Intelligent agent oral hospital diagnosis and treatment model establishment method and device

Abstract

The invention relates to a multi-agent oral hospital diagnosis and treatment model establishment method and device, and the method comprises the steps: carrying out the fine adjustment of an existing large-scale language model through employing an efficient method, and enabling the model to better meet the specific demands of the medical field. And on the basis of fine tuning, a feedback mechanism based on artificial intelligence is further utilized, and the expression of the model is optimized through reinforcement learning, so that the system is more intelligent and high in adaptability when simulating doctor-patient interaction. Through the multi-agent framework, a complex medical interaction process among a patient, an examiner, a medical student and a main doctor can be simulated, so that a training platform close to a real clinical environment is provided. Wide experimental results and comprehensive evaluation prove that the multi-agent oral hospital technology based on role playing can well improve interactivity and practicability of medical education, and meanwhile it is ensured that medical students can learn and practice clinical thinking and decision skills in a safe environment. The invention also relates to a device and a memory.

Description

Technical Field

[0001] The present invention relates to the field of large language models and multi-agent technology, and in particular to a method and device for establishing an intelligent oral hospital diagnosis and treatment model. Background Art

[0002] The orthodontic treatment training model can provide medical students with a variety of virtual cases and simulated doctor-patient interaction scenarios. It provides an innovative way to improve the quality and efficiency of diagnosis and treatment training, especially in the process of simulating the professional thinking of doctors, which can show virtual cases to medical students and provide professional diagnostic suggestions.

[0003] Medical students now have the opportunity to conduct interactive training of clinical thinking in a safe, flexible and highly simulated environment, which not only enriches traditional teaching methods but also helps improve their practical ability. However, there are still some challenges in applying these advanced language models to real-world medical education and clinical practice:

[0004] (1) The difficulty of customizing professional models: Although general large-scale language models have performed well in many fields, the medical field requires more professional models, which need to be trained on a large amount of doctor-patient interaction data. Currently, the collection and organization of such data is not only limited in quantity but also complicated, which makes optimizing and customizing large-scale language models based on limited professional data a major challenge.

[0005] (2) The transparency of the reasoning process: The rigor of medical diagnosis requires that the model not only give a diagnosis result, but also provide a clear reasoning process and a high degree of interpretability. Existing large-scale language models have deficiencies in the control and evaluation of diagnostic evidence, which makes it difficult for them to meet the high standards required in the medical field. Therefore, how to enhance the transparency and interpretability of the model and ensure that its reasoning process meets the strict requirements of medical diagnosis is the second key issue that needs to be addressed. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a method and device for establishing an intelligent dental hospital diagnosis and treatment model.

[0007] In a first aspect, the present invention provides a method for establishing a multi-agent dental hospital diagnosis and treatment model, the method comprising:

[0008] Identify the dental doctor-patient conversation dataset;

[0009] Inputting the dental doctor-patient dialogue data set into a large language model, obtaining a plurality of generated responses of the examiner role and a plurality of generated responses of the chief physician role corresponding to the dental doctor-patient dialogue data set output by the large language model;

[0010] According to the preset preference evaluation criteria, the large language model is trained using multiple generated responses of the examiner role and multiple generated responses of the chief physician role to obtain a multi-agent dental hospital diagnosis and treatment model.

[0011] Based on the above technical solution, the method further includes:

[0012] Real dental doctor-patient conversation data, open source doctor-patient conversation data sets and open source general data sets are collected for preprocessing to obtain the dental doctor-patient conversation data set with consistent style.

[0013] Based on the above technical solution, the method further includes:

[0014] The collected real dental doctor-patient conversation data, open source doctor-patient conversation data set and open source general data set are cleaned at multiple levels and polished using a general large language model, ultimately obtaining the dental doctor-patient conversation data set with a consistent style.

[0015] Based on the above technical solution, the method further includes:

[0016] The large language model is fine-tuned and reinforcement learning based on artificial intelligence feedback is performed, where the reinforcement learning method is the policy gradient-based reinforcement learning PPO algorithm.

[0017] Based on the above technical solution, the patient agent is constructed based on the case information database.

[0018] Based on the above technical solution, the large language model is trained using multiple generated responses of the examiner role and multiple generated responses of the chief physician role according to the preset preference evaluation criteria to obtain a multi-agent dental hospital diagnosis and treatment model, which specifically includes:

[0019] Inputting the dental doctor-patient dialogue dataset into the large language model, and setting a current response role to obtain multiple generated responses of the current response role;

[0020] Calculate the KL divergence between the probability distribution of the multiple generated replies of the current reply role and the probability distribution of the multiple generated replies;

[0021] Based on the preset preference evaluation criteria and the KL divergence of the probability distribution of the plurality of generated replies, the reward information of each generated reply is calculated, and each generated reply is marked according to the reward information of each generated reply;

[0022] Based on the annotation information of the multiple generated responses, the large language model is iteratively trained to obtain the multi-agent dental hospital diagnosis and treatment model.

[0023] Based on the above technical solution, the large language model is iteratively trained based on the annotation information of the multiple generated replies to obtain the multi-agent dental hospital diagnosis and treatment model, including:

[0024] Based on the annotation information of the multiple generated replies, a policy gradient-based reinforcement learning algorithm is used to iteratively optimize the large language model, and the corresponding annotation information is fed back to the model obtained in the current iteration during each iteration.

[0025] In a second aspect, the present invention further provides a device for establishing a multi-agent dental hospital diagnosis and treatment model, the device comprising:

[0026] The first processing module is used to determine a dental doctor-patient dialogue data set;

[0027] A second processing module is used to input the dental doctor-patient dialogue data set into a large language model to obtain a plurality of generated responses of the examiner role and a plurality of generated responses of the chief physician role corresponding to the dental doctor-patient dialogue data set to the patient's questions output by the large language model;

[0028] The third processing module is used to train the large language model according to a preset preference evaluation standard using multiple generated responses of the examiner role and multiple generated responses of the chief physician role to obtain a multi-agent dental hospital diagnosis and treatment model.

[0029] In a third aspect, the present invention further provides a memory storing a plurality of instructions, wherein the instructions are used to implement a method for establishing a multi-agent dental hospital diagnosis and treatment model as described in any one of the above technical solutions.

[0030] In a fourth aspect, the present invention also provides an electronic device comprising a processor and a memory connected to the processor, wherein the memory stores a plurality of instructions which can be loaded and executed by the processor so that the processor can execute a method for establishing a multi-agent dental hospital diagnosis and treatment model as described in any one of the above technical solutions.

[0031] It can be seen from the above technical solutions that, compared with the prior art, the present invention provides a method for establishing a multi-agent dental hospital diagnosis and treatment model, which has the following beneficial effects: an efficient method is used to fine-tune the existing large-scale language model so that it can better adapt to the specific needs of the medical field. On the basis of fine-tuning, the feedback mechanism based on artificial intelligence is further utilized to optimize the performance of the model through reinforcement learning, making it more intelligent and adaptable when simulating doctor-patient interaction. Through this multi-agent framework, the complex medical interaction process between patients, examiners, medical students and chief physicians can be simulated, thereby providing a training platform close to the real clinical environment. Extensive experimental results and comprehensive evaluations have confirmed that the technology of the multi-agent dental hospital based on role-playing of the present invention can well improve the interactivity and practicality of medical education, while ensuring that medical students can learn and practice clinical thinking and decision-making skills in a safe environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0033] In the attached picture:

[0034] Figure 1 A flowchart of a method for establishing a multi-agent dental hospital diagnosis and treatment model disclosed in the present invention;

[0035] Figure 2 Diagnostic schematic diagram disclosed for the example of the present invention;

[0036] Figure 3 This is a system overall structure diagram of a method for establishing a multi-agent dental hospital diagnosis and treatment model disclosed in the present invention.

[0037] Figure 4 It is a flowchart of a method for establishing a multi-agent dental hospital diagnosis and treatment model provided by an embodiment of the present invention;

[0038] Figure 5 It is a module schematic diagram of a multi-agent dental hospital diagnosis and treatment model establishment device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0040] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0041] In application areas such as intelligent dialogue, large-scale language models have made breakthrough progress, opening up a new perspective for orthodontic treatment training. By integrating these advanced models, a more efficient and realistic training system can be built to provide medical students with a variety of virtual cases and simulated doctor-patient interaction scenarios. These models are highly valued by the medical and educational communities because they have the ability to understand and reason about semantics. They provide an innovative way to improve the quality and efficiency of diagnosis and treatment training, especially in simulating the professional thinking of doctors, showing virtual cases to medical students and providing professional diagnostic advice.

[0042] To address these challenges, this application proposes an innovative solution: introducing a multi-agent dental hospital system based on role-playing. This is a multi-agent framework that adopts the following strategies:

[0043] (1) Fine-tuning large language models: This application uses an efficient method to fine-tune the existing large language models so that they can better adapt to the specific needs of the medical field.

[0044] (2) Reinforcement learning: Based on fine-tuning, we further utilize the AI-based feedback mechanism to optimize the model’s performance through reinforcement learning, making it more intelligent and adaptable when simulating doctor-patient interactions.

[0045] (3) Dynamic medical interaction simulation: Through this multi-agent framework, the complex medical interaction process between patients, examiners, medical students and chief physicians can be simulated, thus providing a training platform close to the real clinical environment.

[0046] This system is designed to make medical education more interactive and practical, while ensuring that medical students can learn and practice clinical thinking and decision-making skills in a safe environment. In this way, it is hoped that a richer and more effective learning experience can be provided to medical students, helping them better prepare for their future medical careers.

[0047] In view of this, the present invention provides a method for establishing a multi-agent dental hospital diagnosis and treatment model, which can solve the problems existing in the prior art.

[0048] In order to achieve the above purpose, refer to Figures 1 to 3As shown, the present invention discloses a method for establishing a multi-agent dental hospital diagnosis and treatment model, comprising the following steps:

[0049] S1. Collect real dental doctor-patient conversation data, open source doctor-patient conversation datasets, and open source general datasets and preprocess them to obtain a dental doctor-patient conversation dataset with consistent style.

[0050] S2. Build a large Chinese language model with efficient fine-tuning, and perform reinforcement learning based on artificial intelligence feedback based on fine-tuning.

[0051] S3. Based on the fine-tuned local large model and case information database, construct the patient agent, examiner agent, and chief physician agent, and the medical students are played by humans.

[0052] S4. The medical student first communicates with the patient about the condition, then obtains the medical examination results from the examiner and gives the patient a treatment diagnosis. Finally, the medical student organizes the patient's clinical diagnosis results and submits them to the chief physician for evaluation and scoring.

[0053] The above method, optionally, the collection method in S1 includes at least one of the following: cooperating with medical institutions, conducting simulated conversations with doctors, and recording simulated conversation data; cooperating with medical institutions, reviewing medical records, and collecting medical conversation data in medical records; collecting open source online doctor-patient conversations.

[0054] In the above method, optionally, the preprocessing in S1 includes the following steps: analyzing and processing the collected medical conversation data, and performing multi-level cleaning, such as using ChatGPT to polish the data, and finally obtaining conversation data with a consistent style.

[0055] In the above method, optionally, the Chinese large language model in S2 is an LLaMA3 model.

[0056] In the above method, optionally, the fine-tuning method in S2 is the LoRA method.

[0057] The above method, optionally, the basic category of reinforcement learning in S2 is reinforcement learning based on artificial intelligence feedback.

[0058] In the above method, optionally, the reinforcement learning method in S2 is the PPO algorithm. The PPO (Proximal Policy Optimization) algorithm is a policy gradient-based reinforcement learning algorithm used to optimize the intent recognition model. In the above method, optionally, the intelligence in S3 includes a patient agent, an examiner agent, and a chief physician agent, which belongs to the category of multi-agents.

[0059] In the above method, optionally, the patient agent in S3 is constructed based on the case information database.

[0060] The above method is optional. In the multi-agent communication and collaboration in S4, the medical student first communicates with the patient about the condition, then obtains the medical examination results from the examiner and gives the patient the treatment diagnosis result. Finally, the medical student organizes the patient's clinical diagnosis results and submits them to the chief physician for evaluation and scoring.

[0061] It can be seen from the above technical solutions that, compared with the prior art, the present invention provides a method for establishing a multi-agent dental hospital diagnosis and treatment model, which has the following beneficial effects: the present invention adopts an efficient method to fine-tune the existing large-scale language model so that it can better adapt to the specific needs of the medical field. On the basis of fine-tuning, the feedback mechanism based on artificial intelligence is further utilized to optimize the performance of the model through reinforcement learning, so that it is more intelligent and adaptable when simulating doctor-patient interaction. Through this multi-agent framework, the complex medical interaction process between patients, examiners, medical students and chief physicians can be simulated, thereby providing a training platform close to the real clinical environment. Extensive experimental results and comprehensive evaluations have confirmed that the technology of the multi-agent dental hospital based on role-playing of the present invention can well improve the interactivity and practicality of medical education, while ensuring that medical students can learn and practice clinical thinking and decision-making skills in a safe environment.

[0062] like Figure 4 As shown, the present invention proposes a method for establishing a multi-agent dental hospital diagnosis and treatment model, the method comprising:

[0063] 110. Determine the dental doctor-patient dialogue dataset;

[0064] 120. Input the dental doctor-patient dialogue dataset into a large language model, and obtain a plurality of generated responses of the examiner role and a plurality of generated responses of the chief physician role corresponding to the dental doctor-patient dialogue dataset to the patient's questions output by the large language model;

[0065] 130. According to a preset preference evaluation standard, the large language model is trained using a plurality of generated responses of the examiner role and a plurality of generated responses of the chief physician role to obtain a multi-agent dental hospital diagnosis and treatment model.

[0066] Based on the above technical solution, the method further includes:

[0067] Real dental doctor-patient conversation data, open source doctor-patient conversation data sets and open source general data sets are collected for preprocessing to obtain the dental doctor-patient conversation data set with consistent style.

[0068] Based on the above technical solution, the method further includes:

[0069] The collected real dental doctor-patient conversation data, open source doctor-patient conversation data set and open source general data set are cleaned at multiple levels and polished using a general large language model, ultimately obtaining the dental doctor-patient conversation data set with a consistent style.

[0070] Based on the above technical solution, the method further includes:

[0071] The large language model is fine-tuned and reinforcement learning based on artificial intelligence feedback is performed, where the reinforcement learning method is the policy gradient-based reinforcement learning PPO algorithm.

[0072] Based on the above technical solution, the patient agent is constructed based on the case information database.

[0073] Based on the above technical solution, the large language model is trained using multiple generated responses of the examiner role and multiple generated responses of the chief physician role according to the preset preference evaluation criteria to obtain a multi-agent dental hospital diagnosis and treatment model, which specifically includes:

[0074] Inputting the dental doctor-patient dialogue dataset into the large language model, and setting a current response role to obtain multiple generated responses of the current response role;

[0075] Calculate the KL divergence between the probability distribution of the multiple generated replies of the current reply role and the probability distribution of the multiple generated replies;

[0076] Based on the preset preference evaluation criteria and the KL divergence of the probability distribution of the plurality of generated replies, the reward information of each generated reply is calculated, and each generated reply is marked according to the reward information of each generated reply;

[0077] Based on the annotation information of the multiple generated responses, the large language model is iteratively trained to obtain the multi-agent dental hospital diagnosis and treatment model.

[0078] Based on the above technical solution, the large language model is iteratively trained based on the annotation information of the multiple generated replies to obtain the multi-agent dental hospital diagnosis and treatment model, including:

[0079] Based on the annotation information of the multiple generated replies, a policy gradient-based reinforcement learning algorithm is used to iteratively optimize the large language model, and the corresponding annotation information is fed back to the model obtained in the current iteration during each iteration.

[0080] The present invention provides a method for establishing a multi-agent dental hospital diagnosis and treatment model, which has the following beneficial effects: The present invention adopts an efficient method to fine-tune the existing large-scale language model so that it can better adapt to the specific needs of the medical field. On the basis of fine-tuning, the feedback mechanism based on artificial intelligence is further utilized to optimize the performance of the model through reinforcement learning, so that it is more intelligent and adaptable when simulating doctor-patient interaction. Through this multi-agent framework, the complex medical interaction process between patients, examiners, medical students and chief physicians can be simulated, thereby providing a training platform close to the real clinical environment. Extensive experimental results and comprehensive evaluations have confirmed that the technology of the multi-agent dental hospital based on role-playing of the present invention can well improve the interactivity and practicality of medical education, while ensuring that medical students can learn and practice clinical thinking and decision-making skills in a safe environment.

[0081] like Figure 5 As shown, the present invention also provides a multi-agent oral hospital diagnosis and treatment model establishment device, the device comprising:

[0082] The first processing module is used to determine a dental doctor-patient dialogue data set;

[0083] A second processing module is used to input the dental doctor-patient dialogue data set into a large language model to obtain a plurality of generated responses of the examiner role and a plurality of generated responses of the chief physician role corresponding to the dental doctor-patient dialogue data set to the patient's questions output by the large language model;

[0084] The third processing module is used to train the large language model according to a preset preference evaluation standard using multiple generated responses of the examiner role and multiple generated responses of the chief physician role to obtain a multi-agent dental hospital diagnosis and treatment model.

[0085] The present invention provides a device for establishing a multi-agent dental hospital diagnosis and treatment model, which has the following beneficial effects: The present invention adopts an efficient method to fine-tune the existing large-scale language model so that it can better adapt to the specific needs of the medical field. On the basis of fine-tuning, the feedback mechanism based on artificial intelligence is further utilized to optimize the performance of the model through reinforcement learning, so that it is more intelligent and adaptable when simulating doctor-patient interaction. Through this multi-agent framework, the complex medical interaction process between patients, examiners, medical students and chief physicians can be simulated, thereby providing a training platform close to the real clinical environment. Extensive experimental results and comprehensive evaluations have confirmed that the technology of the multi-agent dental hospital based on role-playing of the present invention can well improve the interactivity and practicality of medical education, while ensuring that medical students can learn and practice clinical thinking and decision-making skills in a safe environment.

[0086] The present invention also provides a memory storing a plurality of instructions, wherein the instructions are used to implement a method for establishing a multi-agent dental hospital diagnosis and treatment model as described in any one of the above technical solutions.

[0087] The present invention also provides an electronic device, comprising a processor and a memory connected to the processor, wherein the memory stores a plurality of instructions, and the instructions can be loaded and executed by the processor so that the processor can execute a method for establishing a multi-agent dental hospital diagnosis and treatment model as described in any one of the above technical solutions.

[0088] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0089] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for establishing a multi-agent dental hospital diagnosis and treatment model, characterized in that: The method comprises: Identify the dental doctor-patient conversation dataset; Inputting the dental doctor-patient dialogue data set into a large language model, obtaining a plurality of generated responses of the examiner role and a plurality of generated responses of the chief physician role corresponding to the dental doctor-patient dialogue data set output by the large language model; According to the preset preference evaluation criteria, the large language model is trained using multiple generated responses of the examiner role and multiple generated responses of the chief physician role to obtain a multi-agent dental hospital diagnosis and treatment model.

2. The method according to claim 1, characterized in that The method further comprises: Real dental doctor-patient conversation data, open source doctor-patient conversation data sets and open source general data sets are collected for preprocessing to obtain the dental doctor-patient conversation data set with consistent style.

3. The method according to claim 2, characterized in that The method further comprises: The collected real dental doctor-patient conversation data, open source doctor-patient conversation data set and open source general data set are cleaned at multiple levels and polished using a general large language model, ultimately obtaining the dental doctor-patient conversation data set with a consistent style.

4. The method according to claim 1, characterized in that The method further comprises: The large language model is fine-tuned and reinforcement learning based on artificial intelligence feedback is performed, where the reinforcement learning method is the policy gradient-based reinforcement learning PPO algorithm.

5. The method according to claim 1, characterized in that The patient agent is constructed based on the case information database.

6. The method according to claim 1, characterized in that The method of training the large language model using multiple generated responses of the examiner role and multiple generated responses of the chief physician role according to the preset preference evaluation criteria to obtain a multi-agent dental hospital diagnosis and treatment model specifically includes: Inputting the dental doctor-patient dialogue dataset into the large language model, and setting a current response role to obtain multiple generated responses of the current response role; Calculate the KL divergence between the probability distribution of the multiple generated replies of the current reply role and the probability distribution of the multiple generated replies; Based on the preset preference evaluation criteria and the KL divergence of the probability distribution of the plurality of generated replies, the reward information of each generated reply is calculated, and each generated reply is marked according to the reward information of each generated reply; Based on the annotation information of the multiple generated responses, the large language model is iteratively trained to obtain the multi-agent dental hospital diagnosis and treatment model.

7. The method according to claim 6, characterized in that The method of iteratively training the large language model based on the label information of the plurality of generated responses to obtain the multi-agent dental hospital diagnosis and treatment model comprises: Based on the annotation information of the multiple generated replies, a policy gradient-based reinforcement learning algorithm is used to iteratively optimize the large language model, and the corresponding annotation information is fed back to the model obtained in the current iteration during each iteration.

8. A multi-agent dental hospital diagnosis and treatment model building device, characterized in that: The device comprises: The first processing module is used to determine a dental doctor-patient dialogue data set; A second processing module is used to input the dental doctor-patient dialogue data set into a large language model to obtain a plurality of generated responses of the examiner role and a plurality of generated responses of the chief physician role corresponding to the dental doctor-patient dialogue data set to the patient's questions output by the large language model; The third processing module is used to train the large language model according to a preset preference evaluation standard using multiple generated responses of the examiner role and multiple generated responses of the chief physician role to obtain a multi-agent dental hospital diagnosis and treatment model.

9. A memory, characterized in that: A plurality of instructions are stored, and the instructions are used to implement the method for establishing a multi-agent dental hospital diagnosis and treatment model as described in any one of claims 1-7.

10. An electronic device, characterized in that: It includes a processor and a memory connected to the processor, the memory stores multiple instructions, and the instructions can be loaded and executed by the processor, so that the processor can execute the method for establishing a multi-agent dental hospital diagnosis and treatment model as described in any one of claims 1-7.

Images