A method and system for generating a clinical research plan for traditional Chinese medicine

Abstract

The present application discloses a method and system for generating a clinical research plan for traditional Chinese medicine, which relates to the field of medicine. The method includes constructing a knowledge graph covering the evidence-based evidence element PICO for clinical research on traditional Chinese medicine; establishing a traditional Chinese medicine prompt word project and an element template based on the evidence-based evidence element PICO covered by the knowledge graph; using the traditional Chinese medicine prompt word project to identify the element information selected in the element template and generate a prompt word feature vector; using a trained language model to retrieve data matching the prompt word feature vector from the knowledge graph; when there are features in the prompt word feature vector that cannot retrieve matching data from the knowledge graph, using a trained language model to automatically generate matching data; integrating the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph to obtain a clinical research plan for traditional Chinese medicine. The present application can effectively improve the production efficiency and quality of clinical research plans for traditional Chinese medicine.

Description

Technical Field

[0001] The present application relates to the medical field, and in particular to a method and system for generating a clinical research plan for traditional Chinese medicine. Background Art

[0002] After thousands of years of evolution and development, Chinese medicine has made great contributions to human health with its unique theoretical system and rich clinical practice. In the new stage of current social development, the intelligentization of clinical research of Chinese medicine is particularly important.

[0003] Clinical research on traditional Chinese medicine can provide more scientific support for traditional Chinese medicine theories. By combining modern scientific and technological means such as advanced medical technology, big data analysis and artificial intelligence to verify and interpret traditional Chinese medicine theories, we can more accurately understand the therapeutic mechanisms and effects of traditional Chinese medicine. This will not only help improve the academic status of traditional Chinese medicine, but also provide stronger theoretical support for its application in contemporary medicine. By using scientific and technological means, we can systematically collect, organize and analyze clinical research data on traditional Chinese medicine, and establish a standardized efficacy evaluation system for traditional Chinese medicine in treating different diseases. This will not only help promote the formulation of international standards for traditional Chinese medicine, but also promote greater recognition and application of traditional Chinese medicine worldwide.

[0004] In addition, intelligent clinical research on traditional Chinese medicine plays an important role in promoting the development of the traditional Chinese medicine industry. Through the use of scientific and technological means, the research and development of new traditional Chinese medicine drugs can be carried out more efficiently, and the innovation level of traditional Chinese medicine products can be improved. At the same time, standardized clinical research can enhance the market competitiveness of traditional Chinese medicine products and provide a solid foundation for the internationalization of the traditional Chinese medicine industry. It will help promote the development of traditional Chinese medicine in the direction of personalized medicine. Through in-depth research on the adaptability differences of traditional Chinese medicine in different populations, combined with individual genetic information and lifestyle and other factors, a personalized traditional Chinese medicine treatment plan can be formulated. This is expected to provide patients with more personalized and precise medical services and improve the actual application effect of traditional Chinese medicine in the field of medical health.

[0005] At present, the design of TCM clinical research programs still relies on a lot of manual labor and has the following shortcomings: 1. Time-consuming and laborious: Manually designing TCM clinical research programs requires a lot of time and expertise, and is easily affected by subjective factors. 2. Experience dependence: The design of the program depends on the experience of researchers, which may lead to inconsistent results. 3. Insufficient information acquisition: In the field of TCM, the existing system still has limited understanding of the rich TCM knowledge and terminology. Summary of the invention

[0006] The purpose of this application is to provide a method and system for generating clinical research plans for traditional Chinese medicine, so as to effectively improve the efficiency and quality of the production of clinical research plans for traditional Chinese medicine.

[0007] To achieve the above objectives, this application provides the following solutions.

[0008] In the first aspect, the present application provides a method for generating a clinical research plan for traditional Chinese medicine, including: constructing a knowledge graph covering the evidence-based evidence element PICO for clinical research on traditional Chinese medicine; the PICO includes a research population, intervention measures, control measures and outcome indicators; using the knowledge graph to train a pre-trained language model; establishing a traditional Chinese medicine prompt word project based on the evidence-based evidence element PICO covered by the knowledge graph; determining an element template based on the evidence-based evidence element PICO covered by the knowledge graph; the element template includes a research population template, an intervention measure template, a control measure template and an outcome indicator template; select element information in the element template according to the TCM clinical research plan to be designed; use the TCM prompt word project to identify the selected element information and generate a prompt word feature vector; use the trained language model to retrieve data matching the prompt word feature vector from the knowledge graph; when there are features in the prompt word feature vector that cannot retrieve matching data from the knowledge graph, use the trained language model to automatically generate matching data; integrate the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph to obtain the TCM clinical research plan.

[0009] Optionally, for TCM clinical research, a knowledge graph covering the evidence-based element PICO is constructed, specifically including: determining the ontology of TCM clinical research using the seven-step ontology method; constructing a corpus of TCM clinical research; aligning and matching the corpus with the ontology of TCM clinical research to obtain the ontology information of the evidence-based element PICO; performing structured processing on the ontology information to obtain PICO raw data; and standardizing and defining the PICO raw data through the seven-step ontology method to generate a knowledge graph covering the evidence-based element PICO.

[0010] Optionally, the pre-trained language model is trained using the knowledge graph, and then the method further includes: receiving a query question for clinical research on traditional Chinese medicine; using the BERT model to identify whether the query question is the content of the clinical research plan for traditional Chinese medicine; if so, retrieving data from the knowledge graph based on the query question using the trained language model; when data cannot be retrieved from the knowledge graph, generating reply data using the trained language model, and marking the reply data as being generated by the model; when data is retrieved from the knowledge graph, feeding back the retrieved data through the trained language model, and marking the retrieved data as being derived from the data knowledge base; if not, replying with feedback that the query question cannot be performed.

[0011] Optionally, based on the evidence-based elements PICO covered by the knowledge graph, a TCM prompt word project is established, specifically including: constructing a prompt word rule framework for TCM clinical research; formulating prompt words according to the inclusion criteria, exclusion criteria, intervention measures and outcome indicators in the knowledge graph corresponding to the evidence-based elements PICO; classifying the prompt words into necessary prompt words and non-essential prompt words according to the importance of TCM clinical research; determining the expression form of the classified prompt words; based on the prompt word rule framework, using the trained language model to debug the prompt words with the determined expression form to obtain the TCM prompt word project.

[0012] Optionally, the TCM prompt word project is used to identify the selected element information and generate a prompt word feature vector, which specifically includes: using the TCM prompt word project to identify the selected element information and feeding back the identified element information; if the fed-back element information is confirmed to be incorrect, returning to the step of "selecting element information from the element template information according to the TCM clinical research plan to be designed"; if the fed-back element information is confirmed to be correct, using the TCM prompt word project to generate a prompt word feature vector based on the confirmed correct element information.

[0013] Optionally, the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph are integrated to obtain a clinical research plan for traditional Chinese medicine, specifically comprising: structuring the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph according to the inclusion criteria, exclusion criteria, intervention measures and outcome indicators in the knowledge graph to obtain a clinical research plan for traditional Chinese medicine.

[0014] Optionally, the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph are integrated to obtain a clinical research plan for traditional Chinese medicine, and then the method also includes: introducing a real-time verification mechanism to verify whether the clinical research plan for traditional Chinese medicine conforms to practical feasibility; introducing an expert evaluation mechanism in the field of traditional Chinese medicine to train the trained language model through reinforced feedback learning.

[0015] In the second aspect, the present application provides a system for generating a clinical research plan for traditional Chinese medicine, including: an element template module, a prompt word engineering control module, a PICO knowledge graph module and a language model module. The PICO knowledge graph module stores a knowledge graph; the knowledge graph is a knowledge graph for clinical research on traditional Chinese medicine, covering the evidence-based evidence element PICO; the PICO includes research population, intervention measures, control measures and outcome indicators; the element template module is connected to the prompt word engineering control module; the element template module is used to accept the element information selected in the element template according to the clinical research plan for traditional Chinese medicine to be designed; the element template includes a research population template, an intervention measure template, a control measure template and an outcome indicator template; the prompt word engineering control module uses the traditional Chinese medicine prompt word engineering to identify the selected elements. The method comprises the following steps: the method comprises: obtaining a prompt word feature vector from the prompt word engineering control module and the PICO knowledge graph module; the language model module is connected to the prompt word engineering control module and the PICO knowledge graph module respectively; the language model module uses the trained language model to retrieve data matching the prompt word feature vector from the PICO knowledge graph module; when there are features in the prompt word feature vector that cannot retrieve matching data from the PICO knowledge graph module, the trained language model is used to automatically generate matching data; and the matching data automatically generated by the trained language model and the matching data retrieved from the PICO knowledge graph module are integrated to obtain a clinical research plan for traditional Chinese medicine.

[0016] Optionally, the knowledge graph in the PICO knowledge graph module includes: inclusion criteria, exclusion criteria, intervention measures and outcome indicators; the inclusion criteria include: basic information of the subject population, gender distribution, Chinese and Western medicine diagnostic criteria for diseases or symptoms, disease course restrictions, characteristic descriptions of patients' diseases, physical sign information of subjects, baseline physiological data and pre-trial review procedures; the exclusion criteria include: exclusion criteria for therapeutic interventions, exclusion time ranges, exclusion criteria for patients' baseline disease status, exclusion criteria for patients with a set course of disease, exclusion criteria for patients who do not meet the suitability standards for traditional Chinese medicine treatment, and exclusion criteria for setting physiological indicators; the intervention measures include: multiple experimental groups and control groups in clinical research on traditional Chinese medicine; the outcome indicators include: the definition of outcome indicators, the time points of outcome indicators and the observation period.

[0017] Optionally, the TCM clinical research plan generation system also includes: a graph database; the graph database is connected to the PICO knowledge graph module; the graph database is used to store the knowledge graph in the PICO knowledge graph module.

[0018] According to the specific embodiments provided in this application, this application discloses the following technical effects.

[0019] The present application provides a method and system for generating a clinical research plan for traditional Chinese medicine. Based on the knowledge graph, a traditional Chinese medicine prompt word project for inclusion criteria, exclusion criteria, intervention measures and outcome indicators is constructed. The element information in the element template can be identified, and a prompt word feature vector can be generated. Then, the language model and the knowledge graph are combined to realize the intelligent design of the clinical trial plan for traditional Chinese medicine, overcoming the existing defects of relying on a large amount of manual labor, being time-consuming and labor-intensive, experience-dependent and insufficient information acquisition, and effectively improving the production efficiency and quality of the clinical research plan for traditional Chinese medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 This is a flowchart of a method for generating a clinical research plan for traditional Chinese medicine in one embodiment of the present application.

[0022] Figure 2 A signal transmission diagram of a system for generating a clinical research plan for traditional Chinese medicine provided in one embodiment of the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0025] In an exemplary embodiment, Figure 1 As shown, a method for generating a clinical research plan for traditional Chinese medicine is provided, including the following steps 101 to 109.

[0026] Step 101: Construct a knowledge graph covering the PICO elements of evidence-based research for traditional Chinese medicine clinical research; the PICO includes research population, intervention measures, control measures and outcome indicators.

[0027] PICO is divided into four parts: Participants / Patients (research population, abbreviated as P), Interventions (intervention measures, abbreviated as I), Control / Comparison (control measures, abbreviated as C), and Outcomes (outcome indicators, abbreviated as O).

[0028] Step 102: Use the knowledge graph to train the pre-trained language model.

[0029] Step 103: Based on the PICO elements of evidence-based knowledge covered by the knowledge graph, establish a TCM prompt word project.

[0030] Step 104: Determine an element template according to the evidence-based elements PICO covered by the knowledge graph; the element template includes a research population template, an intervention measure template, a control measure template and an outcome indicator template.

[0031] Step 105: Select element information in the element template according to the TCM clinical research plan to be designed.

[0032] Step 106: using the TCM prompt word project to identify the selected element information and generate a prompt word feature vector.

[0033] Step 107: Using the trained language model, retrieve data matching the prompt word feature vector from the knowledge graph.

[0034] Step 108: When there are features in the prompt word feature vector that cannot retrieve matching data from the knowledge graph, automatically generate matching data using the trained language model.

[0035] Step 109: Integrate the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph to obtain a clinical research plan for traditional Chinese medicine.

[0036] By implementing the above steps 101 to 109, a TCM prompt word project for inclusion criteria, exclusion criteria, intervention measures and outcome indicators is constructed based on the knowledge graph, which can identify the element information in the element template, generate a prompt word feature vector, and then combine the language model and the knowledge graph to realize the intelligent design of TCM clinical trial plans, overcoming the existing reliance on a large amount of manual labor, the time-consuming and labor-intensive, experience-dependent and insufficient information acquisition defects, and effectively improving the efficiency and quality of TCM clinical research plans.

[0037] In addition, it can be seen from the above steps 108 and 109 that the clinical research plan for traditional Chinese medicine is generated in two ways: language model call + knowledge graph based on prompt words, and intelligent generation of language model based on prompt words.

[0038] In another exemplary embodiment of the present application, in order to construct based on the elements of evidence-based medicine, the ontological information data of the four elements of evidence-based medicine PICO is confirmed, the data is hierarchically processed according to the ontological information of the knowledge graph, and the terminology standard of the evidence-based medicine PICO ontology knowledge graph that conforms to the clinical research of traditional Chinese medicine is constructed, then the above step 101 is replaced by the following steps 201 to 205.

[0039] Step 201: Determine the ontology of clinical research of traditional Chinese medicine using the seven-step ontology method.

[0040] Step 202: Construct a corpus of clinical research on traditional Chinese medicine.

[0041] Step 203: aligning and matching the ontology of TCM clinical research on the corpus to obtain ontology information of the evidence-based evidence element PICO.

[0042] Step 204: Structural processing is performed on the ontology information to obtain PICO original data.

[0043] Step 205: Standardize and define the PICO raw data through the seven-step ontology method to generate a knowledge graph covering the PICO elements of evidence-based evidence.

[0044] The more specific implementation example process of the above steps 201 to 205 is as follows: the ontology-based seven-step method is used to determine the ontology field and scope of TCM clinical research, and the existing ontology system is reviewed, important ontology terms are summarized, ontology definitions and hierarchical levels are defined, attributes are defined, attribute facets are summarized, and the final example is presented. The corpus construction needs to include clinical diagnosis and treatment data based on the real world, high-quality clinical research data based on public publications, data sets based on the field of TCM, and multivariate database information common to diseases. All ontology information is preliminarily screened and structured using manual standards, expert verification, and expert proofreading to construct PICO raw data. The PICO raw data is standardized and defined through the seven-step method of ontology to complete the construction of the data and data structure of the TCM PICO ontology.

[0045] In another exemplary embodiment of the present application, the language model is generally a large language model, and the large language model can use the Qwen-7B model, or other open source large models, which are not limited here. In order to complete the fine-tuning of the large language model, the above step 102 selects the knowledge graph of clinical research on traditional Chinese medicine and covers the evidence-based evidence element PICO as the data source, performs data enhancement expression on the large language model, and completes the data fine-tuning of the large language model. Therefore, the trained language model can also be called a large model after fine-tuning.

[0046] In another exemplary embodiment of the present application, in order to form intelligent "open" questions and answers based on the knowledge graph and combined with the big model technology, after the above step 102, the method may also include the following steps 301 to 306.

[0047] Step 301: Receive a query question about clinical research on traditional Chinese medicine.

[0048] Step 302: Use the BERT model to identify whether the query question is the content of a clinical research plan for traditional Chinese medicine.

[0049] Step 303: If yes, then according to the query question, use the trained language model to retrieve data from the knowledge graph.

[0050] Step 304: When data cannot be retrieved from the knowledge graph, a trained language model is used to generate response data, and the response data is marked as being generated by the model.

[0051] Step 305: When data is retrieved from the knowledge graph, the retrieved data is fed back through the trained language model, and the retrieved data is marked as coming from the data knowledge base.

[0052] Step 306: If not, a feedback is given indicating that the query question cannot be performed.

[0053] It can be seen from the above steps 304 and 305 that intelligent “open” question and answer is performed in two ways: based on the trained language model call + knowledge graph, and based on the trained language model intelligent generation.

[0054] In another exemplary embodiment of the present application, in order to construct a prompt word engineering rule framework that conforms to the clinical research plan of traditional Chinese medicine, the construction of prompt words, the trained language model and the result output of the prompt word framework are completed internally, and the above-mentioned step 103 is replaced by the following steps 401 to 405.

[0055] Step 401: Construct a rule framework of prompt words for clinical research on traditional Chinese medicine.

[0056] Step 402: Formulate prompt words based on the inclusion criteria, exclusion criteria, intervention measures and outcome indicators in the knowledge graph corresponding to the evidence-based evidence element PICO.

[0057] Step 403: Classify prompt words into necessary prompt words and non-essential prompt words according to the importance of clinical research on traditional Chinese medicine.

[0058] Step 404: Determine the presentation form of the classified prompt words.

[0059] Among them, the representation of prompt words is mainly based on intelligent generation of large models and intelligent generation based on large model knowledge graphs. Intelligent generation based on large models refers to the automatic generation of matching data by trained language models, and intelligent generation based on large model knowledge graphs refers to the use of trained language models to retrieve data matching the prompt word feature vector from the knowledge graph.

[0060] Step 405: Based on the prompt word rule framework, the trained language model is used to debug the prompt words with determined expressions to obtain the traditional Chinese medicine prompt word project.

[0061] The above steps 402 to 404 correspond to the construction of prompt words, and step 405 corresponds to the output of the trained language model and prompt word framework. Step 405 can also be understood as: selecting prompt word information that meets the clinical research of traditional Chinese medicine, using the fine-tuned large model to debug the prompt words, completing the construction of the traditional Chinese medicine prompt word rule framework, and obtaining the traditional Chinese medicine prompt word project.

[0062] In another exemplary embodiment of the present application, in order to improve the accuracy of the generated TCM clinical research plan, after selecting the element information in the element template, it is necessary to first undergo recognition by the prompt word engineering, and the recognized information is fed back to the user for multiple rounds of information confirmation. After the user confirms that it is correct, the prompt word feature vector is generated and then fed back to the trained language model. Then the above step 106 can be replaced by the following steps 501 to 503.

[0063] Step 501: using the TCM prompt word project to identify the selected element information, and feeding back the identified element information.

[0064] Step 502: If the fed-back element information is confirmed to be incorrect, return to the step of "selecting element information from the element template information according to the TCM clinical research plan to be designed".

[0065] Step 503: If the fed-back element information is confirmed to be correct, then the TCM prompt word project is used to generate a prompt word feature vector according to the confirmed element information.

[0066] The working principle of the traditional Chinese medicine prompt word project is given below.

[0067] (1) The element template is as follows.

[0068] 1. Study population.

[0069] "Please generate a clinical trial protocol for (×× drug) to treat / prevent / prevent / use for / improve (×× disease).".

[0070] 2. Interventions (determine interventions and combination medications).

[0071] The intervention measures of the experimental group were "×× combined with ××" or "×× combined with ×× combined with ××" or "×× alone"; the intervention measures of the control group were "×× combined with ××" or "×× combined with ×× combined with ××" or "×× alone".

[0072] The intervention time of the experimental group was “×× days / month”, and the follow-up time was “×× days / month / year” / “not involved”.

[0073] The intervention time of the control group was “×× days / month”, and the follow-up time was “×× days / month / year” / “not involved”.

[0074] 3. Outcome indicators.

[0075] The primary outcome indicator requires a “system recommendation” or a “specified indicator domain” or a customized “indicator name”.

[0076] All secondary outcome indicators require "system recommendation" or "specified indicator domain" or customized "indicator name".

[0077] Designated indicator domains: quality of life indicators, biochemical indicators.

[0078] (2) The content of the feedback confirmation is as follows.

[0079] Please generate a clinical trial protocol for the treatment of stable angina pectoris with Compound Danshen Drops.

[0080] The experimental group was treated with compound danshen drops combined with conventional western medicine treatment, the control group 1 was treated with compound danshen drops combined with placebo, the control group 2 was treated with conventional western medicine treatment combined with placebo, and the control group 3 was treated with placebo alone.

[0081] The primary outcome indicator needs to be a quality of life indicator, and the secondary outcome indicator needs to be a systematic recommendation.

[0082] Age was ≥45 years and ≤75 years.

[0083] Gender is "Not Restricted".

[0084] The course of the disease is "1-5 years".

[0085] The intervention time of the experimental group was "21 days". The intervention time of the control group 1 was "21 days", the intervention time of the control group 2 was "21 days", and the intervention time of the control group 3 was "21 days".

[0086] (3) The prompt words are as follows.

[0087] Subject population of the clinical trial program: stable angina pectoris.

[0088] TCM syndrome types of the subjects: Not mentioned.

[0089] Disease severity: Not mentioned.

[0090] Number of experimental groups: 1.

[0091] Number of control groups: 3.

[0092] Intervention measures of the experimental group: Compound Danshen dripping pills; conventional Western medicine treatment.

[0093] Intervention measures for control group 1: Compound Danshen Dropping Pills; placebo.

[0094] Intervention measures for control group 2: conventional western medicine treatment; placebo.

[0095] Control group 3 Intervention: Placebo.

[0096] Intervention method of the experimental group: combined medication.

[0097] Intervention method for control group 1: combined medication.

[0098] Intervention method for control group 2: combined medication.

[0099] Intervention method of control group 3: medication alone.

[0100] The intervention time of the experimental group was 21 days; the follow-up time was not mentioned.

[0101] The intervention time for control group 1 was 21 days; the follow-up time was not mentioned.

[0102] The intervention time for control group 2 was 21 days; the follow-up time was not mentioned.

[0103] The intervention time for control group 3 was 21 days; the follow-up time was not mentioned.

[0104] Sample size of the experimental group: Not mentioned.

[0105] Sample size of control group 1: Not mentioned.

[0106] Sample size of control group 2: Not mentioned.

[0107] Main outcome measures: Quality of life indicators.

[0108] Secondary outcome measures: Systematic recommendations.

[0109] Age was ≥45 years and ≤75 years.

[0110] Gender is "Not Restricted".

[0111] The course of disease is "1-5 years".

[0112] Study implementation time: Not mentioned.

[0113] Time of recruitment of research subjects: Not mentioned.

[0114] Study location: Not mentioned.

[0115] By inputting the text information requirements of the clinical trial protocol, the input text information requirements of the clinical trial protocol are characterized and learned to identify the requirement feature vector of the evidence-based elements. The identified part or all of the evidence-based element information is fed back to the researchers for proofreading and modification. Through multiple rounds of confirmation information replies, the formatted information confirmation is completed, and finally the identifiable prompt word feature vector is clarified.

[0116] In another exemplary embodiment of the present application, in order to realize the intelligent generation of modular clinical research plans for clinical research on traditional Chinese medicine, the method of obtaining the clinical research plan for traditional Chinese medicine in step 109 is specifically as follows: the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph are structured and output according to the inclusion criteria, exclusion criteria, intervention measures and outcome indicators in the knowledge graph to obtain the clinical research plan for traditional Chinese medicine.

[0117] In another exemplary embodiment of the present application, after the above step 109, the method may further include the following steps 601 to 602.

[0118] Step 601: Introduce a real-time verification mechanism to verify whether the TCM clinical research plan is practically feasible. Provide real-time feedback to help users understand whether the selected design is practically feasible.

[0119] Step 602: Introduce an expert evaluation mechanism in the field of traditional Chinese medicine, and train the trained language model through reinforcement feedback learning.

[0120] The beneficial effects of this method are as follows.

[0121] 1. This application can effectively improve the efficiency of making clinical research plans for traditional Chinese medicine by intelligently generating clinical research plans for traditional Chinese medicine based on the four elements of PICO evidence-based medicine, thereby realizing the intelligent design of clinical research plans for traditional Chinese medicine, and effectively solving the digitalization and informatization of clinical research on traditional Chinese medicine. While reducing the manual workload, it can improve the efficiency of designing clinical research plans for traditional Chinese medicine for a certain disease, and better complete the auxiliary decision-making of clinical traditional Chinese medicine.

[0122] 2. This application uses modular design, combines knowledge graphs and large model technologies, and integrates high-quality evidence elements of TCM clinical trials to improve the accuracy of clinical trials. Through the synergy of large language models and knowledge graphs, the accurate identification of the four elements of PICO and the scientific nature of the design of the program are improved, which is conducive to solving the accuracy problem of the current automated design of TCM clinical research and provides interpretable data knowledge support for generative large models.

[0123] 3. The present application constructs knowledge data information for traditional Chinese medicine clinical research through the PICO four elements based on evidence-based medicine rules, which can effectively reduce the experience dependence of clinicians or researchers, and increase the authenticity based on evidence-based data to ensure the applicability and prospectiveness of the research protocol. In the present application, learning based on prompt engineering and expert experience is adopted, and through multiple rounds of expert feedback, the experience dependence of protocol design is gradually reduced, and the professional level of the system is improved.

[0124] Based on the same inventive concept, the embodiment of the present application also provides a traditional Chinese medicine clinical research protocol generation system for implementing the traditional Chinese medicine clinical research protocol generation method involved above. The solution provided by this device for solving problems is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the traditional Chinese medicine clinical research protocol generation system provided below can refer to the limitations on the traditional Chinese medicine clinical research protocol generation method in the above text, and will not be repeated here.

[0125] In an exemplary embodiment, as Figure 2 shown, a traditional Chinese medicine clinical research protocol generation system is provided, including: an element template module, a prompt engineering control module, a PICO knowledge graph module, and a language model module.

[0126] The PICO knowledge graph module stores a knowledge graph; the knowledge graph is a knowledge graph for traditional Chinese medicine clinical research, covering the evidence-based element PICO; PICO includes study population, intervention measures, control measures, and outcome indicators. The element template module is connected to the prompt engineering control module; the language model module is respectively connected to the prompt engineering control module and the PICO knowledge graph module.

[0127] The element template module is used to receive the element information selected in the element template according to the traditional Chinese medicine clinical research protocol to be designed; the element template includes a study population template, an intervention measure template, a control measure template, and an outcome indicator template. The prompt engineering control module uses traditional Chinese medicine prompt engineering to identify the selected element information and generate a prompt feature vector. The language model module uses the trained language model to retrieve data from the PICO knowledge graph module that matches the prompt feature vector; when there are features in the prompt feature vector that cannot be retrieved and matched in the PICO knowledge graph module, the trained language model is used to automatically generate matching data; the matching data automatically generated by the trained language model and the matching data retrieved from the PICO knowledge graph module are integrated to obtain a traditional Chinese medicine clinical research protocol.

[0128] As an optional implementation, the knowledge graph in the PICO knowledge graph module includes: inclusion criteria, exclusion criteria, intervention measures and outcome indicators. Inclusion criteria include: basic information of the subject population, gender distribution, Chinese and Western medicine diagnostic criteria for diseases or symptoms, disease course restrictions, characteristic descriptions of patients' diseases, physical signs information of subjects, baseline physiological data and pre-trial review procedures. Exclusion criteria include: exclusion criteria for therapeutic interventions, exclusion time ranges, exclusion criteria for patients' baseline disease status, exclusion criteria for patients with set disease courses, exclusion criteria for patients who do not meet the suitability criteria for Chinese medicine treatment, and exclusion criteria for setting physiological indicators. Intervention measures include: multiple experimental groups and control groups in clinical studies of Chinese medicine. Outcome indicators include: definition of outcome indicators, time points of outcome indicators and observation periods.

[0129] Specifically, the main functional contents of the inclusion criteria in the knowledge graph include: determining the basic information of the subjects, including age range, to ensure that the research subjects meet the requirements of a specific age group; gender distribution, to determine the ratio of male and female patients in the trial, to ensure the gender representativeness of the research results. Clear diagnostic criteria for diseases or symptoms: meet the diagnostic criteria of traditional Chinese medicine and Western medicine for clear diseases or symptoms to ensure the consistency of research subjects; course of disease restrictions: the length of illness time, so as to include patients at a specific stage. Characteristic description of the patient's disease: describe the patient's symptoms in detail, including the degree of pain, frequency, etc., so as to more accurately include patients who meet the test conditions; physical sign information of the subjects: collect the patient's physical sign information, such as blood pressure, heart rate, etc., to ensure the comparability of clinical characteristics between the test group and the control group. Baseline physiological data: describe the baseline physiological data that needs to be collected before the start of the trial, such as blood test results, biochemical indicators, etc. Ethical review and informed consent: describe the ethical review procedures that need to be conducted before the trial to ensure that the trial complies with ethical standards. Detailed description of the procedures and requirements for patient informed consent to ensure the patient's right to know and voluntary participation.

[0130] The main functional contents of the exclusion criteria in the knowledge graph include: The exclusion criteria of the intelligent generation technology of clinical research plans for traditional Chinese medicine need to construct data graphs based on intervention measures and test populations. It is necessary to clarify the exclusion criteria for other treatment interventions: describe the specific information of other drug treatments, surgical interventions or other therapies that need to be excluded to ensure the consistency of treatment between the test group and the control group. Exclusion time range: the time range of treatment in the exclusion criteria to prevent the influence of past treatment on the research results. Exclusion criteria for patients' baseline disease status: Specific description of disease status: describe the baseline disease status that needs to be excluded to ensure the comparability of the disease status between the test group and the control group. Exclude patients with a specific course of disease: determine patients who need to be excluded, such as patients with a long or short course of disease. Exclusion criteria that do not meet the suitability criteria for traditional Chinese medicine treatment: determine to exclude patients who do not meet the suitability criteria for traditional Chinese medicine treatment to ensure the scientificity and effectiveness of the trial. According to traditional Chinese medicine theory, exclude patients who are not suitable for traditional Chinese medicine treatment. Exclusion criteria for specific physiological indicators: describe patients with abnormal physiological indicators that need to be excluded to ensure the comparability of physiological status between the test group and the control group. Set specific thresholds for physiological indicators to determine whether to exclude specific patients.

[0131] The main functional contents of intervention measures in the knowledge graph include: the intervention measures part includes multiple experimental groups and control groups of clinical studies. First of all, it is necessary to include intelligent selection of multiple groups of medication methods to formulate intervention measures information for traditional Chinese medicine treatment plans. Based on the knowledge graph of traditional Chinese medicine, the system can automatically select applicable intervention measures such as traditional Chinese medicine, western medicine, placebo, etc. Considering the adaptability of medication, the system can automatically adjust the indication population adapted to the intervention measures of traditional Chinese medicine to ensure the personalization and pertinence of the intervention measures. The rationality of the intervention measures, formulate the rationality evaluation standards for traditional Chinese medicine treatment intervention measures, and ensure that the intervention measures are in line with traditional Chinese medicine theory and clinical practice norms. The information such as indications, adapted diseases, dosage, dosage form, category, attributes, etc. of the intervention measures is stored in the form of knowledge graphs. The intervention measures module can make the clinical research plan of traditional Chinese medicine more convenient, thereby promoting the construction of the knowledge graph of intervention measures.

[0132] The main functional contents of the outcome indicators in the knowledge graph include: describing the main endpoint of the study, that is, the efficacy evaluation indicator of the most concern. Determine the specific method of measuring the main outcome indicators to ensure the accuracy and consistency of the results. List other secondary outcome indicators that need attention to comprehensively evaluate the treatment effect. Clarify the measurement method of the secondary outcome indicators. Time point and observation period of the outcome indicators: describe the specific time point for measuring the outcome indicators in order to conduct time series analysis. Set the time range for observation of the outcome indicators to ensure the integrity and continuity of the data. Determine the definition information of the outcome indicators to ensure the clinical interpretability of the outcome indicators. Establish a grading standard for the outcome indicators to ensure the transparency and understandability of the results, provide an explanation of the outcome indicator results, and make the research results more clinically meaningful. Through the above functions, the outcome indicator module can comprehensively and scientifically plan and manage the evaluation system of the research, and improve the diversity of the outcome indicators generated intelligently in clinical research of traditional Chinese medicine.

[0133] As an optional implementation, the TCM clinical research program generation system further includes: a graph database. The graph database is connected to the PICO knowledge graph module; the graph database is used to store the knowledge graph in the PICO knowledge graph module.

[0134] Taking the Neo4j database in the graph database as an example, the data storage and visualization process of the knowledge graph using the Neo4j database is explained: based on the PICO ontology information of the clinical research of traditional Chinese medicine, the data structure is transformed into a triple structure for data storage. The data of the ontology is mapped end-to-end, the ontology is converted into a graph relationship structure, and the entities, relationships and attributes are compiled and designed. Through the mapping of the ontology to the Neo4j knowledge graph, a network knowledge expression based on entities, relationships and attributes is constructed. Node mapping, the instance of the ontology corresponds to the Neo4j node entity. Relationship mapping, the relationship between the categories of the ontology is converted into the edges between Neo4j nodes. Attribute mapping, the instance information description of the ontology is converted into the attribute storage of the node.

[0135] As an optional implementation method, in order to build a visual query system for TCM clinical research plans and obtain a TCM clinical research plan generation system, it is necessary to visualize the population information structure, intervention measure structure, control measure structure and outcome indicator structure of the entities and edges of TCM clinical research, and perform interface feedback interaction based on the input query. The specific design steps are as follows.

[0136] Step 701: Visual display of the knowledge graph, where different entity nodes represent different TCM entity information, and edge relationships illustrate the relationship structure and expression information between two entities.

[0137] Step 702: Visualization of the structure of the research population (P): Displaying the information display structure visualization such as the TCM diagnostic standards, TCM diagnostic definitions, Western medicine diagnostic standards, Western medicine diagnostic definitions, symptoms and signs, disease function classification, and the source of diagnostic standards related to the population.

[0138] Step 703: Intervention measures (I) structural visualization: Display the definition of TCM intervention measures, names of TCM preparations, TCM ingredients, functions and indications related to the intervention measures, etc., and display structural visualization.

[0139] Step 704: Visualization of control measure (C) structure: Visualization of the control group medication method, control group intervention measure definition, applicable conditions, drug categories, etc.

[0140] Step 705: Visualization of the structure of outcome indicators (O): Visualization of the structure of outcome indicator name, primary outcome indicator, secondary outcome indicator, measurement indicator type, indicator domain, indicator measurement tool, etc.

[0141] Step 706: Data visualization interaction and query, based on the input data information (triples), obtain the information vector expression of the input data.

[0142] Step 707: Matching the data information of evidence-based medicine PICO, by identifying the data vector expression of the input data and the node information stored in the knowledge graph for retrieval and data feedback, and displaying the results of the knowledge semantic network and the ranking based on the graph computing algorithm.

[0143] As an optional implementation method, the overall design concept of the TCM clinical research program generation system of the present application is as follows.

[0144] 1. User input interface design: Develop a user input interface for TCM clinical research so that users can easily input basic information about TCM clinical research. Ensure that users can clearly express information such as the purpose of the research and the research population. Users need to make a simple description of the clinical research content to be conducted, and determine the modules they need to generate based on their clinical needs.

[0145] 2. Design a PICO knowledge graph database: Design a database structure to store the various categories of PICO elements and build a knowledge graph. You can use a graph database (such as the Neo4j database) to store the relationships between PICO elements to ensure a clear structure of the data.

[0146] 3. Automatically identify and classify entities in patient input information. Implementation of the PICO element recognition module: Integrate natural language processing technology and use a pre-trained large language model to identify PICO elements in user input and perform data retrieval on the constructed PICO knowledge graph database.

[0147] 4. Construction of a template library for clinical research proposals: Establish a template library containing various TCM clinical research proposals. The modular templates in the template library are formulated based on publicly published guidelines, research, literature or core indicator sets, including inclusion criteria design modules, exclusion criteria design modules, intervention measures design modules and outcome indicator design modules. The template library for the prompt word engineering needs to perform intelligent data matching and recommendation based on the contents of the four modules, and improve the quality of the templates through continuous updating of professional knowledge.

[0148] 5. Development of inclusion criteria design module: Based on the four elements of evidence-based medicine, the inclusion criteria are constructed, including the Western medicine diagnosis basis module, the Western medicine diagnosis standard module, the symptom and sign module, the disease function classification module, the classification standard definition module, the TCM diagnosis basis module, the TCM disease name module, the TCM syndrome type module, and the TCM diagnosis standard module. The intelligent design of the inclusion criteria needs to be comprehensively considered based on the four elements of evidence-based medicine PICO. First, the corresponding prompt words are input according to the input information, and the diagnosis of TCM diseases and Western medicine diseases is performed according to the input prompt words. Then, according to the intervention measures and outcome indicators, the relevant database is called to complete the data supplement of the relevant inclusion criteria, thereby completing the design of the intelligent generation of data for the final inclusion criteria.

[0149] 6. Development of exclusion criteria design module: The intelligent design of exclusion criteria includes the construction of knowledge graphs based on information of "P" (study population), "I" (intervention measures) and "C" (control measures), which mainly include modules for combining serious diseases, medical history, identification of other disease symptoms, abnormal clinical biochemical examinations, and completed or upcoming surgery. It is necessary to retrieve data from the intelligent knowledge graph based on the input disease information, TCM intervention measures and control measures, and finally complete the data generation of each submodule, and then use the pre-trained large model to integrate and optimize the data to complete the automatic generation of the scheme for the exclusion criteria module.

[0150] 7. Development of the Intervention Measure Design Module: Develop a dedicated module that intelligently designs intervention measures, including traditional Chinese medicine (TCM) drug names, Western medicine names, Western medicine categories, dosages, treatment courses, indications, etc., through integration with a medical professional knowledge base. The functions of the intervention measure module mainly include three parts: intervention measures for the experimental group, intervention measures for the control group, and the combination types of intervention measures. First, determine the recommended conventional Western medicine information for the relevant disease by input, then retrieve the corresponding dosage, category, and indications from the knowledge graph, and finally feedback the results for confirmation. The selection of TCM intervention measures requires intelligent retrieval of knowledge graph data based on the functions and indications of TCM. The combination methods of intervention measures for the experimental group and the control group include information such as combination of TCM and conventional Western medicine, single use of TCM, single use of conventional Western medicine, placebo control, etc. Finally, output the information to the large model for processing and automatically generate a recommended plan for the intelligent intervention measures of the intervention measure module.

[0151] 8. Development of the Outcome Indicator Design Module: Based on the relevant technologies of the clinical core indicator set and the data analysis of outcome indicators for TCM clinical research, construct the logical rules for indicator recommendation for various diseases, including indicator domains, indicator names, indicator classifications, indicator quality assessments, etc. First, select the types of outcome indicators involved, including symptom and sign indicator types, TCM indicator types, quality of life indicator types, physical and chemical examination indicator types, long-term prognosis indicator types, and safety indicator types. After inputting the selected intervention measure category, retrieve the relevant data and priority recommendation logical rules from the knowledge graph according to the input disease information for matching, and then intelligently select the corresponding measurement tools, and finally complete the intelligent recommendation generation of outcome indicators.

[0152] 9. Development of the Intelligent Generation Algorithm: Use large language models (such as the Qwen-7B model) and PICO elements in the knowledge graph to design an intelligent generation algorithm. This algorithm can automatically select from the template library or automatically generate a clinical research plan that meets the user's needs.

[0153] 10. Implementation of Automatic Matching and Recommendation: Use deep learning technology to implement a recommendation system based on the information input by the user, which automatically matches and combines templates for inclusion criteria, exclusion criteria, intervention measures, and outcome indicators. It is implemented based on recommendation algorithms such as those of the knowledge graph and large model or collaborative filtering.

[0154] 11. Design of the Modification and Customization Function: Introduce an editable interface in the system to allow users to modify and customize the generated clinical research plan. This may involve interactive elements such as form input and dropdown menu selection.

[0155] 12. Real-time verification and expert feedback: Introduce a real-time verification mechanism to ensure that the generated clinical research plan is scientifically and ethically reasonable. The system can provide real-time feedback to help users understand whether the selected design is practical. Introduce an expert evaluation mechanism in the field of traditional Chinese medicine to train the pre-trained large model through enhanced feedback learning.

[0156] 13. Added export and sharing functions: Provide the function of exporting the generated clinical research plan so that users can save, share or submit it for approval. Support common document formats (such as PDF, Word) and standard data exchange formats.

[0157] 14. Security and privacy protection measures: Implement data encryption, access control and other security measures to ensure that user-entered data and generated solutions are adequately protected and comply with relevant privacy regulations.

[0158] Through the above implementation methods, the intelligent design system for clinical research programs of traditional Chinese medicine based on evidence-based medicine can realize user-friendly interaction and intelligent program generation, and ensure the accuracy and security of generated data according to the prompt word engineering template and expert feedback mechanism. The modular design of the intelligent design system for clinical research programs of traditional Chinese medicine enables the organic connection between various functions to form a coherent and logically clear overall system.

[0159] The main functions of the modular templates in the template library mentioned above include: guidance on function division, guidance on interaction between modules, clear logical relationships and emphasis on function independence.

[0160] Functional division guidance: Modularization prompts help identify and divide different functional modules in the system or software, making the system structure clearer. It can guide developers to organize similar or related functions into independent modules during the design phase, which helps improve the maintainability and scalability of the system.

[0161] Inter-module interaction guidance: It can specify the interface specifications between modules, prompting developers to consider good interaction methods when designing modules, ensuring smooth cooperation between modules. Prompt words help plan the data flow and information transmission methods between modules, ensuring that the transmission and processing of information within the system is orderly.

[0162] Clear logical relationship: Defining the logical relationship between modules can clarify the logical relationship between different modules and help developers clarify the functional division and mutual relationship of each module. Emphasizing coupling and decoupling helps to design low-coupling and high-cohesion modules and improve the flexibility and maintainability of the system.

[0163] Functional independence emphasis: Through the independence of the four modules, it can be emphasized that each module should have independent functions, so that each module can be used or replaced separately in different scenarios. It helps to reduce the dependencies between modules and makes the system easier to maintain and update.

[0164] The main function of the modular template is to guide system designers and developers to carry out modular design, so that the system has a clear structure, independent functional modules and high maintainability.

[0165] Based on the development of the inclusion criteria design module, the inclusion criteria module was designed to solve the problem that the determination of inclusion criteria in traditional research may be subjective and lack of standardization, resulting in poor comparability of research results. The intelligent system can use natural language processing technology to automatically analyze a large amount of medical literature and identify and extract inclusion criteria related to the purpose of the study. The system can regularly retrieve the latest medical literature and guidelines to maintain the timeliness and scientificity of the inclusion criteria. According to the characteristics of the study and user needs, the system can generate personalized inclusion criteria suggestions to improve the pertinence of the research design.

[0166] Based on the development of the exclusion criteria design module, an exclusion criteria module was designed. This solves the problem that manual determination of exclusion criteria may be subjective and easy to miss some important information, affecting the comprehensiveness and credibility of the research. The intelligent system can provide more comprehensive and objective exclusion criteria recommendations by comprehensively considering various factors of the patient population through big data analysis. The system can monitor patient health data and research progress in real time, update exclusion criteria according to changes, and ensure the flexibility and timeliness of the research. According to the actual research situation, the system can intelligently adjust the exclusion criteria to avoid being too rigid and excluding potential applicants.

[0167] Based on the development of the intervention design module, an intervention module was designed. The design of interventions may be limited by the experience and knowledge level of researchers, and there may be deficiencies and unreasonableness. The system integrates professional knowledge in the field of traditional Chinese medicine through knowledge graphs to guide the design of interventions and ensure that the interventions are scientific and consistent with traditional Chinese medicine theory. According to the individual differences of patients and the purpose of the research, the system can generate personalized interventions to improve the customization level of the research. The system can adjust the interventions in real time according to patient feedback and research progress to maintain the flexibility and adaptability of the interventions.

[0168] Based on the development of the outcome indicator design module, an outcome indicator module was designed. This solves the problem that the selection of outcome indicators in traditional studies may lack systematicity and make it difficult to comprehensively evaluate the intervention effect. The intelligent system can comprehensively consider various types of patient data and research results to assist in selecting comprehensive and scientific outcome indicators. Using big data technology, the system can automatically analyze the correlation between various outcome indicators to avoid selecting redundant or independent indicators. The system can monitor the research effect in real time, adjust the outcome indicators according to data changes, and ensure the timeliness and accuracy of the evaluation system.

[0169] This system can promote the intelligence of clinical research on traditional Chinese medicine, solve many problems in traditional research, and provide stronger support for the scientific development and clinical practice of traditional Chinese medicine.

[0170] When using the TCM clinical research program generation system, the TCM clinical research program generation system includes a template end, an input end, and a multi-round feedback output end. The use process is: on the template end, template information based on population, intervention measures, control measures, and outcome indicators are constructed according to PICO. Supplementation of other information on the template end. Summarize the information on the "template end", input it on the input end, and perform semantic analysis through the text information sent by the user. It needs to be recognized by the prompt word project first, and the recognized information is fed back to the user through the multi-round feedback output end for information confirmation. After the confirmation is completed, the multi-round confirmation data is fed back to the large language model, and the large language model completes two tasks based on the prompt word project: 1. Large model call + knowledge graph based on prompt words; 2. Intelligent generation of large models based on prompt words. Finally, data integration will be carried out based on the results of the above two tasks, and all information will be integrated and uniformly summarized and fed back to the user.

[0171] As an optional implementation method, in order to realize intelligent "open" Q&A, the TCM clinical research program generation system of this application can use the "input-output" dialogue mode to input the TCM clinical research data information to be queried at the input end, and output the TCM clinical research Q&A based on the big model and knowledge graph at the output end, and mark the source of the data information. The specific design steps of intelligent "open" Q&A are as follows.

[0172] Step 801: Select the PICO knowledge graph of TCM clinical research as the data source, perform data enhancement expression on the large language model, complete data fine-tuning of the large language model, and call data based on the large language model and the knowledge graph.

[0173] Step 802: Implement control to complete intelligent question and answer of the inclusion criteria module, exclusion criteria module, intervention measures module and outcome indicator module of traditional Chinese medicine.

[0174] Step 803: Generate information by calling the corresponding data information of the inclusion standard module through the large language model, including TCM diagnosis standards, TCM diagnosis basis, Western medicine diagnosis basis, Western medicine diagnosis standards, symptoms and signs, disease function grading standards, definition of grading standards, TCM disease names, etc.

[0175] Step 804: Generate information by calling the data information of the corresponding exclusion criteria module through the large language model, including the situation of combined serious diseases, existing medical history, comorbidities, complications, differential diagnosis of other diseases, abnormal clinical biochemical examinations, previous surgery, upcoming surgery, medication precautions, medication contraindications, etc.

[0176] Step 805: Generate information by calling the data information of the corresponding intervention measures module through the large language model, including medication methods, medication methods of the experimental group, medication methods of the control group, names of Chinese medicine preparations, Chinese medicine ingredients, Chinese medicine properties, Chinese medicine usage and dosage, Chinese medicine specifications, Western medicine used to treat diseases, Western medicine types, Western medicine dosages, applicability of Western medicine, Western medicine category sources, Western medicine dosage sources, applicability sources, etc.

[0177] Step 806: Generate information by calling the data information of the corresponding outcome indicator module through the large language model, including indicator name, indicator domain, indicator pool, core indicator set, main indicator, secondary indicator, indicator English name, indicator measurement tool, indicator development method, etc.

[0178] Step 807: Other parts need to be filled in by the patient according to their needs, including sample size, gender, age, course of disease, comorbidities, complications, etc.

[0179] Step 808: Input the query questions of TCM clinical research input by scientific researchers or clinical researchers, and determine whether they are the content of TCM clinical research plan through the classification and recognition task of the BERT model.

[0180] Step 809: After confirming that the problem is a problem with the TCM clinical research program, execute the next task. Otherwise, reply and feedback that the task cannot be performed, and complete the correlation problem detection.

[0181] Step 810: After confirming the task, use the big model to call the data of the knowledge graph, and perform data retrieval, data feedback and data analysis.

[0182] Step 811: If the data cannot be retrieved from the knowledge graph, the fine-tuned large model will respond automatically, and the labeled data information will be generated intelligently by the model.

[0183] Step 812: The data can be successfully matched from the knowledge graph database, so that the confirmed information is fed back to the questioner through the big model, and the data is marked as coming from the data knowledge base.

[0184] The design content of the above TCM clinical research plan generation system can be summarized into the following design aspects.

[0185] 1. This system is based on the knowledge graph data platform and pre-trained large model technology of evidence-based traditional Chinese medicine. It innovatively constructs prompt word engineering for inclusion criteria, exclusion criteria, intervention measures and outcome indicators, thereby realizing the intelligent generation of modular clinical research plans for clinical research on traditional Chinese medicine.

[0186] 2. Use digital technology to integrate the four elements of evidence-based TCM diagnosis and treatment, "PICO". Through centralized data storage of multi-dimensional data, a unique modular generation method is formed, completing the intelligent generation design of holistic and modular TCM clinical research programs.

[0187] 3. This system includes factor template module, prompt word engineering control module, PICO knowledge graph module and language model module, which solves the key problems of accuracy, effectiveness and intelligent design in clinical research of traditional Chinese medicine.

[0188] 4. The knowledge graph involves the inclusion criteria part, exclusion criteria part, intervention measures part and outcome indicator part; the template involves the construction of a modular template.

[0189] The main feature of this application is the three-stage workflow of "research content inquiry-research program module query-clinical research program design" in the form of human-computer interaction, based on the PICO element knowledge graph of evidence-based medicine, through large-scale language model technology and prompt word engineering technology. The main technologies involved include expert-based construction of the PICO knowledge graph corpus, fine-tuning of large-scale language models, and construction of prompt word engineering rule columns, so as to complete the "open" intelligent question and answer of clinical research programs and "one-stop" intelligent program generation.

[0190] This system covers the "PICO" knowledge network, a key element of evidence-based evidence in evidence-based medicine of traditional Chinese medicine. By constructing relevant TCM prompt word projects, a large language model is used to complete the intelligent experimental plan design. Through standardization and normalization, a clinical research knowledge graph based on "PICO" is constructed. The knowledge graph covers modules such as inclusion criteria, exclusion criteria, intervention measures, and outcome indicators. Based on the PICO clinical research knowledge graph and combined with large model technology, an intelligent "open" question and answer is formed. Combined with the evidence-based medicine prompt word project of each module, the intelligent "one-stop" generation and design of clinical research plans for traditional Chinese medicine are completed. This application solves the problems of non-disclosure, opacity and non-standardization in the current design of clinical research plans for traditional Chinese medicine, and provides clinicians and scientific researchers with an intelligent plan generation technology system for the generation of intelligent evidence for clinical research, thereby improving the efficiency of clinical evidence production and transformation.

[0191] This system uses artificial intelligence technology, evidence-based medicine and traditional Chinese medicine diagnosis and treatment thinking to intelligently generate clinical trial plans, and provides detailed inclusion criteria, exclusion criteria, intervention measures and outcome indicators in an intelligent block-based manner. It can provide clinicians with convenient and efficient plan design tools, promote the development of clinical research on traditional Chinese medicine, promote the integration of clinical practice and scientific research, and provide accuracy and intelligent guarantees for clinical research on traditional Chinese medicine.

[0192] The method and system for generating clinical research plans for traditional Chinese medicine provided in this application, through the combination of large-scale language models and knowledge graphs, adopts the evidence-based medicine modular prompt word engineering technology constructed based on evidence-based medicine characteristics to enhance the intelligent generation technology of clinical trial plans in the field of traditional Chinese medicine, improve the high quality level of clinical trial plans for traditional Chinese medicine, and provide modern technical support and reference for the transformation of clinical evidence of traditional Chinese medicine.

[0193] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0194] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for generating a clinical research plan of traditional Chinese medicine, characterized in that: include: For clinical research on traditional Chinese medicine, a knowledge map covering the PICO elements of evidence-based research is constructed; the PICO includes research population, intervention measures, control measures and outcome indicators; Using the knowledge graph to train a pre-trained language model; Based on the PICO elements of evidence-based knowledge graph, a TCM prompting project was established; Determine element templates according to the evidence-based elements PICO covered by the knowledge graph; the element templates include a research population template, an intervention measure template, a control measure template, and an outcome indicator template; According to the TCM clinical research plan to be designed, selecting element information in the element template; Using the TCM prompt word project to identify the selected element information, generating a prompt word feature vector; Using the trained language model, retrieve data matching the prompt word feature vector from the knowledge graph; When there are features in the prompt word feature vector that cannot retrieve matching data from the knowledge graph, automatically generate matching data using the trained language model; Integrate the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph to obtain a clinical research plan for traditional Chinese medicine; Based on the PICO elements of evidence-based knowledge graph, a TCM prompting project was established, including: To construct a rule framework for prompt words in clinical research of traditional Chinese medicine; According to the inclusion criteria, exclusion criteria, intervention measures and outcome indicators in the knowledge map corresponding to the PICO of evidence-based elements, formulate prompt words; According to the importance of clinical research on traditional Chinese medicine, the prompt words are classified into necessary prompt words and non-essential prompt words; Determine the presentation form of the classified cue words; Based on the prompt word rule framework, a trained language model is used to debug prompt words with a determined expression form to obtain a traditional Chinese medicine prompt word project; The selected element information is identified by the TCM prompt word project to generate a prompt word feature vector, which specifically includes: Using the TCM prompt word project to identify the selected element information, and feeding back the identified element information; If the element information fed back is confirmed to be incorrect, return to step "select element information in the element template according to the TCM clinical research plan to be designed"; If the element information fed back is confirmed to be correct, then the TCM prompt word project is used to generate a prompt word feature vector based on the confirmed element information.

2. The method for generating a clinical research plan of traditional Chinese medicine according to claim 1, characterized in that: For clinical research on traditional Chinese medicine, a knowledge map covering the PICO elements of evidence-based research is constructed, including: The ontology of clinical research of traditional Chinese medicine was determined by adopting the seven-step ontology method; Construct a corpus of clinical research on traditional Chinese medicine; Aligning and matching the ontology of TCM clinical research on the corpus to obtain ontology information of PICO, an evidence-based evidence element; Performing structured processing on the ontology information to obtain PICO original data; The PICO raw data is standardized and defined through the ontology's seven-step method to generate a knowledge graph covering the PICO elements of evidence-based evidence.

3. The method for generating a clinical research plan of traditional Chinese medicine according to claim 1, characterized in that: The pre-trained language model is trained using the knowledge graph, and then the following steps are further included: Receive inquiries regarding clinical research on traditional Chinese medicine; Use the BERT model to identify whether the query question is the content of a clinical research plan for traditional Chinese medicine; If yes, then according to the query question, the trained language model is used to retrieve data from the knowledge graph; When data cannot be retrieved from the knowledge graph, a trained language model is used to generate response data, and the response data is marked as being generated by the model; When data is retrieved from the knowledge graph, the retrieved data is fed back through the trained language model, and the retrieved data is marked as coming from the data knowledge base; If not, the reply is that the query cannot be carried out.

4. The method for generating a clinical research plan of traditional Chinese medicine according to claim 1, characterized in that: Integrate the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph to obtain a clinical research plan for traditional Chinese medicine, specifically including: The matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph are structured and output according to the inclusion criteria, exclusion criteria, intervention measures and outcome indicators in the knowledge graph to obtain a clinical research plan for traditional Chinese medicine.

5. The method for generating a clinical research plan of traditional Chinese medicine according to claim 1, characterized in that: Integrate the matching data automatically generated by the trained language model and the matching data retrieved from the knowledge graph to obtain a clinical research plan for traditional Chinese medicine, and then include: Introduce a real-time verification mechanism to verify whether the TCM clinical research plan is practical and feasible; An expert evaluation mechanism in the field of traditional Chinese medicine is introduced, and the trained language model is trained through reinforced feedback learning.

6. A system for generating a clinical research plan for traditional Chinese medicine, characterized in that: The system for generating clinical research plans of traditional Chinese medicine includes: an element template module, a prompt word engineering control module, a PICO knowledge graph module and a language model module; The PICO knowledge graph module stores a knowledge graph; the knowledge graph is a knowledge graph covering the evidence-based evidence element PICO constructed for clinical research on traditional Chinese medicine; the PICO includes research population, intervention measures, control measures and outcome indicators; The element template module is connected to the prompt word engineering control module; the element template module is used to determine the element template according to the evidence-based evidence element PICO covered by the knowledge graph, and accept the element information selected in the element template according to the clinical research plan of traditional Chinese medicine to be designed; the element template includes a research population template, an intervention measure template, a control measure template and an outcome indicator template; the prompt word engineering control module establishes the traditional Chinese medicine prompt word engineering based on the evidence-based evidence element PICO covered by the knowledge graph, uses the traditional Chinese medicine prompt word engineering to identify the selected element information, and generates a prompt word feature vector; The language model module is connected to the prompt word engineering control module and the PICO knowledge graph module respectively; the language model module trains the pre-trained language model using the knowledge graph, and uses the trained language model to retrieve data matching the prompt word feature vector from the PICO knowledge graph module; when there are features in the prompt word feature vector that cannot retrieve matching data from the PICO knowledge graph module, the trained language model is used to automatically generate matching data; the matching data automatically generated by the trained language model and the matching data retrieved from the PICO knowledge graph module are integrated to obtain a clinical research plan for traditional Chinese medicine; The establishment of the TCM prompt word project specifically includes: To construct a rule framework for prompt words in clinical research of traditional Chinese medicine; According to the inclusion criteria, exclusion criteria, intervention measures and outcome indicators in the knowledge map corresponding to the PICO of evidence-based elements, formulate prompt words; According to the importance of clinical research on traditional Chinese medicine, the prompt words are classified into necessary prompt words and non-essential prompt words; Determine the presentation form of the classified cue words; Based on the prompt word rule framework, a trained language model is used to debug prompt words with a determined expression form to obtain a traditional Chinese medicine prompt word project; The selected element information is identified by the TCM prompt word project to generate a prompt word feature vector, which specifically includes: Using the TCM prompt word project to identify the selected element information, and feeding back the identified element information; If the element information fed back is confirmed to be incorrect, the element template module is called; If the element information fed back is confirmed to be correct, then the TCM prompt word project is used to generate a prompt word feature vector based on the confirmed element information.

7. The system for generating a clinical research plan for traditional Chinese medicine according to claim 6, characterized in that: The knowledge graph in the PICO knowledge graph module includes: inclusion criteria, exclusion criteria, intervention measures and outcome indicators; The inclusion criteria include: basic information of the test population, gender distribution, Chinese and Western medical diagnostic criteria for diseases or symptoms, disease course restrictions, characteristic descriptions of patients' diseases, physical sign information of the subjects, baseline physiological data, and pre-trial review procedures; The exclusion criteria include: exclusion criteria for treatment intervention, exclusion time range, exclusion criteria for patients’ baseline disease status, exclusion criteria for patients with set disease course, exclusion criteria for patients who do not meet the suitability criteria for TCM treatment, and exclusion criteria for set physiological indicators; The interventions include: multiple experimental and control groups in clinical studies of traditional Chinese medicine; The outcome indicators include: the definition of the outcome indicators, the time point of the outcome indicators and the observation period.

8. The system for generating a clinical research plan for traditional Chinese medicine according to claim 6, characterized in that: The TCM clinical research program generation system also includes: a graph database; The graph database is connected to the PICO knowledge graph module; The graph database is used to store the knowledge graph in the PICO knowledge graph module.

Images