System and method for automatically generating hospital medical records by using generative artificial intelligence

Abstract

The present invention relates to a system and a method for automatically generating, using a large language model (LLM), medical records created by medical personnel in a hospital in various formats, the system comprising: a data relay unit which receives, from a medical information system for providing medical information of a patient, a request for generating a medical record in a specific format and unstructured medical records referred to in the creation of the medical record, and which transmits, to the medical information system, a medical record form generated on the basis of the unstructured medical records; a medical record AI generation unit for generating medical record item data included in the medical record form by analyzing the unstructured medical records through the LLM fine-tuned for medical information; and a form generation unit for generating the medical record form by applying the medical record item data generated by the medical record AI generation unit and the unstructured medical records to a specific medical record format.

Description

System and method for automatically generating hospital medical records using generative artificial intelligence

[0001] The present invention relates to a system and method for automatically generating medical records of various forms written by medical staff within a hospital using a Large Language Model (LLM). It is characterized by receiving patient information, clinical observation information, nursing information, prescriptions, surgeries, test information, and surgery schedules, and automatically generating clinical records such as discharge summaries, pre-anesthesia evaluations, admission records, progress records, and discharge summaries through an LLM fine-tuned with medical knowledge.

[0002]

[0003] Recent advancements in Large Language Models (LLMs) are driving rapid growth in Natural Language Processing (NLP) technology and bringing about innovative changes across various industries. In particular, LLMs enable natural language comprehension and generation by learning from massive amounts of text data, demonstrating high performance in automated document creation, summarization, translation, and question-and-answer systems. These technological advancements have high potential for application in the medical field, and LLMs are attracting attention as tools capable of rapidly and accurately processing vast volumes of clinical records. Medical data is often recorded as unstructured descriptive text, making analysis and utilization difficult; however, the application of LLMs allows for the systematic organization and summarization of such data.

[0004] Meanwhile, writing clinical records is a significant burden for medical staff, and issues regarding increased workload and fatigue resulting from this are continuously being raised. Although medical professionals need to focus on patient care and treatment, the significant amount of time required to write clinical records often makes it difficult for them to concentrate on their primary clinical duties. In particular, while summarizing past records and organizing important information is essential during the clinical record-writing process, this is a repetitive and time-consuming task.

[0005] Therefore, there is a need for a method to effectively utilize LLMs with excellent capabilities in analyzing, organizing, and summarizing given data for the preparation of clinical records.

[0006]

[0007] The present invention aims to solve the aforementioned problems by applying a service to be called via an API within a hospital's medical service process, thereby automatically generating clinical records such as discharge summaries, pre-anesthesia evaluations, admission records, progress notes, and discharge summaries through an LLM fine-tuned with medical knowledge based on medical information including patient information, clinical observation information, nursing information, prescriptions, surgeries, test information, and surgery schedules, and enabling the integration and application of automatic generation of medical records via AI within the existing medical process at the necessary time.

[0008]

[0009] A hospital medical record automatic generation system utilizing generative artificial intelligence according to one embodiment of the present invention may include: a data relay unit that receives a request to generate a medical record in a specific format and an unstructured medical record referenced in the creation of said medical record from a medical information system providing medical information of a patient, and transmits a medical record sheet generated based on said unstructured medical record to said medical information system; a medical record AI generation unit that generates medical record item data included in the medical record sheet by analyzing said unstructured medical record through a large-scale language model fine-tuned to be specialized for medical information; and a form generation unit that generates a medical record sheet by applying the medical record item data generated by said medical record AI generation unit and said unstructured medical record to a specific medical record format.

[0010] In addition, the data relay unit can provide an open API to enable other systems to request the creation of medical records in a specific format and to transmit unstructured medical records referenced in the creation of said medical records, and can support real-time medical record creation and batch-based medical record creation through the call of said open API.

[0011] In addition, the data relay unit may verify the data consistency of the received medical record, and when the verification is completed, instruct the data preprocessing unit to convert the medical record into standardized data.

[0012] In addition, the data preprocessing unit can generate FHIR data by converting the requested unstructured medical records to conform to the FHIR (Fast Healthcare Interoperability Resource) standard.

[0013] Additionally, the medical record AI generation unit includes a prompt generation unit that refers to a predefined prompt generation rule, extracts medical items from FHIR data given as input data according to a data analysis rule of a specific requested medical record form, and automatically generates a prompt to be applied to a medical record generation model by combining the extracted medical items according to a prompt creation rule of the medical record form; wherein the prompt generation rule may include a data analysis rule in which the medical items to be extracted per medical record form and FHIR data items mapped to said medical items are defined; and a prompt creation rule in which a rule for integrating the medical items with the prompt template per medical record form is defined.

[0014] In addition, the medical record AI generation unit generates medical record item data of a specific format by applying a prompt generated through the prompt generation unit to a medical record generation model, wherein the medical record generation model is a large-scale language model pre-trained with medical knowledge to be specialized in the medical field, and can generate medical record item data through instruction-based tuning applying the prompt.

[0015] In addition, the above-mentioned form generation unit has an output template for each medical record form, and generates a medical record sheet by inserting medical record data in FHIR format and medical record item data generated through the large-scale language model into the output template by referring to the form mapping rule of the medical record form to be generated, and the form mapping rule may define the medical item inserted into the output template and the data mapped to the medical item.

[0016] Meanwhile, a method for automatically generating hospital medical records using generative artificial intelligence, performed by a system for automatically generating hospital medical records using generative artificial intelligence according to an embodiment of the present invention, wherein the method may include: receiving a request for generating medical records in a specific format and unstructured medical records required for writing medical records from a medical information system; converting the unstructured medical records into FHIR data according to FHIR standards; analyzing the converted FHIR data through a large-scale language model fine-tuned to be specialized for medical information to generate medical record item data to be included in a medical record sheet; integrating the generated medical record item data and the FHIR data into a specific medical format to generate a medical record sheet; and transmitting the generated medical record sheet to the medical information system.

[0017] Additionally, the step of generating the medical record item data may include: a step of generating a prompt that instructs the large-scale language model, which has been fine-tuned according to a prompt generation rule predefined for each medical record format, to generate the medical record item data; and a step of generating the medical record item data by applying the generated prompt to the large-scale language model that has been fine-tuned.

[0018] In addition, the medical information system can provide the generated medical record to a medical patient terminal, and after the medical patient reviews the medical record and transmits approval through the medical patient terminal, it can store the approved medical record including an electronic signature.

[0019]

[0020] By automating the generation of medical records through artificial intelligence, the burden of document preparation on medical staff can be reduced, thereby increasing work efficiency.

[0021] In addition, by managing the rules for extracting medical record data items and combining them required for generating various types of medical records through a prompt generation rule system, prompts are automatically generated. This enables the efficient execution of medical record generation tasks that differ depending on the type of medical record form to be generated, and allows users to easily manipulate the prompt instructions.

[0022] In addition, it provides medical record creation requests and necessary data integration in the form of an Open API, and supports real-time and batch processing, enabling creation tasks to be performed by calling it appropriately at the necessary time within the existing medical process.

[0023]

[0024] Figure 1 is an overall relationship diagram of a hospital medical record automatic generation system utilizing generative artificial intelligence according to one embodiment of the present invention.

[0025] FIG. 2 is a block diagram of the functions of a hospital medical record automatic generation system utilizing generative artificial intelligence according to one embodiment of the present invention.

[0026] FIG. 3 is a diagram showing data in which medical records within a hospital are converted into FHIR standards according to conversion rules in a system for automatically generating hospital medical records using generative artificial intelligence according to an embodiment of the present invention.

[0027] FIG. 4 is a diagram showing a prompt generation rule system in a hospital medical record automatic generation system using generative artificial intelligence according to an embodiment of the present invention.

[0028] FIG. 5 is an example diagram showing a format mapping rule for creating a medical record in a hospital medical record automatic generation system using generative artificial intelligence according to an embodiment of the present invention.

[0029] FIG. 6 is a flowchart illustrating a method for automatically generating hospital medical records using generative artificial intelligence according to an embodiment of the present invention.

[0030] FIG. 7 is a diagram illustrating the process of combining multiple medical records to generate a specific medical record sheet through AI and rule-based mapping operations in a hospital medical record automatic generation system and method utilizing generative artificial intelligence according to an embodiment of the present invention.

[0031] FIG. 8 is a flowchart illustrating a medical record generation process performed through each functional part of a medical record automatic generation system utilizing generative artificial intelligence according to an embodiment of the present invention.

[0032] FIG. 9 is a flowchart illustrating a service embodiment in which medical records are automatically generated in a medical service process within a hospital, in a method for automatically generating hospital medical records using generative artificial intelligence according to an embodiment of the present invention.

[0033]

[0034] Specific embodiments of the present invention will be described in detail below with reference to the drawings. However, the concept of the present invention is not limited to the presented embodiments. Those skilled in the art who understand the concept of the present invention may easily propose other inventions that are inferior or other embodiments included within the scope of the concept of the present invention by adding, changing, or deleting other components within the same scope of the concept, and such are also to be considered to be included within the scope of the concept of the present invention.

[0035] Furthermore, the terms described below are established considering their functions in the present invention; since these may vary depending on the inventor's intent or convention, their definitions should be based on the content throughout this specification. Additionally, if it is determined that a detailed description of known configurations or functions related to the present invention in this specification could obscure the essence of the invention, such detailed description will be omitted.

[0036]

[0037] Hereinafter, a system and method for automatically generating hospital medical records using generative artificial intelligence according to the present invention will be described with reference to the drawings.

[0038]

[0039] Figure 1 is an overall relationship diagram of a hospital medical record automatic generation system utilizing generative artificial intelligence according to one embodiment of the present invention.

[0040] In the present invention, the hospital medical record automatic generation system utilizing generative artificial intelligence (hereinafter referred to as the medical record automatic generation system) is connected to a medical information system (200) and a medical terminal (300) via a network.

[0041] The term "network" as used in the present invention may refer to a core network integrated with a wired public network, a wireless mobile communication network, or the mobile internet, and may refer to a global open computer network structure that provides the TCP / IP protocol and various services existing at the upper layer, such as HTTP (Hyper Text Transfer Protocol), HTTPS (Hyper Text Transfer Protocol Secure), Telnet, FTP (File Transfer Protocol), etc. It is not limited to these examples but comprehensively refers to a data communication network capable of transmitting and receiving data in various forms.

[0042] Furthermore, the server or system referred to in the present invention means a computer on which specific software operates and provides services or data to a terminal or another server via a network. It may include a web server, a database server, a file server, an application server, a cloud server, etc., and is configured to include hardware such as a central processing unit, memory, a hard disk, a user interface, and a network interface.

[0043]

[0044] In the present invention, the medical information system (200) may refer to a system that electronically collects, manages, and provides medical information of patients, such as an electronic medical record (EMR) system and an electronic health record (EHR) system.

[0045] Patient medical information managed in hospitals is written in various medical record formats and stored and managed as digitized documents; this information includes a wide range of clinical records. Here, clinical records refer to documents documenting a patient's medical care, treatment, progress, and other medical-related information, encompassing various forms of records created by medical staff while treating patients. These are important resources that enable the assessment of a patient's health status, the establishment of effective treatment plans, and collaboration among medical professionals.

[0046] Clinical record keeping involves the process of summarizing previous medical records, which may include discharge summaries, pre-anesthesia evaluations, admission records, progress notes, and discharge summaries.

[0047] In addition, the medical record forms mentioned in the present invention refer to types of forms used by each medical institution, such as consultation request forms, test result sheets, emergency records, surgery records, admission and discharge records, progress records, medication records, and procedure records.

[0048] The medical information system (200) can provide a user interface to enable a medical professional to create various medical records using a medical professional terminal (300), and the medical professional can directly create various medical records necessary for treatment and prescription through an application provided to the medical professional terminal (300), or request the creation of a draft medical record through AI.

[0049]

[0050] In the present invention, the automatic medical record generation system (100) receives patient information, clinical observation information, nursing information, prescriptions, surgeries, test information, and surgery schedules from a medical information system (200) and automatically generates clinical records such as discharge summaries, pre-anesthesia evaluations, admission records, progress records, and discharge summaries through a large-scale language model (LLM). At this time, content requiring the opinion or medical judgment of a specialist is excluded from the generation target, and the system performs the task of combining, summarizing, organizing, and outputting a significant amount of previously written medical records according to a format.

[0051] The automatic medical record generation system (100) generates medical records by combining LLM and a rule-based approach. This will be explained in detail later.

[0052] The automatic medical record generation system (100) converts medical data (medical records) received from the medical information system (200) into standardized FHIR data for medical record generation, extracts medical items from the FHIR data by referring to predefined prompt generation rules for each medical record format, and automatically generates prompts to be applied to the medical record generation model based on this. Then, it transmits the prompts to the medical record generation model to generate medical record item data to be included in a specific medical record format, and integrates the generated medical record item data and the medical records received from the medical information system (200) into an output template to generate a draft medical record.

[0053] The generated medical record is transmitted to the medical information system (200), and the medical information system (200) provides the received medical record as a draft medical record so that a medical professional can review it and perform modification, approval, or rejection.

[0054]

[0055] In the present invention, the LLM is built in the form of an artificial intelligence agent and can be provided in the form of an open API so that other systems can generate medical records through artificial intelligence. This allows for natural integration by calling the open API within the process of the medical information system (200) and can be automatically executed in the workflow of a medical professional.

[0056] For example, when a medical professional clicks a button to request the creation of a medical record through a medical professional terminal (300) or instructs a specific prescription, the patient's medical data is automatically transmitted to the LLM and a medical record is created without user intervention, such as searching for the patient's medical records one by one and transmitting them to the LLM.

[0057]

[0058] In the present invention, the medical terminal (300) includes various communication means such as a mobile phone, PC, tablet, or laptop, and can execute various functions provided by the medical information system (200) through a web-based or separate software / application.

[0059]

[0060] FIG. 2 is a block diagram of the functions of a hospital medical record automatic generation system utilizing generative artificial intelligence according to one embodiment of the present invention.

[0061] Referring to FIG. 2, the automatic medical record generation system (100) will be described in detail.

[0062] Referring to FIG. 2, the automatic medical record generation system (100) includes a data relay unit (110), a data preprocessing unit (120), a medical record AI generation unit (130), a form generation unit (140), and a database (150).

[0063] The data relay unit (110) receives a request to create a medical record in a specific format and an unstructured medical record required for creating the medical record from a medical information system (200) that provides medical information of a patient, and transmits the medical record sheet created based on the unstructured medical record to the medical information system (200).

[0064] The data relay unit (110) is responsible for linking with other systems of the medical institution, such as the medical information system (200), and can provide an open API for requesting the creation of medical records, transmitting patient medical record data required for creating medical records, and transmitting the created medical records.

[0065] The open API may be a RESTful-based API and may include APIs that instruct tasks to process medical record generation in real time and tasks to process in batches.

[0066] The data relay unit (110) is responsible for linking with multiple other systems through an open API, performs verification of the consistency of the received data, links the data to each functional unit, converts the received medical record into standard data, requests the creation of a medical record using a medical record creation model, and instructs the creation of a medical record form.

[0067] To this end, each functional unit may be configured as an independent server within the automatic medical record generation system (100).

[0068]

[0069] The data preprocessing unit (120) converts the requested unstructured medical records to the FHIR (Fast Healthcare Interoperability Resource) standard to generate FHIR data.

[0070] Here, FHIR is a standard for medical data interoperability developed by HL7 (Health Level Seven) and is currently adopted worldwide. It supports formats such as RESTful API and JSON, and is designed with a modular structure (Resource-based) to ensure that data can be exchanged in the same way across various institutions and systems.

[0071] FHIR resources are the most core concept in FHIR and are the smallest units representing specific medical data. They include Patient resources containing patient information, Practitioner resources containing medical staff information, Encounter resources containing patient visit (treatment) information, Observation resources containing test results such as blood pressure and body temperature, and Condition resources containing disease diagnosis information.

[0072] The patient's medical records transmitted from the medical information system (200) may include basic patient information, nursing information, clinical observation information, medical record sheets, prescription information, test result information, surgery schedule, etc., and these medical records may be written based on XML to reflect a structured medical record format. However, they are not limited to this and may be written in various markup languages ​​that include tags capable of expressing the format.

[0073] The data preprocessing unit (120) can map medical record forms used within the hospital and medical record information included in the forms to FHIR resources for conversion to FHIR standard data, and can manage conversion rules for converting XML-based medical records into FHIR resources by defining them in advance.

[0074] Conversion rules can be defined by medical professionals or medical information managers handling medical information, and to enable them to define conversion rules, a Custom Domain-Specific Language (DSL) and a user interface and functions for writing conversion rules can be provided.

[0075] FIG. 3 is a diagram showing data in which medical records within a hospital are converted into FHIR standards according to conversion rules in a system for automatically generating hospital medical records using generative artificial intelligence according to an embodiment of the present invention. The format of XML-based medical record data received from a medical information system (200) and the structure of FHIR data converted according to predefined conversion rules can be verified, and the FHIR data can be managed in JSON format as shown in FIG. 3.

[0076]

[0077] Meanwhile, the medical information system (200) can request the creation of a medical record by specifying the type of medical record form to be created when requesting the creation of a medical record, and each type of form can be distinguished by a predefined ID.

[0078] Additionally, the medical information system (200) transmits multiple medical records required to generate a medical record of the requested form to the medical record automatic generation system (100).

[0079] The automatic medical record generation system (100) can generate and manage a request number for a generation request, and multiple medical records transmitted from the medical information system (200) can be managed within the system by corresponding to the request number.

[0080] The data relay unit (110) transmits all data corresponding to the request number during data conversion, medical record creation work through LLM, and medical record sheet creation, and based on this, data analysis, combination, and combination work for medical record creation is performed.

[0081]

[0082] The data relay unit (110) transmits the medical record converted into FHIR data to the medical record AI generation unit (130) to request the generation of medical record item data included in a specific medical record sheet.

[0083] The medical record AI generation unit (130) generates medical record item data included in a specific medical record sheet by analyzing the unstructured medical data through a large-scale language model (LLM) that is fine-tuned to be specialized for medical information according to the instructions of the data relay unit (110).

[0084] At this time, prompt generation rules are provided to automatically generate prompts according to the form in order to generate medical record item data required for various medical record forms.

[0085] The prompt generation rule may include a data analysis rule that defines medical items to be extracted by medical record format and items of FHIR data mapped to said medical items; and a prompt creation rule that defines a rule for integrating the prompt template by medical record format and the extracted medical items.

[0086]

[0087] FIG. 4 is a diagram showing a prompt generation rule system in a hospital medical record automatic generation system using generative artificial intelligence according to an embodiment of the present invention.

[0088] Referring to FIG. 4, the data analysis rule may include a type of medical record creation task (type of form), a medical item to be extracted from medical record data in FHIR format, an extraction path for the medical item, and a tag name (name of the medical item) to be attached to the medical item.

[0089] The prompt creation rules may include the name of the habitat where each extracted medical item is used, the creation order of each habitat within the prompt, and the creation order of each medical item within the habitat.

[0090] Data analysis rules and prompt generation rules are predefined and configured based on the ontology of medical information within the hospital, and can be stored and managed in a database.

[0091]

[0092] The medical record AI generation unit (130) is equipped with a prompt generation unit (131) and a medical record generation AI model unit (132).

[0093] The prompt generation unit (131) refers to a predefined prompt generation rule, extracts medical items from the FHIR data given as input data according to the data analysis rule of the requested specific medical record form, and then combines the extracted medical items according to the prompt generation rule of the medical record form to automatically generate a prompt to be applied to the medical record generation model.

[0094] The prompt generation unit (131) has a prompt template configured for each medical record item to be generated, and generates a prompt by integrating the extracted medical item with the corresponding prompt template.

[0095] Prompt templates include specific instructions that direct given medical information to be formatted, combined, and assembled to fit medical record formats.

[0096]

[0097] The medical record generation AI model unit (132) applies the generated prompt to load and execute the model so that the medical record generation model performs inference.

[0098] The medical record generation model is a large-scale language model (LLM) pre-trained with medical knowledge to be specialized for the medical field, and it generates medical record item data through instruction-based tuning using prompts.

[0099] The medical record generation model may be a compact LLM built to support bilingualism specialized in the medical field by performing fine-tuning with medical knowledge data written in Korean and English.

[0100] By applying a model fine-tuned to understand medical knowledge and bilingualism based on a small LLM, it enables the utilization of artificial intelligence without delay while maintaining the speed of clinical tasks in a medical environment with significant time constraints and cost sensitivity.

[0101] Medical record item data generated through the medical record generation model may include discharge summary, pre-anesthesia evaluation, admission record, progress record, discharge summary, etc.

[0102]

[0103] The form generation unit (140) generates a specific medical record sheet requested by the medical information system (200) using medical record item data such as discharge summary, pre-anesthesia evaluation, admission record, progress record, and discharge summary generated through the medical record generation model, and FHIR data received from the medical information system (200) and converted into an FHIR standard.

[0104] The form generation unit (140) has an output template for each medical record form and, by referring to the form mapping rule of the medical record form to be generated, inserts medical record data in FHIR format and medical record item data generated through the large-scale language model into the output template to generate a medical record sheet.

[0105] Here, the format mapping rule defines the medical items inserted into the output template and the data mapped to the medical items, and can be stored and managed in a database.

[0106] FIG. 5 is an example diagram showing a format mapping rule for creating a medical record in a hospital medical record automatic generation system using generative artificial intelligence according to an embodiment of the present invention.

[0107] Referring to FIG. 5, the format mapping rule consists of mapping information between a medical item (a) included in a medical record sheet of a specific format and a medical record (b) used to configure said medical item. That is, (a) is a medical record item output to a medical record sheet of a specific format, and (b) is a medical record used to configure the data of said medical record item, and the format mapping rule includes information mapping (a) and (b).

[0108] Additionally, the medical record(b) information used to organize medical items included in a specific medical record includes a data path to enable the extraction of mapped items from data in FHIR format.

[0109] If the medical item (a) included in the medical record is a medical item generated through AI, it may be a path on a database where the generated data is stored, or a location on a specific data structure.

[0110] Figure 5 is a diagram illustrating the form mapping rule system, in which form mapping rules are defined for each medical record form and stored and managed in a database.

[0111]

[0112] FIG. 6 is a flowchart illustrating a method for automatically generating hospital medical records using generative artificial intelligence according to an embodiment of the present invention, and FIG. 7 is a diagram illustrating the process of generating a specific medical record by combining multiple medical records through AI and rule-based mapping operations in a system and method for automatically generating hospital medical records using generative artificial intelligence according to an embodiment of the present invention.

[0113] Referring to Figures 6 and 7, a method for automatically generating medical records will be explained in detail.

[0114] The method for automatically generating medical records is performed in the automatic medical record generation system (100).

[0115] First, the automatic medical record generation system (100) performs the step (S610) of receiving a request for medical record generation and unstructured medical records required for medical record creation from the medical information system (200).

[0116] The automatic medical record generation system (100) receives a request to generate a specific medical record from a medical information system (200) that manages and provides medical records, and together with this, receives data regarding patient medical records required for generating a form.

[0117] A request for medical record creation includes the type of form to be created.

[0118] Additionally, patient medical records are transmitted as XML-based text data, and the patient's medical records transmitted from the medical information system (200) may include basic patient information, nursing information, clinical observation information, medical record sheets, prescription information, test result information, surgery schedule, etc., and the medical records transmitted may differ depending on the type of form.

[0119] For example, in the case of a request to generate a surgical consultation request form from the Department of Anesthesiology and Pain Medicine, medical records including patient information, clinical observation records, nursing information, hospitalization records, nursing hospitalization evaluation, specialist consultation records, discharge records, anesthesia records, outpatient treatment records, and test (blood, urine, cardiopulmonary, neurological function, imaging) results can be received from the medical information system (200).

[0120] After step S610, a step (S620) is performed to convert the received XML-based unstructured medical records into FHIR data according to the FHIR standard.

[0121] Step S620 can be performed repeatedly whenever a medical record linked to a received medical record creation request is received.

[0122] Referring to Fig. 7, the received XML-based medical record is converted into FHIR standard-based data according to predefined conversion rules, and the FHIR data is configured in JSON format and used as input data for generating an LLM (medical record generation model) and medical record sheets that are fine-tuned with specific medical information.

[0123] After step S620, a step (S630) is performed to generate medical record item data included in the medical record sheet by analyzing the converted FHIR data through an LLM (medical record generation model) that has been fine-tuned to be specialized for medical information.

[0124] The medical record item data generated through the above LLM is a generation process that analyzes existing clinical records to extract and summarize necessary information, or combines it to output it in a new format.

[0125] Referring to Fig. 7, in the case of a referral for consultation with the Department of Anesthesiology and Pain Medicine, a pre-anesthesia evaluation record can be generated through an LLM that is fine-tuned to be specialized for medical information based on multiple entered medical records.

[0126] In step S630, medical items are extracted from the FHIR data given as input data according to the data analysis rules of the requested specific medical record form by referring to the predefined prompt generation rules, and a prompt to be applied to the LLM is automatically generated according to the prompt generation rules of the medical record form (631), and when the generated prompt is transmitted to the LLM, the LLM generates medical record item data and outputs the result.

[0127] After step S630, a step (S640) is performed to generate a medical record sheet by applying the generated medical record item data and the medical record converted to the FHIR standard to a specific medical form.

[0128] The automatic medical record generation system (100) generates a medical record sheet by referring to the output template and form mapping rules of the requested medical record form, extracting form items from the FHIR data given as input data, and combining the medical record item data generated through the LLM and the patient's medical information received from the medical information system (200) according to the form mapping rules to generate a medical record sheet according to the output template of the requested medical record form.

[0129] At this time, a specific medical record form is generated by referring to form mapping rules that map the items included in each medical record form to FHIR data.

[0130] The automatic medical record generation system (100) performs steps S630 and S640 to perform analysis and generate necessary medical record summaries, and combines or combines data based on rules to finally output medical record paper used within the hospital.

[0131] After step S640, the automatic medical record generation system (100) performs the step (S650) of transmitting the generated medical record to the medical information system (200).

[0132]

[0133] FIG. 8 is a flowchart illustrating a medical record generation process performed through each functional part of a medical record automatic generation system utilizing generative artificial intelligence according to an embodiment of the present invention.

[0134] Referring to FIG. 8, the process of generating a medical record through each functional part of the automatic medical record generation system (100) is described.

[0135] The medical information system (200) requests the creation of a specific medical record through an open API provided by the data relay unit (110). Subsequently, it extracts the patient's medical records required for the creation of the medical record and transmits them through the open API.

[0136] The data relay unit (110) of the automatic medical record system (100) checks whether all medical records required for the creation of the medical record requested by the medical information system (200) have been received in response to an API call, verifies the consistency of the medical records, and then instructs the data preprocessing unit (120) to convert the medical records into FHIR standard data.

[0137] The data preprocessing unit (120) converts the data into FHIR data and transmits it to the data relay unit (110), and the data relay unit (110) transmits the FHIR data and the type of medical record form to be generated to the medical record AI generation unit (130) to instruct the generation of the medical record.

[0138] The prompt generation unit (131) generates a prompt for the requested medical record form and transmits it to the medical record generation AI model unit (132) so that the LLM generates medical record item data according to the prompt.

[0139] The data relay unit (110) receives the generated medical record item data and instructs the form generation unit (140) to create a medical record sheet by applying the form of the requested medical record.

[0140] The habitat generation unit (140) generates the requested medical record sheet by referring to the form mapping rules and then transmits it to the data relay unit (110).

[0141] The data relay unit (110) transmits the generated medical record to the medical information system (200).

[0142]

[0143] FIG. 9 is a flowchart illustrating a service embodiment in which medical records are automatically generated in a medical service process within a hospital, in a method for automatically generating hospital medical records using generative artificial intelligence according to an embodiment of the present invention.

[0144] The process of automatically generating medical records can be integrated into the existing treatment and prescription processes of medical institutions.

[0145] Referring to FIG. 9, a medical professional may instruct the creation of a draft of a specific medical record through a medical professional terminal (300) during the process of performing medical treatment and prescription.

[0146] In response to a request from a medical device terminal (300), the medical information system (200) calls the API of the automatic medical record generation system (100) to request the generation of a specific medical record.

[0147] The automatic medical record generation system (100) generates a specific medical record through AI in response to the request and transmits it to the medical information system (200).

[0148] The medical information system (200) provides a medical record automatically generated through AI to a medical terminal (300), and the medical patient can check and review it, and then modify, save, approve, or reject it.

[0149] Approved medical records are electronically signed and saved as the final electronic medical records.

[0150]

[0151] Each step of the configuration or method described above may be implemented as computer-readable code on a computer-readable recording medium or transmitted via a transmission medium. A computer-readable recording medium is a data storage device capable of storing data that can be read by a computer system.

[0152] Examples of computer-readable recording media include, but are not limited to, databases, ROM, RAM, CD-ROM, DVD, magnetic tape, floppy disk, and optical data storage devices. Transmission media may include carrier waves transmitted via the Internet or various types of communication channels. Additionally, computer-readable recording media may be distributed through networked computer systems so that computer-readable code is stored and executed in a distributed manner.

[0153] In addition, at least one component applied in the present invention may include or be implemented by a processor, such as a central processing unit (CPU) or a microprocessor, that performs a respective function, and two or more of said components may be combined into a single component to perform all operations or functions of the combined two or more components. Furthermore, a part of the at least one component applied in the present invention may be performed by another of these components. Additionally, communication between said components may be performed via a bus.

[0154]

[0155] Although the structure and features of the present invention have been described above based on embodiments according to the present invention, the present invention is not limited thereto. It is understood by those skilled in the art that various changes or modifications can be made within the spirit and scope of the present invention, and thus such changes or modifications fall within the scope of the appended claims.

[0156]

[0157] By automating the extraction of medical record data items required for generating various types of medical records and the rules for combining them, and by automating the generation of medical records through artificial intelligence, enabling integration and application within existing medical processes at the necessary time, the burden of document creation on medical staff can be reduced, thereby increasing work efficiency.

Claims

1. A data relay unit that receives a request to create a medical record in a specific format and an unstructured medical record referenced for the creation of the medical record from a medical information system providing medical information of a patient, and transmits a medical record sheet created based on the unstructured medical record to the medical information system; A medical record AI generation unit that generates medical record item data included in a medical record sheet by analyzing the above unstructured medical records through a large-scale language model fine-tuned to be specialized for medical information; and A form generation unit that generates a medical record sheet by applying the medical record item data generated by the medical record AI generation unit and the unstructured medical record to a specific medical record form; Automatic hospital medical record generation system utilizing generative AI.

2. In Paragraph 1, The above data relay unit is, Provides an open API to request the creation of medical records in a specific format from other systems and to transmit unstructured medical records referenced in the creation of said medical records, and supports real-time medical record creation and batch-based medical record creation through the call of said open API. Automatic hospital medical record generation system utilizing generative AI.

3. In Paragraph 1, The above data relay unit is, Verifying the data consistency of the received medical records, and once the verification is complete, instructing the data preprocessing unit to convert the said medical records into standardized data, Automatic hospital medical record generation system utilizing generative AI.

4. In Paragraph 3, The above data preprocessing unit is, Generating FHIR data by converting requested unstructured medical records according to FHIR (Fast Healthcare Interoperability Resource) standards, Automatic hospital medical record generation system utilizing generative AI.

5. In Paragraph 1, The above medical record AI generation unit is, A prompt generation unit that refers to a predefined prompt generation rule, extracts medical items from FHIR data given as input data according to a data analysis rule of a specific requested medical record form, and automatically generates a prompt to be applied to a medical record generation model by combining the extracted medical items according to the prompt generation rule of the medical record form; The above prompt generation rule is, A data analysis rule defined for medical items to be extracted by medical record form and FHIR data items mapped to said medical items; and a prompt creation rule defined for integrating a prompt template by medical record form and extracted medical items; characterized by including Automatic hospital medical record generation system utilizing generative AI.

6. In Paragraph 5, The above medical record AI generation unit is, Apply the prompt generated through the above-mentioned prompt generation unit to the medical record generation model to generate medical record item data in a specific format, The above medical record generation model is, A large-scale language model pre-trained with medical knowledge to be specialized for the medical field, characterized by generating medical record item data through instruction-based tuning with applied prompts, Automatic hospital medical record generation system utilizing generative AI.

7. In Paragraph 1, The above habitat generation unit is, Provides an output template for each medical record form, and, by referring to the form mapping rules of the medical record form to be created, inserts medical record data in FHIR format and medical record item data generated through the large-scale language model into the output template to create a medical record sheet. The above format mapping rule defines the medical items inserted into the above output template and the data mapped to the medical items, Automatic hospital medical record generation system utilizing generative AI.

8. As a method for automatically generating hospital medical records using generative artificial intelligence performed by a system for automatically generating hospital medical records using generative artificial intelligence, The above method comprises the step of receiving a request to create a medical record of a specific format and unstructured medical records required for creating the medical record from a medical information system; A step of converting the above-mentioned unstructured medical records into FHIR data according to FHIR standards; A step of generating medical record item data included in medical records by analyzing transformed FHIR data through a large-scale language model fine-tuned to be specialized for medical information; A step of generating a medical record sheet by integrating the generated medical record item data and the above FHIR data into a specific medical form; and A step of transmitting the generated medical record to the medical information system; comprising Method for automatically generating hospital medical records using generative artificial intelligence.

9. In Paragraph 8, The step of generating the above medical record item data is, A step of generating a prompt that instructs the above-mentioned large-scale language model, fine-tuned according to predefined prompt generation rules for each medical record format, to generate medical record item data; and A step comprising: generating medical record item data by applying the generated prompt to the above-mentioned large-scale language model that has been fine-tuned; Method for automatically generating hospital medical records using generative artificial intelligence.

10. In Paragraph 8, The above medical information system is, The generated medical record is provided to a medical terminal, and after the medical patient reviews the medical record, when approval is transmitted through the medical terminal, the approved medical record is stored including an electronic signature. Method for automatically generating hospital medical records using generative artificial intelligence.

Images