Case front page coding detection and determination method and device and electronic equipment

Abstract

The application provides a medical record front page coding detection and determination method and device and electronic equipment. The method comprises the following steps: performing entity relationship extraction on the disease descriptions corresponding to all standard codes in the international disease classification; determining commonly used codes according to the number of each original code in a plurality of historical medical record front pages; performing standardized processing on the original diagnosis names corresponding to the commonly used codes, performing entity relationship extraction on all standardized diagnosis names, and constructing a knowledge graph based on the entity relationship extraction result; injecting the knowledge graph into a trained prediction model to obtain a target prediction model; predicting each to-be-detected diagnosis name based on the target prediction model to obtain a plurality of candidate codes corresponding to each to-be-detected diagnosis name; and determining the final code corresponding to each to-be-detected diagnosis name according to the to-be-detected code corresponding to each to-be-detected diagnosis name and the plurality of candidate codes. The application can comprehensively and quickly detect and correct coding errors of the medical record front page.

Description

Technical Field

[0001] This application relates to the field of medical data processing technology, specifically to a method, apparatus, and electronic device for detecting and determining the code of the medical record front page. Background Technology

[0002] Medical records are an important record of medical procedures. The first page of a medical record includes the diagnosis name and its corresponding code. The diagnosis name is filled in by the clinician, while the code is filled in by the coder. Due to differences in the professional level of clinicians, their individual writing habits, and the relatively insufficient clinical knowledge of coders, as well as the large volume of coding requirements, coding errors may occur, resulting in low coding accuracy and poor medical record quality.

[0003] Currently, the correctness of the coding of the medical record cover sheet is usually determined by manually sampling the cover sheet. However, the above method is based on manual sampling and only targets a small portion of the medical record cover sheets, which cannot comprehensively and quickly detect and correct coding errors in the medical record cover sheets. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method, apparatus, and electronic device for detecting and determining the encoding of medical record cover pages, in order to solve the technical problem that the existing method of manually sampling medical record cover pages cannot comprehensively and quickly detect and correct encoding errors in medical record cover pages.

[0005] In a first aspect, embodiments of this application provide a method for detecting and determining the code of a medical record homepage, comprising: extracting entity relations from disease descriptions corresponding to all standard codes in the International Classification of Diseases (ICD) to obtain multiple first triples; constructing an initial knowledge graph based on the multiple first triples and the corresponding standard codes; obtaining multiple original diagnostic names and corresponding original codes from multiple historical medical record homepages; determining commonly used codes based on the number of each original code; standardizing the original diagnostic names corresponding to the commonly used codes to obtain standardized diagnostic names; extracting entity relations from all standardized diagnostic names to obtain multiple second triples; obtaining a knowledge graph based on the multiple second triples and the initial knowledge graph; injecting the knowledge graph into a trained prediction model to obtain a target prediction model; obtaining multiple diagnostic names to be detected and corresponding codes to be detected from the medical record homepage to be detected; predicting each diagnostic name to be detected based on the target prediction model to obtain multiple candidate codes corresponding to each diagnostic name to be detected; and determining the final code corresponding to each diagnostic name to be detected based on the code to be detected and the multiple candidate codes.

[0006] In one possible implementation of the first aspect, determining commonly used codes based on the number of each original code includes: determining the number of each original code and the department corresponding to each original code; each department corresponds to a preset department threshold; if the number of original codes is greater than the preset department threshold corresponding to the original code, then the original code is determined to be a commonly used code.

[0007] In one possible implementation of the first aspect, the original diagnostic names corresponding to commonly used codes are normalized to obtain normalized diagnostic names, and entity relations are extracted from all normalized diagnostic names to obtain multiple second triples, including: normalizing the original diagnostic names corresponding to commonly used codes to obtain normalized diagnostic names; deduplicating the normalized diagnostic names; performing semantic analysis on all normalized diagnostic names after deduplication; and extracting entity relations from all normalized diagnostic names after semantic analysis to obtain multiple second triples.

[0008] In one possible implementation of the first aspect, injecting the knowledge graph into a trained prediction model to obtain a target prediction model includes: calculating entity vectors corresponding to triples in the knowledge graph based on the TransE algorithm; and embedding the entity vectors into the trained prediction model to obtain the target prediction model.

[0009] In one possible implementation of the first aspect, the method of predicting each diagnostic name to be detected based on the target prediction model to obtain multiple candidate codes corresponding to each diagnostic name to be detected includes: normalizing each diagnostic name to be detected to obtain normalized diagnostic names to be detected; and predicting each normalized diagnostic name to be detected based on the target prediction model to obtain multiple candidate codes corresponding to each normalized diagnostic name to be detected.

[0010] In one possible implementation of the first aspect, determining the final code corresponding to each diagnostic name to be tested based on the code to be tested and the corresponding multiple candidate codes includes: for each standard diagnostic name to be tested, performing the following steps: searching the standard diagnostic name to be tested based on the International Classification of Diseases; if a corresponding initial code is found, then the initial code is used as the final code; if no corresponding initial code is found, then performing the following steps: determining whether the code to be tested corresponding to the standard diagnostic name to be tested is consistent with any of the multiple candidate codes; if so, then determining the corresponding code to be tested as the final code; if not, determining whether the corresponding code to be tested has a hierarchical relationship with any of the multiple candidate codes based on the International Classification of Diseases, and taking that candidate code as the final code if a hierarchical relationship exists.

[0011] In one possible implementation of the first aspect, the method further includes: determining the priority of the final code corresponding to each diagnostic name to be tested in the front page of the medical record to be tested according to the priority identifier of the preset standard code; and determining the diagnostic name to be tested corresponding to the final code with the highest priority as the primary diagnosis.

[0012] Secondly, embodiments of this application provide a device for detecting and determining the code of a medical record cover page, comprising:

[0013] The first construction module is used to extract entity relations from the disease descriptions corresponding to all standard codes in the International Classification of Diseases, resulting in multiple first triples, and to construct an initial knowledge graph based on the multiple first triples and the corresponding standard codes.

[0014] The second construction module is used to obtain multiple original diagnosis names and corresponding original codes from multiple historical medical record homepages, determine the commonly used codes based on the number of each original code, perform normalization processing on the original diagnosis names corresponding to the commonly used codes to obtain normalized diagnosis names, extract entity relations from all normalized diagnosis names to obtain multiple second triples, and obtain a knowledge graph based on multiple second triples and the initial knowledge graph.

[0015] The first determining module is used to inject the knowledge graph into the trained prediction model to obtain the target prediction model;

[0016] The second determination module is used to obtain multiple diagnostic names to be tested and their corresponding codes from the front page of the medical record to be tested; predict each diagnostic name to be tested based on the target prediction model to obtain multiple candidate codes corresponding to each diagnostic name to be tested; and determine the final code corresponding to each diagnostic name to be tested based on the code corresponding to each diagnostic name to be tested and the multiple candidate codes.

[0017] Thirdly, embodiments of this application provide an electronic device, including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the medical record cover page code detection and determination method as described in any of the first aspects.

[0018] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for detecting and determining the medical record cover page code as described in any of the first aspects.

[0019] Fifthly, embodiments of this application provide a computer program product that, when run on an electronic device, causes the electronic device to execute the medical record cover page code detection and determination method described in any of the first aspects above.

[0020] It is understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0021] The medical record homepage coding detection, device, and electronic device provided in this application extract entity relations for disease descriptions corresponding to all standard codes in the International Classification of Diseases (ICD) and extract entity relations for standardized diagnostic names corresponding to commonly used codes. Based on the above entity relation extraction, a knowledge graph about medical knowledge is constructed by obtaining triples, and the knowledge graph is injected into a trained prediction model to improve the prediction accuracy of the target prediction model. Then, based on the multiple candidate codes corresponding to each diagnostic name to be detected output by the target prediction model, the final code corresponding to each diagnostic name to be detected is determined. This can comprehensively and quickly detect and correct coding errors on the medical record homepage.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating a method for detecting and determining the code on the front page of a medical record according to an embodiment of this application;

[0025] Figure 2 This is a schematic diagram of the hierarchical structure of the International Classification of Diseases provided in one embodiment of this application;

[0026] Figure 3 This is a schematic diagram of the structure of a medical record cover page coding detection and determination device provided in one embodiment of this application;

[0027] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0028] The present application will be described more clearly below with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the function of the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0029] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0030] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0031] In the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0032] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0033] Furthermore, the term "multiple" mentioned in the embodiments of this application should be interpreted as two or more.

[0034] Figure 1 This is a flowchart illustrating a method for detecting and determining the code on the front page of a medical record, provided in one embodiment of this application. Figure 1 As shown, the method in the embodiments of this application may include:

[0035] Step 101: Extract entity relations from the disease descriptions corresponding to all standard codes in the International Classification of Diseases to obtain multiple first triples. Construct an initial knowledge graph based on the multiple first triples and their corresponding standard codes.

[0036] Optionally, the International Classification of Diseases (ICD) is a classification system published by the World Health Organization (WHO) that covers diseases, injuries, and health problems. It includes standard codes for diseases and descriptions of those diseases. The standard codes consist of letters and numbers. For example, ICD-10 comprises 22 chapters and 5 levels. Each level corresponds to a chapter, section, category, subcategory, and subheading. In the first level of ICD-10, Chapter 1 is coded A00–B99, corresponding to the disease description "certain infectious and parasitic diseases." In the second level, Chapter 1 includes sections coded A00–A09, corresponding to the disease description "intestinal infectious diseases." In the fifth level, subheading A18.202 corresponds to the disease description "tuberous lymph nodes in the buccal lymph nodes." There is a hierarchical relationship between lower-level codes and higher-level codes that contain those lower-level codes. The hierarchical structure of the International Classification of Diseases is as follows: Figure 2 As shown.

[0037] For example, in this embodiment of the application, ICD-10 is used, and entity relations are extracted from the disease descriptions in ICD-10 to obtain multiple first triples. Specifically, semantic analysis is performed on the disease descriptions corresponding to the standard codes, and multi-granularity word segmentation, dependency relation analysis, and entity relation extraction are performed on the processed disease descriptions. Entity relations can include etiology, location, pathology, symptoms, stage, type, and acute / chronic, etc. For example, for the disease description "tuberculous lymphadenitis of the head and neck", triples such as lymphadenitis-location-head and neck, lymphadenitis-etiology / nature-tuberculous, inflammation-location-head, inflammation-location-neck, inflammation-disease / nature-tuberculous, and inflammation-location-lymph node are obtained. An initial knowledge graph is constructed based on the above multiple first triples and the corresponding standard codes, wherein the first triples in the initial knowledge graph are represented in the form of a directed acyclic graph.

[0038] Optionally, the extracted entities can be supplemented with synonyms, aliases, English names, and abbreviations based on a preset medical database to expand the initial knowledge graph. The preset medical database can be a pre-set medical database based on medical textbooks, medical encyclopedias, medical papers, etc.

[0039] Step 102: Obtain multiple original diagnosis names and corresponding original codes from multiple historical medical record homepages; determine commonly used codes based on the number of each original code; standardize the original diagnosis names corresponding to the commonly used codes to obtain standardized diagnosis names; extract entity relations from all standardized diagnosis names to obtain multiple second triples; and obtain a knowledge graph based on the multiple second triples and the initial knowledge graph.

[0040] Optionally, select historical medical record front pages with high writing quality, complete diagnostic names and codes. The historical medical record front page includes a plurality of original diagnostic names filled in by clinicians and corresponding original codes filled in by coders.

[0041] In a possible implementation, determining the common codes according to the quantity of each original code may specifically include: determining the quantity of each original code and the department corresponding to each original code; each department corresponds to a preset department threshold; if the quantity of an original code is greater than the preset department threshold corresponding to the original code, the original code is determined as a common code.

[0042] Optionally, quantity statistics are performed on the obtained plurality of original codes, the quantity of each original code and the corresponding department are determined, and the original codes whose quantity is greater than the preset department threshold corresponding to the original code are taken as common codes. The common codes of each department have high accuracy, and the corresponding original diagnostic names can be integrated into the initial knowledge graph as supplementary knowledge.

[0043] In a possible implementation, standardization processing is performed on the original diagnostic names corresponding to the common codes to obtain standardized diagnostic names, and entity relationship extraction is performed on all the standardized diagnostic names to obtain a plurality of second triplets, which may specifically include:

[0044] S1. Perform standardization processing on the original diagnostic names corresponding to the common codes to obtain standardized diagnostic names.

[0045] S2. Perform deduplication processing on the standardized diagnostic names.

[0046] S3. Perform semantic analysis processing on all the standardized diagnostic names after the deduplication processing.

[0047] S4. Perform entity relationship extraction on all the standardized diagnostic names after the semantic analysis processing to obtain a plurality of second triplets.

[0048] Optionally, the above standardization processing includes writing form standardization processing and homophone standardization processing. The writing form standardization processing mainly includes removing special punctuation marks, converting letters and numbers to lowercase, converting Roman numerals to Arabic numerals, and the like. For example, "type Ⅱ diabetes" is converted to "type 2 diabetes". The writing form standardization processing can be performed by means of regular expression search and replacement. The homophone standardization processing mainly includes error checking and replacement of homophones. For example, "liver stimulative tumor" is converted to "liver secondary tumor". The homophone standardization processing can be performed based on a statistical language model.

[0049] For example, in the homophone standardization process, the original diagnostic names corresponding to commonly used codes are segmented and stop word removed to obtain a word list corresponding to the original diagnostic names. Words in this word list are then retrieved based on a preset set of incorrect words. This preset set of incorrect words can be an existing set, such as a set of incorrect words related to medical knowledge obtained from a website. For any word in the word list, if it can be found in the preset set of incorrect words, it is determined to be an incorrect word and is replaced. If it cannot be found in the preset set of incorrect words, it is determined to be a word to be checked and is judged as a suspected incorrect word. The pkuseg word segmentation tool can be used for this process.

[0050] Optionally, during the suspected misspelling process, the word to be checked is searched based on a preset common dictionary. This preset common dictionary can be an existing dictionary, such as a medical knowledge-related dictionary obtained from a website. If the word to be checked cannot be found in the preset common dictionary, it is added as a misspelling to a preset misspelling set and replaced. If the word to be checked can be found in the preset common dictionary, the original diagnosis name corresponding to the word to be checked is used as input, and the likelihood probability value of the word to be checked is output using an N-Gram language model. The perplexity (PPL) is calculated based on the original diagnosis name corresponding to the word to be checked. If the likelihood probability value of the word to be checked is less than the perplexity, it is added as a misspelling to the preset misspelling set and replaced. If the likelihood probability value of the word to be checked is not less than the perplexity, the next word to be checked is then considered a suspected misspelling.

[0051] The formula for calculating text perplexity is:

[0052]

[0053] Where S is the original diagnostic name corresponding to the word to be tested, n is the length of the original diagnostic name corresponding to the word to be tested, i.e., the number of words contained therein, p(w i ) represents the i-th word w i The probability, p(w) i |w1,w2,w3,…,w i-1 The expression represents the probability of calculating the ith word based on the first i-1 words. Here, text perplexity measures whether the original diagnostic name corresponding to the word being tested is correct; the smaller the text perplexity, the greater the probability that the original diagnostic name corresponding to the word being tested is correct.

[0054] Exemplarily, during the replacement processing, wrong words are replaced based on a preset phonetic similarity dictionary to generate a candidate set. For example, for the wrong word "jiejie" (meaning "combination" in this incorrect context) in "gingival jiejie", the candidate set {jointure, tuberculosis, boundary river, Jiehe (proper noun)} is generated based on the preset phonetic similarity dictionary. The aforementioned preset phonetic similarity dictionary may be an existing phonetic similarity dictionary, for example, a phonetic similarity dictionary related to medical knowledge obtained from a website.

[0055] The statistical language model Kenlm is used to sequentially calculate the likelihood probability of the wrong word and each word in the candidate set in the original diagnosis name corresponding to the wrong word, and the word in the candidate set with a likelihood probability greater than the likelihood probability corresponding to the wrong word and having the maximum value is selected as the replacement word to replace the wrong word. For example, if the replacement word is calculated as "tuberculosis", "combination" is replaced with "tuberculosis".

[0056] Optionally, deduplication processing is performed on the obtained standardized diagnosis names to ensure that each standardized diagnosis name after the deduplication processing is different.

[0057] Exemplarily, the second triplet is fused into the initial knowledge graph to construct a knowledge graph, so as to improve the comprehensiveness of the above knowledge graph.

[0058] Step 103: Inject the knowledge graph into the trained prediction model to obtain a target prediction model.

[0059] In a possible implementation, step 103 may specifically include: calculating the entity vector corresponding to the triplet in the knowledge graph based on the TransE algorithm; embedding the entity vector into the trained prediction model to obtain the target prediction model.

[0060] Optionally, an embedding representation algorithm such as the TransE algorithm, TransH algorithm or TransR algorithm is used to obtain the entity vector corresponding to the triplet in the knowledge graph, and the aforementioned entity vector is embedded into the trained prediction model. For example, the aforementioned trained prediction model may be a BERT (Bidirectional Encoder Representations from Transformers) model. Specifically, the entity vector is embedded into the knowledge layer of the BERT model, so that the trained BERT model is equipped with medical knowledge, and can serve as the target prediction model to better predict the candidate code corresponding to the to-be-detected diagnosis name, thereby improving the prediction accuracy of the target prediction model.

[0061] Step 104: Obtain multiple diagnostic names and corresponding codes from the front page of the medical record to be tested; predict each diagnostic name based on the target prediction model to obtain multiple candidate codes corresponding to each diagnostic name; determine the final code corresponding to each diagnostic name based on the code and multiple candidate codes corresponding to each diagnostic name.

[0062] Optionally, the front page of the medical record to be tested includes multiple diagnostic names to be tested filled in by the clinician and corresponding codes to be tested filled in by the coder.

[0063] In one possible implementation, each diagnostic name to be detected is predicted based on a target prediction model to obtain multiple candidate codes corresponding to each diagnostic name to be detected, which may specifically include:

[0064] The names of the diagnoses to be tested are standardized to obtain standardized names of the diagnoses to be tested.

[0065] Based on the target prediction model, the names of the diagnostic tests to be detected in each standard are predicted, and multiple candidate codes corresponding to the names of the diagnostic tests to be detected in each standard are obtained.

[0066] Optionally, the specific implementation process and principle of standardizing each diagnostic name to be detected can be referred to the description of standardization in step 102 of the aforementioned embodiment, and will not be repeated here.

[0067] For example, the standardized diagnostic name to be detected is input into the target prediction model. After prediction by the target prediction model, it outputs multiple candidate codes corresponding to the standardized diagnostic name to be detected, as well as the matching degree of each candidate code. The hidden layer of the target prediction model can be TextCNN, the output layer can be a fully connected layer, the activation function can be sigmoid, and the loss function can be the binary classification cross-entropy loss function.

[0068] Optionally, the surgical name, pathological diagnosis, etc. in the medical record homepage can be obtained, and the standardized surgical name and pathological diagnosis, along with the standardized diagnosis name to be detected, can be used as the model input of the target prediction model to improve the accuracy of the multiple candidate codes corresponding to the standardized diagnosis name to be detected.

[0069] It should be noted that the multiple candidate codes obtained are arranged in descending order of matching degree. For example, the candidate code ranked first has the highest matching degree.

[0070] In one possible implementation, the final code corresponding to each diagnostic name to be detected is determined based on the code to be detected and multiple candidate codes corresponding to each diagnostic name to be detected, which may specifically include:

[0071] For each diagnostic name to be tested, perform the following steps:

[0072] The standard diagnostic name to be tested is retrieved based on the International Classification of Diseases. If a corresponding initial code is found, the initial code is used as the final code.

[0073] If the corresponding initial code is not found, proceed with the following steps:

[0074] Determine whether the code to be detected corresponding to the diagnostic name to be detected is consistent with any of the multiple candidate codes.

[0075] If yes, then the corresponding code to be detected is determined as the final code; if not, then based on the International Classification of Diseases, it is determined whether the corresponding code to be detected has a hierarchical relationship with any of the multiple candidate codes. If yes, then that candidate code is taken as the final code.

[0076] Optionally, if a standardized diagnostic name can be retrieved based on the disease description in the International Classification of Diseases, the standard code corresponding to the retrieved disease description shall be used as the initial code, and the initial code shall be used as the final code.

[0077] If no standardized diagnostic name is found in the disease description based on the International Classification of Diseases, the final code is determined based on the code to be detected corresponding to the standardized diagnostic name and the corresponding multiple candidate codes.

[0078] Specifically, if the code to be detected corresponding to the diagnostic name to be detected matches any of the multiple candidate codes, then the code to be detected is confirmed to be correct and is used as the final code.

[0079] If the code to be tested corresponding to the diagnostic name to be tested is inconsistent with any of the multiple candidate codes, then the code to be tested is confirmed to be incorrect. In this case, based on the International Classification of Diseases, any candidate code that has a hierarchical relationship with the code to be tested is taken as the final code. The code to be tested can be the higher-order code, and the candidate code can be the lower-order code.

[0080] If the code to be tested corresponding to the diagnostic name to be tested is inconsistent with any of the multiple candidate codes, and there is no hierarchical relationship between the code to be tested and any candidate code, then the diagnostic name to be tested will be marked so that the coder can manually review the diagnostic name to be tested and confirm the final code.

[0081] For example, as mentioned above, the candidate codes are arranged in descending order of matching degree, with the first candidate code having the highest matching degree. To improve the accuracy of the final code, the code can be confirmed as correct and used as the final code only if the code corresponding to the diagnostic name to be tested matches the first candidate code among the multiple candidate codes. If the code corresponding to the diagnostic name to be tested does not match the first candidate code among the multiple candidate codes, and the code to be tested has no hierarchical relationship with any candidate code, the code to be tested is confirmed as incorrect, and the diagnostic name to be tested is marked so that the coder can manually review the diagnostic name to be tested and confirm the final code.

[0082] Optionally, the medical record cover sheet includes multiple diagnostic names, and the primary diagnosis is usually determined manually, which may lead to errors in judgment and a low accuracy rate for the primary diagnosis. To solve the above problem, the medical record cover sheet coding detection and determination method provided in this application embodiment may further include:

[0083] Based on the priority identifier of the preset standard code, the priority of the final code corresponding to each diagnostic name to be tested in the front page of the medical record to be tested is determined; the diagnostic name corresponding to the final code with the highest priority is determined as the primary diagnosis.

[0084] For example, after determining the final code of each diagnostic name to be tested in the front page of the medical record, the priority of each final code is determined based on the priority identifier of the preset standard codes, and the diagnostic name corresponding to the final code with the highest priority is taken as the primary diagnosis. For example, ICD-10 includes 22 chapters. The standard codes corresponding to chapters 15 and 16 are assigned priority identifier 0, the standard codes corresponding to chapters 1, 2, 5, 17 and 19 are assigned priority identifier 1, the standard codes corresponding to chapters 18 and 21 are assigned priority identifier 3, and the standard codes corresponding to the remaining chapters are assigned priority identifier 2. Among them, the priority identifier with the smallest value has the highest priority. And so on, according to the needs, priority identifiers can be assigned to the standard codes corresponding to sections, categories, subcategories and details in ICD-10.

[0085] The primary diagnosis is determined based on the priority identifier of the preset standard code. This can quickly detect whether the original primary diagnosis in the medical record front page is correct, and quickly correct the error in the primary diagnosis when there is an error in the original primary diagnosis.

[0086] This application provides a method for detecting and determining the coding of the medical record homepage. It extracts entity relationships from disease descriptions corresponding to all standard codes in the International Classification of Diseases (ICD) and from standardized diagnostic names corresponding to commonly used codes. Based on the extracted entity relationships, a knowledge graph is constructed using triples. This knowledge graph is then injected into a trained prediction model to obtain a target prediction model. The target prediction model is used to predict each diagnostic name to be detected. The final code corresponding to each diagnostic name is determined based on multiple candidate codes and the corresponding code to be detected. By constructing a knowledge graph about medical knowledge and injecting it into the trained prediction model, the prediction accuracy of the target prediction model is improved. Finally, the final code corresponding to the diagnostic name to be detected is determined based on multiple candidate codes output by the target prediction model. This method can comprehensively and quickly detect and correct coding errors on the medical record homepage.

[0087] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0088] Figure 3 This is a schematic diagram of the structure of a medical record cover page coding detection and determination device provided in one embodiment of this application. Figure 3 As shown, the medical record cover page coding detection and determination device provided in this embodiment may include: a first construction module 201, a second construction module 202, a first determination module 203, and a second determination module 204.

[0089] The first construction module 201 is used to extract entity relations from the disease descriptions corresponding to all standard codes in the International Classification of Diseases, to obtain multiple first triples, and to construct an initial knowledge graph based on the multiple first triples and the corresponding standard codes.

[0090] The second construction module 202 is used to obtain multiple original diagnosis names and corresponding original codes from multiple historical medical record homepages, determine commonly used codes based on the number of each original code, perform normalization processing on the original diagnosis names corresponding to the commonly used codes to obtain normalized diagnosis names, extract entity relations from all normalized diagnosis names to obtain multiple second triples, and obtain a knowledge graph based on multiple second triples and the initial knowledge graph.

[0091] The first determining module 203 is used to inject the knowledge graph into the trained prediction model to obtain the target prediction model.

[0092] The second determining module 204 is used to obtain multiple diagnostic names to be detected and their corresponding codes to be detected in the front page of the medical record to be detected; predict each diagnostic name to be detected based on the target prediction model to obtain multiple candidate codes corresponding to each diagnostic name to be detected; and determine the final code corresponding to each diagnostic name to be detected based on the code to be detected and the multiple candidate codes.

[0093] Optionally, the second construction module 202 is specifically used to: determine the number of each original code and the department corresponding to each original code; each department corresponds to a preset department threshold; if the number of original codes is greater than the preset department threshold corresponding to the original code, then the original code is determined to be a commonly used code.

[0094] Optionally, the second construction module 202 is further specifically used for: standardizing the original diagnostic names corresponding to common codes to obtain standardized diagnostic names; deduplicating the standardized diagnostic names; performing semantic analysis on all standardized diagnostic names after deduplication; and extracting entity relations from all standardized diagnostic names after semantic analysis to obtain multiple second triplets.

[0095] Optionally, the first determining module 203 is specifically used to: calculate the entity vector corresponding to the triple in the knowledge graph based on the TransE algorithm; and embed the entity vector into the trained prediction model to obtain the target prediction model.

[0096] Optionally, the second determining module 204 is specifically used for: standardizing each diagnostic name to be detected to obtain each standardized diagnostic name to be detected; and predicting each standardized diagnostic name to be detected based on the target prediction model to obtain multiple candidate codes corresponding to each standardized diagnostic name to be detected.

[0097] Optionally, the second determining module 204 is further specifically used to: for each standardized diagnostic name to be tested, perform the following steps: search the standardized diagnostic name to be tested based on the International Classification of Diseases; if a corresponding initial code is found, then the initial code is used as the final code; if no corresponding initial code is found, then perform the following steps: determine whether the code to be tested corresponding to the standardized diagnostic name to be tested is consistent with any candidate code among the multiple candidate codes; if yes, then determine the corresponding code to be tested as the final code; if no, determine whether the corresponding code to be tested has a hierarchical relationship with any candidate code among the multiple candidate codes based on the International Classification of Diseases, and if a hierarchical relationship exists, take that candidate code as the final code.

[0098] Optionally, the second determining module 204 is further specifically used to: determine the priority of the final code corresponding to each diagnostic name to be tested in the front page of the medical record to be tested according to the priority identifier of the preset standard code; and determine the diagnostic name to be tested corresponding to the final code with the highest priority as the primary diagnosis.

[0099] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0100] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. For example... Figure 4 As shown, the electronic device 300 of this embodiment includes a processor 310 and a memory 320, wherein the memory 320 stores a computer program 321 that can run on the processor 310. When the processor 310 executes the computer program 321, it implements the steps in any of the above-described method embodiments, for example... Figure 1 Steps 101 to 104 are shown. Alternatively, when processor 310 executes computer program 321, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 3 The functions of modules 201 to 204 are shown.

[0101] For example, computer program 321 may be divided into one or more modules / units, one or more of which are stored in memory 320 and executed by processor 310 to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 321 in electronic device 300.

[0102] Those skilled in the art will understand that Figure 4 This is merely an example of an electronic device and does not constitute a limitation on the electronic device. It may include more or fewer components than shown, or combinations of certain components, or different components, such as input / output devices, network access devices, buses, etc.

[0103] The processor 310 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0104] The memory 320 can be an internal storage unit of the electronic device, such as a hard drive or memory, or an external storage device, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card. The memory 320 can also include both internal and external storage units. The memory 320 is used to store computer programs and other programs and data required by the electronic device. The memory 320 can also be used to temporarily store data that has been output or will be output.

[0105] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0106] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0107] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0108] In the embodiments provided by this invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0109] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0110] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0111] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0112] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for detecting and determining the code on the front page of a medical record, characterized in that, include: Entity relations are extracted from the disease descriptions corresponding to all standard codes in the International Classification of Diseases to obtain multiple first triples. An initial knowledge graph is constructed based on the multiple first triples and the corresponding standard codes. The entity relations include etiology, location, pathology, symptoms, stage, type, and acute / chronic. Multiple original diagnosis names and corresponding original codes are obtained from multiple historical medical record homepages. Common codes are determined based on the number of each original code. The original diagnosis names corresponding to the common codes are normalized to obtain normalized diagnosis names. Entity relations are extracted from all normalized diagnosis names to obtain multiple second triples. A knowledge graph is obtained based on the multiple second triples and the initial knowledge graph. The knowledge graph is injected into the trained prediction model to obtain the target prediction model; Obtain multiple diagnostic names and corresponding codes from the front page of the medical record to be tested; predict each diagnostic name based on the target prediction model to obtain multiple candidate codes corresponding to each diagnostic name; determine the final code corresponding to each diagnostic name based on the code corresponding to each diagnostic name and the multiple candidate codes. The step of predicting each diagnostic name to be detected based on the target prediction model to obtain multiple candidate codes corresponding to each diagnostic name to be detected includes: The names of the diagnoses to be tested are standardized to obtain standardized names of the diagnoses to be tested. Based on the target prediction model, the names of the diagnostic tests to be detected in each standard are predicted, and multiple candidate codes corresponding to the names of the diagnostic tests to be detected in each standard are obtained. The step of determining the final code corresponding to each diagnostic name to be detected based on the code to be detected and the corresponding multiple candidate codes includes: For each of the specified diagnostic names to be tested, perform the following steps: Determine whether the code to be detected corresponding to the diagnostic name to be detected is consistent with any of the multiple candidate codes; If so, then the corresponding code to be detected is determined to be the final code; If not, then based on the International Classification of Diseases, it is determined whether the corresponding code to be detected has a hierarchical relationship with any of the corresponding candidate codes. If so, then that candidate code is taken as the final code. The process of determining commonly used codes based on the quantity of each original code includes: Determine the number of each original code and the department corresponding to each original code; each department corresponds to a preset department threshold; If the number of original codes is greater than the preset department threshold corresponding to the original code, then the original code is determined to be a commonly used code.

2. The method for detecting and determining the medical record cover page code according to claim 1, characterized in that, The original diagnostic names corresponding to the commonly used codes are normalized to obtain normalized diagnostic names. Entity relations are extracted from all normalized diagnostic names to obtain multiple second triplets, including: The original diagnostic names corresponding to the commonly used codes are standardized to obtain standardized diagnostic names; The standardized diagnostic names are deduplicated. Semantic analysis was performed on all standardized diagnostic names after deduplication. Entity relations are extracted from all the standardized diagnostic names after semantic analysis, resulting in multiple second triplets.

3. The method for detecting and determining the medical record cover page code according to claim 1, characterized in that, The step of injecting the knowledge graph into the trained prediction model to obtain the target prediction model includes: The entity vectors corresponding to the triples in the knowledge graph are calculated based on the TransE algorithm. The entity vector is embedded into the trained prediction model to obtain the target prediction model.

4. The method for detecting and determining the medical record cover page code according to any one of claims 1 to 3, characterized in that, The method further includes: Based on the priority identifier of the preset standard code, the priority of the final code corresponding to each diagnosis name to be tested in the front page of the medical record to be tested is determined; The diagnostic name corresponding to the final code with the highest priority is determined as the primary diagnosis.

5. A device for detecting and determining the code on the front page of a medical record, characterized in that, include: The first construction module is used to extract entity relations from the disease descriptions corresponding to all standard codes in the International Classification of Diseases, to obtain multiple first triples, and to construct an initial knowledge graph based on the multiple first triples and the corresponding standard codes; wherein, the entity relations include etiology, location, pathology, symptoms, stage, type and acute and chronic; The second construction module is used to obtain multiple original diagnosis names and corresponding original codes from multiple historical medical record homepages, determine commonly used codes based on the number of each original code, perform normalization processing on the original diagnosis names corresponding to the commonly used codes to obtain normalized diagnosis names, extract entity relations from all normalized diagnosis names to obtain multiple second triples, and obtain a knowledge graph based on the multiple second triples and the initial knowledge graph. The first determining module is used to inject the knowledge graph into the trained prediction model to obtain the target prediction model; The second determining module is used to obtain multiple diagnostic names to be detected and their corresponding codes to be detected from the front page of the medical record to be detected; predict each diagnostic name to be detected based on the target prediction model to obtain multiple candidate codes corresponding to each diagnostic name to be detected; and determine the final code corresponding to each diagnostic name to be detected based on the code to be detected and the multiple candidate codes. The second determining module is further configured to perform normalization processing on each diagnostic name to be detected to obtain each normalized diagnostic name to be detected; and to predict each normalized diagnostic name to be detected based on the target prediction model to obtain multiple candidate codes corresponding to each normalized diagnostic name to be detected. And, for each of the specified diagnostic names to be tested, perform the following steps: Determine whether the code corresponding to the diagnostic name to be tested is consistent with any of the multiple candidate codes; if yes, determine the corresponding code to be tested as the final code; if no, determine whether the corresponding code to be tested has a hierarchical relationship with any of the multiple candidate codes based on the International Classification of Diseases; if yes, take that candidate code as the final code. The second construction module is further configured to determine the number of each original code and the department corresponding to each original code; each department corresponds to a preset department threshold; if the number of the original codes is greater than the preset department threshold corresponding to the original code, then the original code is determined to be a commonly used code.

6. An electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for detecting and determining the medical record cover page code as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for detecting and determining the medical record cover page code as described in any one of claims 1 to 4.

Images