Hospitalization medical record adaptive generation method and device, electronic equipment and storage medium

Abstract

The present application relates to a hospitalization medical record adaptive generation method and device, electronic equipment and storage medium, comprising: obtaining a medical record data set, classifying the medical record data in the medical record data set according to the diagnosis information of the department where the doctor is located, and generating standard medical record text corresponding to each category of medical record data. All medical documents under the standard medical record text are split by configuring scripts, and multiple paragraph contents corresponding to the medical record written by the doctor are extracted. A paragraph association dictionary is constructed according to the paragraph content. The paragraph association dictionary, the paragraph content and the paragraph name list corresponding to the multiple paragraph contents are fused to generate the medical record writing configuration of the doctor individual. The large model is called to iterate the multiple paragraph contents corresponding to the patient medical record data according to the paragraph name list, and the document paragraph conforming to the medical record writing configuration is generated. By analyzing the written medical record of the doctor individual, individualized configuration information is generated, the flexibility of medical record generation is higher, and it is more in line with the writing style of doctors.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical assistance, in particular to a hospitalization medical record adaptive generation method and device, electronic equipment and storage medium. BACKGROUND

[0002] With the development of science and technology, the internal informatization of hospitals is becoming more and more comprehensive, and the writing and management requirements of medical record content are also becoming higher and higher. Usually, a doctor individual receives dozens or even hundreds of patients a day. If detailed medical records are written for each patient, it is difficult to alleviate the pressure of doctor writing despite the management of documents through informatization system. In order to solve this technical problem, more hospital information systems set templates for doctors to templateize the documents, and write by filling in the blank content of the template to improve the writing efficiency.

[0003] However, with the popularization of large models in various industries, there are many medical record generation schemes in the medical industry to help doctors directly generate document content. Compared with the template setting method, the template setting method not only needs to design perfect medical record templates in advance, but also is difficult to adjust the writing content to adapt to some special situations once the template is fixed, thereby reducing the coherence of medical record writing content and reducing the degree of intelligence. In addition, after the popularization of large models in the medical industry, the most common scene is medical record generation, but there are many differences between the output content of the large model and the writing habits of the individual doctors. These differences lead to the illusion of the output content of the large model, and the output flexibility needs to be further improved. SUMMARY

[0004] Therefore, it is necessary to provide a hospitalization medical record adaptive generation method, device, electronic equipment and storage medium with high flexibility and adaptability to different writing habits of individual doctors to solve the above technical problems.

[0005] The present application provides a hospitalization medical record adaptive generation method, which comprises:

[0006] Obtain a set of medical record data to be classified, and classify the medical record data in the set of medical record data according to the diagnosis information of the department where the individual doctor is located, to generate standard medical record texts corresponding to medical record data of each category;

[0007] Split all medical record documents under the standard medical record text through a configuration script, to extract a plurality of paragraph contents corresponding to the medical record written by the individual doctor, and construct a paragraph association dictionary according to the paragraph contents;

[0008] Fuse the paragraph association dictionary, paragraph contents and a paragraph name list corresponding to the plurality of paragraph contents to generate a medical record writing configuration of the individual doctor;

[0009] The large model is called to perform iterative processing on the multiple paragraph contents corresponding to the patient medical record data according to the paragraph name list, so as to generate the medical record paragraphs conforming to the medical record writing configuration.

[0010] In one of the embodiments, the medical record data set to be classified is obtained, and the medical record data in the medical record data set is classified according to the diagnosis information of the department where the individual doctor is located, so as to generate the standard medical record text corresponding to the medical record data of each category, including:

[0011] The patient diagnosis information and the medical record data are obtained, and the medical record data is classified according to the department where the individual doctor corresponding to the patient diagnosis information is located, so as to determine the medical record category corresponding to the patient diagnosis information and the medical record data;

[0012] The medical record data satisfying the preset requirement of medical record completeness is selected from the multiple medical record data under the first medical record category, so as to generate the standard medical record text corresponding to the first medical record category;

[0013] The medical record data satisfying the preset requirement of medical record completeness is the medical record data with the most number of patient admission and discharge records, and the time stamp of the patient discharge record is the closest to the current time.

[0014] In one of the embodiments, the standard medical record text is split by the configuration script to extract the multiple paragraph contents corresponding to the medical record written by the individual doctor, and a paragraph association dictionary is constructed according to the paragraph contents, including:

[0015] An ordered paragraph name list configured is obtained, and all the medical record texts under the standard medical record text are string segmented according to the ordered paragraph name list, to obtain multiple strings with different lengths;

[0016] When the length of the list obtained after segmentation is a first value, it is determined that the paragraph content corresponding to the string is empty.

[0017] In one of the embodiments, the standard medical record text is split by the configuration script to extract the multiple paragraph contents corresponding to the medical record written by the individual doctor, and a paragraph association dictionary is constructed according to the paragraph contents, including:

[0018] When the length of the list obtained after segmentation is a second value, it is determined that the first part of the paragraph content corresponding to the string is the paragraph content of the last non-empty iteration, and the second part is updated to the remaining texts after string segmentation;

[0019] When the length of the list obtained after segmentation exceeds the second value, a large model is called to perform content merging on adjacent strings according to a configured merging prompt dialog box, so as to generate the paragraph content with the string length of the second value.

[0020] The first value and the second value are different values set, and the second value is greater than the first value. The first part is located before the second part in the paragraph content.

[0021] In one embodiment, the method further comprises:

[0022] Obtaining the writing time of each paragraph content after splitting, and calling a large model to determine whether the correlation between the paragraph content after splitting and the medical record document before splitting exceeds a set threshold according to the content of each medical record document before splitting;

[0023] When the correlation between the paragraph content after splitting and the medical record document before splitting exceeds the set threshold, a dependency relationship between the paragraph content after splitting and the medical record document before splitting is constructed to generate the paragraph correlation dictionary.

[0024] In one embodiment, the method further comprises:

[0025] Obtaining patient diagnosis information, and receiving a paragraph generation request when the patient diagnosis information is included in the medical record writing configuration of the doctor;

[0026] In response to the paragraph generation request, calling a large model to obtain reference text and corresponding standard medical record text based on the paragraph correlation dictionary according to the configured ordered paragraph name list;

[0027] Concatenating the reference text, the paragraph content after splitting, and the medical record text into a prompt dialog box to generate the document paragraph.

[0028] In one embodiment, the method further comprises:

[0029] Calling an entity recognition model CRF to perform entity recognition on the document paragraph and the reference text to generate an entity list, and determining whether the entities in the document paragraph completely belong to the entities in the reference text according to the entity list;

[0030] When the entities in the document paragraph do not completely belong to the entities in the document paragraph, calling a correction model based on the reference text to correct the document paragraph to obtain a corrected document paragraph.

[0031] The application further provides an inpatient medical record adaptive generation device for implementing the inpatient medical record adaptive generation method.

[0032] The medical record classification module is configured to obtain a set of medical record data to be classified, and classify medical record data in the set of medical record data according to diagnosis information of a department where a doctor is located, to generate standard medical record texts corresponding to medical record data of each category.

[0033] The medical record document splitting module is configured to split all medical record documents under the standard medical record texts by configuring scripts, to extract a plurality of paragraph contents corresponding to medical records written by the doctor, and construct a paragraph association dictionary according to the paragraph contents.

[0034] The medical record writing configuration module is configured to fuse the paragraph association dictionary, the paragraph contents, and a paragraph name list corresponding to the plurality of paragraph contents, to generate a medical record writing configuration of the doctor.

[0035] The document paragraph generation module is configured to call a large model to perform cyclic iteration on the plurality of paragraph contents corresponding to the patient medical record data according to the paragraph name list, to generate a document paragraph conforming to the medical record writing configuration.

[0036] The application further provides an electronic device including a memory and a processor, the memory storing a computer program, and the processor implementing the inpatient medical record adaptive generation method when executing the computer program.

[0037] The application further provides a computer storage medium storing a computer program, and the computer program implementing the inpatient medical record adaptive generation method when executed by a processor.

[0038] The application further provides a computer program product including a computer program, and the computer program implementing the inpatient medical record adaptive generation method when executed by a processor.

[0039] The hospitalization medical record adaptive generation method, device, electronic equipment and storage medium provided by the present application, by obtaining a set of medical record data to be classified, and classifying the medical record data in the set of medical record data according to the diagnosis information of the department where the doctor is located, to generate the standard medical record text corresponding to each category of medical record data. Then, all medical documents under the standard medical record text are split by configuring the script, to extract a plurality of paragraph contents corresponding to the medical record written by the doctor, and construct a paragraph association dictionary according to the paragraph contents. Then, the paragraph association dictionary, the paragraph contents and the paragraph name list corresponding to the plurality of paragraph contents are fused to generate the medical record writing configuration of the doctor. Finally, in the medical record writing, the large model is called to iterate the plurality of paragraph contents corresponding to the patient medical record data according to the paragraph name list, to generate the document paragraph meeting the medical record writing configuration. The method analyzes the written medical record of the doctor, and generates the configuration information of individualized medical record writing in batches. In the subsequent medical record writing process, the medical record content meeting the writing habit of the doctor can be generated according to the writing configuration of the doctor, which effectively avoids the illusion content output by the large model due to individual differences, and has high flexibility. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0041] Figure 1 One of the flowcharts of the hospitalization medical record adaptive generation method provided by the present application;

[0042] Figure 2 The flowchart of the automatic offline processing of the hospitalization medical record adaptive generation method in the specific embodiment provided by the present application to generate personalized information configuration;

[0043] Figure 3 The flowchart of the personalized writing document information generation of the hospitalization medical record adaptive generation method in the specific embodiment provided by the present application;

[0044] Figure 4 The second flowchart of the hospitalization medical record adaptive generation method provided by the present application;

[0045] Figure 5 The third flowchart of the hospitalization medical record adaptive generation method provided by the present application;

[0046] Figure 6 The fourth flowchart of the hospitalization medical record adaptive generation method provided by the present application;

[0047] Figure 7 Figure 5 is a schematic diagram of the process of the inpatient medical record adaptive generation method provided by the present application;

[0048] Figure 8 Figure 6 is a schematic diagram of the process of the inpatient medical record adaptive generation method provided by the present application;

[0049] Figure 9 Figure 7 is a schematic diagram of the process of the inpatient medical record adaptive generation method provided by the present application;

[0050] Figure 10 Figure 8 is a schematic diagram of the structure of the inpatient medical record adaptive generation device provided by the present application;

[0051] Figure 11 Figure 9 is an internal structure diagram of the computer device provided by the present application. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0053] The inpatient medical record adaptive generation method, device, electronic device and storage medium of the present application will be described below. Figures 1-11 The inpatient medical record adaptive generation method, device, electronic device and storage medium of the present application will be described below.

[0054] As shown in Figure 1 one embodiment, an inpatient medical record adaptive generation method includes the following steps:

[0055] In step S110, a set of medical record data to be classified is obtained, and the medical record data in the set of medical record data is classified according to the diagnosis information of the department where the individual doctor is located, to generate a standard medical record text corresponding to each category of medical record data.

[0056] Specifically, the server obtains a set of medical record data to be classified of all patients in the department where the individual doctor is located, and classifies the medical record data in the set of medical record data according to the diagnosis information of the department where the individual doctor is located corresponding to each medical record data, to generate a medical record writing model corresponding to each category of medical record data, i.e. a standard medical record text.

[0057] As shown in Figure 2 and Figure 3 In a specific embodiment, the inpatient medical record adaptive generation method provided by the present application includes two stages of automatic offline processing and document information generation.

[0058] In the automatic offline processing stage, first, the medical records are classified according to the current diagnosis of the patient, and the category ranges from the common diagnosis information of the department where the doctor is located, for example, the categories related to neurology department include cerebral infarction, cerebral hemorrhage, meningitis and the like. Secondly, the medical records of the same category are screened according to the completeness of the medical records, and one relatively complete medical record in the category is selected as the writing model, that is, the standard medical record text.

[0059] In the medical record screening process, the selected medical record must have a complete medical record writing life cycle, that is, it must have complete admission records and complete discharge records. On the premise of having complete admission records and complete discharge records, according to the number of document types involved in the medical record, each additional document type involved in the medical record adds one point, and the medical record with the highest final score exists in multiple copies. Then, the scores of these medical records are added to the time stamp of the discharge record time, that is, the discharge record closest to the current time is given priority as the model of the same diagnosis category.

[0060] In step S120, all medical record documents under the standard medical record text are split by the configuration script to extract multiple paragraph contents corresponding to the medical record written by the doctor, and a paragraph association dictionary is constructed according to the paragraph contents.

[0061] Specifically, the server splits all medical record documents under the standard medical record text by the configuration script, obtains multiple paragraph contents after splitting, and constructs a paragraph association dictionary integrating the association relationship between each paragraph content and the medical record document according to the multiple paragraph contents after splitting.

[0062] In combination with the embodiments shown in Figure 2 and Figure 3 In a specific embodiment, the hospital medical record self-adaptive generation method provided by the present application splits the paragraph contents of all documents of the model by a script, and the splitting script is part of the process workflow configured for offline extraction of personalized information of the doctor. In the splitting process, first, the document T to be split and the ordered paragraph name list K in the document are obtained, the document T is split according to the paragraph name list (string split), and then the length of the list obtained after splitting is checked. If the length of the list after splitting is 1, it is determined that the information content corresponding to the paragraph is empty; if the length of the list after splitting is 2, the front part of the corresponding paragraph content is the content of the last non-empty iteration paragraph, and the rear part is updated as the remaining document; if the length of the list after splitting is greater than 2, a large model is called to merge the contents, and the merge prompt is used to control the large model to output the only split point of the two continuous paragraphs, and finally the list length is 2. The split content is merged, the front part of which is the content of the last non-iterative paragraph, and the rear part is updated as the remaining document. Finally, the split paragraph information dictionary M is obtained, which has the format {“paragraph name 1”:“paragraph content 1”,“paragraph name 2”:“paragraph content 2”...}.

[0063] In this embodiment, the paragraph content v of the segmented paragraph information dictionary M is processed in sequence, the writing time A of the current paragraph v is obtained first, the medical record document set Z before the time A is obtained, where each document content is z', and the large model is used to judge whether the current paragraph information v refers to z' to build the association relationship one by one, that is, the association relationship dictionary R is obtained, and the specific format is as follows:

[0064] {

[0065] "Paragraph name 1": [["associated document name a", "associated document paragraph name a1"], ["associated document name b", "associated document paragraph name b1"]],

[0066] "Paragraph name 2": [["associated document name c", "associated document paragraph name c1"], ["associated document name c", "associated document paragraph name c2"]] ...

[0067] }

[0068] In this embodiment, it is assumed that the current document to be generated is "case characteristics" of "first course record", and the example content generated by the above process is "chief complaint", "history of present illness", "physical examination" and "auxiliary examination" in "admission record" to obtain dependent information, and the expression of the association relationship dictionary R is:

[0069] {

[0070] "Medical record characteristics": [["admission record", "chief complaint"], ["admission record", "history of present illness"], ["admission record", "physical examination"], ["admission record", "auxiliary examination"]], ...

[0071] }

[0072] Step S130, combine the paragraph association dictionary, paragraph content and paragraph name list corresponding to multiple paragraph contents to generate a medical record writing configuration of a doctor.

[0073] Specifically, the server combines the aforementioned paragraph association dictionary with the split paragraph content and the paragraph name list corresponding to the split multiple paragraph contents, and obtains the medical record writing configuration of the doctor after combination.

[0074] In combination with Figure 2 and Figure 3As shown, in specific embodiments, the hospitalization medical record adaptive generation method provided by the present application combines information document positions and paragraph contents to obtain personalized configurations, that is, each document generation task has an ordered paragraph name list K, a paragraph information dictionary M, and an association relationship dictionary R, which are combined to form the most basic personalized configuration. Multiple document generation tasks have multiple different personalized configurations described above. The basic outline of the personalized configuration information of a single doctor is as follows:

[0075] {

[0076] "diagnosis 1": {"diagnosis 1 generation document task a": [K_1a, M_1a, R_1a], "diagnosis 1 generation document task b": [K_1b, M_1b, R_1b]},

[0077] "diagnosis 2": {"diagnosis 2 generation document task a": [K_2a, M_2a, R_2a], "diagnosis 2 generation document task b": [K_2b, M2_b, R_2b]}, ...

[0078] }

[0079] Step S140, calling the large model to perform cyclic iteration on the multiple paragraph contents corresponding to the patient medical record data according to the paragraph name list to generate document paragraphs conforming to the medical record writing configuration.

[0080] Specifically, the server calls the large model to perform cyclic iteration on the multiple paragraph contents corresponding to the current patient medical record data according to the order of the paragraph name list configured in the configuration script, and finally generates document paragraphs conforming to the medical record writing configuration of the individual doctor.

[0081] In combination with Figure 2 and Figure 3 As shown, in specific embodiments, the hospitalization medical record adaptive generation method provided by the present application, in the document information generation stage, first obtains personalized information configuration based on the department of the patient and the current diagnosis.

[0082] In this embodiment, the personalized configuration generated by the automatic offline processing stage for each doctor is divided based on diagnosis, as follows: Figure 3As shown, first, the patient diagnosis information is obtained, and if it can be obtained and there is such diagnosis information in the doctor's personalized configuration, the information content under this diagnosis is used, that is, the content format is obtained: {"diagnosis 1 generates document task a": [K_1a, M_1a, R_1a], "diagnosis 1 generates document task b": [K_1b, M_1b, R_1b]}; if there is no match, the department-level unified template without dividing disease types is used, (the information format of the unified template is consistent with the doctor's personalized configuration) that is, the configuration cf is obtained, the format is {"base generates document task a": [K_1a, M_1a, R_1a], "base generates document task b": [K_1b, M_1b, R_1b]}.

[0083] After that, it is confirmed whether the current document type task is in the configuration cf, if not, it does not trigger generation; if so, the core configuration cfg is obtained, the format is [K, M, R]. The ordered paragraph name list K is iteratively generated document paragraph gen_section, in the process of generating document paragraph, according to the dependent document association relationship dictionary R, the patient reference information original text content is obtained, and then the sample is obtained from the paragraph information dictionary. Based on the currently available gen_section, content, sample, the stylized prompt is spliced, and the large model is called to generate the paragraph content. Then, the entity recognition is performed on the generated content gen_content and the reference content content. (such as: diagnosis, symptoms, tests, examinations, surgeries, drugs) The entity content of the generated information must appear in the entity content of the reference information, if not, the potential generation problem is recorded.

[0084] In this embodiment, the named entity recognition model mainly automatically identifies and classifies specific entities (such as: name, place name, organization name, time, etc.) in the text. In the medical scene, through the training of medical annotation information, the entities mainly include diagnosis, symptoms, tests, examinations, surgeries, and drugs. The named entity recognition model analyzes the context and semantic features of the text, labels the key information in the sentence, and represents CRF(sentence) as a call to the named entity model, and the result format is: [[word1, type1], [word2, type2], [word3, type3]...].

[0085] Specifically, the content generation call entity recognition model CRF(gen_content) is used to obtain the result gen_content_result, and the reference content call entity recognition model CRF(content) is used to obtain the result content_result. The entity words of “diagnosis”, “symptoms”, “examination”, “inspection”, “surgery” and “drugs” in gen_content_result are obtained to obtain the key list gen_content_key_word_list of the generated content. Similarly, the entity words of “diagnosis”, “symptoms”, “examination”, “inspection”, “surgery” and “drugs” in content_result are obtained to obtain the key list content_key_word_list of the reference content. If the word w1 in gen_content_key_word_list does not appear in content_key_word_list, record the problem in error_message: “w1” does not appear in the patient medical record content, and error_message will collect all potential illusion information.

[0086] Finally, based on the recorded potential illusion information error_message, the medical record reference information content and the first output generated content gent_content, the error correction prompt is constructed, and the large model is called again based on the prompt, and the error corrected generated content is obtained.

[0087] The above-mentioned hospitalization medical record adaptive generation method, by obtaining a set of medical record data to be classified, and classifying the medical record data in the set of medical record data according to the diagnosis information of the department where the doctor individual is located, to generate standard medical record texts corresponding to medical record data of each category. Then, all medical documents under the standard medical record text are split by configuring a script to extract a plurality of paragraph contents corresponding to the medical record written by the doctor individual, and a paragraph association dictionary is constructed according to the paragraph contents. Then, the paragraph association dictionary, the paragraph contents and the paragraph name list corresponding to the plurality of paragraph contents are fused to generate a medical record writing configuration of the doctor individual. Finally, in the medical record writing, the large model is called to iterate the plurality of paragraph contents corresponding to the patient medical record data according to the paragraph name list to generate a document paragraph in accordance with the medical record writing configuration. This method analyzes the written medical record of the doctor individual, and generates batched configuration information of individualized medical record writing. In the subsequent medical record writing process, the medical record content in accordance with the writing habit of the doctor individual can be generated according to the writing configuration of the doctor individual, effectively avoiding the illusion content output by the large model due to individual differences, and has high flexibility.

[0088] As Figure 4As shown, in one embodiment, the adaptive inpatient medical record generation method provided by the present invention includes the following steps in step S110:

[0089] Step S111: Obtain patient diagnosis information and medical record data, and classify the medical record data according to the department of the individual doctor corresponding to the patient diagnosis information to determine the medical record category corresponding to the patient diagnosis information and medical record data.

[0090] Step S112: Select medical record data whose completeness meets the preset requirements from multiple medical record data under the first medical record category, so as to generate the standard medical record text corresponding to the first medical record category.

[0091] Among them, the medical record data that meets the preset requirements for completeness is the medical record data that has patient admission and discharge records, the largest number of medical record categories, and the patient discharge record timestamp is the most recent to the current time.

[0092] like Figure 5 As shown, in one embodiment, the adaptive inpatient medical record generation method provided by the present invention includes the following steps in step S120:

[0093] Step S121: Obtain the configured ordered paragraph name list, and perform string segmentation on all medical record documents under the standard medical record text according to the ordered paragraph name list to obtain multiple strings of different lengths.

[0094] Step S122: When the length of the list obtained after segmentation is the first value, it is determined that the paragraph content corresponding to the string is empty.

[0095] like Figure 6 As shown, in one embodiment, the adaptive inpatient medical record generation method provided by the present invention further includes the following steps in step S120:

[0096] Step S123: When the length of the list obtained after segmentation is the second value, determine that the first part of the paragraph content corresponding to the string is the paragraph content of the most recent non-empty iteration, and update the second part to the remaining text after string segmentation.

[0097] Step S124: When the length of the list obtained after segmentation exceeds the second value, the large model is called to merge the content of adjacent strings according to the configured merge prompt dialog box, so as to generate paragraph content with a string length of the second value.

[0098] The first and second values ​​are different values, and the second value is greater than the first value. The first part is located before the second part in the paragraph content.

[0099] like Figure 7As shown in the figure, in one embodiment, the inpatient medical record adaptive generation method provided by the application specifically further includes the following steps:

[0100] Step S125, the writing time of each paragraph content after splitting is obtained, and whether the correlation degree between the paragraph content after splitting and the medical record before splitting exceeds the set threshold is judged according to the content of each medical record before splitting by calling the large model.

[0101] Step S126, when the correlation degree between the paragraph content after splitting and the medical record before splitting exceeds the set threshold, the dependency relationship between the paragraph content after splitting and the medical record before splitting is constructed to generate a paragraph correlation dictionary.

[0102] As shown in the figure, Figure 8 in one embodiment, the inpatient medical record adaptive generation method provided by the application specifically includes the following steps:

[0103] Step S131, the patient diagnosis information is obtained, and when the patient diagnosis information is included in the medical record writing configuration of the individual doctor, the paragraph generation request is received.

[0104] Step S132, in response to the paragraph generation request, the large model is called to obtain the reference text and the corresponding standard medical record text based on the paragraph correlation dictionary according to the configured ordered paragraph name list.

[0105] Step S133, the reference text, the paragraph content after splitting and the medical record text are spliced into a prompt dialog box to generate a medical record paragraph.

[0106] As shown in the figure, Figure 9 in one embodiment, the inpatient medical record adaptive generation method provided by the application further includes the following steps:

[0107] Step S910, the entity recognition model CRF is called to perform entity recognition on the medical record paragraph and the reference text to generate an entity list, and whether the entity in the medical record paragraph completely belongs to the entity in the reference text is judged according to the entity list.

[0108] Step S920, when the entity in the medical record paragraph does not completely belong to the entity in the medical record paragraph, the medical record paragraph is corrected and modified based on the reference text to obtain a modified medical record paragraph.

[0109] The inpatient medical record adaptive generation device provided by the application is described below, and the inpatient medical record adaptive generation device described below can be correspondingly referred to the inpatient medical record adaptive generation method described above.

[0110] As shown in the figure, Figure 10As shown, in one embodiment, an inpatient medical record adaptive generation device includes a medical record classification module 1010, a medical record document splitting module 1020, a medical record writing configuration module 1030, and a document paragraph generation module 1040.

[0111] The medical record classification module 1010 is configured to obtain a set of medical record data to be classified, and classify the medical record data in the set of medical record data according to diagnosis information of a department where a doctor individual is located, to generate standard medical record texts corresponding to medical record data of each category.

[0112] The medical record document splitting module 1020 is configured to split all medical record documents under the standard medical record texts through a configured script, to extract a plurality of paragraph contents corresponding to the medical record written by the doctor individual, and construct a paragraph association dictionary according to the paragraph contents.

[0113] The medical record writing configuration module 1030 is configured to fuse the paragraph association dictionary, the paragraph contents, and a paragraph name list corresponding to the plurality of paragraph contents, to generate a medical record writing configuration of the doctor individual.

[0114] The document paragraph generation module 1040 is configured to call a large model to perform cyclic iteration on the plurality of paragraph contents corresponding to the patient medical record data according to the paragraph name list, to generate a document paragraph conforming to the medical record writing configuration.

[0115] In the present embodiment, the inpatient medical record adaptive generation device provided by the present application is configured such that the medical record classification module 1010 is specifically configured to:

[0116] obtain patient diagnosis information and medical record data, and classify the medical record data according to a department where a doctor individual corresponding to the patient diagnosis information is located, to determine a medical record category corresponding to the patient diagnosis information and the medical record data.

[0117] select medical record data with a medical record completeness satisfying a preset requirement from a plurality of medical record data under a first medical record category, to generate a standard medical record text corresponding to the first medical record category.

[0118] Among them, the medical record data with the medical record completeness satisfying the preset requirement is medical record data with patient admission and discharge records, the largest number of medical record categories, and the closest time stamp of patient discharge record time to the current time.

[0119] In the present embodiment, the inpatient medical record adaptive generation device provided by the present application is configured such that the medical record document splitting module 1020 is specifically configured to:

[0120] obtain a configured ordered paragraph name list, and perform string segmentation on all medical record documents under the standard medical record texts according to the ordered paragraph name list, to obtain a plurality of strings with different lengths.

[0121] When the length of the list obtained after the segmentation is a first value, it is determined that the paragraph content corresponding to the string is empty.

[0122] In this embodiment, the inpatient medical record adaptive generation device provided by the application is provided, and the medical record document splitting module 1020 is specifically further used for:

[0123] When the length of the list obtained after the segmentation is a second value, it is determined that the first part of the paragraph content corresponding to the string is the paragraph content of the last non-empty iteration, and the second part is updated to the remaining document after the string segmentation.

[0124] When the length of the list obtained after the segmentation exceeds the second value, the adjacent strings are merged according to the configured merging prompt dialog box by calling the large model, so as to generate the paragraph content with the string length of the second value.

[0125] Wherein, the first value and the second value are different values set, and the second value is greater than the first value, and the first part is located before the second part in the paragraph content.

[0126] In this embodiment, the inpatient medical record adaptive generation device provided by the application is provided, and the medical record document splitting module 1020 is specifically further used for:

[0127] Obtain the document writing time corresponding to each paragraph content after splitting, and according to the content of each medical record document before splitting, call the large model to determine whether the correlation degree between the paragraph content after splitting and the medical record document before splitting exceeds the set threshold.

[0128] When the correlation degree between the paragraph content after splitting and the medical record document before splitting exceeds the set threshold, the dependency relationship between the paragraph content after splitting and the medical record document before splitting is constructed to generate a paragraph correlation dictionary.

[0129] In this embodiment, the inpatient medical record adaptive generation device provided by the application is provided, and the medical record writing configuration module 1030 is specifically used for:

[0130] Obtain the patient diagnosis information, and when the patient diagnosis information is included in the medical record writing configuration of the doctor individual, receive a paragraph generation request.

[0131] In response to the paragraph generation request, the large model is called to obtain reference text and corresponding standard medical record text based on the paragraph correlation dictionary according to the configured ordered paragraph name list.

[0132] The reference text, the paragraph content after splitting and the medical record text are spliced into a prompt dialog box to generate a document paragraph.

[0133] In this embodiment, the inpatient medical record adaptive generation device provided by the application further comprises a document paragraph correction module, which is used for:

[0134] The entity recognition model CRF is called to perform entity recognition on the document paragraph and the reference text to generate an entity list, and to determine whether the entity in the document paragraph completely belongs to the entity in the reference text according to the entity list.

[0135] When the entity in the document paragraph does not completely belong to the entity in the document paragraph, error correction of the document paragraph based on the reference text is called to obtain a corrected document paragraph.

[0136] Figure 11 An entity structure diagram of an electronic device is exemplified, which can be a smart terminal, and an internal structure diagram thereof can be as shown in Figure 11 The electronic device includes a processor, a memory and a network interface connected through a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the electronic device is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement a hospitalization medical record adaptive generation method, which includes:

[0137] Obtain a medical record data set to be classified, and classify medical record data in the medical record data set according to diagnosis information of a department where a doctor is located to generate standard medical record texts corresponding to medical record data of each category;

[0138] All medical record documents under the standard medical record text are split by a configuration script to extract a plurality of paragraph contents corresponding to medical record writing of the doctor, and a paragraph association dictionary is constructed according to the paragraph contents;

[0139] The paragraph association dictionary, the paragraph contents and a paragraph name list corresponding to the plurality of paragraph contents are fused to generate a medical record writing configuration of the doctor;

[0140] A large model is called to perform cyclic iteration on the plurality of paragraph contents corresponding to the patient medical record data according to the paragraph name list to generate a document paragraph conforming to the medical record writing configuration.

[0141] Those skilled in the art can understand that Figure 11 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0142] In another aspect, the present application also provides a computer storage medium storing a computer program, the computer program being executed by a processor to implement a hospitalization medical record adaptive generation method, the method comprising:

[0143] obtaining a set of medical record data to be classified, and classifying medical record data in the set of medical record data according to diagnosis information of a department where a doctor individual is located, to generate standard medical record texts corresponding to medical record data of each category;

[0144] splitting all medical record documents under the standard medical record texts through a configuration script, to extract a plurality of paragraph contents corresponding to medical records written by the doctor individual, and constructing a paragraph association dictionary according to the paragraph contents;

[0145] fusing the paragraph association dictionary, the paragraph contents, and a paragraph name list corresponding to the plurality of paragraph contents, to generate a medical record writing configuration of the doctor individual;

[0146] calling a large model to perform cyclic iteration on the plurality of paragraph contents corresponding to the patient medical record data according to the paragraph name list, to generate document paragraphs conforming to the medical record writing configuration.

[0147] In another aspect, a computer program product or computer program is provided, the computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor implements a hospitalization medical record adaptive generation method when executing the computer instructions, the method comprising:

[0148] obtaining a set of medical record data to be classified, and classifying medical record data in the set of medical record data according to diagnosis information of a department where a doctor individual is located, to generate standard medical record texts corresponding to medical record data of each category;

[0149] splitting all medical record documents under the standard medical record texts through a configuration script, to extract a plurality of paragraph contents corresponding to medical records written by the doctor individual, and constructing a paragraph association dictionary according to the paragraph contents;

[0150] fusing the paragraph association dictionary, the paragraph contents, and a paragraph name list corresponding to the plurality of paragraph contents, to generate a medical record writing configuration of the doctor individual;

[0151] calling a large model to perform cyclic iteration on the plurality of paragraph contents corresponding to the patient medical record data according to the paragraph name list, to generate document paragraphs conforming to the medical record writing configuration.

[0152] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory.

[0153] By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0154] The technical features of the above embodiments can be combined in any way. In order to make the description simple, not all possible combinations of the technical features in the above embodiments are described, but as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0155] The above embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the scope of protection of the patent of the present application should be subject to the appended claims.

Claims

1. A method for adaptive generation of inpatient medical record, characterized in that, The method comprises: obtaining a set of medical record data to be classified, and classifying medical record data in the set of medical record data according to diagnosis information of a department where a doctor individual is located, to generate standard medical record texts corresponding to medical record data of each category; splitting all medical record documents under the standard medical record texts through a configuration script, to extract a plurality of paragraph contents corresponding to medical records written by the doctor individual, and constructing a paragraph association dictionary according to the paragraph contents; fusing the paragraph association dictionary, paragraph contents, and a paragraph name list corresponding to the plurality of paragraph contents, to generate medical record writing configurations of the doctor individual; calling a large model to perform cyclic iteration on a plurality of paragraph contents corresponding to patient medical record data according to the paragraph name list, to generate document paragraphs conforming to the medical record writing configurations; wherein the splitting all medical record documents under the standard medical record texts through the configuration script, to extract the plurality of paragraph contents corresponding to the medical records written by the doctor individual, and constructing the paragraph association dictionary according to the paragraph contents, comprises: obtaining a configured ordered paragraph name list, and performing string segmentation on all medical record documents under the standard medical record texts according to the ordered paragraph name list, to obtain a plurality of strings of different lengths; when the length of the list obtained after segmentation is a first value, determining that the paragraph content corresponding to the string is empty; when the length of the list obtained after segmentation is a second value, determining that a first part of the paragraph content corresponding to the string is paragraph content of a last non-empty iteration, and a second part is remaining documents after string segmentation; when the length of the list obtained after segmentation exceeds the second value, calling a large model to perform content merging on adjacent strings according to a configured merging prompt dialog box, to generate paragraph content with a string length of the second value; wherein the first value and the second value are different values set, and the second value is greater than the first value, and the first part is located before the second part in the paragraph content; obtaining a document writing time corresponding to each paragraph content after splitting, and calling a large model to determine whether an association degree between the paragraph content after splitting and the medical record document before splitting exceeds a set threshold value according to the content of each medical record document before splitting; when the association degree between the paragraph content after splitting and the medical record document before splitting exceeds the set threshold value, constructing a dependency relationship between the paragraph content after splitting and the medical record document before splitting, to generate the paragraph association dictionary.

2. The hospital case history self-adaptive generation method according to claim 1, wherein, The obtaining of the set of medical record data to be classified, and the classifying of medical record data in the set of medical record data according to diagnosis information of a department where a doctor individual is located, to generate standard medical record texts corresponding to medical record data of each category, comprises: obtaining patient diagnosis information and medical record data, classifying the medical record data according to a department where a doctor individual corresponding to the patient diagnosis information is located, to determine a medical record category corresponding to the patient diagnosis information and the medical record data; screening medical record data with a medical record completeness degree satisfying a preset requirement from a plurality of medical record data under a first medical record category, to generate a standard medical record text corresponding to the first medical record category; The medical record data meeting the preset requirement of the medical record completeness is medical record data with patient admission and discharge records, the largest number of medical record categories, and the closest time stamp of the patient discharge record to the current time.

3. The method of claim 1, wherein, The calling large model performs cyclic iteration on the multiple paragraph contents corresponding to the patient medical record data according to the paragraph name list to generate a medical record paragraph meeting the medical record writing configuration. The patient diagnosis information is obtained, and when the patient diagnosis information is included in the medical record writing configuration of the individual doctor, a paragraph generation request is received; In response to the paragraph generation request, a large model is called to obtain reference text and corresponding standard medical record text based on the paragraph association dictionary according to the configured ordered paragraph name list; The reference text, the split paragraph content, and the medical record text are spliced into a prompt dialog box to generate the medical record paragraph.

4. The method of claim 3, wherein, The method further comprises: An entity recognition model CRF is called to perform entity recognition on the medical record paragraph and the reference text to generate an entity list, and it is determined whether the entity in the medical record paragraph completely belongs to the entity in the reference text according to the entity list; When the entity in the medical record paragraph does not completely belong to the entity in the medical record paragraph, the medical record paragraph is corrected based on the reference text to obtain a corrected medical record paragraph.

5. A hospitalization medical record self-adaptive generation device, characterized in that, The device for implementing the inpatient medical record adaptive generation method of any one of claims 1 to 4 comprises: A medical record classification module is configured to obtain a set of medical record data to be classified, and classify the medical record data in the set of medical record data according to the diagnosis information of the department where the individual doctor is located to generate standard medical record text corresponding to medical record data of each category. A medical record document splitting module is configured to split all medical record documents under the standard medical record text through a configuration script to extract multiple paragraph contents corresponding to the medical record written by the individual doctor, and construct a paragraph association dictionary according to the paragraph contents. A medical record writing configuration module is configured to fuse the paragraph association dictionary, the paragraph contents, and a paragraph name list corresponding to the multiple paragraph contents to generate a medical record writing configuration of the individual doctor. A medical record paragraph generation module is configured to call a large model to perform cyclic iteration on the multiple paragraph contents corresponding to the patient medical record data according to the paragraph name list to generate a medical record paragraph meeting the medical record writing configuration. 6.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to implement the steps of the inpatient medical record adaptive generation method of any one of claims 1 to 4.

7. A computer storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the inpatient medical record adaptive generation method of any one of claims 1 to 4.

Images