A medical record inconsistency detection method, system, device and storage medium

Abstract

A medical record inconsistency detection method, system, device and storage medium. The method constructs a multi-perspective subject model, in which the patient's medical record is abstracted into an instance, the instance contains a preset medical record document, and the medical record document contains a preset subject word; the specified subject word is inferred according to the multi-perspective subject model, and an instance subject distribution is obtained from each perspective of the instance; a similarity measurement is used to determine whether the instance subject distributions between different perspectives are similar, and if the similarity of the instance subject distributions between different perspectives exceeds a preset range, it is determined that the medical record document corresponding to the specified subject word is inconsistent. The present invention converts the problem of inconsistent medical record document content into a problem of inconsistent subject matter, describes each medical record document as a medical record instance from different perspectives, establishes subject associations between medical record documents, and uses similarity measurement to detect inconsistencies in the content between medical record documents, thereby ensuring the consistency of different medical record documents for the same patient.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and specifically relates to a medical record inconsistency detection method, system, device and storage medium. Background Art

[0002] Currently, in the electronic medical record system, the patient's information is recorded in different medical documents, and the descriptions of the patient's condition in different medical documents need to ensure consistency.

[0003] For example, medical documents include chief complaint documents and current medical history documents. The contents of the chief complaint document are:

[0004] Right shoulder pain, swelling, and limited activity for 1 year;

[0005] The current medical history document is as follows: More than one year ago, the patient developed right knee joint pain, moderate pain, persistent onset, chronic pain, more obvious at night, limited lower limb activity, no muscle strength changes, unrelated to climate change, no long-term fever. The patient has no fever, night sweats, high fever, no migratory joint pain, no symmetrical small joint pain, ... The patient is in good spirits, sleeps well, has normal bowel movements, normal urination, and no significant changes in weight.

[0006] The content of the chief complaint document is extracted from the content of the patient's current medical history document, and the knee pain in the current medical history is inconsistent with the right shoulder pain in the chief complaint.

[0007] In the existing technology, when comparing the consistency of two documents, they are considered consistent only when they are consistent in multiple subject dimensions. For example, "right shoulder pain" and "limited mobility" in the chief complaint document are two clinical symptoms of the same patient. Both require consistent evidence in the current medical history to determine whether the two documents are consistent. However, the existing topic modeling technology cannot directly extract the consistency relationship in multiple dimensions. How to detect the inconsistency defects between documents in the electronic medical record system is a technical problem that needs to be solved urgently. Summary of the invention

[0008] To this end, the present invention provides a medical record inconsistency detection method, system, device and storage medium, which can convert the inconsistency of medical record document content into inconsistency of subject, and solve the problem of detecting inconsistency defects between documents in the electronic medical record system.

[0009] In order to achieve the above object, the present invention provides the following technical solutions: In a first aspect, a method for detecting inconsistency in medical records is provided, comprising the following steps:

[0010] Constructing a multi-perspective subject model, in which the patient medical records are abstracted into instances, the instances contain preset medical record documents, and the medical record documents contain preset subject words;

[0011] Inferring the designated subject words according to the multi-view subject model, and obtaining an instance topic distribution from each view of the instance;

[0012] The similarity measurement is used to determine whether the instance topic distributions between different perspectives are similar. If the instance topic distribution similarity between different perspectives exceeds the preset range, it is determined that the medical record document content corresponding to the specified subject term is inconsistent.

[0013] As a preferred solution of the medical record inconsistency detection method, when constructing a multi-view subject model, different types of features are used as multiple viewpoints to describe the medical record document under the same instance.

[0014] As a preferred solution of the medical record inconsistency detection method, when the designated subject words are inferred according to the multi-view subject model, the designated subject words are derived from the designated viewpoint, and an instance topic distribution is obtained from the corresponding viewpoint using the designated subject words.

[0015] As a preferred solution of the medical record inconsistency detection method, if the similarity of instance subject distribution between different perspectives is within a preset range, it is determined that the medical record document content corresponding to the specified subject word is consistent.

[0016] As a preferred solution of the medical record inconsistency detection method, the medical record documents include a chief complaint document and a current medical history document, and the perspectives include a chief complaint perspective and a current medical history perspective.

[0017] In a second aspect, the present invention provides a medical record inconsistency detection system, comprising:

[0018] A subject model construction module, used to construct a multi-view subject model, in which the patient medical records are abstracted into instances, the instances contain preset medical record documents, and the medical record documents contain preset subject words;

[0019] A subject word inference module, used to infer a specified subject word according to the multi-view subject model, and obtain an instance subject distribution from each view of the instance;

[0020] The consistency judgment module is used to use similarity measurement to determine whether the instance topic distribution between different perspectives is similar. If the similarity of the instance topic distribution between different perspectives exceeds a preset range, it is determined that the medical record document content corresponding to the specified subject word is inconsistent.

[0021] As a preferred solution of the medical record inconsistency detection system, in the subject model construction module, when constructing the multi-view subject model, different types of features are used as multiple viewpoints to describe the medical record document under the same instance;

[0022] In the subject word inference module, when the designated subject word is inferred according to the multi-view subject model, the designated subject word originates from the designated view, and an instance subject distribution is obtained from the corresponding view using the designated subject word.

[0023] As a preferred solution of the medical record inconsistency detection system, in the consistency judgment module, if the similarity of instance subject distribution between different perspectives is within a preset range, it is determined that the medical record document content corresponding to the specified subject word is consistent;

[0024] The medical record documents include chief complaint documents and current medical history documents, and the perspectives include chief complaint perspective and current medical history perspective.

[0025] In a third aspect, a medical record inconsistency detection device is provided, comprising: a memory and a processor; the processor and the memory communicate with each other via a bus; the memory stores program instructions executable by the processor; the processor calls the program instructions to execute the medical record inconsistency detection method of the first aspect or any possible implementation thereof.

[0026] In a fourth aspect, a medical record inconsistency detection storage medium is provided, wherein the storage medium stores a program code of a medical record inconsistency detection method, wherein the program code includes instructions for executing the medical record inconsistency detection method of the first aspect or any possible implementation thereof.

[0027] The present invention has the following advantages: by constructing a multi-perspective subject model, the patient's medical record is abstracted into an instance in the multi-perspective subject model, the instance contains a preset medical record document, and the medical record document contains a preset subject word; the specified subject word is inferred according to the multi-perspective subject model, and an instance topic distribution is obtained from each perspective of the instance; the similarity measurement is used to determine whether the instance topic distributions between different perspectives are similar, and if the similarity of the instance topic distributions between different perspectives exceeds the preset range, it is determined that the medical record document corresponding to the specified subject word is inconsistent. The present invention can convert the problem of inconsistent medical record document content into a problem of inconsistent subject matter, describe each medical record document as a medical record instance from different perspectives, establish the subject association between medical record documents through a multi-perspective subject model, and use the similarity measurement to detect the inconsistency of the content between medical record documents, so as to ensure the consistency of different medical record documents for the same patient. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0029] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0030] Figure 1 A schematic diagram of a flow chart of a method for detecting inconsistency in medical records provided in Example 1 of the present invention;

[0031] Figure 2 A schematic diagram of a multi-view subject model involved in a medical record inconsistency detection method provided in Example 1 of the present invention;

[0032] Figure 3 A schematic diagram of a medical record inconsistency detection device provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0033] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are 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.

[0034] Example 1

[0035] See also Figure 1 Embodiment 1 of the present invention provides a method for detecting inconsistency in medical records, comprising the following steps:

[0036] S1. Construct a multi-perspective subject model, in which the patient's medical record is abstracted into an instance, the instance includes a preset medical record document, and the medical record document includes a preset subject word;

[0037] S2, inferring the designated subject words according to the multi-view subject model, and obtaining an instance topic distribution from each view of the instance;

[0038] S3. Use similarity measurement to determine whether the instance topic distributions between different perspectives are similar. If the instance topic distribution similarities between different perspectives exceed a preset range, it is determined that the medical record document contents corresponding to the specified subject terms are inconsistent.

[0039] In this embodiment, when constructing a multi-perspective subject model, different types of features are used as multiple perspectives to describe the medical record document under the same instance. The medical record document includes a chief complaint document and a current medical history document, and the perspectives include the chief complaint perspective and the current medical history perspective.

[0040] Specifically, in the process of constructing a multi-view subject model, a patient's medical record is abstracted into an instance S, where the instance S contains m medical records, that is, S = {d 1 ,d 2 ,...,d m}, each medical record document k contains n words, d k ={w 1 ,w 2 ,...,w n}.

[0041] The multi-view topic model is an extension of the LDA topic model. The LDA topic model is a Bayesian algorithm model that uses the prior distribution to estimate the likelihood of data and finally obtain the posterior distribution. The LDA topic model assumes that the document topic is a multinomial distribution, and the parameters of the multinomial distribution (prior distribution) follow the Dirichlet distribution.

[0042] Assistance Figure 2 In the multi-view topic model, α is the hyperparameter of the topic distribution, β is the hyperparameter of the keyword distribution under the topic, and θ represents the topic distribution of each instance, and the instance has m different types of features to describe it, that is, m perspectives of an instance.

[0043] The multi-view topic model treats different types of features as multiple perspectives describing the same sample. For example, in the following table:

[0044]

[0045] The chief complaint and history of present illness are two perspectives on the current medical record, and both the chief complaint and history of present illness contain different distributions of subject terms.

[0046] In this embodiment, when the designated subject word is inferred according to the multi-view subject model, the designated subject word originates from the designated view, and an instance topic distribution is obtained from the corresponding view using the designated subject word.

[0047] Specifically, after obtaining the multi-view subject model, we can infer the new data, that is, the new subject words, and calculate the instance topic distribution of the specified subject words. Each view can be regarded as an independent LDA model, that is, each view can obtain an instance topic distribution. The calculation formula is as follows:

[0048]

[0049] Among them, K represents the number of subject words, represents the number of keywords K assigned to all features in instance S, represents the number of all features in instance S, α * is the Dirichlet prior, Represents the value of the topic distribution of m-type features when the topic is k and the feature is x.

[0050] In this embodiment, since different perspectives describe the same instance, the instance subject distributions of different perspectives should be very close, so the similarity between the subject words of multiple perspectives can be determined by similarity measurement, thereby determining whether the medical record document contents corresponding to the specified subject words are consistent.

[0051] Specifically, a two-layer MLP network is used to determine whether two perspectives are similar:

[0052] o = sigmoid(MLP(θ 1 ,...,θ m ))

[0053] Among them, o is the similarity measure and θ represents the topic distribution of each instance.

[0054] Furthermore, a range value of similarity measurement can be given. If the similarity measurement of instance subject distribution between different perspectives exceeds the preset range, it is determined that the medical record document content corresponding to the specified subject word is inconsistent. If the similarity measurement of instance subject distribution between different perspectives is within the preset range, it is determined that the medical record document content corresponding to the specified subject word is consistent.

[0055] In summary, the present invention constructs a multi-perspective subject model, in which the patient's medical record is abstracted into an instance, the instance contains a preset medical record document, and the medical record document contains a preset subject word; the specified subject word is inferred according to the multi-perspective subject model, and an instance topic distribution is obtained from each perspective of the instance; a similarity metric is used to determine whether the instance topic distributions between different perspectives are similar, and if the similarity of the instance topic distributions between different perspectives exceeds a preset range, it is determined that the contents of the medical record documents corresponding to the specified subject words are inconsistent. The present invention can convert the problem of inconsistent medical record content into a problem of inconsistent themes, describe each medical record document as a medical record instance from different perspectives, establish thematic associations between medical record documents through a multi-perspective subject model, and use similarity metrics to detect inconsistencies in the contents between medical record documents, so as to ensure the consistency of different medical record documents for the same patient.

[0056] Example 2

[0057] See also Figure 3Embodiment 2 of the present invention further provides a medical record inconsistency detection system, including:

[0058] A subject model construction module 1 is used to construct a multi-view subject model, in which the patient medical records are abstracted into instances, the instances contain preset medical record documents, and the medical record documents contain preset subject words;

[0059] A subject word inference module 2, configured to infer a designated subject word according to the multi-view subject model, and obtain an instance subject distribution from each view of the instance;

[0060] The consistency judgment module 3 is used to use similarity measurement to judge whether the instance subject distribution between different perspectives is similar. If the instance subject distribution similarity between different perspectives exceeds a preset range, it is determined that the medical record document content corresponding to the specified subject word is inconsistent.

[0061] In this embodiment, in the subject model building module 1, when building a multi-view subject model, different types of features are used as multiple viewpoints to describe the medical record document under the same instance;

[0062] In the subject word inference module 2, when the designated subject word is inferred according to the multi-view subject model, the designated subject word originates from the designated viewpoint, and an instance subject distribution is obtained from the corresponding viewpoint using the designated subject word.

[0063] In this embodiment, in the consistency judgment module 3, if the similarity of instance subject distribution between different perspectives is within a preset range, it is determined that the medical record document content corresponding to the specified subject word is consistent;

[0064] The medical record documents include chief complaint documents and current medical history documents, and the perspectives include chief complaint perspective and current medical history perspective.

[0065] It should be noted that the information interaction, execution process and other contents between the modules of the above-mentioned system are based on the same concept as the method embodiment in Example 1 of the present application, and the technical effects they bring are the same as those of the method embodiment of the present application. For specific contents, please refer to the description in the method embodiment shown in the previous part of the present application, and will not be repeated here.

[0066] Example 3

[0067] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, in which a program code of a medical record inconsistency detection method is stored. The program code includes instructions for executing the medical record inconsistency detection method of embodiment 1 or any possible implementation thereof.

[0068] The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0069] Example 4

[0070] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0071] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the medical record inconsistency detection method of embodiment 1 or any possible implementation thereof.

[0072] Specifically, the processor can be implemented by hardware or by software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor implemented by reading software codes stored in a memory. The memory can be integrated in the processor or can be located outside the processor and exist independently.

[0073] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center.

[0074] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0075] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for detecting inconsistency in medical records, characterized in that: The following steps are involved: Constructing a multi-perspective subject model, in which the patient medical records are abstracted into instances, the instances contain preset medical record documents, and the medical record documents contain preset subject words; Inferring the designated subject words according to the multi-view subject model, and obtaining an instance topic distribution from each view of the instance; The similarity measurement is used to determine whether the instance topic distributions between different perspectives are similar. If the instance topic distribution similarity between different perspectives exceeds the preset range, it is determined that the medical record document content corresponding to the specified subject word is inconsistent; When constructing a multi-view subject model, different types of features are used as multiple viewpoints to describe the medical record document under the same instance; When the specified subject word is inferred according to the multi-view subject model, the specified subject word originates from the specified view, and an instance subject distribution is obtained from the corresponding view using the specified subject word; If the similarity of instance subject distribution between different perspectives is within a preset range, it is determined that the medical record document content corresponding to the specified subject term is consistent; The medical record documents include the chief complaint document and the current medical history document, and the perspectives include the chief complaint perspective and the current medical history perspective; The medical record of a patient is abstracted into an instance S, where instance S contains m medical records, that is, S = {d1, d2, ..., d m }, each medical record document k contains n words, d k ={w1,w2,...,w n }; After obtaining the multi-view subject model, the new data, that is, the new subject words, are inferred to calculate the instance topic distribution of the specified subject words. Each view can be regarded as an independent LDA model, that is, each view can obtain an instance topic distribution. The calculation formula is as follows: Among them, K represents the number of subject words, represents the number of keywords K assigned to all features in instance S, represents the number of all features in instance S, α * is the Dirichlet prior, The value of the topic distribution of m-type features when the topic is k and the feature is x; Use a two-layer MLP network to determine whether two views are similar: o=sigmoid(MLP(θ 1 ,...,θ m )) Among them, o is the similarity measure and θ represents the topic distribution of each instance.

2. A medical record inconsistency detection system, characterized in that: include: A subject model construction module, used to construct a multi-view subject model, in which the patient medical records are abstracted into instances, the instances contain preset medical record documents, and the medical record documents contain preset subject words; A subject word inference module, used to infer a specified subject word according to the multi-view subject model, and obtain an instance subject distribution from each view of the instance; The consistency judgment module is used to use similarity measurement to judge whether the instance subject distribution between different perspectives is similar. If the instance subject distribution similarity between different perspectives exceeds a preset range, it is determined that the medical record document content corresponding to the specified subject word is inconsistent; In the subject model building module, when building a multi-view subject model, different types of features are used as multiple viewpoints to describe the medical record document under the same instance; In the subject word inference module, when the specified subject word is inferred according to the multi-view subject model, the specified subject word originates from the specified view, and an instance subject distribution is obtained from the corresponding view using the specified subject word; In the consistency judgment module, if the similarity of instance subject distribution between different perspectives is within a preset range, it is determined that the medical record document content corresponding to the specified subject word is consistent; The medical record documents include the chief complaint document and the current medical history document, and the perspectives include the chief complaint perspective and the current medical history perspective; The medical record of a patient is abstracted into an instance S, where instance S contains m medical records, that is, S = {d1, d2, ..., d m }, each medical record document k contains n words, d k ={w1,w2,...,w n }; After obtaining the multi-view subject model, the new data, that is, the new subject words, are inferred to calculate the instance topic distribution of the specified subject words. Each view can be regarded as an independent LDA model, that is, each view can obtain an instance topic distribution. The calculation formula is as follows: Among them, K represents the number of subject words, represents the number of keywords K assigned to all features in instance S, represents the number of all features in instance S, α * is the Dirichlet prior, The value of the topic distribution of m-type features when the topic is k and the feature is x; Use a two-layer MLP network to determine whether two views are similar: o=sigmoid(MLP(θ 1 ,...,θ m )) Among them, o is the similarity measure and θ represents the topic distribution of each instance.

3. A medical record inconsistency detection device, comprising: Memory and processor; The processor and the memory communicate with each other via a bus; The memory stores program instructions that can be executed by the processor; characterized in that the processor calls the program instructions to execute the medical record inconsistency detection method described in claim 1.

4. A medical record inconsistency detection storage medium, characterized in that: The storage medium stores a program code of a medical record inconsistency detection method, and the program code includes instructions for executing the medical record inconsistency detection method according to claim 1.

Images