CN111370086A

CN111370086A - Electronic case detection method, electronic case detection device, computer equipment and storage medium

Info

Publication number: CN111370086A
Application number: CN202010123452.4A
Authority: CN
Inventors: 杨志专
Original assignee: Ping An International Smart City Technology Co Ltd
Current assignee: Shenzhen Ping An Smart Healthcare Technology Co ltd
Priority date: 2020-02-27
Filing date: 2020-02-27
Publication date: 2020-07-03
Also published as: WO2021169099A1

Abstract

The application relates to machine learning and provides an electronic case detection method, an electronic case detection device, computer equipment and a storage medium. The method comprises the following steps: acquiring electronic case information to be detected and target electronic case information; identifying the electronic case information to be detected and the target electronic case information entity to obtain an electronic case entity to be detected and a target electronic case entity; determining corresponding entity types according to the electronic case entity to be detected and the target electronic case entity, and determining a corresponding entity matching model according to each entity type; inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into corresponding entity matching models for matching to obtain entity similarity corresponding to each entity type; acquiring the weight of each entity type, and calculating the similarity of the electronic case according to the entity type weight and the entity similarity; and determining similar electronic cases according to the electronic case similarity. The method can improve the accuracy of obtaining similar electronic cases.

Description

Electronic case detection method, electronic case detection device, computer equipment and storage medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to an electronic case detection method, an electronic case detection apparatus, a computer device, and a storage medium.

Background

With the development of electronic case technology, finding similar electronic cases has great value in the application fields of clinical diagnosis, clinical scientific research and the like, doctors can diagnose and make prescriptions based on the similar cases, and the working time of the clinicians is greatly saved.

At present, searching for similar electronic cases usually includes vectorizing electronic cases, and performing similarity calculation according to the vectorized electronic cases to determine similar electronic cases, however, the similarity calculation is performed only through the vectorized electronic cases, and there is a problem that the accuracy of the determined similar electronic cases is low.

Disclosure of Invention

In view of the above, it is necessary to provide an electronic case detection method, an apparatus, a computer device, and a storage medium capable of improving the accuracy of similar electronic case determination in view of the above technical problems.

An electronic case detection method, the method comprising:

receiving an electronic case similarity detection instruction, and acquiring electronic case information to be detected and target electronic case information according to the electronic case similarity detection instruction;

carrying out entity recognition on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity;

determining corresponding entity types according to the electronic case entities to be detected and the target electronic case entities, and determining corresponding entity matching models according to the entity types;

inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into corresponding entity matching models for matching to obtain entity similarity corresponding to each entity type;

acquiring entity type weights, and calculating to obtain electronic case similarity of the electronic case information to be detected and the target electronic case information according to the entity type weights and the entity similarity;

and when the similarity of the electronic case exceeds a preset threshold value, determining the electronic case information to be detected as the similar electronic case of the target electronic case information.

In one embodiment, the entity recognition of the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity includes:

inputting electronic case information to be detected into an established named entity recognition model for recognition to obtain a first recognition result, inputting the electronic case information to be detected into an established medical knowledge map for matching to obtain a first matching result, and obtaining each electronic case entity to be detected according to the first recognition result and the first matching result;

and inputting the target electronic case information into the established named entity recognition model for recognition to obtain a second recognition result, inputting the target electronic case information into the established medical knowledge map for matching to obtain a second matching result, and obtaining each target electronic case entity according to the second recognition result and the second matching result.

In one embodiment, inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into corresponding entity matching models for matching to obtain entity similarity corresponding to each entity type, including:

inputting the electronic case entity to be detected and the target electronic case entity of the name entity type into the established medical name classification tree to obtain a to-be-detected entity node corresponding to the electronic case entity to be detected and a target entity node corresponding to the target electronic case entity;

calculating the path length between the entity node to be detected and the target entity node, and determining the name similarity of the electronic case entity to be detected and the target electronic case entity of the name entity type according to the path length between the entity node to be detected and the target entity node.

In one embodiment, determining the name similarity between the electronic case entity to be detected and the target electronic case entity of the name entity type according to the path length between the entity node to be detected and the target entity node includes:

acquiring a root entity node of the established medical name classification tree, acquiring a common father node of an entity node to be detected and a target entity node from the established medical name classification tree, and calculating a second path length between the root entity node and the common father node;

and acquiring the target path length, and calculating the name similarity between the electronic case entity to be detected of the name entity type and the target electronic case entity according to the path length, the second path length and the target path length.

and inputting the electronic case entity to be detected of the age entity type and the target electronic case entity into corresponding nonlinear distance calculation modules for calculation to obtain the age similarity between the electronic case entity to be detected of the age entity type and the target electronic case entity.

acquiring each electronic case entity to be detected and each target electronic case entity of the text entity type, vectorizing each electronic case entity to be detected and each target electronic case entity of the text entity type to obtain each vector to be detected and each target vector, and calculating the similarity of each vector to be detected and each target vector;

and determining the text similarity between the text to be detected composed of the electronic case entities to be detected and the target text composed of the target electronic case entities according to the similarity between the vectors to be detected and the target vectors.

In one embodiment, determining the text similarity between the text to be detected composed of the electronic case entities to be detected and the target text composed of the target electronic case entities according to the similarity between the vectors to be detected and the target vectors includes:

acquiring a text to be detected composed of each electronic case entity to be detected and a target text composed of each target electronic case entity;

calculating the number of the electronic case entities to be detected in the text to be detected and the number of the target electronic case entities in the target text;

determining each similarity to be detected corresponding to each vector to be detected according to the similarity of each vector to be detected and each target vector, and obtaining comprehensive similarity to be detected according to each similarity to be detected and the number of electronic case entities to be detected;

determining the similarity of each target corresponding to each target vector according to the similarity of each vector to be detected and each target vector, and obtaining the comprehensive similarity of the targets according to the similarity of each target and the number of target electronic case entities;

and obtaining the text similarity between the text to be detected and the target text according to the comprehensive similarity to be detected and the target comprehensive similarity.

An electronic case detection device, the device comprising:

the information acquisition module is used for receiving the electronic case similarity detection instruction and acquiring the electronic case information to be detected and the target electronic case information according to the electronic case similarity detection instruction;

the entity obtaining module is used for carrying out entity recognition on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity;

the model determining module is used for determining corresponding entity types according to the electronic case entities to be detected and the target electronic case entities and determining corresponding entity matching models according to the entity types;

the entity similarity obtaining module is used for inputting the electronic case entity to be detected and the target electronic case entity with the consistent entity types into corresponding entity matching models for matching to obtain entity similarities corresponding to the entity types;

the similarity calculation module is used for acquiring the weight of each entity type and calculating the electronic case similarity of the electronic case information to be detected and the target electronic case information according to the entity type weight and the entity similarity;

and the case determining module is used for determining the electronic case information to be detected as the similar electronic case of the target electronic case information when the similarity of the electronic cases exceeds a preset threshold value.

A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:

According to the electronic case detection method, the electronic case detection device, the computer equipment and the storage medium, each electronic case entity to be detected and each target electronic case entity are obtained through identification. And determining corresponding entity types according to the electronic case entities to be detected and the target electronic case entities, and determining corresponding entity matching models according to the entity types. And inputting the electronic case entity to be detected and the target electronic case entity with the consistent entity types into corresponding entity matching models for matching to obtain entity similarity corresponding to each entity type. And then, calculating the electronic case similarity of the electronic case information to be detected and the target electronic case information according to the entity type weight and the corresponding entity similarity, and taking the electronic case information to be detected with the electronic case similarity exceeding a preset threshold value as a similar electronic case of the target electronic case information. The similarity between entities corresponding to each entity type is calculated through different entity matching models, and the electronic case similarity is further determined according to the weight of each entity type and the entity similarity of each entity type, so that the accuracy of the obtained electronic case similarity is improved. And then, similar electronic cases are determined according to the similarity of the electronic cases, so that the accuracy of obtaining the similar electronic cases can be improved.

Drawings

FIG. 1 is a diagram illustrating an exemplary electronic medical record carrier;

FIG. 2 is a schematic flow chart diagram of an electronic case detection method in one embodiment;

FIG. 3 is a flow diagram illustrating entity identification in one embodiment;

FIG. 4 is a flow diagram illustrating the determination of name similarity in one embodiment;

FIG. 5 is a flowchart illustrating the determination of name similarity in another embodiment;

FIG. 6 is a diagram of an established medical name classification tree, under an embodiment;

FIG. 7 is a flow diagram that illustrates the determination of text similarity, according to one embodiment;

FIG. 8 is a flowchart illustrating the determination of text similarity according to another embodiment;

FIG. 9 is a block diagram of the electronic medical record monitor of one embodiment;

FIG. 10 is a diagram showing an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The electronic case detection method provided by the application can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The server 104 receives an electronic case similarity detection instruction sent by the terminal 102, and obtains electronic case information to be detected and target electronic case information according to the electronic case similarity detection instruction; the server 104 performs entity identification on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity; the server 104 determines corresponding entity types according to the electronic case entities to be detected and the target electronic case entities, and determines corresponding entity matching models according to the entity types; the server 104 inputs the electronic case entity to be detected and the target electronic case entity with the consistent entity types into corresponding entity matching models for matching, and entity similarity corresponding to each entity type is obtained; the server 104 acquires the entity type weight, and calculates the electronic case similarity of the electronic case information to be detected and the target electronic case information according to the entity type weight and the entity similarity; when the similarity of the electronic case exceeds the preset threshold, the server 104 determines that the electronic case information to be detected is a similar electronic case of the target electronic case information, and returns the similar electronic case to the terminal 102 for display. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and the server 104 may be implemented by an independent server or a server cluster formed by a plurality of servers.

In one embodiment, as shown in fig. 2, an electronic case testing method is provided, which is exemplified by the application of the method to the server in fig. 1, and comprises the following steps:

s202, receiving the electronic case similarity detection instruction, and acquiring the electronic case information to be detected and the target electronic case information according to the electronic case similarity detection instruction.

The electronic case information to be detected refers to specific information of electronic cases which have been diagnosed and prescribed, and the information can include disease names, ages, sexes, chief complaint texts, current medical history texts, personal history texts, differential diagnosis texts, prescription information and the like. The target electronic case information refers to specific information of electronic cases that have not been prescribed through diagnosis, including disease name, age, sex, chief complaint text, current medical history text, personal history text, differential diagnosis text, and the like.

Specifically, the server receives an electronic case similarity detection instruction sent by a doctor through the terminal, and acquires the electronic case information to be detected from the historical database according to the electronic case similarity detection instruction and acquires the target electronic case information uploaded by the doctor through the terminal. The target electronic case information may also be retrieved from a database.

And S204, carrying out entity recognition on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity.

The electronic case information to be detected is named entity identification to obtain the electronic case information to be detected. The named entity recognition algorithm can use CRF (conditional random field) and bilSTM (Bi-directional Long Short-Term Memory neural network) target electronic case entities, which are obtained after named entity recognition is carried out on target electronic case information.

Specifically, the server inputs the electronic case information to be detected and the target electronic case information into the established named entity recognition model for recognition, and each output electronic case entity to be detected and each target electronic case entity are obtained. The entity may include the name of the disease, age, gender, and entity words in the respective text. The established named entity recognition model is a model trained using CRF and bilSTM algorithms based on existing medically relevant data.

S206, determining corresponding entity types according to the electronic case entities to be detected and the target electronic case entities, and determining corresponding entity matching models according to the entity types.

The entity type refers to a preset type of each entity. The entity types include a disease name entity type, an age entity type, a gender entity type, and a text entity type, among others. The entity type is used to determine a corresponding entity matching model. The entity matching model is used for carrying out similarity calculation on each electronic case entity to be detected and each target electronic case entity. And each entity type is preset with a corresponding entity matching model.

Specifically, the server determines corresponding entity types according to the electronic case entities to be detected and the target electronic case entities, and determines corresponding entity matching models according to the entity types. For example, the corresponding age entity type is determined according to the age entity, and the corresponding age similarity calculation model is obtained according to the age entity type, and may be established by using a nonlinear distance algorithm.

And S208, inputting the electronic case entity to be detected and the target electronic case entity with the consistent entity types into corresponding entity matching models for matching to obtain the entity similarity corresponding to each entity type.

The entity similarity refers to the similarity between the electronic case entity to be detected and the target electronic case entity with the same entity type.

Specifically, the server inputs the electronic case entity to be detected and the target electronic case entity corresponding to each entity type into the corresponding entity matching models for similarity calculation to obtain the entity similarity corresponding to each entity type, for example, the age entity in the electronic case entity to be detected and the age entity in the target electronic case entity are input into the corresponding age similarity calculation models for similarity calculation to obtain the age entity similarity corresponding to the age entity type. And inputting the sex entity in the electronic case entity to be detected and the sex entity in the target electronic case entity into corresponding sex similarity calculation models for similarity calculation to obtain the sex entity similarity corresponding to the sex entity type, wherein the sex similarity calculation models can be established by using a distance similarity calculation method.

S210, obtaining the weight of each entity type, and calculating the electronic case similarity of the electronic case information to be detected and the target electronic case information according to the entity type weight and the entity similarity.

The entity type weight is a preset weight corresponding to each entity type and can be configured by a doctor. The electronic case similarity refers to the similarity between the electronic case information to be detected and the target electronic case information.

Specifically, the server obtains each entity type weight, and calculates the electronic case similarity of the electronic case information to be detected and the target electronic case information according to each entity type weight and the entity similarity of the corresponding entity type. For example, the electronic case similarity can be calculated using formula (1).

Wherein S (electronic case) refers to electronic case similarity, n refers to entity type total number, and w is Stelon type weight. s refers to entity similarity. w (i) s (i) refers to the product of the ith entity type weight and the corresponding ith entity similarity.

S212, when the similarity of the electronic cases exceeds a preset threshold value, determining the electronic case information to be detected as the similar electronic case of the target electronic case information.

The preset threshold refers to a preset similarity threshold.

Specifically, when the electronic case similarity exceeds a preset threshold, the electronic case information to be detected is used as the electronic case with the similarity of the target electronic case information, and then the electronic case information to be detected can be returned to the terminal for displaying, so that a doctor can process the target electronic case information according to the displayed electronic case information to be detected.

In one embodiment, the similarity detection between the plurality of pieces of electronic case information to be detected and the target electronic case information can be acquired from the historical database until the electronic case information to be detected which is most similar to the target electronic case information is found, and the most similar electronic case information to be detected is returned to the terminal for display.

In the electronic case detection method, the similarity between the entities corresponding to each entity type is calculated through different entity matching models, and the electronic case similarity is further determined according to the weight of each entity type and the entity similarity of each entity type, so that the accuracy of the obtained electronic case similarity is improved. And then, similar electronic cases are determined according to the similarity of the electronic cases, so that the accuracy of obtaining the similar electronic cases can be improved.

In one embodiment, as shown in fig. 3, in step S204, performing entity identification on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity, including the steps of:

s302, inputting the electronic case information to be detected into the established named entity recognition model for recognition to obtain a first recognition result, inputting the electronic case information to be detected into the established medical knowledge map for matching to obtain a first matching result, and obtaining each electronic case entity to be detected according to the first recognition result and the first matching result.

The first recognition result refers to an entity obtained by recognizing the electronic case information to be detected through the established named entity recognition model. The established medical knowledge map is a knowledge map which is established in advance according to professional medical data. The first matching result is an entity which is matched and consistent and is obtained by matching the electronic case information to be detected in the established medical knowledge graph.

Specifically, the electronic case information to be detected is input into an established named entity recognition model for recognition to obtain a first recognition result, and the electronic case information to be detected is input into an established medical knowledge map for matching to obtain a first matching result. And comparing the first matching result with the first identification result, searching the entities which are in the first matching result and are not in the first identification result, and taking the searched entities and the entities in the first identification result as the entities of the electronic cases to be detected.

S304, inputting the target electronic case information into the established named entity recognition model for recognition to obtain a second recognition result, inputting the target electronic case information into the established medical knowledge map for matching to obtain a second matching result, and obtaining each target electronic case entity according to the second recognition result and the second matching result.

And the second recognition result refers to an entity obtained by recognizing the target electronic case information through the established named entity recognition model. The second matching result is an entity which is matched and consistent and is obtained by matching the target electronic case information in the established medical knowledge graph.

Specifically, the target electronic case information is input into an established named entity identification model for identification to obtain a second identification result, the target electronic case information is input into an established medical knowledge graph for matching to obtain a second matching result, the second matching result is compared with the second identification result, entities which are in the second matching result and not in the second identification result are found, and the found entities and the entities in the second identification result are used as the target electronic case entities.

In the embodiment, the entity recognition is carried out through the named entity recognition model and the medical knowledge map, so that the accuracy of each obtained electronic case entity to be detected and each obtained target electronic case entity is improved.

In one embodiment, as shown in fig. 4, step S208, inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into the corresponding entity matching model for matching, so as to obtain entity similarity corresponding to each entity type, includes the steps of:

s402, inputting the electronic case entity to be detected and the target electronic case entity of the name entity type into the established medical name classification tree to obtain the entity node to be detected corresponding to the electronic case entity to be detected and the target entity node corresponding to the target electronic case entity.

The name entity type refers to an entity type corresponding to the disease name entity, and the medical name classification tree is constructed according to the established medical knowledge map and is used for classifying medical data. The entity node to be detected refers to an entity matched with a disease name entity in the electronic case entity in the medical name classification tree. The target entity node refers to an entity matched with a disease name entity in the target electronic case entity in the medical name classification tree.

Specifically, the server inputs disease name entities in the electronic case entity to be detected and the target electronic case entity into the established medical name classification tree for matching, so as to obtain a to-be-detected entity node corresponding to the disease name entity in the electronic case entity to be detected and a target entity node corresponding to the disease name entity in the target electronic case entity.

S404, calculating the path length between the entity node to be detected and the target entity node, and determining the name similarity of the electronic case entity to be detected and the target electronic case entity of the name entity type according to the path length between the entity node to be detected and the target entity node.

The path length refers to the number of edges passing from the entity node to be detected to the target entity node in the medical name classification tree, that is, the path length is increased by one every time the number of the edges passing is increased by one. The path length may be the minimum number of edges that the entity node to be detected passes through to the target entity node.

Specifically, the server calculates the path length between the entity node to be detected and the target entity node, and then determines the name similarity of the electronic case entity to be detected and the target electronic case entity of the name entity type according to the path length between the entity node to be detected and the target entity node.

In the embodiment, the similarity calculation is performed on the entities of the disease name types through the established medical name classification tree, so that the accuracy of the similarity calculation is improved.

In one embodiment, as shown in fig. 5, the step S404 of determining the name similarity between the electronic case entity to be detected and the target electronic case entity of the name entity type according to the path length between the entity node to be detected and the target entity node includes the steps of:

s502, acquiring a root entity node of the established medical name classification tree, acquiring a common father node of the entity node to be detected and the target entity node from the established medical name classification tree, and calculating a second path length between the root entity node and the common father node.

The root entity node refers to the ancestors of all entity nodes except the root entity node in the medical name classification tree, and has no father entity node. The common father node refers to a common father entity node which is closest to the entity node to be detected and the target entity node. The second path length refers to the minimum number of edges that pass from the root entity node to the common parent node in the established medical name classification tree.

Specifically, the server determines a root entity node and a common parent node of the entity node to be detected and the target entity node in the established medical name classification tree, and then calculates the number of edges passing from the root entity node to the common parent node, so as to obtain the second path length.

S504, the target path length is obtained, and the name similarity between the electronic case entity to be detected and the target electronic case entity of the name entity type is calculated according to the path length, the second path length and the target path length.

The target path length refers to the maximum path length between the entity node to be detected and the target entity node, and the maximum path length is twice the depth of the established medical name classification tree.

Specifically, when the server obtains the target path length, the path length and the second path length, the name similarity between the electronic case entity to be detected of the name entity type and the target electronic case entity is calculated by using a formula (2). Equation (2) is as follows:

wherein c1 is the entity node to be detected, c2 is the target entity node, Consim (c)₁,c₂) The name similarity between the electronic case entity to be detected and the target electronic case entity. dis (c)₁,c₂) And indicating the path length between the entity node to be detected and the target entity node, wherein the LCA is a common father node of the entity node to be detected and the target entity node. root is the root entity node. dis (root, LCA) denotes the second path length. H is the depth of the established medical name classification tree. 2H is the target path length. Fig. 6 is a schematic diagram of the medical name classification tree that has been built. When H is 10, dis (root, LCA) is 6, dis (c)₁,c₂) When the name is 4, the similarity of the obtained name is

In the embodiment, the name similarity between the electronic case entity to be detected and the target electronic case entity of the name entity type is calculated through the path length, the second path length and the target path length, so that the accuracy of obtaining the name similarity is improved, the accuracy of the similarity between the calculated case information to be detected and the calculated target case information is improved, and the obtained similar electronic case is more accurate.

In one embodiment, step S404, inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into the corresponding entity matching model for matching, so as to obtain entity similarity corresponding to each entity type, including the steps of:

The nonlinear distance calculation module is a module which performs calculation by using a nonlinear distance calculation algorithm. The age entity type refers to a type corresponding to an age entity in the electronic case entity to be detected and the target electronic case entity. The age similarity refers to the similarity between the age entities in the electronic case entity to be detected and the target electronic case entity.

Specifically, the server inputs the electronic case entity to be detected of the age entity type and the target electronic case entity into the corresponding nonlinear distance calculation module for calculation, so as to obtain the age similarity between the electronic case entity to be detected of the age entity type and the target electronic case entity. The nonlinear distance calculation module may use formula (3) to perform calculation, where formula (3) is as follows:

wherein p1 refers to an age entity in each electronic case entity to be detected, and p2 refers to an age entity in each target electronic case entity. S (p1, p2) refers to age similarity. max (p1, p2) refers to the older age entity of p1 and p 2.

For example: the age entity in each electronic case entity to be detected is 25 years old. The age entity in each target electronic case entity is 30 years old. Then

The age similarity between age entity 25 years and age entity 30 years is calculated as

In the embodiment, the non-linear distance calculation module is used for calculating the electronic case entity to be detected and the target electronic case entity of the age entity type, so that the accuracy of the calculated age similarity is improved, the accuracy of the similarity of the calculated case information to be detected and the target case information is improved, and the obtained similar electronic case is more accurate.

In one embodiment, as shown in fig. 7, step S404, inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into the corresponding entity matching model for matching, so as to obtain entity similarity corresponding to each entity type, includes the steps of:

s702, obtaining each electronic case entity to be detected and each target electronic case entity of the text entity type, vectorizing each electronic case entity to be detected and each target electronic case entity of the text entity type to obtain each vector to be detected and each target vector, and calculating the similarity of each vector to be detected and each target vector.

The text entity type refers to an entity type of the electronic case information belonging to the text. For example, the entity is obtained according to a main complaint text, a present history text, a personal history text, a differential diagnosis text, and the like. The vector to be detected is obtained by vectorizing the electronic case entity to be detected of the text entity type, and the target vector is obtained by vectorizing the target electronic case entity of the text entity type.

Specifically, the server obtains each electronic case entity to be detected and each target electronic case entity of the text entity type, and each text has a plurality of corresponding entities. Vectorizing each electronic case entity to be detected and each target electronic case entity of the text entity type to obtain each vector to be detected and each target vector. For example, a trained Neural Network Language Model (NNLM) may be used to vectorize each electronic case entity to be detected and each target electronic case entity. The word embedding vector can also be obtained by vectorizing each electronic case entity to be detected and each target electronic case entity by using a trained Skip-gram Model or/CBOW (Continuous Bag-of-Words Model) Model. At this time, the similarity between each vector to be detected and each target vector may be calculated using a cosine similarity algorithm.

In one embodiment, when the similarity between each vector to be detected and each target vector is calculated, the word reflecting the entity degree in the text can be obtained, and the word reflecting the entity degree in the text is encoded to obtain the degree parameter. For example, for the "red and swollen" entity in the text, the word reflecting the degree of the entity may include the presence or absence of the type of degree modifier and the degree of severity modifier, for example: "none", "significant", "slight", "present", and the like. And performing polarity coding on the type-existence degree modifiers to obtain polarity parameters, for example, whether the codes are-1, not-1, having 1 and the like can be coded. The modifiers of the severity are linearly coded to obtain a severity parameter, such as slightly 0.5. A bit of 0.5. The severity was 1. Is remarkably 1. Then, when calculating the similarity between the vector to be detected and the target vector, the calculation can be performed using equation (4). The formula (4) is as follows:

s (a1, a2) ═ sim (a1, a2) ═ b1 × (b 2) formula (4)

Wherein a1 is the vector to be detected, a2 is the target vector, b1 is the severity parameter, b2 is the polarity parameter, sim (a1, a2) is the cosine similarity between the vector to be detected and the target vector. By introducing the degree parameter, the accuracy of the similarity between the obtained vector to be detected and the target vector is improved.

S704, determining text similarity between the text to be detected composed of the electronic case entities to be detected and the target text composed of the target electronic case entities according to the similarity between the vectors to be detected and the target vectors.

The text to be detected refers to a text composed of electronic case entities to be detected. The text includes a chief complaint text, a present history text, a personal history text, and a differential diagnosis text, etc. For example, each electronic case entity to be detected comprises a cold and a prescription, and the formed text to be detected is a main complaint text "cold prescription". The target text refers to the text composed of each target electronic case entity, and comprises a chief complaint text, a current medical history text, a personal history text, a differential diagnosis text and the like. For example, each target electronic case entity includes cough, how, and medication, and the composed target text is the complaint text "how to take cough".

Specifically, the maximum similarity between the vector to be detected and the target vector is determined according to the similarity between each vector to be detected and each target vector, and the text similarity between the text to be detected and the target text is determined according to the maximum similarity and the number of entities.

In the embodiment, each to-be-detected electronic case entity and each target electronic case entity of the text entity type are vectorized to obtain each to-be-detected vector and each target vector, and the text similarity between the to-be-detected text and the target text is determined according to the similarity between each to-be-detected vector and each target vector, so that the accuracy of obtaining the text similarity is improved.

In one embodiment, as shown in fig. 8, the step S704 of determining text similarity between the text to be detected composed of each electronic case entity to be detected and the target text composed of each target electronic case entity according to the similarity between each vector to be detected and each target vector includes the steps of:

s802, obtaining texts to be detected composed of the electronic case entities to be detected and target texts composed of the target electronic case entities.

S804, the number of the electronic case entities to be detected in the text to be detected and the number of the target electronic case entities in the target text are calculated.

Specifically, the server obtains a text to be detected composed of each electronic case entity to be detected and a target text composed of each target electronic case entity, and calculates the number of the electronic case entities to be detected in the text to be detected and the number of the target electronic case entities in the target text.

S806, determining each similarity to be detected corresponding to each vector to be detected according to the similarity of each vector to be detected and each target vector, and obtaining the comprehensive similarity to be detected according to each similarity to be detected and the number of the electronic case entities to be detected.

Specifically, the server calculates the similarity between the vector to be detected and each target vector by using a cosine similarity algorithm, and then determines the maximum similarity from the similarities between the vector to be detected and each target vector to obtain the maximum similarity corresponding to the vector to be detected. Calculating the similarity between each vector to be detected and each target vector to obtain the maximum similarity corresponding to each vector to be detected as each similarity to be detected, calculating the sum of the similarities to be detected, and calculating the ratio of the sum of the similarities to be detected and the number of the electronic case entities to be detected to obtain the comprehensive similarity to be detected.

And S808, determining the similarity of each target corresponding to each target vector according to the similarity of each vector to be detected and each target vector, and obtaining the comprehensive similarity of the targets according to the similarity of each target and the number of target electronic case entities.

Specifically, the server calculates the similarity between the target vector and each vector to be detected by using a cosine similarity algorithm, and then determines the maximum similarity from the similarities between the target vector and each vector to be detected to obtain the maximum similarity corresponding to the target vector. And calculating the similarity between each target vector and each vector to be detected to obtain the maximum similarity corresponding to each target vector, taking the maximum similarity as the similarity of each target, calculating the sum of the similarities of the targets, and calculating the ratio of the sum of the similarities of the targets to the number of the electronic case entities of the targets to obtain the comprehensive similarity of the targets.

And S810, obtaining the text similarity between the text to be detected and the target text according to the comprehensive similarity to be detected and the target comprehensive similarity.

Specifically, the server calculates the average similarity according to the comprehensive similarity to be detected and the target comprehensive similarity to obtain the text similarity between the text to be detected and the target text. For example, the text similarity may be calculated using formula (5). Equation (5) is as follows:

wherein S1 refers to the text to be detected, and S2 refers to the target text. The SIM (S1, S2) indicates a text similarity between the text to be detected and the target text. nums (S1) represents the number of entities in the text to be detected, and nums (S2) represents the number of entities in the target text. w1i refers to a vector to be detected corresponding to an electronic case entity to be detected in a text to be detected, and w2j refers to a target vector corresponding to a target electronic case entity in a target text. sim (w1i, w2j) refers to the similarity between the vector to be detected and the respective target vector, max [ sim (w1i, w2j)]The maximum similarity corresponding to each vector to be detected is obtained, and the similarity to be detected is obtained. sim (w2j, w1i) refers to the similarity between the target vector and each vector to be detected. max [ sim (w2j, w1i)]The maximum similarity corresponding to each target vector is obtained, namely the similarity of each target is obtained ∑_w1imax[sim(w1i,w2j)]Refers to the sum of the respective similarity degrees to be detected, ∑_w2jmax[sim(w2j,w1i)]The sum of the similarity of the respective objects.

For example, the following steps are carried out: the text to be detected is 'cold prescription'. The target text is "how to take cough". There are 2 entities in the text to be detected, including "cold" and "prescription". There are 3 entities in the target text, including "cough", "how" and "medication". Among them, the entity most similar to "cold" is "cough", and the similarity is 0.9. The most similar entity to the prescription is medication, with a similarity of 0.7. The calculated overall similarity to be determined was sim1 ═ 0.9+0.7)/2 ═ 0.8. Conversely, the most similar entity to "cough" is "cold", with a similarity of 0.9. The entity most similar to "how" is "prescription" and the similarity is 0.2. The most similar entity to "medication" is "prescription" with a similarity of 0.7. The target integrated similarity was calculated to be sim2 ═ (0.9+0.2+0.7)/3 ═ 0.6. The similarity between the text to be detected and the target text is as follows: SIM ═ (0.8+0.6)/2 ═ 0.7.

It should be understood that although the steps in the flowcharts of fig. 2-5, 7-8 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2-5, 7-8 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performing the sub-steps or stages is not necessarily sequential, but may be performed alternately or alternatingly with other steps or at least some of the sub-steps or stages of other steps.

In one embodiment, as shown in fig. 9, there is provided an electronic case detection apparatus 900 comprising: an information acquisition module 902, an entity obtaining module 904, a model determining module 906, an entity similarity obtaining module 908, a similarity calculating module 910, and a case determining module 912, wherein:

the information acquisition module 902 is used for receiving the electronic case similarity detection instruction and acquiring the electronic case information to be detected and the target electronic case information according to the electronic case similarity detection instruction;

an entity obtaining module 904, configured to perform entity identification on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity;

a model determining module 906, configured to determine each corresponding entity type according to each electronic case entity to be detected and each target electronic case entity, and determine a corresponding entity matching model according to each entity type;

an entity similarity obtaining module 908, configured to input the electronic case entity to be detected and the target electronic case entity, which have the same entity type, into corresponding entity matching models for matching, so as to obtain entity similarities corresponding to the entity types;

the similarity calculation module 910 is configured to obtain each entity type weight, and calculate an electronic case similarity between the electronic case information to be detected and the target electronic case information according to the entity type weight and the entity similarity;

the case determining module 912 is configured to determine, when the similarity of the electronic case exceeds a preset threshold, a similar electronic case in which the electronic case information to be detected is the target electronic case information.

In one embodiment, the entity derivation module 904 includes:

the entity obtaining unit to be detected is used for inputting the electronic case information to be detected into the established named entity recognition model for recognition by the entity to obtain a first recognition result, inputting the electronic case information to be detected into the established medical knowledge map for matching to obtain a first matching result, and obtaining each electronic case entity to be detected according to the first recognition result and the first matching result;

and the target entity obtaining unit is used for inputting the target electronic case information into the established named entity recognition model for recognition to obtain a second recognition result, inputting the target electronic case information into the established medical knowledge map for matching to obtain a second matching result, and obtaining each target electronic case entity according to the second recognition result and the second matching result.

In one embodiment, the entity similarity obtaining module 908 includes:

the node determining unit is used for inputting the electronic case entity to be detected and the target electronic case entity of the name entity type into the established medical name classification tree to obtain a to-be-detected entity node corresponding to the electronic case entity to be detected and a target entity node corresponding to the target electronic case entity;

and the name similarity calculation unit is used for calculating the path length between the entity node to be detected and the target entity node, and determining the name similarity of the electronic case entity to be detected and the target electronic case entity of the name entity type according to the path length between the entity node to be detected and the target entity node.

In one embodiment, the name similarity calculation unit is further configured to obtain a root entity node of the established medical name classification tree, obtain a common parent node of the entity node to be detected and the target entity node from the established medical name classification tree, and calculate a second path length between the root entity node and the common parent node; and acquiring the target path length, and calculating the name similarity between the electronic case entity to be detected of the name entity type and the target electronic case entity according to the path length, the second path length and the target path length.

In one embodiment, the entity similarity obtaining module 908 includes:

and the age similarity calculation unit is used for inputting the electronic case entity to be detected of the age entity type and the target electronic case entity into the corresponding nonlinear distance calculation module for calculation to obtain the age similarity between the electronic case entity to be detected of the age entity type and the target electronic case entity.

In one embodiment, the entity similarity obtaining module 908 includes:

the vector similarity calculation unit is used for acquiring each electronic case entity to be detected and each target electronic case entity of the text entity type, vectorizing each electronic case entity to be detected and each target electronic case entity of the text entity type to obtain each vector to be detected and each target vector, and calculating the similarity of each vector to be detected and each target vector;

and the text similarity determining unit is used for determining the text similarity between the text to be detected composed of the electronic case entities to be detected and the target text composed of the target electronic case entities according to the similarity between the vectors to be detected and the target vectors.

In one embodiment, the text similarity determining unit is further configured to obtain a text to be detected composed of each electronic case entity to be detected and a target text composed of each target electronic case entity; calculating the number of the electronic case entities to be detected in the text to be detected and the number of the target electronic case entities in the target text; determining each similarity to be detected corresponding to each vector to be detected according to the similarity of each vector to be detected and each target vector, and obtaining comprehensive similarity to be detected according to each similarity to be detected and the number of electronic case entities to be detected; determining the similarity of each target corresponding to each target vector according to the similarity of each vector to be detected and each target vector, and obtaining the comprehensive similarity of the targets according to the similarity of each target and the number of target electronic case entities; and obtaining the text similarity between the text to be detected and the target text according to the comprehensive similarity to be detected and the target comprehensive similarity.

For the specific limitations of the electronic medical record detection device, reference may be made to the above limitations of the electronic medical record detection method, which are not described herein again. The various modules in the electronic case finding apparatus described above may be implemented in whole or in part by software, hardware, and combinations thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 10. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store electronic case data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement an electronic case detection method.

Those skilled in the art will appreciate that the architecture shown in fig. 10 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In an embodiment, there is provided a computer device comprising a memory storing a computer program and a processor implementing the steps of the electronic case detection method of any of the above embodiments when the processor executes the computer program.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which, when executed by a processor, performs the steps described in any of the above embodiments for electronic case detection.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. An electronic case detection method, the method comprising:

performing entity recognition on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity;

acquiring entity type weights, and calculating to obtain the electronic case similarity of the electronic case information to be detected and the target electronic case information according to the entity type weights and the entity similarity;

and when the similarity of the electronic case exceeds a preset threshold value, determining that the electronic case information to be detected is a similar electronic case of the target electronic case information.

2. The method according to claim 1, wherein the performing entity recognition on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity comprises:

inputting the electronic case information to be detected into an established named entity recognition model for recognition to obtain a first recognition result, inputting the electronic case information to be detected into an established medical knowledge map for matching to obtain a first matching result, and obtaining each electronic case entity to be detected according to the first recognition result and the first matching result;

inputting the target electronic case information into the established named entity recognition model for recognition to obtain a second recognition result, inputting the target electronic case information into the established medical knowledge map for matching to obtain a second matching result, and obtaining each target electronic case entity according to the second recognition result and the second matching result.

3. The method according to claim 1, wherein the inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into the corresponding entity matching model for matching to obtain the entity similarity corresponding to each entity type comprises:

and calculating the path length between the entity node to be detected and the target entity node, and determining the name similarity of the electronic case entity to be detected and the target electronic case entity of the name entity type according to the path length between the entity node to be detected and the target entity node.

4. The method according to claim 3, wherein determining the name similarity of the electronic case entity to be detected and the target electronic case entity of the name entity type according to the path length between the entity node to be detected and the target entity node comprises:

acquiring a root entity node of the established medical name classification tree, acquiring a common father node of the entity node to be detected and the target entity node from the established medical name classification tree, and calculating a second path length between the root entity node and the common father node;

5. The method according to claim 1, wherein the inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into the corresponding entity matching model for matching to obtain the entity similarity corresponding to each entity type comprises:

6. The method according to claim 1, wherein the inputting the electronic case entity to be detected and the target electronic case entity with the same entity type into the corresponding entity matching model for matching to obtain the entity similarity corresponding to each entity type comprises:

acquiring each electronic case entity to be detected and each target electronic case entity of a text entity type, vectorizing each electronic case entity to be detected and each target electronic case entity of the text entity type to obtain each vector to be detected and each target vector, and calculating the similarity of each vector to be detected and each target vector;

7. The method according to claim 6, wherein determining text similarity between the text to be detected composed of the electronic case entities to be detected and the target text composed of the target electronic case entities according to the similarity between the vectors to be detected and the target vectors comprises:

acquiring texts to be detected composed of the electronic case entities to be detected and target texts composed of the target electronic case entities;

determining each similarity to be detected corresponding to each vector to be detected according to the similarity of each vector to be detected and each target vector, and obtaining comprehensive similarity to be detected according to each similarity to be detected and the number of the electronic case entities to be detected;

determining each target similarity corresponding to each target vector according to the similarity of each vector to be detected and each target vector, and obtaining target comprehensive similarity according to each target similarity and the target electronic case entity number;

8. An electronic medical record testing device, characterized in that the device comprises:

the information acquisition module is used for receiving an electronic case similarity detection instruction and acquiring electronic case information to be detected and target electronic case information according to the electronic case similarity detection instruction;

the entity obtaining module is used for carrying out entity identification on the electronic case information to be detected and the target electronic case information to obtain each electronic case entity to be detected and each target electronic case entity;

the entity similarity obtaining module is used for inputting the electronic case entity to be detected with the consistent entity type and the target electronic case entity into corresponding entity matching models for matching to obtain entity similarity corresponding to each entity type;

the similarity calculation module is used for acquiring entity type weights and calculating the electronic case similarity of the electronic case information to be detected and the target electronic case information according to the entity type weights and the entity similarity;

and the case determining module is used for determining that the electronic case information to be detected is a similar electronic case of the target electronic case information when the similarity of the electronic case exceeds a preset threshold value.

9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.