Artificial intelligence traditional Chinese medicine intelligent report analysis and generation method and system

Abstract

The invention relates to an artificial intelligence traditional Chinese medicine intelligent report analysis and generation method and system, and the method comprises the steps: generating an initial diagnosis report based on the symptom information inputted by a patient through a preset symptom analysis model, and transmitting the initial diagnosis report to a corresponding doctor terminal; when a report interaction request sent by a physician terminal is received, the symptom analysis model collects interaction data sent by the physician terminal in real time; identifying associated items of the interaction data based on a feature reasoning strategy; generating associated data of the associated item according to the interaction data, and mapping the associated data to a corresponding position of the initial diagnosis report; and when a report confirmation instruction of the doctor terminal is received, generating a collaborative diagnosis report, generating an interaction change log based on the interaction data and the associated data, and sending the interaction change log to the symptom analysis model as a training sample. The method has the effect of improving the accuracy of traditional Chinese medicine intelligent report analysis.

Description

Technical Field

[0001] This application relates to the technical field of intelligent interaction, and in particular to an artificial intelligence-based method and system for generating and analyzing TCM intelligent reports. Background Technology

[0002] Existing AI-powered TCM diagnostic reports are typically static reports output by models and knowledge graphs, which are rather rigid. The models struggle to adapt to subtle individual differences among patients and the physician's on-the-spot decisions. If the AI's conclusions differ from the physician's judgment, the physician will discard the report, while the AI ​​model fails to recognize the discrepancy, creating a negative cycle that leads to a decline in the accuracy of AI-generated TCM diagnostic reports. Therefore, improvements are needed. Summary of the Invention

[0003] To enable interaction and feedback between AI-generated TCM medical reports and physicians, and to improve the accuracy of TCM intelligent report analysis, this application provides an AI-based TCM intelligent report analysis and generation method and system.

[0004] The above-mentioned objective of this application is achieved through the following technical solution:

[0005] An AI-powered method for generating intelligent reports on Traditional Chinese Medicine (TCM), comprising the following steps:

[0006] The preset symptom analysis model generates an initial diagnostic report based on the symptom information input by the patient, and sends the initial diagnostic report to the corresponding physician terminal;

[0007] When a report interaction request is received from the physician's terminal, the symptom analysis model collects the interaction data sent by the physician's terminal in real time.

[0008] Based on a feature-based reasoning strategy, the associated items in the interactive data are identified.

[0009] Based on the interaction data, generate associated data for the related items, and map the associated data to the corresponding location in the initial diagnostic report;

[0010] When a report confirmation instruction is received from the physician's terminal, a collaborative diagnosis report is generated, and an interaction change log is generated based on the interaction data and related data. The interaction change log is then sent to the symptom analysis model as a training sample.

[0011] By adopting the above technical solution, patients undergo AI-based consultations using a pre-trained symptom analysis model. The model outputs an initial diagnostic report based on symptom information, which is then sent to the corresponding physician's terminal. This initial report includes automatically generated diagnostic, treatment, and prescription information. Furthermore, the physician's terminal can issue interactive requests to the initial diagnostic report, allowing for editing and modification. Physicians can add or delete items, strengthen or weaken elements, adjust medication relationships, and modify the diagnostic description style, generating interactive data. Crucially, during the editing process, the physician, through feature reasoning strategies, can identify... The system identifies and generates associated data based on the interactive data, meaning it can recognize and infer the content that needs to be changed related to the content edited by the physician's terminal. For example, if the physician's terminal modifies diagnostic information, it can infer the addition or deletion of a certain herb in the corresponding prescription. This helps the physician's terminal efficiently and accurately complete the interactive request for the initial diagnostic report, generating a more accurate and reliable collaborative diagnostic report. Finally, both the interactive data and the generated associated data are sent back to the symptom analysis model for training. This targeted enhancement strengthens and improves the symptom analysis model's learning based on different physician diagnostic styles, improving diagnostic accuracy and enabling interaction and feedback between AI-powered TCM consultation reports and physicians, thereby improving the accuracy of TCM intelligent report analysis.

[0012] Optionally: The preset symptom analysis model generates an initial diagnostic report based on the symptom information input by the patient, and sends the initial diagnostic report to the corresponding physician terminal.

[0013] When the symptom analysis model receives symptom information input by the patient through the user terminal, it identifies the symptom type described in the symptom information;

[0014] Matching the corresponding physician terminal based on symptom type;

[0015] Retrieve the historical collaborative diagnosis reports and change logs associated with the matched physician terminal, and generate an initial diagnosis report based on the historical collaborative diagnosis reports and change logs.

[0016] By adopting the above technical solution, patients input symptom information through the user terminal. The symptom information includes text descriptions, images, or videos to demonstrate the patient's symptoms. The symptom analysis model determines the symptom type based on the symptom information. By matching the symptom type, it can match the corresponding physician in the relevant department and locate the physician's terminal. Furthermore, the symptom analysis model will retrieve the historical collaborative diagnosis reports and change logs of the physician's terminal and use the changed interaction data as a reference to improve the initial diagnosis report output by the symptom analysis model so that it is more consistent with the diagnostic style of the corresponding physician's terminal, and the generation of the initial diagnosis report is more accurate.

[0017] Optionally: the step of the symptom analysis model collecting the interaction data sent by the physician terminal in real time when a report interaction request is received from the physician terminal includes:

[0018] When a report interaction request is received from the physician's terminal, the physician's terminal is allowed to enter the editing interface of the initial diagnostic report;

[0019] Real-time acquisition of editing data from physicians' terminals on the editing interface, and identification of the intent information in the editing data;

[0020] Interaction data is obtained by associating edit data with intent information.

[0021] By adopting the above technical solution, after the physician's terminal enters the editing interface of the initial diagnostic report, it can adjust different types of data in the initial diagnostic report, including adjusting the order of primary and secondary diagnostic elements, adjusting the confidence level of symptoms, adding and deleting natural language annotations, and adding or deleting prescription drug names. Through the above operations, an intent understanding and analysis engine is used to determine the intent of the edited data. For example, the adjustment of the order of primary and secondary elements reflects the physician's emphasis on or weakening of a certain diagnostic result, and the semantic analysis is used to understand the physician's intent for natural language annotations. Finally, the edited data and intent information are associated to obtain interactive data, which facilitates the subsequent judgment and identification of related items.

[0022] Optionally: The step of identifying associated items in the interaction data based on the feature reasoning strategy includes:

[0023] Obtain the intent information of the edited data in the interaction data, and input the intent information into the pre-constructed knowledge reasoning graph;

[0024] In the knowledge reasoning graph, it is determined whether there are knowledge nodes associated with the intent information, and all associated knowledge nodes are filtered out.

[0025] Determine the mutual verification relationship between the currently input intent information and other intent information and symptom information. Define the confidence level of each knowledge node based on the mutual verification relationship, and filter out knowledge nodes with confidence levels higher than the threshold confidence level as associated items.

[0026] By adopting the above technical solution, a pre-constructed knowledge reasoning graph is used to link the three treatment processes: symptoms, diagnosis, and prescription. Intent information from interactive data is input into the knowledge reasoning graph, which then filters out knowledge nodes that permeate these processes. For example, if a physician adds a description of a symptom, the corresponding diagnosis and medication information nodes may need adjustment. Therefore, the corresponding knowledge nodes are first filtered out. Finally, the symptom information in the initial diagnostic report is cross-verified with the intent information from other interactive data in the same initial diagnostic report, assigning confidence levels to different knowledge nodes. This involves adjusting the weight of knowledge nodes based on consistency comparisons between different intents, thereby revealing the true purpose of the physician's editing and adjustments. Knowledge nodes with confidence levels higher than a threshold are selected as associated items, and the data for these associated items is automatically modified based on the interactive data.

[0027] Optional: The mutual verification relationship includes positive feedback and negative feedback relationships. The step of determining the mutual verification relationship between the currently input intent information and other intent information and symptom information, defining the confidence level of each knowledge node based on the mutual verification relationship, and filtering out knowledge nodes with confidence levels higher than a threshold confidence level as associated items includes:

[0028] Obtain the knowledge nodes associated with the intent information and match them with the basic confidence level. Different knowledge nodes correspond to different basic confidence levels.

[0029] Obtain symptom information from the current initial diagnostic report and identify whether there is any other intent information in the current initial diagnostic report;

[0030] When other intent information exists, it is determined whether there is a positive feedback relationship between the symptom information, other intent information and the current intent information. If there is, the knowledge node that plays the role of the positive feedback relationship is identified and the confidence of the knowledge node is increased.

[0031] When other intent information exists, it is determined whether there is a negative feedback relationship between the symptom information, other intent information and the current intent information. If so, the knowledge node that the negative feedback relationship plays a role is identified, and the confidence of the knowledge node is lowered.

[0032] By adopting the above technical solution, after the knowledge nodes are filtered out, a basic confidence level is preset for different knowledge nodes that jump from the intent information. The different basic confidence levels are pre-set based on the historical experience of intent information. Furthermore, based on the symptom information and the positive and negative feedback relationships between the other intent information and the current intent information, it reflects which knowledge node the physician's terminal's intention to modify and adjust the edited data is focused on, thereby realizing the adjustment of the confidence level of each knowledge node. For example, if the physician's terminal adds symptom b to the edited data based on symptom a, the intent information is determined to be based on the symptom information in the initial diagnosis report, and the physician believes that the patient also has symptom b. When this occurs, the knowledge reasoning graph generates multiple possible diagnostic nodes and medication information nodes related to symptom b. At the same time, the physician's terminal annotates the cause of symptom b in the initial diagnostic report. Therefore, through the positive feedback relationship between different intention information, the confidence of diagnostic nodes related to the cause can be increased and the confidence of diagnostic nodes unrelated to the cause can be decreased among the multiple possible diagnostic nodes of symptom b. Finally, the knowledge node containing the pathological name with the highest confidence is selected as the associated item. Furthermore, in the medication information node, the knowledge node containing the pathological name with the highest confidence can also be selected as the associated item based on the corresponding medication information node with the highest confidence.

[0033] Optionally: The step of generating associated data for the associated items based on the interaction data and mapping the associated data to the corresponding location in the initial diagnostic report includes:

[0034] Obtain the initial data from the initial diagnostic report of the related projects;

[0035] Based on the intent information in the interaction data, infer the direction of data adjustment for the initial data;

[0036] Based on the data adjustment direction, the corresponding related data is filtered out from the database of the corresponding knowledge node and matched and mapped to the position of the corresponding initial data in the initial diagnostic report.

[0037] Using the above technical solution, after the associated project is confirmed, the direction of data adjustment is confirmed through the initial data and intent information on the initial diagnostic report. The direction of data adjustment includes adding, deleting, replacing, and adjusting parameter values. Specifically, the corresponding associated data is selected from the database of the corresponding knowledge node for adding and replacing. Deletion is direct deletion. Parameter value adjustment is to select the corresponding adjustment rules from the database to adjust the parameters. For example, the corresponding disease information and drug name are selected from the database for adding, or the corresponding adjustment rules are selected from the database to adjust the dosage information after adding the drug, thereby achieving accurate generation of associated data.

[0038] Optionally: the step of generating an interaction change log based on interaction data and related data, and sending the interaction change log to the symptom analysis model as a training sample, includes:

[0039] Obtain the patient's symptom information from the initial diagnostic report that generated the interaction data, and perform anonymization processing on the symptom information, which includes the patient's identity information;

[0040] The symptom information, interaction data, and corresponding generated related data are associated and packaged to obtain the interaction change log;

[0041] The interaction change log is sent to the symptom analysis model. The symptom analysis model filters out the identity information of the corresponding physician based on the interaction change log and sends the interaction change log as a training sample to the model adaptation layer of the corresponding physician for training.

[0042] By adopting the above technical solution, the symptom information includes the patient's identity information, avatar and other private information. Therefore, for the purpose of training the symptom analysis model, the symptom information needs to be desensitized to protect the patient's privacy and security. Furthermore, the symptom information, interaction data and related data are packaged to obtain the interaction change log, which provides a closed-loop and accurate training dataset for the symptom analysis model. This improves the symptom analysis model's ability to output more accurate pathological diagnoses and medication recommendations when faced with different physician terminals and different symptom information.

[0043] The second objective of this invention is achieved through the following technical solution:

[0044] An AI-powered intelligent report analysis and generation system for Traditional Chinese Medicine (TCM) includes:

[0045] The initial diagnosis module is used to generate an initial diagnosis report based on the symptom information input by the patient using a preset symptom analysis model, and then send the initial diagnosis report to the corresponding physician terminal.

[0046] The interactive acquisition module is used to collect interactive data sent by the physician terminal in real time when a report interaction request is received from the physician terminal.

[0047] The association reasoning module is used to identify associated items in the interaction data based on a feature reasoning strategy.

[0048] The associated data module is used to generate associated data for the associated items based on the interaction data, and map the associated data to the corresponding position in the initial diagnostic report;

[0049] The collaborative acquisition module is used to generate a collaborative diagnosis report when it receives a report confirmation instruction from the physician's terminal, and to generate an interaction change log based on the interaction data and related data, and send the interaction change log to the symptom analysis model as a training sample.

[0050] The above-mentioned objective three of this application is achieved through the following technical solution:

[0051] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described artificial intelligence-based TCM intelligent report analysis and generation method.

[0052] The fourth objective of this application is achieved through the following technical solution:

[0053] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned artificial intelligence-based TCM intelligent report analysis and generation method.

[0054] In summary, this application includes at least one of the following beneficial technical effects:

[0055] 1. The key is that during the process of editing the initial diagnostic report, the physician can identify related items and generate related data through feature reasoning strategies. That is, it can identify and infer the content that needs to be changed related to the content edited by the physician's terminal. For example, if the physician modifies the diagnostic information, it can infer the addition or deletion of a certain herb in the corresponding prescription. This helps the physician's terminal to efficiently and accurately complete the interaction request of the initial diagnostic report, generate a more accurate and reliable collaborative diagnostic report, and finally send the interaction data and the generated related data back to the symptom analysis model for training. This targeted strengthens and improves the symptom analysis model's learning based on different physician diagnostic styles, improves the accuracy of diagnosis, realizes the interaction and feedback between AI TCM medical reports and physicians, and improves the accuracy of TCM intelligent report analysis.

[0056] 2. The symptom analysis model will retrieve the doctor's terminal's historical collaborative diagnosis reports and change logs, and use the changed interaction data as a reference to improve the initial diagnosis report output by the symptom analysis model so that it is more in line with the diagnostic style of the corresponding doctor's terminal, and the generation of the initial diagnosis report is more accurate.

[0057] 3. By verifying the symptom information in the initial diagnostic report and the intent information of other interactive data in the same initial diagnostic report, confidence levels are assigned to different knowledge nodes. That is, by comparing the consistency between different intents, the weight of knowledge nodes is adjusted, thereby obtaining the true purpose of the physician's terminal editing and adjustment, and selecting knowledge nodes with confidence levels higher than the threshold confidence level as associated items. The data of associated items will be automatically modified according to the interactive data.

[0058] 4. Based on the initial data and intent information in the initial diagnostic report, confirm the direction of data adjustment. The direction of data adjustment includes adding, deleting, replacing, and adjusting parameter values. Specifically, this involves selecting the corresponding related data from the database of the corresponding knowledge node for adding or replacing, deleting directly, and adjusting parameter values ​​by selecting the corresponding adjustment rules from the database to adjust the parameters. For example, selecting the corresponding disease information or drug name from the database for adding, or selecting the corresponding adjustment rules from the database to adjust the dosage information after adding a drug, thereby achieving accurate generation of related data. Attached Figure Description

[0059] Figure 1 This is a flowchart of an embodiment of an artificial intelligence-based method for generating and analyzing TCM reports according to this application;

[0060] Figure 2 This is a flowchart of step S20 in an embodiment of an artificial intelligence-based TCM intelligent report analysis and generation method of this application;

[0061] Figure 3 This is a flowchart of step S33 in an embodiment of an artificial intelligence-based TCM intelligent report analysis and generation method of this application;

[0062] Figure 4 This is a schematic block diagram of a computer device according to this application. Detailed Implementation

[0063] The following is in conjunction with the appendix Figure 1-4 This application will be described in further detail.

[0064] In one embodiment, such as Figure 1 As shown, this application discloses an artificial intelligence-based method for generating intelligent reports on traditional Chinese medicine, which specifically includes the following steps:

[0065] S10: The preset symptom analysis model generates an initial diagnostic report based on the symptom information input by the patient and sends the initial diagnostic report to the corresponding physician terminal;

[0066] In this embodiment, the symptom information includes data such as symptom text information, symptom images or videos input by the patient. The symptom analysis model adopts the training language model in the field of traditional Chinese medicine, including Hua Tuo GPT. By inputting multi-source heterogeneous symptom information from various patients and performing feature vector transformation, the BiLSTM+CRF model is used to extract standardized symptom terms, and application network fusion is used to perform new associations, outputting the pathological diagnosis information and medication information corresponding to the patient's symptoms.

[0067] The initial diagnostic report includes the patient's identity information, symptom information recorded in text, images, and / or video, corresponding physician information, and medication prescription information. The initial diagnostic report is an electronic report. The physician terminal is a PC or mobile terminal used by physicians and linked to their identity.

[0068] Specifically, step S10 includes the following steps:

[0069] S11: When the symptom analysis model receives symptom information input by the patient through the user terminal, it identifies the symptom type described in the symptom information;

[0070] S12: Matching the corresponding physician terminal based on symptom type;

[0071] S13: Retrieve the historical collaborative diagnosis reports and change logs associated with the matched physician terminal, and generate an initial diagnosis report based on the historical collaborative diagnosis reports and change logs.

[0072] Among them, the symptom type is a pre-set pathological type. Based on the symptom type, it is matched with the corresponding physician terminal. When the initial diagnosis report is generated by matching in the model adaptation layer, it also refers to the historical collaborative reports of the same symptom information and the change log of the corresponding physician terminal, and assigns them corresponding reference weight values, so that the generated initial diagnosis report is more in line with the medication and diagnosis style of the corresponding physician terminal.

[0073] S20: When a report interaction request is received from the physician terminal, the symptom analysis model collects the interaction data sent by the physician terminal in real time.

[0074] In this embodiment, the physician can send an interaction request to edit the initial diagnosis report by clicking on the interaction request area of ​​the initial diagnosis report displayed on the physician terminal. Each time the physician terminal modifies a section of the initial diagnosis report, it is considered to send one interaction data. Modifications to different sections are considered to send multiple interaction data. The section classification is a custom setting.

[0075] Specifically, refer to Figure 2 Step S20 includes the following steps:

[0076] S21: When a report interaction request is received from the physician terminal, the physician terminal is allowed to enter the editing interface of the initial diagnostic report;

[0077] S22: Real-time acquisition of editing data from the physician's terminal on the editing interface, and identification of the intent information of the editing data;

[0078] S23: Associate the edit data with the intent information to obtain the interaction data.

[0079] The editing interface is for the initial diagnostic report and contains multiple editing areas, each representing a section. Editing data includes drag-and-drop sorting, slider adjustment, natural language annotation, and adding, deleting, and modifying knowledge units. Drag-and-drop sorting allows users to adjust the order of syndrome elements in the diagnosis by dragging the cards; slider adjustment allows users to directly adjust the confidence or severity of "spleen deficiency" from 0.8 to 0.9; natural language annotation allows users to highlight text at any location and input voice or text: "This patient is depressed and should also have liver-soothing symptoms"; adding, deleting, and modifying knowledge units allows users to delete "Codonopsis pilosula" and drag "Codonopsis pilosula" from the pop-up recommended drug library; or directly add a new symptom, "distension in the hypochondrium".

[0080] Intent understanding employs an intent understanding engine for real-time classification and parsing, including operations such as strengthening or weakening elements, adding or deleting elements, adjusting relationships, and modifying expression style. For natural language annotations, a semantic parsing module is used to understand the intent of the natural language annotations. Associating the edit data in a section with the corresponding intent information yields interaction data. The physician terminal generates one or more interaction data in a single interaction request, thus generating multiple different intent information.

[0081] S30: Based on the feature reasoning strategy, identify the associated items of the interactive data;

[0082] In this embodiment, the feature inference strategy refers to inferring the related changes in the content modified by the physician based on the pathological relationships of symptoms, diagnosis, and medication within a pre-constructed knowledge graph for pathological analysis. Related items refer to data in the same or other sections of the initial diagnostic report that require synchronized modification with the edited data.

[0083] Specifically, step S30 includes:

[0084] S31: Obtain the intent information of editing data from the interactive data, and input the intent information into the pre-built knowledge reasoning graph;

[0085] S32: In the knowledge reasoning graph, determine whether there are any knowledge nodes associated with the intent information, and filter out all associated knowledge nodes;

[0086] S33: Determine the mutual verification relationship between the currently input intent information and other intent information and symptom information, define the confidence level of each knowledge node based on the mutual verification relationship, and filter out knowledge nodes with confidence levels higher than the threshold confidence level as associated items.

[0087] The knowledge reasoning graph includes three different types of nodes: symptom information nodes, disease name nodes, and medication information nodes. Intent information is input into the knowledge reasoning graph to select the starting position for navigation. Based on the intent at the starting position, navigation proceeds to other types of related knowledge nodes. Mutual verification relationships refer to the relationships between different intent information that mutually verify each other through features to prove they belong to the same intent. This is used to adjust the confidence level of knowledge nodes and improve the accuracy of judging related items. For example, symptom 'a' includes three disease diagnosis nodes: node 1, node 2, and node 3. Guided by the patient's symptoms in the initial diagnostic report or other intent information from the doctor's terminal, it can be determined that symptom 'a' is caused by the disease at node 2. Node 2 is then designated as the related item.

[0088] Furthermore, refer to Figure 3 The mutual verification relationship includes positive feedback and negative feedback relationships. Step S33 includes the following steps:

[0089] S331: Obtain the knowledge nodes associated with the intent information and match them with the basic confidence level. Different knowledge nodes correspond to different basic confidence levels.

[0090] S332: Obtain symptom information from the current initial diagnostic report and identify whether there is any other intent information in the current initial diagnostic report;

[0091] S333: When other intent information exists, determine whether there is a positive feedback relationship between the symptom information, other intent information and the current intent information. If so, identify the knowledge node that plays the role of the positive feedback relationship and increase the confidence of the knowledge node.

[0092] S334: When other intent information exists, determine whether there is a negative feedback relationship between the symptom information, other intent information and the current intent information. If so, identify the knowledge node that the negative feedback relationship is effective and lower the confidence level of the knowledge node.

[0093] In this embodiment, the basic confidence level of different knowledge nodes is defined based on the historical relationship between specific intent information and the corresponding knowledge nodes. Positive feedback relationships can be weighted and increased to enhance the confidence of knowledge nodes, while negative feedback relationships decrease their confidence. For example, if a physician adds symptom b to symptom a in the edited data of the physician's terminal, the intent information is determined to be based on the symptom information in the initial diagnostic report, indicating that the physician believes the patient also has symptom b. At this point, the knowledge reasoning graph generates multiple possible diagnostic nodes and medication information nodes related to symptom b. Simultaneously, the physician's terminal annotates the cause of symptom b in the initial diagnostic report. Therefore, through the positive feedback relationship between different intent information, the confidence level of diagnostic nodes related to the cause can be increased, while the confidence level of diagnostic nodes unrelated to the cause can be decreased among the multiple possible diagnostic nodes for symptom b. Ultimately, the knowledge node containing the pathological name with the highest confidence level is selected as the associated item.

[0094] S40: Based on the interaction data, generate associated data for the associated items and map the associated data to the corresponding location in the initial diagnostic report;

[0095] In this embodiment, the data types of the associated data and the edited data are consistent. For example, the addition of a symptom leads to the generation of a new symptom in the associated data. Mapping to the corresponding position in the initial diagnostic report refers to mapping to the editable position of each section in the initial diagnostic report.

[0096] Specifically, step S40 includes the following steps:

[0097] S41: Obtain the initial data from the initial diagnostic report of the related project;

[0098] S42: Based on the intent information in the interaction data, infer the direction of data adjustment for the initial data;

[0099] S43: Based on the data adjustment direction, filter out the corresponding related data in the database of the corresponding knowledge node, and match and map it to the position of the corresponding initial data in the initial diagnostic report.

[0100] The data adjustment methods include adding, deleting, and replacing data; strengthening and weakening numerical parameters; and adjusting specific parameter values. The corresponding knowledge node databases include databases storing information on multiple diseases, databases storing information on multiple prescriptions and dosages, and databases storing information on multiple symptom manifestations.

[0101] S50: When a report confirmation instruction is received from the physician terminal, a collaborative diagnosis report is generated, and an interaction change log is generated based on the interaction data and related data. The interaction change log is then sent to the symptom analysis model as a training sample.

[0102] In this embodiment, the interaction request is considered complete only when the physician's terminal issues a report confirmation command. The interaction change log records the interaction data made by the physician's terminal in response to the patient's specific symptom information.

[0103] Specifically, step S50 includes the following steps:

[0104] S51: Obtain the patient's symptom information from the initial diagnostic report that generated the interaction data, and perform desensitization processing on the symptom information, which includes the patient's identity information;

[0105] S52: Combine and package the symptom information, interaction data, and corresponding generated related data to obtain the interaction change log;

[0106] S53: The interaction change log is sent to the symptom analysis model. The symptom analysis model filters out the identity information of the corresponding physician based on the interaction change log and sends the interaction change log as a training sample to the model adaptation layer of the corresponding physician for training.

[0107] In this process, anonymization is used to protect the patient's identity and image information. For video data, keyframes are extracted and anonymized, making the training of the symptom analysis model more lightweight.

[0108] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0109] In one embodiment, an AI-based intelligent TCM report analysis and generation system is provided, which corresponds to the AI-based intelligent TCM report analysis and generation method described in the previous embodiment. This AI-based intelligent TCM report analysis and generation system includes:

[0110] The initial diagnosis module is used to generate an initial diagnosis report based on the symptom information input by the patient using a preset symptom analysis model, and then send the initial diagnosis report to the corresponding physician terminal.

[0111] The interactive acquisition module is used to collect interactive data sent by the physician terminal in real time when a report interaction request is received from the physician terminal.

[0112] The association reasoning module is used to identify associated items in the interaction data based on a feature reasoning strategy.

[0113] The associated data module is used to generate associated data for the associated items based on the interaction data, and map the associated data to the corresponding position in the initial diagnostic report;

[0114] The collaborative acquisition module is used to generate a collaborative diagnosis report when it receives a report confirmation instruction from the physician's terminal, and to generate an interaction change log based on the interaction data and related data, and send the interaction change log to the symptom analysis model as a training sample.

[0115] Optionally, the initial diagnostic module includes:

[0116] The type identification submodule is used to identify the symptom type when the symptom analysis model receives symptom information input by the patient through the user terminal;

[0117] The terminal matching submodule is used to match the corresponding physician terminal based on the symptom type;

[0118] The report generation submodule is used to retrieve the historical collaborative diagnosis reports and change logs associated with the matched physician terminal, and generate an initial diagnosis report based on the historical collaborative diagnosis reports and change logs.

[0119] Optionally, the interactive data acquisition module includes:

[0120] The interaction request submodule is used to allow the physician terminal to enter the editing interface of the initial diagnostic report when a report interaction request is received from the physician terminal.

[0121] The intent analysis submodule is used to collect editing data from the physician's terminal in the editing interface in real time and identify the intent information of the editing data;

[0122] The association submodule is used to associate edit data with intent information to obtain interactive data.

[0123] Optional, the associative reasoning module includes:

[0124] The intent input submodule is used to obtain intent information from the interactive data and input the intent information into the pre-built knowledge reasoning graph.

[0125] The node filtering submodule is used to determine whether there are any knowledge nodes associated with intent information in the knowledge reasoning graph, and to filter out all associated knowledge nodes.

[0126] The confidence submodule is used to determine the mutual verification relationship between the currently input intent information and other intent information and symptom information. Based on the mutual verification relationship, the confidence of each knowledge node is defined, and knowledge nodes with confidence higher than the threshold confidence are selected as associated items.

[0127] Optionally, the mutual verification relationship includes positive feedback and negative feedback relationships, and the confidence submodule includes:

[0128] The confidence matching unit is used to obtain the knowledge nodes associated with the intent information and match them with the basic confidence level. Different knowledge nodes correspond to different basic confidence levels.

[0129] The mutual verification relationship unit is used to obtain symptom information in the current initial diagnosis report and identify whether there is other intent information in the current initial diagnosis report; when there is other intent information, it is determined whether there is a positive feedback relationship between the symptom information, other intent information and the current intent information; if so, the knowledge node that the positive feedback relationship plays a role is identified and the confidence of the knowledge node is increased.

[0130] When other intent information exists, it is determined whether there is a negative feedback relationship between the symptom information, other intent information and the current intent information. If so, the knowledge node that the negative feedback relationship plays a role is identified, and the confidence of the knowledge node is lowered.

[0131] Optional, the associated data module includes:

[0132] The initial data acquisition submodule is used to acquire the initial data of related projects in the initial diagnostic report;

[0133] The data adjustment direction submodule is used to infer the data adjustment direction of the initial data based on the intent information in the interactive data;

[0134] The data mapping submodule is used to adjust the direction based on data, filter out the corresponding related data in the database of the corresponding knowledge node, and match and map it to the position of the corresponding initial data in the initial diagnostic report.

[0135] Optionally, the collaborative data acquisition module includes:

[0136] The desensitization submodule is used to obtain the patient's symptom information from the initial diagnostic report that generates the interactive data, and to desensitize the symptom information, which includes the patient's identity information.

[0137] The Interaction Change Log submodule is used to associate and package symptom information, interaction data, and corresponding generated related data to obtain the interaction change log.

[0138] The training set sending submodule is used to send the interaction change log to the symptom analysis model. The symptom analysis model filters out the identity information of the corresponding physician based on the interaction change log and sends the interaction change log as a training sample to the model adaptation layer of the corresponding physician for training.

[0139] Specific limitations regarding an AI-based intelligent TCM report analysis and generation system can be found in the above description of the limitations of an AI-based intelligent TCM report analysis and generation method, and will not be repeated here. Each module in the aforementioned AI-based intelligent TCM report analysis and generation system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0140] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an artificial intelligence-based method for generating intelligent reports in Traditional Chinese Medicine.

[0141] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements an artificial intelligence-based method for generating intelligent reports on traditional Chinese medicine.

[0142] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements an artificial intelligence-based method for generating and analyzing TCM intelligent reports.

[0143] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0144] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0145] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for generating intelligent TCM reports using artificial intelligence, characterized in that, Including the following steps: The preset symptom analysis model generates an initial diagnostic report based on the symptom information input by the patient, and sends the initial diagnostic report to the corresponding physician terminal; When a report interaction request is received from the physician's terminal, the symptom analysis model collects the interaction data sent by the physician's terminal in real time. Based on a feature-based reasoning strategy, the associated items in the interactive data are identified. Based on the interaction data, generate associated data for the related items, and map the associated data to the corresponding location in the initial diagnostic report; When a report confirmation instruction is received from the physician's terminal, a collaborative diagnosis report is generated, and an interaction change log is generated based on the interaction data and related data. The interaction change log is then sent to the symptom analysis model as a training sample.

2. The method for generating and analyzing intelligent TCM reports using artificial intelligence according to claim 1, characterized in that, The steps of generating an initial diagnostic report based on the patient's input symptom information using the preset symptom analysis model, and sending the initial diagnostic report to the corresponding physician terminal are as follows: When the symptom analysis model receives symptom information input by the patient through the user terminal, it identifies the symptom type described in the symptom information; Matching the corresponding physician terminal based on symptom type; Retrieve the historical collaborative diagnosis reports and change logs associated with the matched physician terminal, and generate an initial diagnosis report based on the historical collaborative diagnosis reports and change logs.

3. The method for generating intelligent TCM reports using artificial intelligence according to claim 1, characterized in that, The step of the symptom analysis model collecting the interaction data sent by the physician terminal in real time when a report interaction request is received from the physician terminal includes: When a report interaction request is received from the physician's terminal, the physician's terminal is allowed to enter the editing interface of the initial diagnostic report; Real-time acquisition of editing data from physicians' terminals on the editing interface, and identification of the intent information in the editing data; Interaction data is obtained by associating edit data with intent information.

4. The method for generating intelligent TCM reports using artificial intelligence according to claim 3, characterized in that, The step of identifying associated items in the interaction data based on the feature reasoning strategy includes: Obtain the intent information of the edited data in the interaction data, and input the intent information into the pre-constructed knowledge reasoning graph; In the knowledge reasoning graph, it is determined whether there are knowledge nodes associated with the intent information, and all associated knowledge nodes are filtered out. Determine the mutual verification relationship between the current input intent information and other intent information and symptom information. Define the confidence level of each knowledge node based on the mutual verification relationship. Select knowledge nodes with confidence levels higher than the threshold confidence level as associated items.

5. The method for generating intelligent TCM reports using artificial intelligence according to claim 4, characterized in that, The mutual verification relationship includes positive feedback and negative feedback relationships. The step of determining the mutual verification relationship between the currently input intent information and other intent information and symptom information, defining the confidence level of each knowledge node based on the mutual verification relationship, and filtering out knowledge nodes with confidence levels higher than a threshold confidence level as associated items includes: Obtain the knowledge nodes associated with the intent information and match them with the basic confidence level. Different knowledge nodes correspond to different basic confidence levels. Obtain symptom information from the current initial diagnostic report and identify whether there is any other intent information in the current initial diagnostic report; When other intent information exists, it is determined whether there is a positive feedback relationship between the symptom information, other intent information and the current intent information. If there is, the knowledge node that plays the role of the positive feedback relationship is identified and the confidence of the knowledge node is increased. When other intent information exists, it is determined whether there is a negative feedback relationship between the symptom information, other intent information and the current intent information. If so, the knowledge node that the negative feedback relationship plays a role is identified, and the confidence of the knowledge node is lowered.

6. The method for generating intelligent TCM reports using artificial intelligence according to claim 3, characterized in that, The step of generating associated data for the associated items based on the interaction data and mapping the associated data to the corresponding location in the initial diagnostic report includes: Obtain the initial data from the initial diagnostic report of the related projects; Based on the intent information in the interaction data, infer the direction of data adjustment for the initial data; Based on the data adjustment direction, the corresponding related data is filtered out from the database of the corresponding knowledge node and matched and mapped to the position of the corresponding initial data in the initial diagnostic report.

7. The method for generating intelligent TCM reports using artificial intelligence according to claim 1, characterized in that, The step of generating an interaction change log based on interaction data and related data, and sending the interaction change log to the symptom analysis model as a training sample, includes: Obtain the patient's symptom information from the initial diagnostic report that generated the interaction data, and perform anonymization processing on the symptom information, which includes the patient's identity information; The symptom information, interaction data, and corresponding generated related data are associated and packaged to obtain the interaction change log; The interaction change log is sent to the symptom analysis model. The symptom analysis model filters out the identity information of the corresponding physician based on the interaction change log and sends the interaction change log as a training sample to the model adaptation layer of the corresponding physician for training.

8. An artificial intelligence-based intelligent report analysis and generation system for traditional Chinese medicine, characterized in that, include: The initial diagnosis module is used to generate an initial diagnosis report based on the symptom information input by the patient using a preset symptom analysis model, and then send the initial diagnosis report to the corresponding physician terminal. The interactive acquisition module is used to collect interactive data sent by the physician terminal in real time when a report interaction request is received from the physician terminal. The association reasoning module is used to identify associated items in the interaction data based on a feature reasoning strategy. The associated data module is used to generate associated data for the associated items based on the interaction data, and map the associated data to the corresponding position in the initial diagnostic report; The collaborative acquisition module is used to generate a collaborative diagnosis report when it receives a report confirmation instruction from the physician's terminal, and to generate an interaction change log based on the interaction data and related data, and send the interaction change log to the symptom analysis model as a training sample.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the artificial intelligence-based TCM intelligent report analysis and generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the artificial intelligence-based TCM intelligent report analysis and generation method as described in any one of claims 1 to 7.

Images