Electronic medical record labeling and analyzing method based on Bert platform

Abstract

The invention provides an electronic medical record labeling and analyzing method based on a Bert platform, and the method comprises the following steps: obtaining an electronic medical record text, and developing an interactive labeling tool based on the Bert platform; labeling the electronic medical record text by using an interactive labeling tool, and extracting a relationship between entities in the electronic medical record; performing deep learning on the electronic medical record text by using the BERT model, automatically extracting key information, and completing training of the BERT model; constructing a medical record knowledge graph by combining the entities in the electronic medical record text, the relationship between the entities and the key information; a graph database query language is introduced into the medical record knowledge graph, an electronic medical record text is analyzed, and then the trained BERT model is used for labeling and analyzing newly-added data of the electronic medical record text. According to the method, richer and more accurate medical information can be extracted, manual annotation intervention is reduced, and the efficiency and reliability of electronic medical record annotation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to an electronic medical record annotation and parsing method based on a Bert platform. Background Art

[0002] Electronic Medical Record (EMR) is a digital record of patient health information, usually created and maintained by medical institutions. It contains information such as the patient's medical history, diagnosis, treatment plan, drug allergies, test results, imaging data, etc. Electronic medical records are an important part of medical informatization. Labeling and parsing electronic medical records are of great significance to the quality and efficiency of medical services. At present, the existing electronic medical record (EMR) labeling methods mainly rely on manual operations, which are inefficient and error-prone. In addition, traditional machine learning models face the challenge of feature engineering when processing complex medical text data, and it is difficult to parse deep semantic information. Summary of the invention

[0003] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide an electronic medical record annotation and parsing method based on the Bert platform, which can extract richer and more accurate medical information, reduce manual annotation intervention, and improve the efficiency and reliability of electronic medical record annotation.

[0004] To achieve the above object, the present invention provides the following solution: an electronic medical record annotation and parsing method based on the Bert platform, comprising the following steps:

[0005] Obtain electronic medical record texts and develop an interactive annotation tool based on the Bert platform;

[0006] Annotating the electronic medical record text using the interactive annotation tool to extract entities in the electronic medical record text and define relationships between entities;

[0007] Selecting a BERT model, and using the BERT model to perform deep learning on the electronic medical record text, automatically extracting key information from the electronic medical record text, and completing the training of the BERT model;

[0008] Combining entities, relationships between entities and key information in the electronic medical record text, constructing a medical record knowledge graph;

[0009] A graph database query language is introduced into the medical record knowledge graph to parse the diagnosis and treatment data in the electronic medical record text, and then the trained BERT model is used to annotate and parse the newly added data in the electronic medical record text.

[0010] Optionally, obtain the electronic medical record text and develop an interactive annotation tool based on the Bert platform, including: analyzing the objectives and scope of the electronic medical record text and the work content of relevant medical personnel, based on the analysis results, using the Bert platform for front-end customization, and developing an interactive annotation tool for entities and entity relationships based on a Web UI.

[0011] Optionally, a BERT model is selected, and the BERT model is used to perform deep learning on the electronic medical record text, key information in the electronic medical record text is automatically extracted, and training of the BERT model is completed, including:

[0012] Using the selected BERT model to classify each word in the electronic medical record text to obtain multiple entity types;

[0013] The binary cross entropy loss function is used to calculate the loss value of each entity type to obtain a first loss value, and the classification corresponding to the lowest first loss value is selected as the entity type;

[0014] Pairing entities in pairs to obtain multiple comprehensive relationships, calculating the loss value of the comprehensive relationship using a binary cross entropy loss function to obtain a second loss value, and selecting the comprehensive relationship corresponding to the lowest second loss value as the relationship type;

[0015] The entity location and type are predicted according to the entity type and the relationship type, key information in the electronic medical record text is automatically extracted, and the training of the BERT model is completed.

[0016] Optionally, combining entities, relationships between entities and key information in the electronic medical record text, a medical record knowledge graph is constructed, including:

[0017] Using the trained BERT model to identify the electronic medical record text, extract entities and entity relationships, and obtain multiple triples; the triples include source entities, target entities, and entity relationships;

[0018] The triples are inserted into a graph database, the electronic medical record text is converted into a structured medical representation, and a medical record knowledge graph is constructed.

[0019] The present invention discloses the following technical effects by providing an electronic medical record annotation and parsing method based on the Bert platform:

[0020] 1. This paper significantly improves the efficiency of EMR annotation and the accuracy of intelligent analysis by combining the interactive annotation of the Brat platform and the deep learning of the BERT model.

[0021] 2. The constructed knowledge graph helps to achieve deep mining and multi-dimensional analysis of EMR data, providing richer information support for clinical decision-making.

[0022] 3. The application of intelligent analytical methods (such as graph database query language) makes the information extracted from EMR more accurate and comprehensive, which helps to improve the quality and efficiency of medical services.

[0023] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0025] Figure 1 A schematic diagram of a method flow chart provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] like Figure 1 As shown, the present invention provides an electronic medical record annotation and parsing method based on the Bert platform, comprising the following steps:

[0029] 1. Obtain electronic medical record text and develop an interactive annotation tool based on the BERT platform.

[0030] Including: analyzing the objectives and scope of the electronic medical record text and the work content of relevant medical personnel, based on the analysis results, using the Bert platform for front-end customization, and developing an interactive annotation tool for entities and entity relationships based on a Web UI.

[0031] Based on the Web UI, select the text to be annotated, mark the corresponding entity, select two entities, and then mark their relationship. Compared with ordinary DOM-based markup technology, brat uses SVG technology, which has the advantages of better visualization and higher performance.

[0032] 2. Use the interactive annotation tool to annotate the electronic medical record text to extract entities in the electronic medical record text and define the relationships between entities. The specific entities and relationships are shown in Table 1 below:

[0033] Table 1 Entity types and relationships

[0034]

[0035]

[0036] 3. Select a BERT model, and use the BERT model to perform deep learning on the electronic medical record text, automatically extract key information from the electronic medical record text, and complete the training of the BERT model. Including:

[0037] The selected BERT model is used to classify each word in the electronic medical record text to obtain multiple entity types.

[0038] The binary cross entropy loss function is used to calculate the loss value of each entity type, and the first loss value is obtained. The classification corresponding to the lowest first loss value is selected as the entity type. The loss function formula is:

[0039]

[0040] Where N is the number of samples, y i is the actual label of the i-th sample, which takes a value of 0 or 1. is the predicted probability of the i-th sample, usually output by a sigmoid function to ensure that its range is between 0 and 1.

[0041] The entities are paired in pairs to obtain multiple comprehensive relationships, and the loss value of the comprehensive relationship is calculated using the binary cross entropy loss function to obtain a second loss value, and the comprehensive relationship corresponding to the lowest second loss value is selected as the relationship type.

[0042] The entity location and type are predicted according to the entity type and the relationship type, key information in the electronic medical record text is automatically extracted, and the training of the BERT model is completed.

[0043] 4. Combine the entities, relationships between entities and key information in the electronic medical record text to build a medical record knowledge graph. Including:

[0044] The trained BERT model is used to recognize the electronic medical record text, extract entities and entity relationships, and obtain multiple triples; the triples include source entities, target entities, and entity relationships.

[0045] Insert the triples into a graph database (such as neo4j), convert the electronic medical record text into a structured medical representation, and construct a medical record knowledge graph.

[0046] 5. Introduce a graph database query language into the medical record knowledge graph to parse the diagnosis and treatment data in the electronic medical record text, and then use the trained BERT model to annotate and parse the newly added data in the electronic medical record text.

[0047] Therefore, the present invention provides an electronic medical record annotation and parsing method based on the Bert platform, which can extract richer and more accurate medical information, reduce manual annotation intervention, and improve the efficiency and reliability of electronic medical record annotation.

[0048] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0049] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for marking and parsing electronic medical records based on the Bert platform, characterized in that: The following steps are involved: Obtain electronic medical record texts and develop an interactive annotation tool based on the Bert platform; Annotating the electronic medical record text using the interactive annotation tool to extract entities in the electronic medical record text and define relationships between entities; Selecting a BERT model, and using the BERT model to perform deep learning on the electronic medical record text, automatically extracting key information from the electronic medical record text, and completing the training of the BERT model; Combining entities, relationships between entities and key information in the electronic medical record text, constructing a medical record knowledge graph; A graph database query language is introduced into the medical record knowledge graph to parse the diagnosis and treatment data in the electronic medical record text, and then the trained BERT model is used to annotate and parse the newly added data in the electronic medical record text.

2. According to claim 1, a method for electronic medical record annotation and analysis based on the Bert platform is characterized in that: Obtain electronic medical record text and develop an interactive annotation tool based on the Bert platform, including: analyzing the objectives and scope of the electronic medical record text and the work content of relevant medical personnel, based on the analysis results, using the Bert platform for front-end customization, and developing an interactive annotation tool for entities and entity relationships based on the Web UI.

3. The electronic medical record annotation and parsing method based on the Bert platform according to claim 2 is characterized in that: Selecting a BERT model, and using the BERT model to perform deep learning on the electronic medical record text, automatically extracting key information from the electronic medical record text, and completing the training of the BERT model, including: Using the selected BERT model to classify each word in the electronic medical record text to obtain multiple entity types; The binary cross entropy loss function is used to calculate the loss value of each entity type to obtain a first loss value, and the classification corresponding to the lowest first loss value is selected as the entity type; Pairing entities in pairs to obtain multiple comprehensive relationships, calculating the loss value of the comprehensive relationship using a binary cross entropy loss function to obtain a second loss value, and selecting the comprehensive relationship corresponding to the lowest second loss value as the relationship type; The entity location and type are predicted according to the entity type and the relationship type, key information in the electronic medical record text is automatically extracted, and the training of the BERT model is completed.

4. The method for electronic medical record annotation and analysis based on the Bert platform according to claim 3, characterized in that: Combining the entities, relationships between entities and key information in the electronic medical record text, a medical record knowledge graph is constructed, including: Using the trained BERT model to identify the electronic medical record text, extract entities and entity relationships, and obtain multiple triples; the triples include source entities, target entities, and entity relationships; The triples are inserted into a graph database, the electronic medical record text is converted into a structured medical representation, and a medical record knowledge graph is constructed.

Images