Interrogation teaching method and system based on large language model

Abstract

The invention discloses an inquiry teaching method and system based on a large language model, and the method comprises the steps: randomly extracting to-be-recombined data from a pre-constructed patient play background library, a patient disease question and answer library and a disease picture library, recombining each piece of data to be recombined to generate a virtual digital patient object for interacting with the user; in the interrogation teaching process, performing similarity analysis on the interrogation question output by the user and the patient disease question and answer library, and screening and determining a matching result corresponding to the interrogation question from the patient disease question and answer library; and obtaining reference inquiry reply data according to a matching result, and inputting the reference inquiry reply data into the constructed big language inquiry teaching model to obtain a reply result which is output by the big language inquiry teaching model and meets the inquiry teaching standard so as to realize virtual inquiry teaching. Doctor-patient flexible interaction can be carried out, and the independent diagnosis ability of doctors and students is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical simulation teaching, and in particular to a large language model-based consultation teaching method and system. Background Art

[0002] In medical education, students not only need to lay a solid theoretical foundation, but also be exposed to a large number of real clinical cases. However, traditional case-based learning methods often face a serious shortage of real clinical case resources. In addition, due to the scarcity of clinical mentors and limited practice opportunities, students often miss out on key interactions with real patients. This gap hinders them from improving important clinical skills such as history taking, differential diagnosis, and patient management strategy formulation.

[0003] Based on this, the use of standardized patients (SP) for clinical teaching was proposed. This method allows medical students to learn clinical skills independently of real cases. However, the use of SP requires specialized training for individuals to shape the SP, which affects the efficiency of medical training and leads to high teaching costs, and cannot completely solve the problem of limited clinical education resources. In addition, the existing consultation teaching process often uses fixed teaching templates and lacks interactive functions with patients, which makes the effect of medical simulation consultation poor and is not conducive to cultivating students' independent thinking and diagnostic abilities.

[0004] It can be seen that how to provide a scientific and effective medical consultation teaching method has become a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0005] The present invention provides a large language model-based medical consultation teaching method and system to solve the problem of the inability to perform doctor-patient interaction in the existing medical teaching process, so as to improve the independent diagnosis ability of medical students.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a large language model questioning teaching method, including:

[0007] Randomly extracting data to be reorganized from a pre-built target patient role-playing background library, a target patient disease question-and-answer library, and a target disease picture library, and reorganizing each data to be reorganized to generate a virtual digital patient object for interacting with a user;

[0008] During the process of teaching the medical consultation, a similarity analysis is performed between the medical consultation questions output by the user and the target patient disease question-answer database, and a matching result corresponding to the medical consultation question is screened and determined from the target patient disease question-answer database;

[0009] Reference consultation answer data is obtained based on the matching results, and the reference consultation answer data is input into the constructed large language consultation teaching model to obtain a response result that meets the consultation teaching standard output by the large language consultation teaching model, so as to realize virtual consultation teaching.

[0010] Furthermore, the method randomly extracts the data to be reorganized from the pre-constructed target patient role-playing background library, the target patient disease question-and-answer library, and the target disease picture library, and reorganizes each of the data to be reorganized to generate a virtual digital patient object for interacting with the user, including:

[0011] Randomly extract the target patient role-playing background database, the target patient disease question and answer database, the target disease picture database, and the target patient family history database for basic patient information, patient disease picture information, and patient disease question and answer information, and reorganize them to obtain corresponding disease characteristic information;

[0012] The corresponding virtual digital patient object is generated according to the disease characteristic information.

[0013] Further, the similarity analysis is performed on the diagnosis question output by the user and the target patient disease question-answer database, and a matching result corresponding to the diagnosis question is screened and determined from the target patient disease question-answer database, including:

[0014] Calculating the similarity between the medical question and the standard preset question in the target patient disease question and answer library using an embedded vector similarity algorithm;

[0015] According to the similarity calculation results, matching screening is performed in the target patient disease question and answer database to obtain corresponding reference diagnosis questions.

[0016] Furthermore, the reference consultation answer data is input into the constructed large language consultation teaching model to obtain a response result output by the large language consultation teaching model that meets the consultation teaching standard, including:

[0017] Constructing a prompt based on the reference consultation answer data; the prompt reflects the identity information and disease information of the virtual digital patient object;

[0018] The prompt is input into the large language inquiry teaching model to obtain the response result.

[0019] Furthermore, it also includes:

[0020] In the process of conducting the medical consultation teaching, the medical consultation questions output by the user within a specified time are classified according to the medical history type;

[0021] Assigning weights to each type of the medical questions according to pre-set criteria;

[0022] An importance analysis is performed based on the weights corresponding to the medical questions, and the medical questions are scored based on the analysis results.

[0023] Another embodiment of the present invention provides a large language model consultation teaching system, including:

[0024] A virtual digital patient generation module is used for the virtual digital patient to randomly extract data to be reorganized from a pre-built target patient role-playing background library, a target patient disease question-and-answer library, and a target disease picture library, and to reorganize each data to be reorganized to generate a virtual digital patient object for interacting with the user;

[0025] A question matching module is used to perform similarity analysis between the question and answer database of the target patient's disease and select and determine matching results corresponding to the question and answer database of the target patient's disease during the process of the question and answer teaching;

[0026] The answer generation module is used to obtain reference consultation answer data according to the matching results, input the reference consultation answer data into the constructed large language consultation teaching model, and obtain the answer result output by the large language consultation teaching model that meets the consultation teaching standards, so as to realize virtual consultation teaching.

[0027] Furthermore, the virtual digital patient generation module is specifically used for:

[0028] Randomly extract the target patient role-playing background database, the target patient disease question and answer database, the target disease picture database, and the target patient family history database for basic patient information, patient disease picture information, and patient disease question and answer information, and reorganize them to obtain corresponding disease characteristic information;

[0029] The corresponding virtual digital patient object is generated according to the disease characteristic information.

[0030] Furthermore, the question matching module is specifically used to:

[0031] Calculating the similarity between the medical question and the standard preset question in the target patient disease question and answer library using an embedded vector similarity algorithm;

[0032] According to the similarity calculation results, matching screening is performed in the target patient disease question and answer database to obtain corresponding reference diagnosis questions.

[0033] Furthermore, the reply generation module is specifically used to:

[0034] Reference consultation answer data is obtained based on the matching results, and the reference consultation answer data is input into the constructed large language consultation teaching model to obtain a response result that meets the consultation teaching standard output by the large language consultation teaching model, so as to realize virtual consultation teaching.

[0035] Furthermore, the system also includes an importance analysis module, which is specifically used to:

[0036] In the process of conducting the medical consultation teaching, the medical consultation questions output by the user within a specified time are classified according to the medical history type;

[0037] Assigning weights to each type of the medical questions according to pre-set criteria;

[0038] An importance analysis is performed based on the weights corresponding to the medical questions, and the medical questions are scored based on the analysis results.

[0039] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0040] The embodiment of the present invention uses a multidimensional knowledge base to assist a large-scale language model (LLM) in simulating a virtual digital patient. Through a multidimensional standard patient knowledge base and the interaction between the user and the virtual digital patient, the LLM is provided with relevant answer information to assist medical students in learning to consult and realize one-to-one consultation with virtual digital patients. By constructing a digital standard patient simulation technology, it is possible to simulate real-time communication and dialogue between doctors and patients, thereby improving the learning ability of medical students and the reliability of virtual medical teaching. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flowchart of a large language model-based medical consultation teaching method in one embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of an interface of a large language model-based medical consultation teaching system in one embodiment of the present invention;

[0043] Figure 3 It is a structural diagram of a medical consultation teaching system based on a large language model in one embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0046] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0047] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0048] With the advancement of digital and multimedia technologies, virtual patients (VPs) have emerged and are expected to address these important challenges. VPs can stimulate students' intrinsic motivation, promote the mastery of clinical knowledge, and enhance the development of clinical reasoning skills. Recently, the emergence of large language models (LLMs), such as ChatGPT (Open AI, San Francisco, California, USA), represents a major advancement in the field of healthcare, especially in the improvement of medical education. Their established effectiveness in medical consultations has laid the foundation for exploring their impact on medical education. By processing large amounts of medical literature and electronic health records (EHRs), LLMs can be used to create digital twin patients (DTPs) - improved virtual patient models that simulate real patients for educational purposes. Based on this, an embodiment of the present invention provides a consultation teaching method based on a large language model. For details, please refer to Figure 1 , Figure 1 The figure shows a flow chart of a large language model-based teaching method for medical consultation in one embodiment of the present invention, which includes the following steps:

[0049] S1. Randomly extract the data to be reorganized from the pre-constructed target patient role-playing background library, target patient disease question-and-answer library, and target disease picture library, and reorganize each data to be reorganized to generate a virtual digital patient object for interacting with the user.

[0050] Traditional methods of collecting medical history cases rely on manual entry, which is often inconsistent and time-consuming. In addition, due to the cumbersome nature of this process, the same medical history is often reused, making it challenging to provide medical students with diverse case studies.

[0051] Therefore, in order to solve the above problems, the embodiments of the present invention generate a virtual digital patient based on a combination of a patient role-playing background library, a patient disease picture library and a patient disease question-and-answer library. In some embodiments of the present invention, the patient role-playing background library contains basic information such as the patient's age, gender, occupation, and medical history; the patient disease picture library provides pictures related to the patient's disease that can intuitively reflect the patient's disease condition, and the patient disease question-and-answer library contains a series of questions and answers related to eye diseases, which are used to simulate the patient's actual question-and-answer scenario.

[0052] Specifically, the target patient role-playing background database, the target patient disease question and answer database, the target disease picture database, and the target patient family history database are randomly extracted and reorganized to obtain the corresponding disease characteristic information, and the corresponding virtual digital patient object is generated according to the disease characteristic information. That is, the information randomly extracted by the system will be combined to generate a virtual digital patient with a unique background and condition.

[0053] It is worth noting that after the user / medical student fills in the corresponding student number on the system, the system will begin to randomly extract relevant information to generate corresponding virtual digital patients for interaction, so as to provide medical students with a simulated consultation object suffering from the corresponding disease, making the consultation teaching closer to actual clinical conditions.

[0054] The multi-library fusion operation in this step makes the generated virtual digital patient not just a collection of data, but an individual with personality and background, which can cover a wider range of disease types, severity and patient backgrounds. For example, real disease manifestation pictures can enhance the visual authenticity and credibility of virtual patients, while the patient disease question and answer library can enhance the interactivity and authenticity of virtual patients and medicine by simulating conversations with real patients.

[0055] S2. During the process of teaching the medical consultation, a similarity analysis is performed between the medical consultation questions output by the user and the target patient disease question-answer database, and a matching result corresponding to the medical consultation question is screened and determined from the target patient disease question-answer database.

[0056] Medical students start the simulated consultation process based on the background information of the generated virtual digital patient. They may ask questions about the patient's symptoms, medical history, lifestyle habits, etc. to obtain more information about the patient's condition.

[0057] In some embodiments of the present invention, when medical students begin to ask questions, the questions they ask may be more colloquial, or may be slightly different in expression from the standard questions in the question-and-answer database. Therefore, the system will match the questions asked by the students through an embedding similarity algorithm to find the most relevant questions in the patient's disease question-and-answer database as reference questions. The system will also obtain reference answers to the corresponding questions as supplementary knowledge input to the LLM. Specifically,

[0058] It can be understood that this embodiment will use the embedded vector similarity algorithm to calculate the similarity between the medical questions and the standard preset questions in the target patient disease question and answer library; based on the similarity calculation results, matching screening is performed in the target patient disease question and answer library to obtain corresponding reference medical questions.

[0059] When the most relevant reference medical question is found, the system will obtain the reference response corresponding to the question and input it into LLM as supplementary knowledge. By continuously introducing new and relevant medical knowledge, LLM can continuously update its knowledge base and keep pace with the times. This helps improve the accuracy and performance of LLM when dealing with new diseases and new symptoms.

[0060] S3. Obtain reference consultation answer data according to the matching result, input the reference consultation answer data into the constructed large language consultation teaching model, and obtain the answer result that meets the consultation teaching standard output by the large language consultation teaching model to realize virtual consultation teaching.

[0061] After obtaining the reference answer corresponding to the question based on the obtained reference medical question, this embodiment will construct a prompt based on the reference medical question answer data; and the prompt reflects the identity information and disease information of the virtual digital patient, mainly including two important parts: one is to tell the LLM model the background and current situation of the virtual digital patient; the other is the prompt of the user role, which will include the questions asked by the medical students and the corresponding reference answers.

[0062] Next, the constructed prompt is input into the large language consultation teaching model, and the large language consultation teaching model will generate a colloquial response result that conforms to the background and current situation of the virtual digital patient based on the information in the prompt and its internal knowledge base.

[0063] Since in the process of medical consultation teaching, the interaction between medical students and virtual digital patients is a multi-round question-and-answer process, that is, the above-mentioned process of asking questions and obtaining reference replies is repeated. Therefore, in this process, the results output by the large language medical consultation teaching model will be closer and closer to the actual medical consultation teaching, that is, the generated reply results are in line with the medical consultation teaching standards.

[0064] It is worth noting that in order to provide feedback on students' learning during the teaching process, in some embodiments of the present invention, it is necessary to score the questions asked by medical students according to their importance. Specifically,

[0065] During the course of the medical consultation teaching, the medical consultation questions output by the user within the prescribed time are classified according to the medical history type; and each type of medical consultation question is weighted according to a preset standard. In some embodiments, for example, the prescribed time may be set to 15 minutes.

[0066] For different types of medical questions (such as "history of present illness", "past history", "personal history", etc.), the system will set different weights for each type of question based on the importance of medical education and actual clinical needs. These weights reflect the relative importance of each type of question in the diagnosis process.

[0067] Importance analysis is performed based on the weights corresponding to the medical questions, and the medical questions are scored based on the analysis results. After the scoring is completed, it is detected whether there are any missed questions, and the unmentioned questions are generated one by one, and the medical consultation ideas are fed back. Furthermore, in some embodiments, the system will record the time, number of times, medical questions and results of each consultation to form a corresponding medical consultation learning progress; based on the medical consultation learning progress, a corresponding medical consultation teaching report is generated. Based on the medical consultation teaching report, effective feedback can be provided on the learning situation of medical students.

[0068] In order to describe the above process in detail, the present invention provides an embodiment of a specific medical consultation teaching system. Specifically, please refer to Figure 2 , Figure 2 The diagram is a schematic diagram of the interface of the large language model-based consultation teaching system in one embodiment of the present invention. It can be seen that this embodiment conducts consultation interaction for a virtual digital patient with an eye disease. Furthermore, the virtual consultation process specifically includes the following steps:

[0069] 1. After medical students fill in their student numbers in the system, a virtual digital patient is randomly generated based on the patient role-playing background library, eye disease image library, and patient disease question-and-answer library.

[0070] The random factors are the patient's basic information, the patient's eye disease pictures, the patient's special conditions, and the corresponding eye disease question and answer database. Specifically, the patient role-playing background database is shown in Table 1 below:

[0071] Table 1: Example of patient role-playing background library

[0072]

[0073]

[0074] After random selection, the patient role-playing background will be written into the system prompt when accessing the LLM model to determine the LLM imitation role background, as shown below:

[0075] "messages":[{"role":"system","content":"You are a patient with itchy eyes, but you don't know what disease it is. You have a leg injury and are unable to move. The doctor is now examining you. Please answer the doctor's questions in a colloquial tone."},{"role":"user","content":prompt}]

[0076] However, illustratively, the patient role-playing background library is shown in Table 2 below:

[0077] Table 2: Example of patient disease Q&A database - disease type [Cataract]

[0078]

[0079] After random selection, the patient's disease Q&A database will serve as the knowledge base that LLM needs to refer to for each round of dialogue. After determining the patient's role-playing background and selecting the patient's disease Q&A database, the image of the virtual digital patient will also be determined and generated based on the patient's basic information, relevant disease pictures, and relevant information in the patient's family history database.

[0080] 2. Medical students begin to examine the generated virtual digital patients.

[0081] The questions asked will be matched by the embedding similarity model to find the most relevant questions in the patient disease question and answer database, and obtain the reference answers to the corresponding questions and input them into the LLM model. For example,

[0082] Medical student: What’s wrong with your eyes?

[0083] Question matching: Where is your eye uncomfortable? ----What is wrong with your eyes?

[0084] Reference answer matching: What's wrong with your eyes? ----I can't see clearly now

[0085] LLM prompts consist of:

[0086] "messages":[{"role":"system","content":"You are a patient with itchy eyes, but you don't know what the disease is. You have recently injured your leg and have difficulty moving. The doctor is now examining you. Please reply to the doctor's questions in a colloquial tone."},{"role":"user","content":"Please answer the doctor's questions based on the reference reply below.\nDoctor's question: Where is your eye uncomfortable?\nReference reply: I can't see clearly now"}]

[0087] If the patient asks about a specific examination or eye disease, an eye disease image will be displayed during the conversation, and the possible types of eye diseases will be identified through a neural network model. The corresponding eye disease symptoms will be obtained from the eye disease-symptom knowledge base and supplemented in the conversation.

[0088] 3. The LLM model generates a response that matches the role based on the prompt. For example:

[0089] Medical student: What’s wrong with your eyes?

[0090] LLM: I have had trouble seeing clearly lately, and my eyes are a little itchy. I feel a burning sensation when I wake up in the morning, and occasionally there is a lot of secretion, which makes my eyes feel like I am looking at a frosted glass, sticky.

[0091] 4. Repeat steps 1 to 2 above. At the same time, the system records the types of questions asked by medical students in each round. If all questions in the same category, such as "history of current illness", are asked, the prompt box on the left of the system will be marked as completed.

[0092] 5. Medical students complete as many and as complete questions as possible within the specified time (usually 15 minutes), and the system will score them based on the importance of the questions they ask. After the scoring is completed, the system will check whether there are any missed questions. If there are any missed questions, the system will generate feedback for each missed question.

[0093] In summary, a virtual digital patient for doctor-patient interaction is generated based on the patient role-playing background library, disease image library and patient disease question-and-answer library. The user asks questions to this virtual digital patient and analyzes and diagnoses the answers output by the LLM model, thereby realizing virtual consultation teaching of a one-to-one consultation system to assist medical students in their consultation learning. In the process of consultation teaching, the importance of the consultation questions raised by the users is also evaluated, which can help them better grasp the key points of the consultation, so as to help medical students improve their consultation skills and clinical thinking ability.

[0094] An embodiment of the present invention provides a large language model-based teaching system for diagnosis. Figure 3 , Figure 3 The diagram is a schematic diagram of the structure of a large language model-based medical consultation teaching system in one embodiment of the present invention, including:

[0095] A virtual digital patient generating module M1 is used for the virtual digital patient to randomly extract data to be reorganized from a pre-built target patient role-playing background library, a target patient disease question-and-answer library, and a target disease picture library, and to reorganize each data to be reorganized to generate a virtual digital patient object for interacting with the user;

[0096] The question matching module M2 is used to perform similarity analysis between the question and answer database of the target patient's disease and select and determine the matching result corresponding to the question and answer database in the process of the question and answer teaching;

[0097] The answer generation module M3 is used to obtain reference consultation answer data based on the matching results, input the reference consultation answer data into the constructed large language consultation teaching model, and obtain the answer result output by the large language consultation teaching model that meets the consultation teaching standards, so as to realize virtual consultation teaching.

[0098] This embodiment generates a virtual digital patient object for interaction through a virtual digital patient generation module M1. Specifically, the basic information of the target patient role-playing background library, the target patient disease question and answer library, the target disease picture library, and the target patient family history database, the patient's disease picture information, and the patient's disease question and answer information are randomly extracted and reorganized to obtain the corresponding disease characteristic information; the corresponding virtual digital patient object is generated according to the disease characteristic information.

[0099] Furthermore, this embodiment matches the medical questions raised by the user through the question matching module M2 to obtain reference medical questions. Specifically, the similarity between the medical questions and the standard preset questions in the target patient disease question and answer library is calculated using the embedded vector similarity algorithm;

[0100] According to the similarity calculation results, matching screening is performed in the target patient disease question and answer database to obtain corresponding reference diagnosis questions.

[0101] Furthermore, this embodiment generates a reply result of the consultation teaching standard through the reply generation module M3. Specifically, a prompt is constructed according to the reference consultation reply data; the prompt reflects the identity information and disease information of the virtual digital patient object; the prompt is input into the large language consultation teaching model to obtain the reply result.

[0102] The technical features and technical effects of the medical consultation teaching system based on a large language model proposed in the embodiment of the present invention are the same as the technical features and technical effects of the medical consultation teaching method based on a large language model proposed in the embodiment of the present invention, and will not be elaborated here.

[0103] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A medical consultation teaching method based on a large language model, characterized in that: include: Randomly extracting data to be reorganized from a pre-built target patient role-playing background library, a target patient disease question-and-answer library, and a target disease picture library, and reorganizing each data to be reorganized to generate a virtual digital patient object for interacting with a user; During the process of teaching the medical consultation, a similarity analysis is performed between the medical consultation questions output by the user and the target patient disease question-answer database, and a matching result corresponding to the medical consultation question is screened and determined from the target patient disease question-answer database; Reference consultation answer data is obtained based on the matching results, and the reference consultation answer data is input into the constructed large language consultation teaching model to obtain a reply result that meets the consultation teaching standard output by the large language consultation teaching model, so as to realize virtual consultation teaching.

2. The large language model-based medical consultation teaching method according to claim 1, characterized in that: The method comprises randomly extracting data to be reorganized from the pre-built target patient role-playing background library, the target patient disease question-and-answer library, and the target disease picture library, and reorganizing each data to be reorganized to generate a virtual digital patient object for interacting with the user, including: Randomly extract the target patient role-playing background database, the target patient disease question and answer database, the target disease picture database, and the patient basic information, patient disease picture information, and patient disease question and answer information from the target patient family history database and reorganize them to obtain corresponding disease characteristic information; The corresponding virtual digital patient object is generated according to the disease characteristic information.

3. The large language model-based medical consultation teaching method according to claim 1, characterized in that: The performing of similarity analysis on the diagnosis question output by the user and the target patient disease question-answer database, and screening and determining a matching result corresponding to the diagnosis question from the target patient disease question-answer database, includes: Calculating the similarity between the medical question and the standard preset question in the target patient disease question and answer library using an embedded vector similarity algorithm; According to the similarity calculation results, matching screening is performed in the target patient disease question and answer database to obtain corresponding reference diagnosis questions.

4. The large language model-based medical consultation teaching method according to claim 1, characterized in that: The step of inputting the reference consultation answer data into the constructed large language consultation teaching model to obtain a response result output by the large language consultation teaching model that meets the consultation teaching standard includes: Constructing a prompt based on the reference consultation answer data; the prompt reflects the identity information and disease information of the virtual digital patient object; The prompt is input into the large language inquiry teaching model to obtain the response result.

5. The large language model-based medical consultation teaching method according to claim 1, characterized in that: Also includes: In the process of conducting the medical consultation teaching, the medical consultation questions output by the user within a specified time are classified according to the medical history type; Assigning weights to each type of the medical questions according to pre-set criteria; An importance analysis is performed based on the weights corresponding to the medical questions, and the medical questions are scored based on the analysis results.

6. A medical consultation teaching system based on a large language model, characterized in that: A virtual digital patient generation module is used for the virtual digital patient to randomly extract data to be reorganized from a pre-built target patient role-playing background library, a target patient disease question-and-answer library, and a target disease picture library, and to reorganize each data to be reorganized to generate a virtual digital patient object for interacting with the user; A question matching module is used to perform similarity analysis between the question and answer database of the target patient's disease and select and determine matching results corresponding to the question and answer database of the target patient's disease during the process of the question and answer teaching; The answer generation module is used to obtain reference consultation answer data according to the matching results, input the reference consultation answer data into the constructed large language consultation teaching model, and obtain the answer result output by the large language consultation teaching model that meets the consultation teaching standards, so as to realize virtual consultation teaching.

7. The large language model-based medical consultation teaching system according to claim 6, characterized in that: The virtual digital patient generation module is specifically used for: Randomly extract the target patient role-playing background database, the target patient disease question and answer database, the target disease picture database, and the patient basic information, patient disease picture information, and patient disease question and answer information from the target patient family history database and reorganize them to obtain corresponding disease characteristic information; The corresponding virtual digital patient object is generated according to the disease characteristic information.

8. The large language model-based medical consultation teaching system according to claim 6, characterized in that: The question matching module is specifically used for: Calculating the similarity between the medical question and the standard preset question in the target patient disease question and answer library using an embedded vector similarity algorithm; According to the similarity calculation results, matching screening is performed in the target patient disease question and answer database to obtain corresponding reference diagnosis questions.

9. The large language model-based medical consultation teaching system as described in claim 6, characterized in that. The reply generation module is specifically used for: Reference consultation answer data is obtained based on the matching results, and the reference consultation answer data is input into the constructed large language consultation teaching model to obtain a reply result that meets the consultation teaching standard output by the large language consultation teaching model, so as to realize virtual consultation teaching.

10. The large language model-based medical consultation teaching system according to claim 6, characterized in that: The system also includes an importance analysis module, which is specifically used to: In the process of conducting the medical consultation teaching, the medical consultation questions output by the user within a specified time are classified according to the medical history type; Assigning weights to each type of the medical questions according to pre-set criteria; An importance analysis is performed based on the weights corresponding to the medical questions, and the medical questions are scored based on the analysis results.

Images