge graph of diagnosis and treatment events in the present disclosure is not limited.

[0042] It should be noted that the relevant information (including but not limited to diagnosis and treatment event information, patient personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data that have been authorized by the patient or fully authorized by various medical institutions. Moreover, the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions, and corresponding operation entrances are provided for users to choose to authorize or refuse. For example, an interface is set between this system and relevant users or institutions. Before obtaining relevant information, a request for acquisition needs to be sent to the aforementioned users or institutions through the interface, and relevant information can be obtained after receiving the consent information feedback from the aforementioned users or institutions.

[0043] In related technologies, there are mainly three technical methods for event extraction. The first is event knowledge extraction based on pattern matching, which is used to identify and extract certain types of events under the guidance of some patterns, including supervised and weakly supervised pattern matching. However, this method usually requires a large amount of human resources for knowledge extraction and has poor accuracy. Especially when migrating to new disease data, patterns may need to be re-mined. The second is the machine learning method, which models the event knowledge extraction task as a multi-classification problem and is the mainstream method for current event extraction. However, its accuracy is usually affected by feature engineering and requires a large-scale refined corpus, and is easily affected by situations such as unbalanced corpus categories and long-tail data. The third is the neural network method, which models event extraction as an end-to-end system, uses word vectors containing rich language features as input, and automatically extracts features and classifies them through a neural network for event extraction. It can avoid a lot of feature engineering and can also adapt to context information, but there is still a certain gap between its accuracy and the complexity of model training and the requirements of scientific research and clinical practice.

[0044] In related technologies, there are mainly three technical methods for event relation extraction. The first is the pipeline learning method, which conducts relation extraction based on the completion of entity extraction. Therefore, the quality of the relation extraction result is directly related to the result of entity extraction, and it ignores the connection between the relation and the entity, and makes predictions based on pairwise combinations, with low efficiency. The second is joint learning, and the means are divided into two categories. Parameter sharing means that the named entity recognition model and the relation classification model are jointly trained through a shared layer, and the joint annotation strategy means using an extended annotation strategy to complete both the entity recognition and relation extraction tasks simultaneously. The joint learning method is affected by the selection of the shared layer, and an inappropriate shared layer may lead to low accuracy in event relation extraction. The third is distant supervision learning. A certain association between entities in the text will definitely be displayed in some form. Based on distant supervision, entity pairs of sentences with relations are first extracted from the text, and then the sentences are used as training data to be put into the model for relation extraction. The accuracy of this method does not meet the requirements of clinical research.

[0045] The following embodiments of the present invention can be applied to various systems / applications / devices that need to construct a knowledge graph of medical treatment events and a knowledge graph of natural language events. To solve the drawbacks of knowledge graph construction in related technologies, the present invention can achieve more accurate, rich, and semantically deep problem-solving by means of an event graph, integrating information in multiple dimensions such as entities, time, and operations, having stronger knowledge expression ability, effectively breaking data islands, and better serving AI in real medical scenarios.

[0046] The present invention can also track and quickly capture various medical events in the real world, enrich event attributes, create associations between events, and form network connections with events as basic units, enabling people to more quickly discover key events, better understand the context and cause-and-effect relationships of many related events, and helping to make more accurate next-step clinical decisions in medical work.

[0047] The present invention will be described in detail below in conjunction with each embodiment.

[0048] Embodiment 1

[0049] According to an embodiment of the present invention, a method for constructing a knowledge graph of medical treatment events is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0050] Figure 1 is a flowchart of an optional method for constructing a knowledge graph of medical treatment events according to an embodiment of the present invention, as Figure 1 shown, the method includes the following steps:

[0051] Step S101, obtain the medical treatment text to be processed.

[0052] Step S102, extract the event types in the medical treatment text using an event ontology model, and extract the event key elements of the medical treatment events included under the event types using an element extraction model, to obtain a set of medical treatment event types and a set of medical treatment event elements.

[0053] Step S103, generate a medical treatment event sequence based on the set of medical treatment event types and the set of medical treatment event elements.

[0054] Step S104, fuse the medical treatment event sequence to obtain a medical treatment event graph that associates different event types.

[0055] Through the above steps, the medical treatment text to be processed can be obtained first. The event types in the medical treatment text are extracted using the event ontology model, and the event key elements of the medical treatment events included under the event types are extracted using the element extraction model, obtaining a set of medical treatment event types and a set of medical treatment event elements. Based on the set of medical treatment event types and the set of medical treatment event elements, a medical treatment event sequence is generated, and the medical treatment event sequence is fused to obtain a medical treatment event graph that associates different event types. In the embodiments of the present invention, when constructing the event knowledge graph, by automatically extracting the event types in the medical treatment text and the event key elements of the medical treatment events, and then based on the set of medical treatment event types and the set of medical treatment event elements, a medical treatment event sequence is generated, and when fusing the medical treatment event sequence, a medical treatment event graph that associates different event types can be obtained. In the present disclosure, when constructing the event knowledge graph, the corresponding knowledge graph can be accurately constructed for each type of medical event, and the event knowledge graph of the patient's medical treatment process can be constructed, which can more effectively improve the efficiency of medical research work, thus solving the technical problems in the related art that a large amount of human resources are required and the accuracy is not good when constructing the knowledge graph.

[0056] The embodiments of the present invention will be described in detail below in combination with the above steps.

[0057] In the embodiments of the present invention, in order to facilitate the element extraction of the collected events in real time, it is necessary to pre-construct an event ontology model. The construction of the event ontology model will be described below.

[0058] Optionally, when constructing the event ontology model, it includes: constructing an entity model based on the pre-classified entity objects, where the types of entity objects include at least one of the following: patient, disease, consulting room, drug, symptom, examination, test, gene, surgery, part, time; constructing an event ontology model based on the entity model and the real-time patient medical treatment information, where each type of event in the event ontology model is defined using multi-tuple elements.

[0059] Optionally, the event type is at least one of the following: diagnosis event, visit event, test event, examination event, surgery event, medication event, death event, infection event, hyperplasia event, myogenesis event, and rehabilitation event, and the event key elements include at least one of the following: data source, medical treatment object, diagnosis result, medical treatment operation, and medical treatment start time, end time.

[0060] It should be noted that to build an entity model, it is necessary to classify entity objects in advance. In this embodiment, there are 11 categories in total for illustration: patients, diseases, consulting rooms, drugs, symptoms, examinations, tests, genes, surgeries, body parts, and time. Among them, diseases, consulting rooms, and drugs can be mapped using various standard codes. For example, the ICD-10 medical standard code is used for mapping. Examinations and tests refer to the names of specific items or indicators done by patients, and time refers to the time when relevant events occur.

[0061] Furthermore, except for test events and examination events, the above-mentioned various types of events can be described by the following multi-tuples:

[0062] Event := <Subj, Trigger, Obj, Time>, where Subj is the event subject, Trigger is the event trigger word, Obj is the event object, and Time is the event occurrence time.

[0063] Test events and examination events have one more attribute, and the multi-tuple expression is:

[0064] Event := <Subj, Trigger, Obj, Time, Attr> where Subj is the event subject, Trigger is the event trigger word, Obj is the event object, Time is the event occurrence time, and Attr is the attribute value of this event.

[0065] Step S101, obtain the medical text to be processed.

[0066] The medical text to be processed can refer to the text transmitted by a physician or a staff member assisting the physician. The object of the medical text can be various types of texts determined by the classification of consulting rooms in each clinic or hospital. For example, a burn medical record text.

[0067] It should be noted that this embodiment can collect and classify data from different hospital information systems, form a data center, and then extract relevant patient data to complete data collection as the medical text to be processed. This medical text refers to a text similar to an electronic medical record. Extracting events and event relationships based on multi-source medical data can prepare for constructing an event knowledge graph of the entire diagnosis and treatment process of patients, expand the application scenarios of the event knowledge graph in the field of clinical medicine, and can more effectively assist clinicians in their daily work and improve the efficiency of clinicians' research work.

[0068] Optionally, after obtaining the medical text to be processed, it further includes: parsing the medical text; extracting punctuation marks, characters, and stop words from the medical text based on a preset standard term library; performing text cleaning on the medical text to complete the preprocessing of the medical text, where text cleaning includes: deleting punctuation marks, deleting characters, and deleting stop words.

[0069] For the medical treatment text to be processed input by the staff, which is the diagnostic information of an electronic medical record, except for the standard diagnostic terms corresponding to the ICD codes diagnosed by the medical record room in the front page of the medical record, the rest are free texts manually entered by clinicians. Due to personal habits, punctuation marks such as commas and full stops are often added after the diagnosis results. On the other hand, different writing habits may also bring different disease descriptions, and the same disease may have multiple synonyms. In addition, due to different writing methods, some meaningless special identification symbols are likely to appear, and these characters will have a certain impact on subsequent event mapping and event relationship sorting. Therefore, this embodiment needs to perform certain preprocessing and cleaning on these data.

[0070] Optionally, the preprocessing steps further include: 1. Simplified and traditional Chinese conversion: converting traditional Chinese to simplified Chinese or vice versa; 2. Full-width and half-width unification: unifying full-width characters (such as symbols, numbers) and half-width characters; 3. Entity normalization: normalizing various forms that entities (such as important information like names and addresses) may appear in; 4. Replacing synonyms: replacing and correcting the problem of different forms of synonyms caused by different spelling formats or noise introduction, etc., with a unified expression; 5. Other processing methods include but are not limited to removing numbers, merging the beginning and end, etc.

[0071] Optionally, after completing the preprocessing of the medical treatment text, perform structured processing on the preprocessed medical treatment text, including: extracting the original medical treatment words in the medical treatment text based on a preset standard term library; analyzing the similarity between the original medical treatment words in the medical treatment text and each standard diagnostic word in the preset medical classification library; regarding all original medical treatment words with a similarity greater than the preset similarity threshold as synonyms of the corresponding standard diagnostic words; mapping all the original medical treatment words defined as synonyms to the corresponding standard diagnostic words to complete the structured processing of the medical treatment text.

[0072] It should be noted that the preset medical classification library can refer to any one of the following: ICD-10, MeSH, SNOMED CT, ICF, etc. For example, taking ICD-10 as the preset medical classification library, known standard medical diagnostic term words can be determined through this classification library.

[0073] Structuring the medical treatment text means extracting information related to the disease process such as symptoms, diseases, and examinations in the text, so as to achieve the structuring and standardization of the medical treatment text; for the situation where there are a large number of synonyms for the same disease caused by different writing habits, in this embodiment, after word segmentation through a medical standard thesaurus, the similarity between them and the IDC-10 standard diagnostic words is judged, and the synonyms that meet the similarity are mapped to the standard diagnostic words, thereby realizing data standardization.

[0074] Optionally, the method for analyzing the similarity between the original diagnosis and treatment term and each standard diagnosis term may be as follows: Define the original diagnosis and treatment term vector corresponding to the original diagnosis and treatment term and the standard diagnosis term vector corresponding to the standard diagnosis term, calculate the cosine value between the original diagnosis and treatment term vector and each standard diagnosis term vector, and determine the similarity between the original diagnosis and treatment term and each standard diagnosis term based on the cosine value.

[0075] This embodiment does not limit the method for defining the word vector, and any of the following methods can be used: 1. Word2Vec encoding, using a neural network model to map each word to a vector of a fixed dimension, which can represent the semantics of the word; 2. One-hot encoding, representing each word as a vector with n elements, where only one element is 1 and the others are 0, and the size of n is the total number of different words in the entire corpus; 3. GloVe encoding, using matrix factorization method to map each word to a vector of a fixed dimension, which can represent the semantics of the word; 4. Hash vector model, a text vectorization representation method based on the Hash function, which converts natural language text into a vector of a fixed length with low dimensions.

[0076] It should be noted that the calculation formula for the cosine value between the original diagnosis term vector and each standard diagnosis term vector in this embodiment may be: where V 1 and V 2 are two vectors of the same dimension. Based on the cosine value to measure the similarity between V1 and V2, the larger the cosine value, the higher the similarity.

[0077] Step S102: Extract the event types in the diagnosis and treatment text using the event ontology model, and extract the event key elements of the diagnosis and treatment events included under the event types using the element extraction model, to obtain the diagnosis and treatment event type set and the diagnosis and treatment event element set.

[0078] Based on the event ontology model, parse the cleaned diagnosis and treatment data, detect the event types of the diagnosis and treatment text data, obtain the corresponding event types, and extract each key element of different types of events using the preset event element extraction model.

[0079] It should be noted that the methods of element extraction mainly include the following two: the entity-relationship joint extraction algorithm based on the multi-head selection mechanism. Suppose there are three entities A, B, and C in a sentence. After extracting entities A, B, and C, they are combined in pairs and sent to the relationship classifier. The NER loss + BCE loss is used to jointly train the model to infer the relationship between each pair of entities. In the inference stage, there is a preset threshold for each relationship. When the model calculation result is greater than this threshold, this relationship is considered valid; NER. Deep learning and transfer learning use low-dimensional, real-valued, and dense vector forms to represent words, phrases, and sentences, and then use deep networks such as RNN / CNN / attention mechanism to obtain text feature representations.

[0080] Optionally, all processes from the onset of the patient's condition to hospital diagnosis and treatment are composed of a series of events. The key element extraction process includes: detecting according to the burn event ontology model to obtain the event types corresponding to the burn diagnosis and treatment text data: diagnosis event, visit event, test event, examination event, surgery event, medication event, death event, infection event, hyperplasia event, myogenesis event, and rehabilitation event; after identifying the events, extract the key element information of the relevant events according to the determined event types. For example, the key elements of the visit event: data source, treatment object, diagnosis result, treatment operation, and treatment start time and end time.

[0081] Step S103: Generate a diagnosis and treatment event sequence based on the set of diagnosis and treatment event types and the set of diagnosis and treatment event elements.

[0082] Optionally, step S103 includes: using each treatment object in the set of treatment event elements as a graph node to obtain treatment object nodes; extracting the treatment event types associated with each treatment object node and the treatment events corresponding to each treatment event type to obtain all treatment events associated with the treatment object nodes; determining the association relationships of the treatment object nodes based on all treatment events of the treatment object nodes; establishing different case connection nodes for the treatment objects based on the association relationships; using the time order as the sequence axis, generating a treatment event sequence associated with the treatment object based on the case connection nodes, where the treatment event sequence includes: treatment time, treatment object, event type, and at least one event key element.

[0083] For example, after obtaining the diagnosis and treatment event types and treatment element information, they are sequentially converted into a structured event relationship sequence, the entity objects in each event description information are extracted, the association relationships of different diagnosis and treatment event nodes are determined, and different case connection nodes for the treatment objects are established based on the association relationships to generate the corresponding treatment event sequence (treatment time, treatment object, event type, key element 1, key element 2,...), so as to obtain what events occurred to the treatment object at a certain time point.

[0084] Step S104: Fuse the diagnosis and treatment event sequences to obtain a diagnosis and treatment event graph that associates different event types.

[0085] Optionally, after determining the event sequences corresponding to all the burn diagnosis and treatment data of the diagnosis and treatment object, the event sequences can be fused, so as to establish the change relationship between events based on the common features between different event types, and thus obtain a diagnosis and treatment event graph constructed across time periods and regions.

[0086] Optionally, step S104 includes: fusing the diagnosis and treatment event sequences to obtain the common features between different event types; based on the common features between different event types, establishing the event change relationship between diagnosis and treatment events to obtain a diagnosis and treatment event graph.

[0087] Through the above embodiments, the problem is solved with higher precision, richness and semantic depth by means of the event graph, integrating information in multiple dimensions such as entities, time, operations, etc., having stronger knowledge expression ability, effectively breaking data islands, and better serving AI in real medical scenarios.

[0088] Through the above embodiments, it is also possible to track and quickly capture events in the real world, enrich event attributes, create associations between events, and form network connections with events as basic units, enabling people to discover key events faster, better understand the context and cause-and-effect relationships of many related events, and helping to make more accurate next-step clinical decisions in medical work.

[0089] The following describes the present invention in conjunction with another specific implementation manner.

[0090] An embodiment of the present invention proposes a real-world-based electronic case data mining method for automatically extracting the time and several of its attributes that occur during a patient's visit. This embodiment is illustrated by taking burn diagnosis and treatment events as an example, and a burn event knowledge graph is constructed.

[0091] Figure 2 It is a flowchart framework of an optional real-world-based electronic case data mining method according to an embodiment of the present invention. As Figure 2 shown, the process includes the following steps:

[0092] S1: Data collection and processing, obtain the diagnosis and treatment data to be processed, and clean and structure the data based on the medical standard terms of ICD-10;

[0093] S2: Extract the text content of the burn medical record data, parse the burn diagnosis and treatment data, and extract each key element of the event in the text;

[0094] S3. Determine the association relationships of the diagnosis and treatment object nodes, and establish different case connection nodes for the diagnosis and treatment objects based on the association relationships to generate corresponding diagnosis and treatment event sequences;

[0095] S4. Fuse the diagnosis and treatment event sequences to obtain a comprehensive burn diagnosis and treatment event map.

[0096] It should be noted that before collecting and organizing the data in this embodiment, a burn event ontology model needs to be pre-constructed. The specific construction method is as follows:

[0097] First, construct an entity model, and divide the entities into 11 categories: patients, diseases, consulting rooms, drugs, symptoms, examinations, tests, genes, surgeries, body parts, and time. Among them, diseases, consulting rooms, and drugs can be mapped using the ICD-10 medical standard coding. Examinations and tests refer to the names of the indicators of the examinations and tests performed by the patients, and time refers to the time when the relevant events occur.

[0098] Furthermore, based on the entity model and the actual medical treatment situation of patients, construct an event ontology model. Generally, events can be divided into general events and special events. General events refer to some events that occur during the process of patients seeing a doctor; special events are events that can only be triggered when some special patients seek medical treatment. General events can be divided into four types: diagnosis events, consultation events, test events, and examination events. Special events are divided into: infection events, hyperplasia events, myogenesis events, and rehabilitation events.

[0099] Except for test events and examination events, the above events can be described by the following multi-tuples:

[0100] Event: = <Subj, Trigger, Obj, Time>, where Subj is the event subject, Trigger is the event trigger word, Obj is the event object, and Time is the event occurrence time.

[0101] Test events and examination events have additional attributes, and the multi-tuple representation is: Event: = <Subj, Trigger, Obj, Time, Attr>, where Subj is the event subject, Trigger is the event trigger word, Obj is the event object, Time is the event occurrence time, and Attr is the attribute value of this event.

[0102] Figure 3 It is an optional entity model relationship diagram constructed according to the data and the clinical experience of hospital doctors in the embodiment of the present invention. Its main entities include: symptoms, diseases, operations, departments (or consulting rooms), drugs; in addition, it also includes the clinical manifestations from diseases to symptoms, the indication relationships from drugs to diseases, the treatment methods from diseases to operations, the relationships from tests, patients to diagnoses, diseases, and the relationships from diseases to the affiliated departments.

[0103] Figure 4 It is an optional event ontology model diagram constructed according to the entity model and the actual patient medical treatment situation in an embodiment of the present invention.

[0104] Specific events include: general events such as examination events, diagnosis events, and test events pointed to by medical treatment events, and also include special events such as death events, infection events, hyperplasia events, myogenesis events, and rehabilitation events; in addition to these, it also includes two types of events: surgical events and medication events.

[0105] After the burn event ontology model is constructed, the above Figure 2 shown process steps are carried out, and the specific implementation manners are as follows.

[0106] S1, data collection and processing, obtain the medical treatment data to be processed, and clean and structure the data based on the medical standard terms of ICD-10.

[0107] Figure 5 It is a schematic diagram of an optional data collection and arrangement according to an embodiment of the present invention. As Figure 5 shown, for the collection of clinical big data, first judge whether there is a clinical data center in the hospital where the target data is located. If there is a clinical data center in the hospital, dock with the relevant clinical data center to extract electronic medical record data to complete the collection of electronic medical record data. If there is no clinical data center in the hospital, collect and classify the data from different hospital information systems ( Figure 5 The HIS system, LIS system, CIS system, PACS system and other information systems existing in the hospital are shown in the figure), form a clinical big data center and then obtain the electronic medical records, and then extract the relevant patient data to complete the data collection.

[0108] For the diagnostic information of the electronic medical record, except for the standard diagnostic terms corresponding to the ICD codes diagnosed by the medical record room in the medical record homepage, the rest are free texts manually entered by clinicians. Due to the personal habits of doctors, punctuation marks such as commas and full stops are often added after the diagnosis results; on the other hand, different writing habits may also bring different disease descriptions, and the same disease may have multiple synonyms; in addition, because of different writing methods, some meaningless special identification symbols are likely to be brought, and these characters will have a certain impact on subsequent event mapping and event relationship sorting. Therefore, these data need to be preprocessed and cleaned to a certain extent.

[0109] Structurize the electronic medical records, extract information related to the disease process such as symptoms, diseases, and examinations, so as to realize the structurization and standardization of large sections of text in the electronic medical records; for the situation where there are a large number of synonyms for the same disease caused by different writing habits, we perform word segmentation through a medical standard thesaurus and then judge the similarity between them and the IDC-10 standard diagnosis words, and map the synonyms that meet the similarity to the standard diagnosis words, thereby realizing data standardization.

[0110] S2. Extract the text content of the burn case data, analyze the burn diagnosis and treatment data, and extract each key element of the event in the text;

[0111] Based on the burn event ontology model, analyze the cleaned burn diagnosis and treatment data, detect the event types of the burn diagnosis and treatment text data, obtain the corresponding event types, and according to the corresponding event types, use a preset event element extraction model to extract each key element of different types of events.

[0112] Figure 6 It is an optional schematic diagram of key element extraction according to an embodiment of the present invention. As Figure 6 shown, all processes from the patient's onset of illness to hospital diagnosis and treatment are composed of a series of events. The key element extraction process is as follows: According to the detection of the burn event ontology model, obtain the event types corresponding to the burn diagnosis and treatment text data: diagnosis event, visit event, inspection event, examination event, surgery event, medication event, death event, infection event, hyperplasia event, myogenesis event, and rehabilitation event. After identifying the events, extract the key element information of the relevant events according to the determined event types. For example, the key elements of the visit event: data source, treatment object, diagnosis result, treatment operation, and treatment start time and end time.

[0113] S3. Determine the association relationship of the treatment object nodes, and establish different case connection nodes for the treatment objects according to the association relationship, and generate corresponding treatment event sequences.

[0114] After obtaining the treatment event types and treatment element information, convert them into a structured event relationship sequence in turn, extract the entity objects in each event description information, determine the association relationship of different treatment event nodes, and establish different case connection nodes for the treatment objects according to the association relationship, and generate corresponding treatment event sequences (treatment time, treatment object, event type, key element 1, key element 2...), so as to obtain what events occurred to the treatment object at a certain time point.

[0115] S4. Integrate the treatment event sequences to obtain a comprehensive burn treatment event map of burns.

[0116] After determining the event sequence corresponding to all burn diagnosis and treatment data of the diagnosis and treatment object, the event sequence can be fused to establish the change relationship between events based on the common features between different event types, so as to obtain a burn diagnosis and treatment event map constructed across time periods and regions.

[0117] Through the above implementation method, it is possible to create a knowledge map of diagnosis and treatment events related to burn patients, integrating information in multiple dimensions such as entities, time, and operations, with stronger knowledge expression ability, effectively breaking the data silos, and better serving AI in real medical scenarios.

[0118] The following describes the present invention in conjunction with another optional embodiment.

[0119] Embodiment 2

[0120] According to an embodiment of the present invention, there is provided a device for constructing a knowledge map of diagnosis and treatment events, which includes a plurality of implementation units, and each implementation unit corresponds to each implementation step in the above Embodiment 1.

[0121] Figure 7 It is a schematic diagram of an optional device for constructing a knowledge map of diagnosis and treatment events according to an embodiment of the present invention. As Figure 7 shown, the device for constructing a knowledge map of diagnosis and treatment events may include: an acquisition unit 71, an extraction unit 72, a generation unit 73, and a fusion unit 74.

[0122] Among them, the acquisition unit 71 is used to acquire the diagnosis and treatment text to be processed;

[0123] The extraction unit 72 is used to extract the event types in the diagnosis and treatment text by using an event ontology model, and extract the event key elements of the diagnosis and treatment events included under the event types by using an element extraction model, so as to obtain a set of diagnosis and treatment event types and a set of diagnosis and treatment event elements;

[0124] The generation unit 73 is used to generate a diagnosis and treatment event sequence based on the set of diagnosis and treatment event types and the set of diagnosis and treatment event elements;

[0125] The fusion unit 74 is used to fuse the diagnosis and treatment event sequence to obtain a diagnosis and treatment event map that associates different event types.

[0126] The device for constructing the diagnosis and treatment event knowledge graph can first obtain the diagnosis and treatment text to be processed through the acquisition unit 71; adopt the event ontology model through the extraction unit 72 to extract the event types in the diagnosis and treatment text, and adopt the element extraction model to extract the event key elements of the diagnosis and treatment events included under the event types, so as to obtain the diagnosis and treatment event type set and the diagnosis and treatment event element set; generate the diagnosis and treatment event sequence through the generation unit 73 based on the diagnosis and treatment event type set and the diagnosis and treatment event element set; and fuse the diagnosis and treatment event sequence through the fusion unit 74 to obtain the diagnosis and treatment event graph associating different event types.

[0127] In the embodiment of the present invention, when constructing the event knowledge graph, by automatically extracting the event types in the diagnosis and treatment text and the event key elements of the diagnosis and treatment events, and then generating the diagnosis and treatment event sequence based on the diagnosis and treatment event type set and the diagnosis and treatment event element set, and fusing the diagnosis and treatment event sequence, a diagnosis and treatment event graph associating different event types can be obtained. In the present disclosure, when constructing the event knowledge graph, the corresponding knowledge graphs can be accurately constructed for various types of medical events, and the event knowledge graph of the patient's diagnosis and treatment process can be constructed, which can more effectively improve the efficiency of medical research work, thus solving the technical problems in the related art that a large amount of human resources are required and the accuracy is not good when constructing the knowledge graph.

[0128] Optionally, the device for constructing the diagnosis and treatment event knowledge graph further includes: a parsing sub-module, configured to parse the diagnosis and treatment text after obtaining the diagnosis and treatment text to be processed; a first extraction sub-module, configured to extract punctuation marks, characters, and stop words in the diagnosis and treatment text based on a preset standard term library; and a text cleaning sub-module, configured to perform text cleaning on the diagnosis and treatment text to complete the preprocessing of the diagnosis and treatment text, where the text cleaning includes: deleting punctuation marks, deleting characters, and deleting stop words.

[0129] Optionally, the device for constructing the diagnosis and treatment event knowledge graph further includes: a second extraction sub-module, configured to extract the original diagnosis and treatment words in the diagnosis and treatment text based on a preset standard term library after completing the preprocessing of the diagnosis and treatment text; an analysis sub-module, configured to analyze the similarity between the original diagnosis and treatment words in the diagnosis and treatment text and each standard diagnosis word in a preset medical classification library; a definition sub-module, configured to use all the original diagnosis and treatment words with a similarity greater than a preset similarity threshold as synonyms of the corresponding standard diagnosis words; and a mapping sub-module, configured to map all the original diagnosis and treatment words defined as synonyms to the corresponding standard diagnosis words to complete the structured processing of the diagnosis and treatment text.

[0130] Optionally, the construction device of the diagnosis and treatment event knowledge graph further includes: a first construction module, configured to construct an entity model based on pre-classified entity objects, where the types of entity objects include at least one of the following: patient, disease, consulting room, drug, symptom, examination, test, gene, surgery, body part, time; a second construction module, configured to construct an event ontology model according to the entity model and real-time patient medical treatment information, where each type of event in the event ontology model is defined by a multi-tuple element.

[0131] Optionally, the event types mentioned in the construction device of the diagnosis and treatment event knowledge graph are at least one of the following: diagnosis event, medical treatment event, test event, examination event, surgery event, medication event, death event, infection event, hyperplasia event, myogenesis event, and rehabilitation event, and the event key elements include at least one of the following: data source, medical treatment object, diagnosis result, medical treatment operation, and medical treatment start time, end time.

[0132] Optionally, the generation unit includes: a first determination module, configured to use each medical treatment object in the medical treatment event element set as a graph node to obtain a medical treatment object node; an extraction module, configured to extract the medical treatment event types associated with each medical treatment object node and the medical treatment events corresponding to each medical treatment event type to obtain all the medical treatment events associated with the medical treatment object node; a second determination module, configured to determine the association relationship of the medical treatment object node based on all the medical treatment events of the medical treatment object node; a building module, configured to build different case connection nodes of the medical treatment object based on the association relationship; a generation module, configured to use the time sequence as the sequence axis to generate a medical treatment event sequence associated with the medical treatment object based on the case connection nodes, where the medical treatment event sequence includes: medical treatment time, medical treatment object, event type, and at least one event key element.

[0133] Optionally, the fusion unit includes: a fusion module, configured to fuse the medical treatment event sequences to obtain the common features between different event types; a building module, configured to establish an event change relationship between medical treatment events based on the common features between different event types to obtain a medical treatment event graph.

[0134] The above-mentioned construction device of the diagnosis and treatment event knowledge graph may further include a processor and a memory. The above-mentioned word segmentation unit 71, measurement unit 72, model generation unit 73, clustering and merging unit 74, etc. are all stored in the memory as program units, and the processor executes the above program units stored in the memory to implement corresponding functions.

[0135] The above-mentioned processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set. By adjusting the kernel parameters, the medical treatment text to be processed is obtained. The event ontology model is used to extract the event types in the medical treatment text, and the element extraction model is used to extract the event key elements of the medical treatment events included under the event types. Based on the medical treatment event type set and the medical treatment event element set, a medical treatment event sequence is generated, and the medical treatment event sequence is fused to obtain a medical treatment event graph that associates different event types.

[0136] The above-mentioned memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flash RAM). The memory includes at least one memory chip.

[0137] The present application also provides a computer program product, which, when executed on a data processing device, is adapted to execute a program initialized with the following method steps: obtaining the medical treatment text to be processed; using the event ontology model to extract the event types in the medical treatment text, and using the element extraction model to extract the event key elements of the medical treatment events included under the event types, to obtain a medical treatment event type set and a medical treatment event element set; generating a medical treatment event sequence based on the medical treatment event type set and the medical treatment event element set; and fusing the medical treatment event sequence to obtain a medical treatment event graph that associates different event types.

[0138] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium. The computer-readable storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the construction method of any one of the above-mentioned medical treatment event knowledge graphs.

[0139] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including one or more processors and a memory. The memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the construction method of any one of the above-mentioned medical treatment event knowledge graphs.

[0140] Figure 8 is a hardware structure block diagram of an electronic device (or mobile device) for the construction method of a medical treatment event knowledge graph according to an embodiment of the present invention. As Figure 8 shown, the electronic device may include one or more ( Figure 8802a, 802b, ……, 802n are used to illustrate) a processor 802 (the processor 802 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), and a memory 804 for storing data. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply, and / or a camera. Those of ordinary skill in the art can understand that Figure 8 The structure shown is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the electronic device may further include more or fewer components than those shown in Figure 8 or have a different configuration from that shown in Figure 8

[0141] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0142] In the above embodiments of the present invention, the descriptions of the respective embodiments have their own emphases. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0143] In several embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of units or modules can be electrical or other forms.

[0144] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0145] In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0146] ​When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs.

[0147] The foregoing are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for constructing a medical treatment event knowledge graph, characterized in that, it includes: Obtain the medical treatment text to be processed; Use an event ontology model to extract the event types in the medical treatment text, and use an element extraction model to extract the event key elements of the medical treatment events included under the event types, to obtain a set of medical treatment event types and a set of medical treatment event elements; Generate a medical treatment event sequence based on the set of medical treatment event types and the set of medical treatment event elements; Fuse the medical treatment event sequence to obtain a medical treatment event graph that associates different event types.

2. The method for constructing a medical treatment event knowledge graph according to claim 1, characterized in that, after obtaining the medical treatment text to be processed, it includes: Parse the medical treatment text; Extract punctuation marks, characters, and stop words in the medical treatment text based on a preset standard term library; Perform text cleaning on the medical treatment text to complete the preprocessing of the medical treatment text, wherein the text cleaning includes: deleting punctuation marks, deleting characters, and deleting stop words.

3. The method for constructing a medical treatment event knowledge graph according to claim 2, characterized in that, after completing the preprocessing of the medical treatment text, perform structured processing on the preprocessed medical treatment text, including: Extract the original medical treatment words in the medical treatment text based on the preset standard term library; Analyze the similarity between the original medical treatment words in the medical treatment text and each standard diagnosis word in the preset medical classification library; All the original medical treatment words with a similarity greater than the preset similarity threshold are used as synonyms of the corresponding standard diagnosis words; Map all the original medical treatment words defined as synonyms to the corresponding standard diagnosis words to complete the structured processing of the medical treatment text.

4. The method for constructing a medical treatment event knowledge graph according to claim 1, characterized in that, the event ontology model is pre-constructed, and when constructing the event ontology model, it includes: Based on the pre-classified entity objects, construct an entity model, wherein the types of the entity objects include at least one of the following: patient, disease, consulting room, drug, symptom, examination, test, gene, surgery, part, time; Construct the event ontology model based on the entity model and the real-time patient medical treatment information, wherein each type of event in the event ontology model is defined by a multi-tuple element.

5. The method for constructing a medical treatment event knowledge graph according to any one of claims 1 to 4, characterized in that, the event types are at least one of the following: diagnosis event, medical treatment event, test event, examination event, surgery event, medication event, death event, infection event, hyperplasia event, myogenesis event, and rehabilitation event, and the event key elements include at least one of the following: data source, medical treatment object, diagnosis result, medical treatment operation, and medical treatment start time, end time.

6. The method for constructing a medical treatment event knowledge graph according to claim 1, characterized in that, the step of generating a medical treatment event sequence based on the set of medical treatment event types and the set of medical treatment event elements includes: Each diagnosis and treatment object in the set of diagnosis and treatment event elements is used as a graph node to obtain diagnosis and treatment object nodes; Extract the diagnosis and treatment event types associated with each diagnosis and treatment object node and the diagnosis and treatment events corresponding to each diagnosis and treatment event type to obtain all the diagnosis and treatment events associated with the diagnosis and treatment object nodes; Based on all the diagnosis and treatment events of the diagnosis and treatment object nodes, determine the association relationships of the diagnosis and treatment object nodes; Based on the association relationships, establish different case connection nodes for the diagnosis and treatment objects; Using the time sequence as the sequence axis, generate a diagnosis and treatment event sequence associated with the diagnosis and treatment object based on the case connection nodes, where the diagnosis and treatment event sequence includes: diagnosis and treatment time, diagnosis and treatment object, event type, and at least one of the event key elements.

7. The method for constructing a diagnosis and treatment event knowledge graph according to claim 1, wherein, The step of fusing the diagnosis and treatment event sequences to obtain a diagnosis and treatment event graph associated with different event types includes: Fusing the diagnosis and treatment event sequences to obtain common features between different event types; Based on the common features between different event types, establish an event change relationship between diagnosis and treatment events to obtain the diagnosis and treatment event graph.

8. A device for constructing a diagnosis and treatment event knowledge graph, wherein, It includes: An acquisition unit for acquiring the diagnosis and treatment text to be processed; An extraction unit for extracting the event types in the diagnosis and treatment text using an event ontology model and extracting the event key elements of the diagnosis and treatment events included under the event types using an element extraction model to obtain a set of diagnosis and treatment event types and a set of diagnosis and treatment event elements; A generation unit for generating a diagnosis and treatment event sequence based on the set of diagnosis and treatment event types and the set of diagnosis and treatment event elements; A fusion unit for fusing the diagnosis and treatment event sequences to obtain a diagnosis and treatment event graph associated with different event types.

9. An electronic device, wherein, It includes: A processor; and A memory for storing executable instructions of the processor; wherein, the processor is configured to execute the method for constructing a diagnosis and treatment event knowledge graph according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium, wherein, The computer-readable storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for constructing a diagnosis and treatment event knowledge graph according to any one of claims 1 to 7.

Images