Medical support system, medical support method, and medical support program

Abstract

A medical care support system, a medical care support method, and a medical care support program are provided that can provide highly useful medical information. [Solution] The medical support system 11 includes a medical information processing unit 48 configured to process medical information included in text data TD by inputting drug information and text data TD into an artificial intelligence model M. The medical information processing unit 48 is configured to extract drug name text information that is presumed to be a drug name from the text data TD, search for candidate drug name information that matches at least a portion of the drug name text information from the drug name information in the drug information, and, if the candidate drug name information does not match the drug name text information, correct the drug name text information to the candidate drug name information or output the candidate drug name information.

Description

[Technical Field]

[0001] The present invention relates to a medical assistance system, a medical assistance method, and a medical assistance program. [Background technology]

[0002] Conventionally, systems that use voice recognition technology to conduct medical interviews to understand a patient's condition are known. For example, a system described in Patent Document 1 records voice information of conversations between multiple speakers, including medical, caregiving, or nursing professionals, and subjects who are the targets of medical diagnosis, caregiving, or nursing, and converts the voice information into text data by voice recognition. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2018-206055 Summary of the Invention [Problem to be solved by the invention]

[0004] Recently, the workload of doctors in the medical field has become increasingly severe, making it difficult for them to check evidence and accurate information about drugs in order to make reliable treatment suggestions during consultations.

[0005] Therefore, it is possible to process the text data converted from voice information using artificial intelligence using LLM (Large Language Models) to suggest reliable evidence or information about drugs.

[0006] However, if the transcription of drug names is inaccurate during the speech recognition stage, there is a risk that inaccurate drug names will be output. Furthermore, if the text data contains terms that LLM has not learned, there is a risk of hallucination, where LLM generates false information based on uncertain information. Thus, there is room for improvement in the provision of medical information using artificial intelligence.

[0007] The present invention has been made in view of the above background, and aims to provide a medical support system, a medical support method, and a medical support program that can provide highly useful medical information. [Means for solving the problem]

[0008] A first aspect of the present invention is a drug information acquisition unit configured to periodically or irregularly acquire drug information including drug name information from a drug information database; a drug information storage unit configured to store the drug information; a text data acquisition unit configured to acquire text data including a drug name; It is a large-scale language model Artificial intelligence model to perform the desired action at the system prompt. a medical information processing unit configured as The predetermined processing is Input into the artificial intelligence model extracting drug name text information estimated to be a drug name from the text data; Input into the artificial intelligence model Searching for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information; The candidate drug name information and the drug name text information Part of In this case, the drug name text information is corrected to the candidate drug name information, or the candidate drug name information is output. 、 It is in the medical support system.

[0009] A second aspect of the present invention is a medical literature acquisition unit configured to periodically or irregularly acquire a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage unit configured to store a plurality of the medical literatures; a text data acquisition unit configured to acquire text data including medical information related to the medical treatment of a patient; It is a large-scale language model Artificial intelligence model to perform the desired action at the system prompt. a medical information processing unit configured as The predetermined processing is Input into the artificial intelligence model extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data; Input into the artificial intelligence model From multiple medical literature 、 Searching at least one evidence-based medical literature containing information corresponding to the symptom text information or the disease name text information; Output the evidence-based medical literature 、 It is in the medical support system.

[0010] A third aspect of the present invention is The processor a drug information acquisition step of periodically or irregularly acquiring drug information including drug name information from a drug information database; a drug information storage step of storing the drug information in a memory; a text data acquisition step of acquiring text data including a drug name; It is a large-scale language model Artificial intelligence model a medical information processing step of executing predetermined processing of a system prompt by the medical information processing step; The predetermined processing is extracting drug name text information estimated to be a drug name from the text data; Searching for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information; The candidate drug name information and the drug name text information Part ofIn this case, the medical care support method corrects the drug name text information to the candidate drug name information or outputs the candidate drug name information.

[0011] A fourth aspect of the present invention is The processor a medical literature acquisition step of periodically or irregularly acquiring a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage step of storing a plurality of the medical literatures in a memory; a text data acquisition step of acquiring text data including medical information related to the medical treatment of a patient; It is a large-scale language model Artificial intelligence model a medical information processing step of executing predetermined processing of a system prompt by the medical information processing step; The predetermined processing is Input into the artificial intelligence model extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data; Input into the artificial intelligence model From multiple medical literature 、 Searching at least one evidence-based medical literature containing information corresponding to the symptom text information or the disease name text information; The medical support method includes outputting the evidence-based medical literature.

[0012] A fifth aspect of the present invention is A medical assistance program executed by at least one processor, a drug information acquisition step of periodically or irregularly acquiring drug information including drug name information from a drug information database; a drug information storage step of storing the drug information in a memory; a text data acquisition step of acquiring text data including a drug name; It is a large-scale language model Artificial intelligence model and performing a predetermined processing of the system prompts in response to the medical information processing step; The predetermined processing is Input into the artificial intelligence modelextracting drug name text information estimated to be a drug name from the text data; Input into the artificial intelligence model Searching for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information; The candidate drug name information and the drug name text information Part of In this case, the medical assistance program corrects the drug name text information to the candidate drug name information or outputs the candidate drug name information.

[0013] A sixth aspect of the present invention is A medical assistance program executed by at least one processor, a medical literature acquisition step of periodically or irregularly acquiring a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage step of storing a plurality of the medical literatures in a memory; a text data acquisition step of acquiring text data including medical information related to the medical treatment of a patient; It is a large-scale language model Artificial intelligence model and performing a predetermined processing of the system prompts in response to the medical information processing step; The predetermined processing is Input into the artificial intelligence model extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data; Input into the artificial intelligence model From multiple medical literature 、 Searching at least one evidence-based medical literature containing information corresponding to the symptom text information or the disease name text information; The medical support program outputs the evidence-based medical literature. [Effects of the Invention]

[0014] According to the first, third and fifth aspects of the present invention, accurate drug names can be output, making it possible to provide highly useful medical information.

[0015] According to the second, fourth, and sixth aspects of the present invention, the artificial intelligence model is prevented from making decisions based on uncertain information, thereby suppressing hallucination, thereby providing highly useful medical information.

[0016] As described above, according to the above aspects, it is possible to provide a medical assistance system, a medical assistance method, and a medical assistance program that can provide highly useful medical information. [Brief explanation of the drawings]

[0017] [Figure 1] FIG. 1 is a configuration diagram of a medical support system according to a first embodiment. [Figure 2] FIG. 2 is a diagram showing hardware of a terminal device according to the first embodiment. [Figure 3] FIG. 2 is a diagram showing text data according to the first embodiment. [Figure 4] FIG. 2 is a diagram showing cache data in FIG. 1; [Figure 5] FIG. 1 is a diagram showing the process of generating medical literature translation information, medical literature summary information, medical literature citation information, summary translation information, and citation translation information from medical literature in the first embodiment. [Figure 6] FIG. 1 is a functional block diagram of a medical support system according to a first embodiment. [Figure 7] FIG. 10 is a diagram for explaining the processing to be performed when the medicine name text information is incorrect in the first embodiment. [Figure 8] FIG. 2 is a diagram for explaining audio data, divided text data, and summarized text data in the first embodiment. [Figure 9] FIG. 2 is a diagram for explaining the operation of the medical support system in the first embodiment. [Figure 10] FIG. 10 is a configuration diagram of a medical support system according to a second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0018] (Embodiment 1) 1.Configuration of Medical Support System 11 The configuration of a medical assistance system 11 according to embodiment 1 will be described with reference to Fig. 1. The medical assistance system 11 supports doctors in making medical decisions by outputting information related to medical treatment based on text data TD generated or input when the doctors treat patients.

[0019] 1, the medical care support system 11 includes a terminal device 12, an artificial intelligence computer 13, and a storage 14. The terminal device 12, the artificial intelligence computer 13, and the storage 14 form a network. The terminal device 12, the artificial intelligence computer 13, and the storage 14 are configured to be able to communicate with a drug information database 15 and a medical literature database 16 via the network.

[0020] [Terminal Device 12] The terminal device 12 includes a tablet terminal, a smartphone, a dedicated terminal, a desktop computer, a laptop computer, etc. As shown in Fig. 2, the terminal device 12 includes a processor 21, a memory 22, an input device 23, an output device 24, a communication interface 25, etc.

[0021] The processor 21 is configured by a CPU (Central Processing Unit) etc. The processor 21 realizes a predetermined function by executing a program stored in the memory 22, a program input from the input device 23, or a program input via the communication interface 25.

[0022] The memory 22 includes a volatile memory and a non-volatile memory. The volatile memory can be appropriately selected from RAM (Random Access Memory), static memory, etc. The non-volatile memory can be appropriately selected from ROM (Read Only Memory), a hard disk drive, flash memory, optical disk, etc. The non-volatile memory may be configured to store the medical assistance program. However, the medical assistance program may also be configured to be stored in a server (not shown) connected to a network.

[0023] The memory 22 may store medical documents such as electronic medical records. The medical assistance system 11 may be configured to generate electronic medical records and store them in the memory 22, or to add new information to electronic medical records already stored in the memory 22.

[0024] The input device 23 of the first embodiment includes a microphone 23a, a touch panel 23b, etc. The microphone 23a can capture speech uttered by a doctor or patient as audio data. The touch panel 23b can input letters, numbers, symbols, etc. The input device 23 is not limited to the above, and any input device such as a keyboard or joystick may be appropriately selected.

[0025] The output device 24 of the first embodiment includes a screen 24a, a speaker 24b, and the like. The screen 24a can be appropriately selected from a liquid crystal display, an organic EL (Electro-Luminescence), and the like. The output device 24 can output any information by displaying characters, numbers, symbols, images, and the like on the screen 24a. The output device 24 can output any information by sound using the speaker 24b.

[0026] The communication interface 25 exchanges information between the processor 21 and external devices via a network.

[0027] The terminal device 12 may be configured to be connected to the network via, for example, a WebSocket-based real-time agent.

[0028] [Artificial Intelligence Computer 13] The artificial intelligence computer 13 shown in FIG. 1 stores an artificial intelligence model M. The artificial intelligence model M may be a natural language model, or may include a general-purpose natural language processing trained model such as Large Language Models (LLM) trained on a huge amount of data. The artificial intelligence model M is a type of generative AI. The artificial intelligence model M is not particularly limited, and can be appropriately selected from models provided by services such as GPT (registered trademark) and GPT-4o from OpenAI, Caude from Anthropic, and Gemini (registered trademark) from Google.

[0029] The artificial intelligence model M may be configured to realize its functions from the terminal device 12 via a network, or may be configured to be downloadable from the artificial intelligence computer 13 to the terminal device 12.

[0030] In the medical assistance system 11, a system prompt is set in advance when the artificial intelligence model M is used. The system prompt serves as a specification for the medical assistance system 11. In addition, in the medical assistance system 11, a user prompt is input to the artificial intelligence model M. The user prompt is an instruction input by a user who uses the medical assistance system 11. The medical assistance system 11 outputs a response to the instructions given by the system prompt and the user prompt.

[0031] [Storage 14] The storage 14 shown in FIG. 1 stores cache data CD. The storage 14 is an example of a memory. The cache data CD is acquired regularly or irregularly from a drug information database 15 and a medical literature database 16 (described later) and stored in the storage 14. This allows the cache data CD to be updated regularly or irregularly. The period for updating the cache data CD is not particularly limited and can be set to any period, such as every hour, every few hours, every day, every few days, every week, every few weeks, every month, or every few months. Furthermore, the period for updating the cache data CD is not limited to a fixed period. For example, the cache data CD may be acquired from the drug information database 15 or the medical literature database 16 on the condition that the terminal device 12 receives a notification that the drug information database 15 or the medical literature database 16 has been updated.

[0032] The storage 14 is not particularly limited, and a database such as MySQL (registered trademark) can be used. For example, Pinecone (registered trademark) can be suitably used as a Vector DB.

[0033] The information acquired from the drug information database 15 and the information acquired from the medical literature database 16 are not limited to being stored in one storage 14, but may be stored in separate servers. Furthermore, the information acquired from the drug information database 15 and the information acquired from the medical literature database 16 may be stored in the memory 22 of the terminal device 12.

[0034] [Drug Information Database 15] 1 is a database that stores pharmaceutical information. For example, any database such as that of the Pharmaceuticals and Medical Devices Agency (PDMA), an independent administrative institution, can be used as the pharmaceutical information database 15. Pharmaceutical information can be obtained, for example, from a search site for package inserts of prescription pharmaceuticals on the PDMA homepage.

[0035] [Medical Literature Database 16] 1 is a database that stores a plurality of medical literature. The medical literature database 16 is not particularly limited, and any database such as PubMed, Today's Clinical Support (registered trademark), or Today's Practice (registered trademark), all of which are operated by the U.S. National Library of Medicine, can be used as appropriate.

[0036] Medical literature includes both medical literature and healthcare literature. Medical literature is academic literature related to medicine. Medical literature is reliable literature related to healthcare (e.g., official literature). Medical literature includes any literature related to medicine or healthcare, such as medical papers, medical textbooks, case reports, and guidelines issued by public organizations such as the World Health Organization and the Ministry of Health, Labor and Welfare. However, medical literature includes both physically printed and published literature and electronic files generated solely online.

[0037] PubMed can be used efficiently by using Entrez Programming Utilities (E-utilities), a Web API provided by the National Center for Biotechnology Information (NCBI).

[0038] For example, by filtering or crawling highly reliable medical literature such as Systematic Reviews and Meta-analyses from PubMed and storing them as medical literature information in storage 14, it is possible to store medical literature with a high level of evidence.

[0039] 2. Text Data TD The text data TD includes a drug name or medical information related to the patient's medical treatment. The medical information related to the patient's medical treatment includes at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease. The text data TD may be generated based on voice data uttered by a doctor or a patient. The text data TD may be generated by the medical support system 11 converting voice data, or may be generated by conversion by a device different from the medical support system 11. The text data TD may also be input by a doctor via an input device 23 such as a keyboard. As shown in FIG. 3, the text data TD includes, for example, drug name text information, drug efficacy text information, symptom text information, and disease name text information.

[0040] The drug name text information is information included in the text information and is information that is presumed to be the drug name. The drug name text information may, for example, match or partially match the drug name listed on the package insert of the prescription drug obtained from the PDMA. The drug name may be the brand name or the generic name.

[0041] The drug efficacy text information is information included in the text information and is information that is estimated to be a drug efficacy. The drug efficacy text information may be the same as or different from the drug efficacy described in the package insert of the prescription drug obtained from the PDMA, for example.

[0042] The symptom text information is information included in the text data TD and is presumed to be the patient's symptom. The disease name text information is information included in the text data TD and is presumed to be the patient's disease name.

[0043] 3. Cache Data CD 4, the cache data CD includes drug information acquired from the drug information database 15 and medical literature acquired from the medical literature database 16. The cache data CD further includes a plurality of pieces of information generated from the medical literature.

[0044] The drug information is information acquired from the drug information database 15. As shown in Fig. 4, the drug information includes drug name information, drug efficacy information, side effect information, and caution information regarding concomitant use with other drugs. The drug information may further include other information.

[0045] The drug information may also be configured such that a plurality of pieces of information extracted from the information acquired from the drug information database 15 are associated with each other and stored in a structured state in the storage 14. In this case, the drug name information, drug efficacy information, side effect information, and information on precautions for use in combination with other drugs are associated with each other and structured.

[0046] The drug name information is information relating to the name of a drug. The drug name information may be a brand name or a generic name. The efficacy information is information relating to the efficacy or effect of a drug.

[0047] The side effect information is information about the side effects of the drug. The caution information for concomitant use with other drugs is information about precautions to be taken when the drug in question is used in combination with other drugs.

[0048] As shown in Figure 4, the cache data CD further includes medical literature, medical literature summary information, medical literature citation information, medical literature translation information, summary translation information, citation translation information, and tag information. The cache data CD may also include other information. The medical literature, medical literature summary information, medical literature citation information, medical literature translation information, summary translation information, citation translation information, and tag information are stored in storage 14 in a structured state by being associated with each other. However, the medical literature summary information, medical literature citation information, medical literature translation information, summary translation information, citation translation information, and tag information may be omitted.

[0049] The medical literature is an electronic file of medical literature obtained from the medical literature database 16. The medical literature can be written in any language, such as English or Japanese.

[0050] Medical literature summary information is information that summarizes medical literature, generated by inputting medical literature into an artificial intelligence model M. As shown in FIG. 5, medical literature summary information may be generated by summarizing the entire text of the medical literature, or may be generated by extracting the abstract of the medical literature. Medical literature summary information can be written in any language. There is no particular limit to the number of characters in the medical literature summary information.

[0051] The medical literature citation information is generated by inputting medical literature into the artificial intelligence model M. The medical literature citation information is generated by extracting descriptions related to at least one of symptom text information and disease name text information from the evidence medical literature described below. The medical literature citation information can be written in any language. There is no particular limit to the number of characters in the medical literature citation information.

[0052] Translated medical literature information is information translated into a language different from the language in which the medical literature was written by inputting the medical literature into an artificial intelligence model M. Translated medical literature information may be generated by generating one or more translated medical literature information written in one or more languages ​​from medical literature written in one language. For example, translated medical literature information written in Japanese may be generated from medical literature written in English, or translated medical literature information written in Japanese and translated medical literature information written in Chinese may be generated from medical literature written in English. Translated medical literature information is generated by translating the main text and abstract of the medical literature, as shown in FIG. 5.

[0053] The translated summary information is generated by inputting medical literature into an artificial intelligence model M. The translated summary information is information equivalent to the summary of medical literature, translated into a language different from the language in which the medical literature was written. The translated summary information may be generated by summarizing already translated medical literature translation information, as shown in Figure 5, or by extracting an abstract from the translated medical literature information. The translated summary information may also be generated by translating untranslated medical literature summary information.

[0054] Citation translation information is generated by inputting medical literature into the artificial intelligence model M. Citation translation information is information equivalent to the above-mentioned medical literature citation information, translated into a language different from the language in which the medical literature was written. Citation translation information may be generated by extracting descriptions related to at least one of symptom text information and disease name text information from already translated medical literature translation information, as shown in Figure 5. Citation translation information may also be generated by translating untranslated medical literature citation information.

[0055] The tag information is extracted from terms contained in each of the multiple medical documents and is information associated with each of the multiple medical documents. The tag information can be any term appropriately selected from terms contained in the medical documents, such as disease names such as "influenza" or "atopic dermatitis," drug names such as "Devigo" or "amlodipine," organ names such as "ear" or "heart," or symptom names such as "headache" or "itch." The multiple medical documents are stored in a state associated with each other by tag information commonly contained in each medical document. Using tag information can improve the efficiency of searching medical documents.

[0056] 4. Functions of the Medical Support System 11 The functional configuration of the medical assistance system 11 and the processing method performed by the medical assistance system 11 will be described with reference to Fig. 6. As shown in Fig. 6, the medical assistance system 11 includes a voice data acquisition unit 41, a text data conversion unit 42, a text data acquisition unit 43, a drug information acquisition unit 44, a drug information storage unit 45, a medical literature acquisition unit 46, a medical literature storage unit 47, and a medical information processing unit 48. In the first embodiment, the elements 41 to 48 are configured by the processor 21 in Fig. 2.

[0057] 6, steps in parentheses in each block indicate steps executed by the processor 21 in each of the elements 41 to 48. In other words, the medical support method by the medical support system 11 corresponds to the execution of each step in each of the elements 41 to 48.

[0058] [Audio data acquisition section 41] The voice data acquiring unit 41 shown in Fig. 6 executes a step of acquiring voice data uttered by a doctor or a patient (voice data acquiring step) using the processor 21 shown in Fig. 2. The voice data may be acquired by the microphone 23a of the terminal device 12, or may be acquired by a microphone separate from the terminal device 12.

[0059] [Text data conversion part 42] 6 executes a step of converting voice data into text data TD (text data conversion step) using the processor 21 shown in FIG. 2. The text data conversion unit 42 may convert the voice data acquired by the voice data acquisition unit 41 into text data TD. However, the text data conversion unit 42 may be omitted. In this case, the text data TD may be converted from the voice data by a device different from the terminal device 12.

[0060] [Text data acquisition section 43] The text data acquisition unit 43 shown in Fig. 6 executes a step of acquiring text data TD (text data acquisition step) using the processor 21 shown in Fig. 2. The text data acquisition unit 43 may be configured to acquire text data TD generated by the text data conversion unit 42. The text data acquisition unit 43 may also be configured to acquire text data TD generated by a device different from the terminal device 12. The text data acquisition unit 43 may also be configured to acquire text data TD input by a doctor via the input device 23, such as the touch panel 23b or a keyboard.

[0061] [Drug Information Acquisition Department 44] 2, the drug information acquisition unit 44 shown in Fig. 6 executes a step (drug information acquisition step) of periodically or irregularly acquiring drug information including drug name information from the drug information database 15. This reduces the frequency of acquiring drug information from the drug information database 15, thereby suppressing communication delays when accessing the drug information database 15 via a network.

[0062] [Drug information storage unit 45] The drug information storage unit 45 shown in FIG. 6 executes a step of storing drug information in the storage 14 by the processor 21 shown in FIG. 2 (drug information storage step).

[0063] [Medical Literature Collection Department 46] The medical literature acquisition unit 46 shown in Fig. 6 executes a step (medical literature acquisition step) of periodically or irregularly acquiring a plurality of medical literature from the medical literature database 16 using the processor 21 shown in Fig. 2. This reduces the frequency with which medical literature is acquired from the medical literature database 16, thereby suppressing communication delays when accessing the medical literature database 16 via a network. Furthermore, since the frequency of API access can be reduced, communication costs can be reduced and access restrictions can be avoided.

[0064] [Medical Literature Memory Section 47] The medical literature storage unit 47 shown in FIG. 6 executes a step of storing a plurality of medical literatures in the storage 14 by the processor 21 shown in FIG. 2 (medical literature storage step).

[0065] [Medical Information Processing Department 48] The medical information processing unit 48 shown in Figure 6 executes a process (medical information processing process) of processing medical information contained in the text data TD by inputting pharmaceutical information and text data TD into the artificial intelligence model M using the processor 21 shown in Figure 2.

[0066] The medical information processing unit 48 extracts drug name text information that is estimated to be a drug name from the text data TD. When the text data TD is obtained by converting voice data spoken by a doctor or a patient into text data TD, the text data TD may contain an incorrect drug name if the doctor or patient speaks the drug name incorrectly due to a mistake or the like. Furthermore, even if the doctor or patient speaks the drug name correctly, the text data TD may contain an incorrect drug name if the voice data is converted incorrectly into text data TD. For this reason, the drug name text information may contain an incorrect drug name.

[0067] Therefore, the medical information processing unit 48 searches for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the drug information. Since the drug information is acquired from the drug information database 15, the drug name information of the drug information is accurate. If the searched candidate drug name information and the drug name text information completely match, the drug name text information is presumed to be accurate information. On the other hand, if the candidate drug name information and the drug name text information are completely different, Part of In this case, the drug name text information is presumed to be incorrect.

[0068] The process of searching for candidate drug name information may be performed for all drug name text information. Alternatively, the process of searching for candidate drug name information may be configured to be performed when the drug name text information differs from a drug name already stored in the artificial intelligence model M. In this case, the frequency of searching for candidate drug name information can be reduced, thereby improving overall search efficiency.

[0069] As described above, the candidate drug name information, the drug name text information, and Part of In this case, it is estimated that the drug name text information is incorrect. Therefore, the medical information processing unit 48 compares the candidate drug name information with the drug name text information. Part of In this case, the drug name text information is corrected to candidate drug name information, or the candidate drug name information is output.

[0070] By correcting the drug name text information to candidate drug name information, it is possible to prevent the inclusion of an incorrect drug name in the text data TD.

[0071] Furthermore, by outputting the candidate drug name information, it is possible to compare the drug name text information written in the text data TD with the candidate drug name information. This allows the doctor to confirm and correct the correct drug name. The candidate drug name information to be output may be one or multiple.

[0072] Furthermore, the medical information processing unit 48 extracts the drug effect text information estimated as the drug effect from the text data TD, and combines the candidate drug name information and the drug name text information. Part of In this case, if the efficacy information and the efficacy text information match, the drug name text information is corrected to the candidate drug name information, or the candidate drug name information is output. As a result, by comparing the efficacy text information with the efficacy information associated with the drug name information, it is possible to further prevent the text data TD from including an incorrect drug name.

[0073] The process relating to the name of a drug will be described with reference to Fig. 7. First, a doctor utters, "I will prescribe Dayvigo as a drug for treating insomnia," and voice data with the above content is generated.

[0074] Next, text data TD is generated based on the generated speech data. At this time, suppose that what should have been converted to "Deibigo" is mistakenly converted to "Deibigo." In this case, the text data TD is generated as "I will prescribe Deibigo as a treatment for insomnia." If the text data TD contains an incorrect drug name, there is a risk that the doctor will make an incorrect diagnosis. Furthermore, even if the doctor notices the error in the drug name, investigation and confirmation are required to confirm that the drug name is incorrect, which reduces the efficiency of medical treatment.

[0075] When the text data TD is input into the artificial intelligence, the artificial intelligence model M searches for pharmaceutical information, extracts the correct drug name information "Deevigo" from the pharmaceutical information, which at least partially matches "Deevigo" contained in the text data TD, and extracts the correct medicinal effect, "insomnia."

[0076] The medical information processing unit 48 determines that "Deibigo" included in the text data TD is incorrect and that "Deebigo" extracted from the drug information is correct.

[0077] Next, the medical information processing unit 48 may correct the text data TD to read, "I will prescribe Dayvigo as a treatment for insomnia." Furthermore, the medical information processing unit 48 may output, for example, "Dayvigo" in a drop-down list format for the portion of the text data TD that contains "Dayvigo," and have the doctor confirm and decide.

[0078] Furthermore, the medical information processing unit 48 outputs a medical document based on the text data TD. As described above, even if the drug name included in the text data TD is incorrect, it is corrected to the correct drug name by the medical information processing unit 48 or by a doctor. Therefore, it is possible to prevent an incorrect drug name from being written in a medical document. The medical document is not particularly limited, and may include an electronic medical record, a lifestyle-related disease treatment plan, a prescription, etc.

[0079] The medical document may be output in any format, such as audible output from the speaker 24b of the terminal device 12, displayed on the screen 24a of the terminal device 12, or printed out from a printer (not shown).The medical document may also be stored as an electronic file in the memory 22. The same applies below.

[0080] The electronic medical record contains subjective information, objective information, assessment, and plan. The medical information processing unit 48 extracts the subjective information, objective information, assessment, and plan from the text data TD and records them in the electronic medical record. The medical information processing unit 48 creates a new electronic medical record for a first-time patient and records the subjective information, objective information, assessment, and plan, and adds the new subjective information, objective information, assessment, and plan to the already created electronic medical record for a repeat patient.

[0081] A lifestyle-related disease treatment plan is a document in which doctors and nurses provide specific instructions and treatment goals tailored to each patient for patients with lifestyle-related diseases such as hypertension, dyslipidemia, diabetes, etc. The medical information processing unit 48 extracts information on treatment goals, dietary therapy, exercise therapy, etc. from the text data TD and outputs the lifestyle-related disease treatment plan.

[0082] A prescription is a document in which a doctor writes down the type, amount, and dosage of medicine required for the patient's treatment. The medical information processing unit 48 extracts information about the type, amount, and dosage of medicine required for the patient's treatment from the text data TD and outputs the prescription.

[0083] Furthermore, the medical information processing unit 48 extracts symptom text information regarding the patient's symptoms from the text data TD, searches for drug name information corresponding to the symptom text information from the pharmaceutical information based on the symptom text information and at least one of the efficacy information, side effect information, and combined use precautions information, and outputs the searched drug name information.

[0084] Symptom text information is information about a patient's symptoms, such as "I have a headache" or "I feel dizzy," spoken by a doctor or a patient.

[0085] The medical information processing unit 48 can output drug name information that alleviates the patient's symptoms contained in the symptom text information based on the symptom text information and the drug efficacy information. For example, if the symptom text information is "I have a headache," the medical information processing unit 48 searches the pharmaceutical information for headache medications that alleviate the headache and outputs drug name information for the headache medications found. This makes it possible to provide doctors with highly useful information.

[0086] Furthermore, the medical information processing unit 48 can output information on the name of a drug that improves the patient's symptoms contained in the symptom text information based on the symptom text information and the side effect information. For example, if the symptom text information is "I feel dizzy," the medical information processing unit 48 searches the pharmaceutical information for information on the name of a drug that may cause the side effect "dizziness," and outputs the searched information. This makes it possible to provide doctors with highly useful information.

[0087] Furthermore, the medical information processing unit 48 can output information on the names of drugs that can be used in combination with drugs already prescribed to the patient, based on the symptom text information and the concomitant use warning information. For example, if the symptom text information is "I have a headache," the medical information processing unit 48 can output information on headache medications that can be used in combination with drugs already prescribed to the patient. This makes it possible to provide doctors with highly useful information.

[0088] In addition, the medical information processing unit 48 shown in Figure 6 executes a process of generating medical information about a patient (medical information processing process) by inputting multiple medical literature and text data TD into the artificial intelligence model M using the processor 21 shown in Figure 2.

[0089] The medical information processing unit 48 extracts at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease from the text data TD. The symptom text information is not particularly limited, and is information related to the patient's symptoms, such as, for example, "I have a headache" or "I can't sleep at night." There is no limitation on who spoke the symptom text information, and it may be the patient, a doctor, a nurse, or someone accompanying the patient. The disease name text information is not particularly limited, and is information related to the patient's disease, such as, for example, "migraine" or "insomnia." There is no limitation on who spoke the disease name text information, and it may be the patient, a doctor, a nurse, or someone accompanying the patient.

[0090] The medical information processing unit 48 searches a plurality of medical literatures for at least one evidential medical literature that includes information corresponding to the symptom text information or disease name text information. The medical literature obtained from the medical literature database 16 contains various medical or healthcare information. Among this medical information, the medical literature that includes information corresponding to the symptom text information or disease name text information is extracted as the evidential medical literature.

[0091] The information contained in the medical literature does not have to exactly match the symptom text information or disease name text information. For example, if the symptom text information is "I have trouble falling asleep at night," "I can't fall asleep easily," or "I stay awake for hours after getting into bed," and the medical literature contains a statement that "I have difficulty falling asleep," it is determined that the medical literature contains information corresponding to the symptom text information, and the medical literature is extracted as evidence-based medical literature.

[0092] Evidence-based medical literature is medical literature that contains information that can serve as the basis for medical decisions and medical procedures made by doctors, such as diagnosing a patient's illness and prescribing medication.

[0093] The medical information processing unit 48 outputs the evidence-based medical literature in any output format, and may output the evidence-based medical literature as audio from the speaker 24b of the terminal device 12, as a display on the screen 24a of the terminal device 12, or as a printout from a printer (not shown).

[0094] The occurrence of hallucination based on inaccurate evidence can be prevented by outputting evidence-based medical literature from the medical information processing unit 48. Furthermore, the output of evidence-based medical literature can provide doctors with highly useful information.

[0095] Furthermore, the medical information processing unit 48 generates a plurality of pieces of medical literature summary information that summarize each of the plurality of medical literatures. The medical information processing unit 48 associates the medical literature with the medical literature summary information that summarizes the medical literature, and stores the medical literature summary information in the storage 14.

[0096] The medical information processing unit 48 outputs the evidence-based medical literature and medical literature summary information that summarizes the evidence-based medical literature. By checking the medical literature summary information, doctors can easily understand the content of the evidence-based medical literature.

[0097] Furthermore, the medical information processing unit 48 extracts medical literature citation information, which is a description related to at least one of the symptom text information and the disease name text information, from the evidence medical literature. The medical information processing unit 48 associates the medical literature with the medical literature citation information extracted from the medical literature and stores them in the storage 14.

[0098] The medical literature citation information is a specific description in the extracted evidence-based medical literature that corresponds to at least one of the symptom text information and the disease name text information. The medical literature citation information is not particularly limited and may be a phrase, a section, one or more sentences, or one or more paragraphs. The medical literature citation information may also be one or more extracted passages from the evidence-based medical literature.

[0099] The medical information processing unit 48 outputs the evidence-based medical literature and medical literature citation information extracted from the evidence-based medical literature. By checking the medical literature citation information, doctors can easily confirm the basis for their own medical decisions and medical procedures.

[0100] Furthermore, the medical information processing unit 48 generates a plurality of pieces of translated medical literature information into which each of the plurality of medical literatures has been translated. The medical information processing unit 48 associates the medical literature with the translated medical literature information into which the medical literature has been translated and stores the associated information in the storage 14. The medical information processing unit 48 outputs the evidential medical literature and the translated medical literature information into which the evidential medical literature has been translated. By checking the translated medical literature information, doctors can easily confirm the basis for their own medical decisions and medical procedures, even if the evidential medical literature is not written in the doctor's native language.

[0101] Furthermore, the medical information processing unit 48 may generate summary translation information by translating the medical literature summary information. The medical information processing unit 48 may also generate summary translation information by summarizing the medical literature translation information. The medical information processing unit 48 associates the medical literature with the summary translation information generated from the medical literature and stores them in the storage 14. The medical information processing unit 48 outputs the evidence-based medical literature and the summary translation information generated from the evidence-based medical literature. By checking the summary translation information, doctors can easily confirm the basis for their medical decisions and medical procedures, even if the evidence-based medical literature is not written in the doctor's native language.

[0102] Furthermore, the medical information processing unit 48 may generate citation translation information by translating the medical literature citation information. The medical information processing unit 48 may also extract citation translation information, which is a description related to at least one of the symptom text information and the disease name text information, from the medical literature translation information. The medical information processing unit 48 associates the medical literature with the citation translation information generated from the medical literature and stores them in the storage 14. The medical information processing unit 48 outputs the evidence-based medical literature and the citation translation information generated from the evidence-based medical literature. By checking the citation translation information, doctors can easily confirm the basis for their medical decisions and medical procedures, even if the evidence-based medical literature is not written in the doctor's native language.

[0103] Furthermore, the medical information processing unit 48 generates tag information that is extracted from terms contained in each of the plurality of medical documents and associated with each of the plurality of medical documents. The medical information processing unit 48 associates the medical documents with the tag information and stores them in the storage 14. The medical information processing unit 48 searches the evidence-based medical documents using the tag information and at least one of the symptom text information and the disease name text information. This improves the efficiency of searching for evidence-based medical documents from among the plurality of medical documents.

[0104] Furthermore, the text data conversion unit 42 generates a plurality of divided text data DD by converting the audio data into text data TD at predetermined time intervals. The divided text data DD is included in the text data TD. The predetermined time is not particularly limited, and is preferably between 1 second and 5 minutes, more preferably between 5 seconds and 3 minutes, and even more preferably between 8 seconds and 1 minute. In the first embodiment, it is set to 10 seconds.

[0105] The medical information processing unit 48 inputs a predetermined number of pieces of divided text data DD into the artificial intelligence model M, thereby generating summarized text data SD in which the predetermined number of pieces of divided text data DD are summarized. The summarized text data SD is included in the text data TD. The predetermined number is not particularly limited, and is preferably between 1 and 20, more preferably between 2 and 15, and even more preferably between 3 and 10. In the first embodiment, it is set to 4.

[0106] The medical information processing unit 48 inputs the divided text data DD and the summarized text data SD into the artificial intelligence model M at predetermined time intervals, thereby extracting at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease name from the divided text data DD and the summarized text data SD. The predetermined time is not particularly limited, and is preferably from 1 second to 5 minutes, more preferably from 5 seconds to 3 minutes, and even more preferably from 8 seconds to 1 minute. In the first embodiment, it is set to 10 seconds.

[0107] The audio data AD, divided text data DD, and summarized text data SD will be described with reference to Fig. 8. However, the times shown in Fig. 8 are merely an example and do not limit the invention.

[0108] 8(a) shows the voice data AD, the divided text data DD, and the summary text data SD 50 seconds after the doctor started the examination. The top part of FIG. 8(a) shows the state in which 50 seconds of voice data AD has been acquired.

[0109] The lower part of Figure 8(a) shows divided text data DD converted from 10 seconds of audio data AD from 30 to 40 seconds after the start of the consultation, and summarized text data SD obtained by summarizing four pieces of divided text data DD. The divided text data DD contains information obtained by converting audio data AD spoken by a doctor or patient into text data TD. The summarized text data SD contains information obtained by summarizing four pieces of divided text data DD, i.e., from 0 to 40 seconds after the start of the consultation. The medical information processing unit 48 inputs the divided text data DD and summarized text data SD shown in Figure 8(a) into the artificial intelligence model M. This allows the artificial intelligence model M to obtain information related to a total of 50 seconds of audio data AD.

[0110] 8(b) shows the voice data AD, the divided text data DD, and the summary text data SD 60 seconds after the doctor started the examination. The upper part of FIG. 8(b) shows the state in which 60 seconds of voice data AD has been acquired.

[0111] The lower part of Figure 8(b) shows divided text data DD converted from 10 seconds of voice data AD from 50 to 60 seconds after the start of the consultation, and summarized text data SD summarizing four of the divided text data DD. The divided text data DD contains information obtained by converting the voice data AD spoken by the doctor or patient into text data TD. The summarized text data SD contains information summarizing four of the summarized text data SD, i.e., from 10 to 50 seconds after the start of the consultation. The medical information processing unit 48 inputs the divided text data DD and summarized text data SD shown in Figure 8(b) into the artificial intelligence model M. This allows the artificial intelligence model M to obtain information related to a total of 50 seconds of voice data AD.

[0112] 8(c) shows the voice data AD, the divided text data DD, and the summary text data SD 70 seconds after the doctor started the examination. The top part of FIG. 8(c) shows the state in which 70 seconds of voice data AD have been acquired.

[0113] The lower part of Figure 8(c) shows divided text data DD converted from 10 seconds of audio data AD from 60 to 70 seconds after the start of the consultation, and summarized text data SD obtained by summarizing four pieces of divided text data DD. The divided text data DD contains information obtained by converting audio data AD spoken by a doctor or patient into text data TD. The summarized text data SD contains information obtained by summarizing four pieces of divided text data DD, i.e., information obtained by summarizing the four pieces of divided text data DD from 20 to 60 seconds after the start of the consultation. The medical information processing unit 48 inputs the divided text data DD and summarized text data SD shown in Figure 8(c) into the artificial intelligence model M. This allows the artificial intelligence model M to obtain information related to a total of 50 seconds of audio data AD.

[0114] Similarly, the medical information processing unit 48 inputs the divided text data DD and the summarized text data SD to the artificial intelligence model M. As a result, the artificial intelligence model M acquires the divided text data DD corresponding to the most recent speech data AD and the summarized text data SD obtained by summarizing the multiple divided text data DD, and can therefore accurately extract at least one of the symptom text information and the disease name text information.

[0115] 5. Operation of Medical Support System 11 Next, an outline of the operation of the medical assistance system 11 according to the first embodiment will be described with reference to FIG.

[0116] A system prompt is set in advance in the artificial intelligence model M. The system prompt is not particularly limited, but can be set, for example, as follows: "You are an artificial intelligence that supports doctors. You listen to patients or doctors and make suggestions to doctors. The tools given to you are drug information search and medical literature search. For drug information search, you can use PDMA, for example. For medical literature search, you can use PubMed, for example." The system prompt is not limited to the above, and any system prompt can be set in the artificial intelligence model M.

[0117] When the medical assistance system 11 is started, the drug information acquisition unit 44 in Fig. 6 acquires drug information from the drug information database 15. The drug information storage unit 45 in Fig. 6 stores the acquired drug information as cache data CD in the storage 14 in Fig. 1.

[0118] 6 acquires medical literature from the medical literature database 16. The medical literature storage unit 47 in Fig. 6 stores the acquired medical literature as cache data CD in the storage 14 in Fig. 1.

[0119] The medical information processing unit 48 in FIG. 6 extracts the drug name information, efficacy information, side effect information, and caution information for concomitant use shown in FIG. 4 from the pharmaceutical information, associates them with each other, and stores them as cache data CD in the storage 14 in FIG.

[0120] The medical information processing unit 48 in Figure 6 generates medical literature summary information, medical literature citation information, medical literature translation information, summary translation information, citation translation information, and tag information shown in Figure 4 from medical literature, associates them with each other, and stores them as cache data CD in the storage 14 in Figure 1.

[0121] Next, when the patient's examination begins and the patient or doctor speaks, the voice data acquisition unit 41 in Fig. 6 acquires voice data AD from the patient's or doctor's speech. The text data conversion unit 42 in Fig. 6 converts the voice data AD into text data TD. The text data acquisition unit 43 in Fig. 6 acquires the text data TD.

[0122] The medical information processing unit 48 in FIG. 6 inputs the text data TD to the artificial intelligence model M.

[0123] A user prompt is input as text data TD to the artificial intelligence model M. The user prompt is not particularly limited, and may be in the form of an instruction or question to the artificial intelligence model M, such as a doctor's utterance such as "Please correct the drug name," "Please tell me the correct drug name," or "Please tell me the medical literature that serves as evidence." The doctor may also input the text data TD using the touch panel 23b or a keyboard.

[0124] Furthermore, a user prompt may be configured to be input to the artificial intelligence model M when a predetermined keyword is included in the text data. For example, when the text data TD includes drug name text information, a user prompt such as "Search for candidate drug name information that matches at least a portion of the drug name text information from the drug name information in the pharmaceutical information" may be input to the artificial intelligence model M. Furthermore, when the text data TD includes symptom text information or disease name text information, a user prompt such as "Search for at least one evidence-based medical literature from multiple medical literature" may be input to the artificial intelligence model M.

[0125] Also, for example, a system prompt may be set in advance as "If the text data contains a drug name, search for candidate drug names that match at least part of the drug name text information from the drug name information in the pharmaceutical information." If the text data TD contains drug name text information, the artificial intelligence model M may use this drug name text information as a user prompt to search for candidate drug names.

[0126] The medical information processor 48 of FIG. 6 causes the artificial intelligence model M to perform various steps based on user prompts.

[0127] The medical information processing unit 48 in FIG. 6 extracts drug name text information that is estimated to be a drug name from the input text data TD, searches for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information, and compares the candidate drug name information with the drug name text information. Part of In this case, the drug name text information is corrected to candidate drug name information, or the candidate drug name information is output.

[0128] 6 extracts at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease from the text data TD, and inputs the plurality of medical literatures, the symptom text information, and at least one of the disease name text information to the artificial intelligence model M. As a result, the medical information processing unit 48 searches the plurality of medical literatures for at least one evidential medical literature that includes information corresponding to the symptom text information or the disease name text information, and outputs the evidential medical literature.

[0129] The medical information processing unit 48 in Figure 6 may output only the evidence medical literature, or may output the evidence medical literature and at least one selected from the group consisting of medical literature summary information, medical literature citation information, medical literature translation information, summary translation information, and citation translation information.

[0130] Of the information output by AI model M, the information enclosed by the dashed line is checked by the doctor during the patient's medical examination and is used to determine the patient's medical treatment plan. In other words, the doctor seamlessly obtains information from AI model M during the examination and, while referring to the obtained information, makes a final decision on the patient's medical treatment plan. This saves the time required to investigate the basis for the medical treatment plan, thereby improving medical treatment efficiency.

[0131] When the doctor confirms the information output by the artificial intelligence model M, the medical information processing unit 48 in Figure 6 outputs medical documents. The medical documents include electronic medical records, lifestyle-related disease treatment plans, prescriptions, etc. This completes the operation of the medical care support system 11.

[0132] 6. Effects of Embodiment 1 Next, a description will be given of the effects of embodiment 1. The medical care support system 11 of embodiment 1 includes a drug information acquisition unit 44, a drug information storage unit 45, a text data acquisition unit 43, and a medical information processing unit .

[0133] The drug information acquisition unit 44 is configured to acquire drug information including drug name information from the drug information database 15 on a regular or irregular basis. The drug information storage unit 45 is configured to store drug information. The text data acquisition unit 43 is configured to acquire text data TD including drug names. The medical information processing unit 48 is configured to process the medical information included in the text data TD by inputting the drug information and the text data TD into the artificial intelligence model M.

[0134] The medical information processing unit 48 extracts drug name text information estimated to be a drug name from the text data TD, searches for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information, and compares the candidate drug name information with the drug name text information. Part of In this case, the drug name text information is corrected to candidate drug name information, or the candidate drug name information is output.

[0135] In the medical care support method of the first embodiment, the processor 21 executes a drug information acquisition step, a drug information storage step, a text data acquisition step, and a medical information processing step.

[0136] The drug information acquisition step periodically or irregularly acquires drug information including drug name information from the drug information database 15. The drug information storage step stores the drug information in the memory 22. The text data acquisition step acquires text data TD including drug names. The medical information processing step processes the medical information included in the text data TD by inputting the drug information and the text data TD into the artificial intelligence model M.

[0137] The medical information processing step extracts drug name text information estimated to be a drug name from the text data TD, searches for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information, and compares the candidate drug name information with the drug name text information. Part of In this case, the drug name text information is corrected to candidate drug name information, or the candidate drug name information is output.

[0138] The medical support program of embodiment 1 is a medical support program executed by at least one processor 21, and causes at least one processor 21 to execute a drug information acquisition process, a drug information storage process, a text data acquisition process, and a medical information processing process.

[0139] The drug information acquisition step periodically or irregularly acquires drug information including drug name information from the drug information database 15. The drug information storage step stores the drug information in the storage 14. The text data acquisition step acquires text data TD including drug names. The medical information processing step processes the medical information included in the text data TD by inputting the drug information and the text data TD into the artificial intelligence model M.

[0140] The medical information processing step extracts drug name text information estimated to be a drug name from the text data TD, searches for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information, and compares the candidate drug name information with the drug name text information. Part of In this case, the drug name text information is corrected to candidate drug name information, or the candidate drug name information is output.

[0141] According to the first embodiment, accurate drug names can be output, making it possible to provide doctors with highly useful medical information.

[0142] According to the first embodiment, a doctor can confirm the correct name of a drug while examining a patient. As a result, the doctor can save time and effort in checking literature or searching databases to confirm the correct name of a drug, thereby improving the efficiency of medical practice.

[0143] The drug information according to the first embodiment further includes efficacy information associated with the drug name information. The medical information processing unit 48 according to the first embodiment further extracts efficacy text information estimated as the drug efficacy from the text data TD, and combines the candidate drug name information and the drug name text information. Part of In this case, and when the efficacy information and the efficacy text information match, the drug name text information is corrected to candidate drug name information, or the candidate drug name information is output.

[0144] According to the first embodiment, the accuracy of a drug name can be determined using efficacy information and efficacy text information, thereby improving the accuracy of the drug name, thereby further improving the usefulness of medical information.

[0145] The medical information processing unit 48 according to the first embodiment is further configured to output a medical document based on the text data TD and the drug name text information corrected to the candidate drug name information. According to the first embodiment, the medical document can be output based on the accurate drug name, which further improves the usefulness of the medical information.

[0146] Moreover, the medical document according to the first embodiment is an electronic medical record. Therefore, according to the first embodiment, it is possible to improve the utility value of the electronic medical record.

[0147] The drug information according to the first embodiment further includes at least one of efficacy information, side effect information, and caution information regarding concomitant use with other drugs, all of which are associated with the drug name information. The medical information processing unit 48 is further configured to extract symptom text information regarding the patient's symptoms from the text data TD, search for drug name information corresponding to the symptom text information from the drug information based on the symptom text information and at least one of efficacy information, side effect information, and caution information regarding concomitant use with other drugs, and output the searched drug name information.

[0148] According to the first embodiment, it is possible to search for correct drug name information from at least one of efficacy information, side effect information, and caution information on concomitant use with other drugs. This allows the correct drug name to be searched for based on information other than the drug name. This makes it possible to perform a so-called reverse lookup search, further improving the usefulness of medical information.

[0149] The medical support system 11 according to the first embodiment further includes a medical literature acquisition unit 46 and a medical literature storage unit 47.

[0150] The medical literature acquisition unit 46 is configured to periodically or irregularly acquire a plurality of medical literature from the medical literature database 16 in which medical literature is stored. The medical literature storage unit 47 is configured to store a plurality of medical literature. The text data TD further includes medical information relating to the medical treatment of the patient.

[0151] The medical information processing unit 48 is configured to generate medical information related to a patient by inputting a plurality of medical literatures and text data TD into the artificial intelligence model M. The medical information processing unit 48 is configured to extract at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease name from the text data TD, search the plurality of medical literatures for at least one evidential medical literature containing information corresponding to the symptom text information or disease name text information, and output the evidential medical literature.

[0152] The medical support system 11 according to the first embodiment includes a medical literature acquisition unit 46, a medical literature storage unit 47, a text data acquisition unit 43, and a medical information processing unit .

[0153] The medical literature acquisition unit 46 is configured to periodically or irregularly acquire a plurality of medical literatures from the medical literature database 16 that stores medical literatures. The medical literature storage unit 47 is configured to store a plurality of medical literatures. The text data acquisition unit 43 is configured to acquire text data TD including medical information related to the medical treatment of a patient. The medical information processing unit 48 is configured to generate medical information related to a patient by inputting the plurality of medical literatures and the text data TD into the artificial intelligence model M.

[0154] The medical information processing unit 48 is configured to extract at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease name from the text data TD, search for at least one evidence-based medical literature containing information corresponding to the symptom text information or disease name text information from multiple medical literatures, and output the evidence-based medical literature.

[0155] The medical support method according to the first embodiment further includes a medical literature acquisition step and a medical literature storage step.

[0156] The medical literature acquisition step periodically or irregularly acquires a plurality of medical literatures from a medical literature database, and the medical literature storage step stores the plurality of medical literatures.

[0157] The medical information processing step generates medical information about a patient by inputting a plurality of medical literatures and text data TD into an artificial intelligence model. The medical information processing step is configured to extract at least one of symptom text information about the patient's symptoms and disease name text information about the patient's disease name from the text data TD, search the plurality of medical literatures for at least one evidential medical literature containing information corresponding to the symptom text information or disease name text information, and output the evidential medical literature.

[0158] The medical care support method according to the first embodiment includes a medical literature acquisition step, a medical literature storage step, a text data acquisition step, and a medical information processing step, which are executed by the processor 21.

[0159] The medical literature acquisition step periodically or irregularly acquires a plurality of medical literatures from the medical literature database 16. The medical literature storage step stores a plurality of medical literatures in the memory 22. The text data acquisition step acquires text data TD including medical information related to the treatment of a patient. The medical information processing step generates medical information related to the patient by inputting a plurality of medical literatures and the text data TD into the artificial intelligence model M.

[0160] The medical information processing process extracts at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease name from the text data TD, and inputs a plurality of medical literatures and at least one of the symptom text information and disease name text information into the artificial intelligence model M, thereby searching for at least one evidential medical literature that contains information corresponding to the symptom text information or disease name text information from the plurality of medical literatures, and outputting the evidential medical literature.

[0161] The medical assistance program according to the first embodiment further causes at least one processor to execute a medical literature acquisition step and a medical literature storage step.

[0162] The medical literature acquisition step periodically or irregularly acquires a plurality of medical literatures from a medical literature database, and the medical literature storage step stores the plurality of medical literatures.

[0163] The medical information processing step generates medical information about a patient by inputting a plurality of medical literatures and text data TD into an artificial intelligence model. The medical information processing step is configured to extract at least one of symptom text information about the patient's symptoms and disease name text information about the patient's disease name from the text data TD, search the plurality of medical literatures for at least one evidential medical literature containing information corresponding to the symptom text information or disease name text information, and output the evidential medical literature.

[0164] The medical support program of embodiment 1 is a medical support program executed by at least one processor 21, and causes at least one processor 21 to execute a medical literature acquisition process, a medical literature storage process, a text data acquisition process, and a medical information processing process.

[0165] The medical literature acquisition step acquires multiple medical literatures from the medical literature database 16 on a regular or irregular basis. The medical literature storage step stores multiple medical literatures in the storage 14. The text data acquisition step acquires text data TD including medical information related to the treatment of a patient. The medical information processing step generates medical information related to the patient by inputting multiple medical literatures and the text data TD into the artificial intelligence model M.

[0166] The medical information processing process extracts at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease name from the text data TD, and inputs a plurality of medical literatures and at least one of the symptom text information and disease name text information into the artificial intelligence model M, thereby searching for at least one evidential medical literature that contains information corresponding to the symptom text information or disease name text information from the plurality of medical literatures, and outputting the evidential medical literature.

[0167] According to the first embodiment, the artificial intelligence model M is prevented from making decisions based on uncertain information, thereby suppressing hallucination, thereby providing highly useful medical information.

[0168] The medical information processing unit 48 according to the first embodiment is further configured to generate a plurality of medical literature summary information summarizing each of a plurality of medical literatures, and to output the medical literature summary information summarizing the evidence-based medical literature. This allows doctors to check not only the evidence-based medical literature itself but also its summaries, making it easier to understand the content of the evidence-based medical literature. Furthermore, this reduces the time required to determine a treatment policy, thereby improving the overall efficiency of medical procedures.

[0169] The medical information processing unit 48 according to the first embodiment is further configured to extract medical literature citation information, which is a description related to at least one of the symptom text information and the disease name text information, from the evidence-based medical literature, and to output the medical literature citation information extracted from the evidence-based medical literature.

[0170] According to the first embodiment, by checking the medical literature citation information, doctors can clearly understand the basis for their own medical decisions and medical procedures. In addition, the time required to decide on a medical treatment policy can be shortened, thereby improving the efficiency of medical treatment overall.

[0171] The medical information processing unit 48 according to the first embodiment is further configured to generate a plurality of medical literature translation information in which each of a plurality of medical literatures is translated, and to output the medical literature translation information in which the evidence medical literature is translated.

[0172] According to the first embodiment, by checking the medical literature translation information, doctors can easily confirm the basis for their own medical decisions and medical procedures, even if the evidence-based medical literature is not written in the doctor's native language. In addition, the time required to decide on a medical treatment policy can be reduced, thereby improving the overall efficiency of medical treatment.

[0173] According to the first embodiment, the medical information processing unit 48 is further configured to generate summary translation information and output the summary translation information generated from medical literature summary information that summarizes evidence-based medical literature. By checking the summary translation information, doctors can easily confirm the basis for their own medical decisions and medical procedures, even if the evidence-based medical literature is not written in the doctor's native language. Furthermore, the time required to determine a medical treatment policy can be shortened, thereby improving the overall efficiency of medical treatment.

[0174] The medical information processing unit 48 according to the first embodiment is further configured to generate citation translation information and output the citation translation information generated from medical literature citation information extracted from evidence-based medical literature. By checking the citation translation information, doctors can easily confirm the basis for their own medical decisions and medical procedures, even if the evidence-based medical literature is not written in their native language. Furthermore, the time required to determine a medical treatment policy can be shortened, thereby improving the overall efficiency of medical treatment.

[0175] In the first embodiment, the medical information processing unit 48 is further configured to generate tag information extracted from terms contained in each of the plurality of medical documents and associated with each of the plurality of medical documents, and to search for evidence-based medical documents using the tag information and at least one of symptom text information and disease name text information. This improves the efficiency of searching for evidence-based medical documents from among the plurality of medical documents. As a result, the efficiency of medical procedures can be improved.

[0176] In the first embodiment, the medical assistance system 11 further includes a text data conversion unit 42 that converts voice data AD uttered by a doctor or a patient into text data TD. The text data conversion unit 42 generates a plurality of divided text data DD by converting the voice data AD into text data TD at predetermined time intervals.

[0177] The medical information processing unit 48 is further configured to generate summarized text data SD in which a predetermined number of divided text data DD are summarized by inputting a predetermined number of divided text data DD into the artificial intelligence model M, and to extract at least one of symptom text information related to the patient's symptoms and disease name text information related to the patient's disease name from the divided text data DD and the summarized text data SD by inputting the divided text data DD and the summarized text data SD into the artificial intelligence model M at predetermined time intervals.

[0178] According to the first embodiment, the artificial intelligence model M acquires the divided text data DD and the summarized text data SD, and therefore can extract at least one of the symptom text information and the disease name text information with high accuracy.

[0179] (Embodiment 2) Next, a medical assistance system 11 according to a second embodiment will be described with reference to Fig. 10. Note that, among the symbols used in the second and subsequent embodiments, the same symbols as those used in the previous embodiments represent the same components as those in the previous embodiments unless otherwise specified.

[0180] In the second embodiment, the artificial intelligence model M and the cache data CD are stored in the memory 22 of the terminal device 12. In the second embodiment, the artificial intelligence computer and storage are omitted.

[0181] According to the second embodiment, the medical assistance system 11 functions stably regardless of the communication state of the network, thereby further improving the efficiency of medical treatment.

[0182] The present invention is not limited to the above-described embodiments, and can be applied to various embodiments within the scope of the present invention. [Explanation of symbols]

[0183] 11 Medical Support System 14. Storage 15 Drug Information Database 16 Medical Literature Database 21 processors 22 Memory 43 Text data acquisition unit 44 Drug Information Acquisition Department 45 Drug information storage unit 46 Medical Literature Acquisition Department 47 Medical Literature Storage Department 48 Medical Information Processing Department AD audio data DD divided text data M Artificial Intelligence Model SD Summary Text Data TD text data

Claims

1. a drug information acquisition unit configured to periodically or irregularly acquire drug information including drug name information from a drug information database; a drug information storage unit configured to store the drug information; a text data acquisition unit configured to acquire text data including a drug name; a medical information processing unit configured to perform predetermined processing of system prompts using an artificial intelligence model that is a large-scale language model; The predetermined processing is extracting drug name text information estimated to be a drug name from the text data input to the artificial intelligence model; Searching for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information input to the artificial intelligence model; A medical support system that, when a portion of the candidate drug name information and the drug name text information match, corrects the drug name text information to the candidate drug name information or outputs the candidate drug name information.

2. The medical assistance system according to claim 1 , wherein the text data is generated based on voice data uttered by a doctor or a patient.

3. The pharmaceutical information further includes: The drug name information includes drug effect information associated with the drug name information, The predetermined processing further includes: extracting text information on drug efficacy estimated as drug efficacy from the text data input to the artificial intelligence model; A medical support system as described in claim 1 or 2, wherein when the candidate drug name information and the drug name text information partially match, and when the efficacy information and the efficacy text information match, the drug name text information is corrected to the candidate drug name information or the candidate drug name information is output.

4. The medical support system of claim 1 or 2, wherein the predetermined processing further outputs a medical document based on the text data input to the artificial intelligence model and the drug name text information corrected to the candidate drug name information.

5. The medical support system according to claim 4 , wherein the medical document is an electronic medical record.

6. The pharmaceutical information input into the artificial intelligence model further comprises: At least one of information on efficacy, information on side effects, and information on precautions for use in combination with other drugs is included, which are associated with the drug name information; The predetermined processing further includes: extracting symptom text information relating to a patient's symptoms from the text data input to the artificial intelligence model; searching for the drug name information corresponding to the symptom text information from the pharmaceutical information based on the symptom text information and at least one of the efficacy information, the side effect information, and the combined use precautions information; The medical support system according to claim 1 or 2, wherein the searched drug name information is output.

7. The medical support system further comprises: a medical literature acquisition unit configured to periodically or irregularly acquire a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage unit configured to store a plurality of the medical literatures; the text data input to the artificial intelligence model further includes clinical information related to a patient's medical care; The predetermined processing is extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data input to the artificial intelligence model; Searching at least one evidence-based medical literature that includes information corresponding to the symptom text information or the disease name text information from the plurality of medical literatures input to the artificial intelligence model; The medical support system according to claim 1 or 2, wherein the evidence-based medical literature is output.

8. The predetermined processing further includes: generating a plurality of medical literature summary information summarizing each of the plurality of medical literatures input to the artificial intelligence model; The medical support system according to claim 7 , wherein the medical literature summary information in which the evidence-based medical literature is summarized is output.

9. The predetermined processing further includes: extracting medical literature citation information, which is a description related to at least one of the symptom text information and the disease name text information, from the evidence-based medical literature; The medical support system according to claim 7 , wherein the medical literature citation information extracted from the evidence-based medical literature is output.

10. The predetermined processing further includes: generating a plurality of medical literature translation information pieces in which each of the plurality of medical literature pieces input to the artificial intelligence model is translated; The medical support system according to claim 7 , wherein the medical literature translation information obtained by translating the evidence-based medical literature is output.

11. The predetermined processing further includes: generating a translated summary of the medical literature summary information; The medical support system according to claim 8 , wherein the translation information is generated from the medical literature summary information in which the evidence-based medical literature is summarized.

12. The predetermined processing further includes: generating a summary translation of the medical literature translation information; The medical support system according to claim 10, wherein the summary translation information generated from the medical literature translation information obtained by translating the evidence-based medical literature is output.

13. The predetermined processing further includes: generating translation information of the medical literature citation; The medical support system according to claim 9, wherein the translation of the citation information is generated from the medical literature citation information extracted from the evidence-based medical literature.

14. The predetermined processing further includes: extracting, from the medical literature translation information, citation translation information that is a description related to at least one of the symptom text information and the disease name text information; The medical support system according to claim 10, wherein the citation translation information generated from the medical literature translation information obtained by translating the evidence-based medical literature is output.

15. The predetermined processing further includes: generating tag information extracted from terms contained in each of the plurality of medical documents input to the artificial intelligence model and associated with each of the plurality of medical documents; The medical support system according to claim 7 , wherein the evidence-based medical literature is searched using the tag information and at least one of the symptom text information and the disease name text information.

16. The medical support system further comprises: a text data conversion unit that converts voice data uttered by a doctor or the patient into text data; The text data conversion unit generating a plurality of divided text data by converting the voice data into the text data at predetermined time intervals; The predetermined processing further includes: generating summarized text data by summarizing a predetermined number of the divided text data input to the artificial intelligence model; 8. The medical support system according to claim 7, wherein at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name is extracted from the divided text data input to the artificial intelligence model and the summarized text data input to the artificial intelligence model at each predetermined time.

17. a medical literature acquisition unit configured to periodically or irregularly acquire a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage unit configured to store a plurality of the medical literatures; a text data acquisition unit configured to acquire text data including medical information related to a medical treatment of a patient; a medical information processing unit configured to perform predetermined processing of system prompts using an artificial intelligence model that is a large-scale language model; The predetermined processing is extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data input to the artificial intelligence model; Searching at least one evidence-based medical literature that includes information corresponding to the symptom text information or the disease name text information from the plurality of medical literatures input to the artificial intelligence model; A medical support system that outputs the evidence-based medical literature.

18. The processor a drug information acquisition step of periodically or irregularly acquiring drug information including drug name information from a drug information database; a drug information storage step of storing the drug information in a memory; a text data acquisition step of acquiring text data including a drug name; a medical information processing step of executing predetermined processing of system prompts using an artificial intelligence model, which is a large-scale language model; The predetermined processing is extracting drug name text information estimated to be a drug name from the text data input to the artificial intelligence model; Searching for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information input to the artificial intelligence model; A medical support method in which, when a portion of the candidate drug name information and the drug name text information matches, the drug name text information is corrected to the candidate drug name information or the candidate drug name information is output.

19. The medical assistance method further comprises: a medical literature acquisition step of periodically or irregularly acquiring a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage step of storing a plurality of said medical literatures; the text data further includes medical information relating to the patient's medical care; The predetermined processing is extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data input to the artificial intelligence model; Searching at least one evidence-based medical literature that includes information corresponding to the symptom text information or the disease name text information from the plurality of medical literatures input to the artificial intelligence model; The medical support method according to claim 18, wherein the evidence-based medical literature is output.

20. The processor a medical literature acquisition step of periodically or irregularly acquiring a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage step of storing a plurality of the medical literatures in a memory; a text data acquisition step of acquiring text data including medical information related to the medical treatment of a patient; a medical information processing step of executing predetermined processing of system prompts using an artificial intelligence model, which is a large-scale language model; The predetermined processing is extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data input to the artificial intelligence model; Searching at least one evidence-based medical literature that includes information corresponding to the symptom text information or the disease name text information from the plurality of medical literatures input to the artificial intelligence model; A medical support method that outputs the evidence-based medical literature.

21. A medical assistance program executed by at least one processor, a drug information acquisition step of periodically or irregularly acquiring drug information including drug name information from a drug information database; a drug information storage step of storing the drug information in a memory; a text data acquisition step of acquiring text data including a drug name; and a medical information processing step of performing predetermined processing of system prompts using an artificial intelligence model that is a large-scale language model, The predetermined processing is extracting drug name text information estimated to be a drug name from the text data input to the artificial intelligence model; Searching for candidate drug name information that matches at least a part of the drug name text information from the drug name information of the pharmaceutical product information input to the artificial intelligence model; A medical support program that, when a portion of the candidate drug name information and the drug name text information match, corrects the drug name text information to the candidate drug name information or outputs the candidate drug name information.

22. The medical assistance program further comprises: a medical literature acquisition step of periodically or irregularly acquiring a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage step of storing a plurality of said medical literatures by said at least one processor; the text data further includes medical information relating to the patient's medical care; The predetermined processing is extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data input to the artificial intelligence model; Searching at least one evidence-based medical literature that includes information corresponding to the symptom text information or the disease name text information from the plurality of medical literatures input to the artificial intelligence model; The medical support program according to claim 21 , wherein the medical literature on evidence-based medicine is output.

23. A medical assistance program executed by at least one processor, a medical literature acquisition step of periodically or irregularly acquiring a plurality of medical literature from a medical literature database in which medical literature is stored; a medical literature storage step of storing a plurality of the medical literatures in a memory; a text data acquisition step of acquiring text data including medical information related to the medical treatment of a patient; and a medical information processing step of performing predetermined processing of system prompts using an artificial intelligence model that is a large-scale language model, The predetermined processing is extracting at least one of symptom text information relating to the patient's symptoms and disease name text information relating to the patient's disease name from the text data input to the artificial intelligence model; Searching at least one evidence-based medical literature that includes information corresponding to the symptom text information or the disease name text information from the plurality of medical literatures input to the artificial intelligence model; A medical support program that outputs the evidence-based medical literature.

Images