A traditional Chinese medicine data record management platform and method using knowledge graph

Abstract

The present invention discloses a TCM data record management platform and method using a knowledge graph, specifically relating to the field of data management technology. The platform collects clinical cases of multiple famous TCM doctors and constructs a TCM image thinking feature dataset. A fuzzy clustering algorithm is used to extract continuous image feature patterns, an image feature pattern library is established, and symbolic mapping is performed to form a preliminary knowledge graph. An enhanced knowledge graph is constructed through non-discretization enhancement processing, and an iterative adjustment model is constructed based on mapping distortion analysis to optimize and adjust the enhanced knowledge graph structure. Finally, a TCM data management knowledge graph is constructed, which can achieve semantic mapping and intelligent reasoning of new patient diagnosis and treatment data, effectively improving the digital expression capability of traditional image thinking and the level of intelligent support for diagnosis and treatment decision-making.

Description

Technical Field

[0001] The present invention relates to the field of data management technology, and more specifically, to a traditional Chinese medicine data record management platform and method using a knowledge graph. Background Art

[0002] Traditional Chinese Medicine (TCM) diagnosis, especially by renowned TCM practitioners, widely employs a "Xiang Thinking" model, characterized by analogy and inference, based on external observation and inference. This model relies heavily on the physician's intuitive perception and comprehensive judgment of the holistic, nonlinear, and ambiguous nature of disease characteristics. In modern TCM data management and digital representation, knowledge graphs, as a structured tool based on symbolic logic, struggle to effectively express the holistic and continuous nature of "Xiang Thinking." This leads to semantic distortion and loss of key information during the transformation of diagnostic and treatment knowledge from holistic and fuzzy semantics to symbolic logic, severely limiting the accuracy and credibility of graph-based reasoning in the TCM context.

[0003] In order to solve the above problems, a technical solution is now provided. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a traditional Chinese medicine data record management platform and method using a knowledge graph to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for managing traditional Chinese medicine data records using a knowledge graph comprises the following steps:

[0007] Collect and structure the diagnosis and treatment experience data from multiple clinical cases of famous Chinese medicine practitioners to construct a dataset of Chinese medicine image thinking characteristics;

[0008] Based on the TCM Xiang thinking feature dataset, we conduct a continuous feature analysis of Xiang concepts, use fuzzy clustering algorithms to determine the continuous feature patterns of Xiang features in the diagnosis and treatment experience data, and output the Xiang feature pattern library.

[0009] The object feature pattern library is symbolically mapped and the initial TCM diagnosis and treatment knowledge map is formed using the ontology construction method;

[0010] The nonlinear association relationship between node entities in the preliminary TCM diagnosis and treatment knowledge graph is processed by non-discretization enhancement to form an enhanced knowledge graph;

[0011] Perform mapping distortion analysis on the enhanced knowledge graph and the image feature pattern library, and build an iterative adjustment model;

[0012] Based on the iterative adjustment model, the ontology node and relationship edge parameters of the enhanced knowledge graph are optimized and output as a knowledge graph for TCM data management.

[0013] Based on the knowledge graph of traditional Chinese medicine data management, the newly input patient diagnosis and treatment data are mapped and inferred in real time to generate diagnosis and treatment decision recommendations that match the thinking logic of famous Chinese medicine practitioners.

[0014] In a preferred embodiment, the diagnosis and treatment experience data from multiple clinical cases of famous Chinese medicine practitioners are collected and structured to construct a dataset of Chinese medicine image thinking characteristics, specifically:

[0015] Collect diagnostic description information and treatment medication information from multiple clinical cases of famous Chinese medicine practitioners;

[0016] Classify and mark the diagnostic description information according to symptoms, tongue description, pulse description, and disease change characteristics;

[0017] Structurally mark the therapeutic drug information according to drug type, drug dosage, preparation method, and drug compatibility relationship;

[0018] Establish a clinical case database of famous Chinese medicine practitioners based on the classified and labeled diagnostic description information and the structured and labeled therapeutic medication information;

[0019] Based on the clinical case database of famous Chinese medicine practitioners, the correlation data of diagnosis and treatment elements are extracted to form a dataset of the characteristics of Chinese medicine image thinking.

[0020] In a preferred embodiment, based on the TCM Xiang thinking feature dataset, the continuity feature analysis of the Xiang concept is performed, and the fuzzy clustering algorithm is used to determine the continuity feature pattern of the Xiang feature in the diagnosis and treatment experience data, and the Xiang feature pattern library is output, specifically:

[0021] The diagnosis and treatment element association data in the TCM Xiang Si thinking feature dataset is vectorized and coded to construct a diagnosis and treatment feature vector set;

[0022] The fuzzy clustering algorithm based on membership calculation is used to perform feature pattern recognition on the diagnosis and treatment feature vector set to generate multiple initial clusters of image features;

[0023] Calculate the aggregate membership index of each diagnosis and treatment feature vector in each initial cluster of image features;

[0024] Determine the continuity boundary between image features based on the aggregation membership index;

[0025] The continuity boundary is subjected to fuzzy constraint optimization processing, and an image feature pattern set is output to form an image feature pattern library.

[0026] In a preferred embodiment, the object feature pattern library is symbolically mapped and an ontology construction method is used to form a preliminary TCM diagnosis and treatment knowledge map, specifically:

[0027] Assign a unique symbol identifier to each image feature pattern in the image feature pattern library to generate an image feature symbol table;

[0028] Determine the arrangement rules of the image concept node set and the image concept hierarchy relationship according to the image feature symbol table;

[0029] Determine the node entity set and relationship edge set based on the diagnosis and treatment element association data;

[0030] Adopting the knowledge domain ontology construction rules, the ontology structure framework is established according to the hierarchical relationship arrangement rules of image concepts, node entity sets, and relationship edge sets;

[0031] The concept node set, node entity set and relationship edge set are written into the ontology structure framework to generate a preliminary TCM diagnosis and treatment knowledge graph.

[0032] In a preferred embodiment, the nonlinear association relationship between node entities in the preliminary TCM diagnosis and treatment knowledge graph is subjected to non-discretization enhancement processing to form an enhanced knowledge graph, specifically:

[0033] Construct a node entity association matrix based on the node entities and relationship edges in the preliminary TCM diagnosis and treatment knowledge graph;

[0034] Perform continuous weight assignment processing on the association relationships in the node entity association matrix to form a node entity continuous association matrix;

[0035] Establish a node-entity multidimensional relationship network based on the node-entity continuous association matrix;

[0036] The strength of the association relationship in the multidimensional relationship network of node entities is weighted and optimized to form a non-discretized enhanced network structure of node entities;

[0037] According to the non-discretized enhanced network structure of node entities, the continuity weights of node entities and relationship edges in the preliminary TCM diagnosis and treatment knowledge graph are updated to generate an enhanced knowledge graph.

[0038] In a preferred embodiment, a mapping distortion analysis is performed on the enhanced knowledge graph and the image feature pattern library to construct an iterative adjustment model, specifically:

[0039] Generate enhanced knowledge graph structure data based on the enhanced knowledge graph, and generate image feature pattern benchmark data based on the image feature pattern library;

[0040] Calculate the mapping error values ​​of node entities and relationship edges based on the enhanced knowledge graph structure data and the image feature pattern benchmark data;

[0041] Determine the overall mapping distortion of the enhanced knowledge graph based on the mapping error value between node entities and relationship edges;

[0042] The mapping error feedback function is established using the overall mapping distortion of the enhanced knowledge graph;

[0043] Based on the mapping error feedback function, iterative adjustment rules for node entities and relationship edges are constructed to form an iterative adjustment model.

[0044] In a preferred embodiment, the ontology nodes and relationship edge parameters of the knowledge graph are optimized and enhanced according to the iterative adjustment model, and the TCM data management knowledge graph is output, specifically:

[0045] According to the iterative adjustment rules of the node entities in the iterative adjustment model, the definition attributes and hierarchical structure of the ontology nodes in the enhanced knowledge graph are modified;

[0046] According to the iterative adjustment rules of the relationship edges in the iterative adjustment model, the weight parameters and association directions of the relationship edges in the enhanced knowledge graph are modified;

[0047] Perform consistency check on the revised ontology nodes and relationship edges to generate a structured updated dataset that conforms to semantic logic;

[0048] The structure-updated dataset is applied to the enhanced knowledge graph to form a knowledge graph for traditional Chinese medicine data management.

[0049] In a preferred embodiment, based on the TCM data management knowledge graph, newly input patient diagnosis and treatment data is mapped and reasoned in real time to generate diagnosis and treatment decision suggestions that match the thinking logic of famous TCM practitioners, specifically:

[0050] Extract diagnosis and treatment elements from newly input patient diagnosis and treatment data to obtain diagnosis and treatment feature items;

[0051] Semantically match the diagnosis and treatment feature items with the ontology node set in the TCM data management knowledge graph to determine the semantic relevance with each ontology node;

[0052] Based on the semantic correlation, a multi-hop correlation path set is constructed;

[0053] Generate a diagnosis and treatment decision reasoning path sequence based on the multi-hop association path set, and extract the diagnosis recommendation nodes and medication recommendation nodes in the diagnosis and treatment decision reasoning path sequence;

[0054] The diagnosis and treatment recommendations corresponding to the diagnosis and treatment recommendation nodes are combined to form treatment decision recommendations.

[0055] On the other hand, the present invention provides a traditional Chinese medicine data record management platform using a knowledge graph, including a data construction module, a pattern clustering module, an ontology construction module, a graph enhancement module, an iterative adjustment module, a graph optimization module, and an intelligent decision-making module;

[0056] The data construction module collects and structures the diagnosis and treatment experience data from multiple clinical cases of famous Chinese medicine practitioners to construct a dataset of Chinese medicine image thinking characteristics;

[0057] The pattern clustering module analyzes the continuity characteristics of the Xiang concept based on the TCM Xiang thinking feature dataset, uses the fuzzy clustering algorithm to determine the continuity characteristic pattern of the Xiang feature in the diagnosis and treatment experience data, and outputs the Xiang feature pattern library;

[0058] The ontology construction module object feature pattern library is symbolically mapped, and the ontology construction method is used to form a preliminary TCM diagnosis and treatment knowledge map;

[0059] The graph enhancement module performs non-discretization enhancement processing on the nonlinear association relationship between node entities in the preliminary TCM diagnosis and treatment knowledge graph to form an enhanced knowledge graph;

[0060] The iterative adjustment module performs mapping distortion analysis on the enhanced knowledge graph and the image feature pattern library to build an iterative adjustment model;

[0061] The graph optimization module optimizes the ontology nodes and relationship edge parameters of the enhanced knowledge graph based on the iterative adjustment model, and outputs the TCM data management knowledge graph;

[0062] The intelligent decision-making module is based on the knowledge graph of traditional Chinese medicine data management, and performs real-time mapping and reasoning on newly input patient diagnosis and treatment data to generate diagnosis and treatment decision recommendations that match the thinking logic of famous Chinese medicine practitioners.

[0063] The technical effects and advantages of the TCM data record management platform and method using knowledge graphs provided by the present invention are as follows:

[0064] By constructing an image feature pattern library and conducting continuity feature analysis, the fuzzy clustering algorithm is used to effectively maintain the integrity and fuzziness of the image thinking and treatment characteristics of traditional Chinese medicine; through the non-discretization enhancement and mapping distortion analysis of the knowledge graph, the information loss in the conversion process of image features to symbolic logic is reduced, and the risk of semantic distortion caused by traditional graph expression is reduced; through the iterative adjustment model to dynamically optimize the knowledge graph structure, the high degree of matching between the knowledge graph of traditional Chinese medicine data management and the actual clinical diagnosis and treatment logic of famous Chinese medicine practitioners is guaranteed, the accuracy and interpretability of diagnosis and treatment decisions are improved, which is conducive to the effective inheritance and digital management of traditional Chinese medicine clinical experience, and improves the accuracy of digital expression of traditional Chinese medicine diagnosis and treatment knowledge and the efficiency of clinical reasoning. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a schematic diagram of a method for managing traditional Chinese medicine data records using a knowledge graph according to the present invention;

[0066] Figure 2 This is a structural diagram of a traditional Chinese medicine data record management platform using a knowledge graph according to the present invention. DETAILED DESCRIPTION

[0067] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention. Example

[0068] Figure 1 The present invention provides a method for managing traditional Chinese medicine data records using a knowledge graph, which includes the following steps:

[0069] Collect and structure the diagnosis and treatment experience data from multiple clinical cases of famous Chinese medicine practitioners to construct a dataset of Chinese medicine image thinking characteristics;

[0070] Based on the TCM Xiang thinking feature dataset, we conduct a continuous feature analysis of Xiang concepts, use fuzzy clustering algorithms to determine the continuous feature patterns of Xiang features in the diagnosis and treatment experience data, and output the Xiang feature pattern library.

[0071] The object feature pattern library is symbolically mapped and the initial TCM diagnosis and treatment knowledge map is formed using the ontology construction method;

[0072] The nonlinear association relationship between node entities in the preliminary TCM diagnosis and treatment knowledge graph is processed by non-discretization enhancement to form an enhanced knowledge graph;

[0073] Perform mapping distortion analysis on the enhanced knowledge graph and the image feature pattern library, and build an iterative adjustment model;

[0074] Based on the iterative adjustment model, the ontology node and relationship edge parameters of the enhanced knowledge graph are optimized and output as a knowledge graph for TCM data management.

[0075] Based on the knowledge graph of traditional Chinese medicine data management, the newly input patient diagnosis and treatment data are mapped and inferred in real time to generate diagnosis and treatment decision recommendations that match the thinking logic of famous Chinese medicine practitioners.

[0076] Specifically, we collected and structured the diagnosis and treatment experience data from multiple famous TCM clinical cases to construct a TCM image thinking feature dataset, including:

[0077] Collect diagnostic description information and treatment medication information from multiple clinical cases of famous Chinese medicine practitioners;

[0078] Specifically, by connecting with the information systems of multiple TCM medical institutions or hospitals, we collect the complete diagnosis and treatment records of multiple famous TCM doctors in their actual clinical work. The collected data types include two parts:

[0079] The first part is the diagnostic description information, including but not limited to the patient's initial complaint, symptom evolution, tongue characteristics, pulse characteristics, diagnostic conclusion, etc. All information is presented in natural language;

[0080] The second part is the treatment medication information, which includes but is not limited to the prescribed Chinese medicine prescriptions, specifically including the name, dosage, preparation method, compatibility relationship, and dosage cycle and method of each Chinese medicine.

[0081] Diagnostic and therapeutic medication information is sourced from digitized records in electronic medical record systems and paper medical records. To ensure the consistency and legitimacy of the sources of these information, qualified Traditional Chinese Medicine professionals conduct data review and preliminary screening to eliminate incomplete or ambiguous diagnostic and therapeutic medication information.

[0082] Classify and mark the diagnostic description information according to symptoms, tongue description, pulse description, and disease change characteristics;

[0083] Specifically, diagnostic description information is usually described in natural language, which has certain semantic ambiguity and individual differences in expression. In order to ensure the uniformity of data structure processing, all diagnostic description information needs to be semantically annotated and labeled in different categories. The specific classification tags are as follows:

[0084] Classify symptoms into clear subjective or objective clinical symptoms, such as dry mouth and throat, chest tightness and shortness of breath;

[0085] Decompose the tongue image description into dimensions such as tongue color, tongue coating color, tongue coating thickness, and tongue shape, and label them, such as pale and fat tongue, white and smooth coating, thick and greasy coating, etc.

[0086] The pulse description is marked with dimensions such as pulse position, pulse shape, and pulse strength, such as thin and rapid pulse, slippery pulse, deep and slow pulse, etc.

[0087] The characteristics of disease changes are marked as state variables such as symptom persistence, remission, and emergence of new symptoms.

[0088] The above classification tags are completed through a combination of manual verification and regular word segmentation to ensure that the labeled diagnostic description information and treatment medication information have a unified semantic standard, which facilitates feature extraction and vectorization processing.

[0089] Structurally mark the therapeutic drug information according to drug type, drug dosage, preparation method, and drug compatibility relationship;

[0090] Specifically, when extracting treatment medication information, in order to take into account the individualized medication expressions in different TCM styles, each treatment medication information is broken down into the following four basic medication elements and structured:

[0091] The drug type corresponds to the name of each Chinese medicine, such as Astragalus, Angelica, Bupleurum, etc., and the standardized names in the standard Chinese medicine name dictionary are uniformly adopted;

[0092] The basic unit of drug dosage is grams;

[0093] The processing method is the preparation method corresponding to each medicinal ingredient, which is clearly marked in each medicinal property;

[0094] The drug compatibility relationship is marked in the form of combination numbers, for example, drug three and drug four form synergy, and the compatibility type is marked at the same time, such as mutual assistance, mutual use, etc.

[0095] After the structured tagging is completed, each prescription will be converted into a set of multi-field structured record tables containing drug types, dosages, preparation methods and compatibility logic to form a standardized treatment data set.

[0096] Establish a clinical case database of famous Chinese medicine practitioners based on the classified and labeled diagnostic description information and the structured and labeled therapeutic medication information;

[0097] Specifically, the diagnostic description information that has been processed through classification and labeling and the therapeutic medication information that has been processed through structured labeling are associated and integrated to form a unified clinical case database of famous Chinese medicine practitioners.

[0098] In the famous TCM clinical case database, each famous TCM clinical case has a unique clinical case identification number. The famous TCM clinical cases corresponding to the clinical case identification number include:

[0099] The classification mark data of the diagnosis description information includes the classification mark data of symptom manifestations, the classification mark data of tongue description, the classification mark data of pulse description and the classification mark data of disease change characteristics.

[0100] The structured labeled data of therapeutic drug information includes structured labeled data of drug types, structured labeled data of drug dosages, structured labeled data of preparation methods, and structured labeled data of drug compatibility relationships.

[0101] In the famous Chinese medicine clinical case database, the diagnostic description information classification mark data and the treatment medication information structured mark data of each clinical case identification number are stored using a structured mapping association, that is, the diagnostic description information classification mark data and the treatment medication information structured mark data are one-to-one corresponded through a unique clinical case identification number, forming a complete and standardized clinical diagnosis and treatment case record.

[0102] The database of clinical cases of famous Chinese medicine practitioners is stored in a unified data table, in which each row represents a complete clinical case of a famous Chinese medicine practitioner, and each column represents the classification labeling data of diagnostic description information or treatment medication information.

[0103] Extract the associated data of diagnosis and treatment elements based on the clinical case database of famous Chinese medicine practitioners to form a dataset of TCM image thinking characteristics;

[0104] Specifically, based on the classified labeling data of diagnosis description information and the structured labeling data of treatment medication information contained in the clinical case database of famous Chinese medicine practitioners, the extraction operation of the associated data of diagnosis and treatment elements is carried out to form a dataset of Chinese medicine image thinking characteristics.

[0105] The diagnostic description information classification label data and therapeutic medication information structured label data corresponding to each clinical case identification number are extracted one by one from the famous Chinese medicine clinical case database, and the association relationship between the two types of data is structured matching analysis and clear labeling to generate diagnostic and treatment element association data with a clear structure. For example, in the famous Chinese medicine clinical case database, the "symptom manifestations" in the diagnostic description information classification label data include dry mouth and insomnia, the tongue is described as red tongue with little coating, the pulse is described as a thin and rapid pulse, and the change in the condition is characterized by worsening symptoms at night; the corresponding therapeutic medication information structured label data includes the drug types of raw rehmannia, ophiopogon, and jujube seed, the drug dosage is 15 grams per day for raw rehmannia, 9 grams per day for ophiopogon, and 12 grams per day for jujube seed, the processing method is to use the stir-frying method for jujube seed, and the drug compatibility relationship is that raw rehmannia and ophiopogon are mutually dependent, and ophiopogon and jujube seed are mutually beneficial.

[0106] Through data matching analysis, a set of clearly structured diagnostic and treatment factor-related data was formed, including the symptoms of dry mouth and insomnia corresponding to the types of raw Rehmannia and Ophiopogon japonicus, the red tongue with little coating corresponding to the daily dosage of nine grams of Ophiopogon japonicus, the thin and rapid pulse corresponding to the daily dosage of fifteen grams of raw Rehmannia, the disease change characteristics of worsening symptoms at night corresponding to the frying and processing method of Ziziphus jujuba seeds, etc.

[0107] Each set of clearly structured diagnosis and treatment element associated data is saved using a unified associated data storage structure. Each piece of diagnosis and treatment element associated data clearly indicates the matching relationship between the diagnosis and treatment elements, that is, the structured matching relationship between the diagnosis description information classification label data and the treatment medication information structured label data.

[0108] All the data related to diagnosis and treatment elements extracted from the clinical case database of famous Chinese medicine practitioners are summarized to form a dataset of Chinese medicine image thinking characteristics.

[0109] Specifically, based on the TCM Xiang thinking feature dataset, we conduct a continuous feature analysis of the Xiang concept, use a fuzzy clustering algorithm to determine the continuous feature pattern of the Xiang feature in the diagnosis and treatment experience data, and output the Xiang feature pattern library, including:

[0110] The diagnosis and treatment element association data in the TCM Xiang Si thinking feature dataset is vectorized and coded to construct a diagnosis and treatment feature vector set;

[0111] Specifically, the TCM Xiang Si thinking feature dataset consists of data related to multiple diagnosis and treatment elements. This data includes information on the association between symptoms and medication types, the matching relationship between tongue image descriptions and medication dosages, and the linkage between pulse descriptions and medication compatibility. This data is mostly in text format and needs to be vectorized and encoded.

[0112] Vectorized encoding processing includes:

[0113] Establish a standard coding dictionary for diagnosis and treatment elements and map TCM terms into unified coding labels;

[0114] A set of attribute vectors is constructed for each set of diagnosis and treatment element associated data. Each attribute vector contains multiple dimensions, where each dimension corresponds to the position of a diagnosis and treatment element coding label, and the value of each dimension indicates whether the diagnosis and treatment element coding label appears in the diagnosis and treatment element associated data; the diagnosis and treatment element coding label uses a standardized label selected from the unified coding label set to represent a specific diagnosis and treatment element.

[0115] Each set of diagnosis and treatment element-related data is converted into a corresponding vector form to form a set of diagnosis and treatment feature vectors. Each diagnosis and treatment feature vector is a high-dimensional discrete vector that expresses the combination state of the diagnosis and treatment elements in the corresponding case.

[0116] The fuzzy clustering algorithm based on membership calculation is used to perform feature pattern recognition on the diagnosis and treatment feature vector set to generate multiple initial clusters of image features;

[0117] Specifically, a fuzzy clustering method based on membership calculation is used to classify the diagnosis and treatment feature vector set. The clustering process is mainly divided into:

[0118] Set the expected number of clusters to be generated and initialize the cluster center vector set;

[0119] The membership degree is calculated based on the Euclidean distance between each vector in the diagnosis and treatment feature vector set and each center vector. The membership degree indicates the likelihood that each diagnosis and treatment feature vector belongs to different clusters.

[0120] Update the center vector according to the membership value, and repeat the above calculation and update process until the membership matrix converges;

[0121] Based on the membership matrix results, the diagnosis and treatment feature vector set is divided into several initial clusters of Xiang features. Each initial cluster of Xiang features represents a combination of diagnosis and treatment elements that have semantic similarity or clinical co-occurrence under the TCM Xiang thinking logic.

[0122] For example, when "red tongue with little coating, thin and rapid pulse, and worsening symptoms at night" are repeatedly combined with "raw Rehmannia root, Ophiopogon japonicus, and Ziziphus jujuba seed" in multiple cases, the above combination may be automatically classified into the same initial cluster of elephant features.

[0123] Calculate the aggregate membership index of each diagnosis and treatment feature vector in each initial cluster of image features;

[0124] Specifically, after completing the initial clustering of the image features, continue to calculate the aggregation membership index of the diagnosis and treatment feature vector within each initial cluster of the image features. The aggregation membership index reflects the closeness between the diagnosis and treatment feature vector and the cluster center, and its calculation method is:

[0125] For each diagnosis and treatment feature vector, extract its membership value in each initial cluster of image features;

[0126] Standardize and normalize each membership value;

[0127] The normalized membership value is the aggregate membership index of the diagnosis and treatment feature vector in the initial cluster of each feature;

[0128] For each initial cluster of image features, the average membership value of all diagnosis and treatment feature vectors in the initial cluster of image features is counted, and the diagnosis and treatment feature vectors with membership values ​​exceeding the average membership value are marked as the core expression components of the corresponding initial cluster of image features.

[0129] For example, if there are ten groups of eigenvectors in an initial cluster of image features, and the average membership value is 0.6, then the eigenvectors with membership values ​​exceeding 0.6 will be regarded as the main representatives of the initial cluster of image features and included in the continuity analysis.

[0130] Determine the continuity boundary between image features based on the aggregation membership index;

[0131] Specifically, after obtaining the aggregation membership index of each initial cluster of image features, the intersection and fusion between different initial clusters of image features are analyzed to determine the continuity boundaries between the image features.

[0132] The method for determining the continuity boundary is:

[0133] Compare the number of common diagnosis and treatment feature vectors and the difference in aggregated membership between two initial clusters of image features; the difference in aggregated membership is the absolute value of the difference in membership values ​​of the same diagnosis and treatment feature vector in different initial clusters of image features;

[0134] If there are multiple common diagnosis and treatment feature vectors between two initial clusters of image features, and the membership values ​​of the common diagnosis and treatment feature vectors in both initial clusters of image features are greater than or equal to the preset membership threshold, then it is judged that there is significant image feature continuity between the two initial clusters of image features; otherwise, it is judged to be discontinuous.

[0135] Perform fuzzy constraint optimization processing on the continuity boundary, output the image feature pattern set, and form the image feature pattern library;

[0136] Specifically, for the initial clusters of image features that are determined to be continuous boundaries, a fuzzy constraint function is constructed, and the fuzzy constraint function weights the aggregated membership index;

[0137] The membership values ​​of the diagnosis and treatment feature vectors on both sides of the continuous boundary are readjusted through the fuzzy constraint function, and the center positions of the two initial clusters of image features are adjusted at the same time to make the center positions closer to the boundary;

[0138] The membership adjustment and center update are performed cyclically until the fuzzy constraint function converges;

[0139] After the fuzzy constraint function converges, the initial clusters of image features with continuous semantics are merged into one image feature pattern, and the initial clusters of image features with discrete semantics are kept independent, and finally a set of image feature patterns is output.

[0140] Specifically, the object feature pattern library is symbolically mapped, and the ontology construction method is used to form a preliminary TCM diagnosis and treatment knowledge map, including:

[0141] Assign a unique symbol identifier to each image feature pattern in the image feature pattern library to generate an image feature symbol table;

[0142] Specifically, each object feature pattern in the object feature pattern library performs numbering and symbol abstraction operations.

[0143] Each characteristic pattern represents a combination structure of diagnosis and treatment elements with semantic consistency and structural stability in the diagnosis and treatment experience of famous Chinese medicine practitioners. The combination structure of diagnosis and treatment elements includes several diagnosis and treatment features classified by fuzzy clustering;

[0144] In order to achieve structured identification and computational processing, each image feature pattern is assigned a non-repeatable coding symbol identifier, such as a combination of pinyin abbreviations and numbers;

[0145] Coding rules must ensure consistency and readability with terminology in the field of traditional Chinese medicine diagnosis;

[0146] The mapping relationship between all coding identifiers and corresponding image feature patterns is uniformly stored in the image feature symbol table, which is the reference basis for constructing graph nodes and establishing relationships.

[0147] For example, if an image feature pattern includes "fine and rapid pulse, red tongue, symptoms aggravated at night, and the combination of Ophiopogon japonicus and Ziziphus jujuba seeds", it can be assigned a unique code "ZLP001", and the semantic labels and feature lists of all its constituent items can be recorded in the image feature symbol table.

[0148] Determine the arrangement rules of the image concept node set and the image concept hierarchy relationship according to the image feature symbol table;

[0149] Specifically, the components of all encoded image feature patterns in the image feature symbol table are extracted and divided into different image concept categories according to semantic labels, such as tongue image category, pulse image category, medicine category, disease change category, etc.;

[0150] Each type of image feature has subcategories, for example, tongue image has red tongue and white tongue coating; pulse image has slow pulse and thin pulse, etc.

[0151] According to the order from concept categories to concept subcategories and then to image feature patterns, the hierarchical arrangement rules of image concept nodes are constructed;

[0152] All image concept nodes are independent of each other at the semantic level and have a clear nested structure;

[0153] The hierarchical arrangement results serve as the basis for the definition and inheritance of concept nodes in the ontology structure.

[0154] The final set of image concept nodes and hierarchical arrangement rules will directly determine the ontology structure hierarchical framework of the knowledge graph.

[0155] Determine the node entity set and relationship edge set based on the diagnosis and treatment element association data;

[0156] Specifically, based on the constructed diagnosis and treatment element association data and image feature patterns, diagnosis and treatment entities and the semantic relationships between them are extracted and standardized.

[0157] The construction of the node entity set is based on the diagnostic and treatment elements contained in the TCM Xiang Si thinking feature dataset, such as red tongue with little coating, thin and rapid pulse, and worsening at night.

[0158] Each node entity has a clearly defined semantic meaning of TCM terminology;

[0159] After determining the node entity set, the relationship edge set is constructed by combining the structural affiliation between different diagnosis and treatment elements in the image feature pattern;

[0160] The relationship edge set defines the semantic connection types between different node entities, including symptoms-direction-drug, tongue image-hint-constitution type, pulse condition-guidance-dose, etc.

[0161] Each relationship edge contains a clear relationship type label, starting node, and ending node, and is uniformly encoded to construct the edge attribute structure of the graph.

[0162] For example, if an image feature pattern includes red tongue with little coating - combined with Rehmannia root and Radix Ophiopogonis, the relationship edge should be recorded as: red tongue with little coating (node ​​entity) - association - Ophiopogon japonicus (node ​​entity), and the relationship type is tongue image - combined with guidance.

[0163] Adopting the knowledge domain ontology construction rules, the ontology structure framework is established according to the hierarchical relationship arrangement rules of image concepts, node entity sets, and relationship edge sets;

[0164] Specifically, we import the concept node set and hierarchical structure and build a semantic classification tree of the ontology structure;

[0165] Under each concept node of the ontology structure, the corresponding diagnosis and treatment node entity is attached;

[0166] Establish a relationship edge index table between all node entities, and define edge attributes, directions, and label types;

[0167] The ontology structure should include a semantic consistency verification mechanism to ensure that the concept nodes to which each node entity belongs are correct and non-repetitive;

[0168] All construction structures are organized using graph database standard structures to form an ontology structure framework.

[0169] Write the concept node set, node entity set and relationship edge set into the ontology structure framework to generate a preliminary TCM diagnosis and treatment knowledge graph;

[0170] Specifically, the image concept node set, diagnosis and treatment node entity set, and entity relationship edge set are written into the ontology structure framework in sequence to form a preliminary TCM diagnosis and treatment knowledge graph.

[0171] Each Xiang concept node is loaded into the semantic layer of the preliminary TCM diagnosis and treatment knowledge graph as a basis for structural indexing and semantic classification;

[0172] Each diagnosis and treatment node entity is written into the entity layer of the preliminary TCM diagnosis and treatment knowledge graph, serving as the main information source for query and reasoning of the preliminary TCM diagnosis and treatment knowledge graph;

[0173] Each relationship edge is written into the connection layer of the preliminary TCM diagnosis and treatment knowledge graph, providing semantic paths between entities and traversability of the preliminary TCM diagnosis and treatment knowledge graph;

[0174] During the writing process, semantic consistency detection, entity uniqueness check, and edge type constraint verification are performed to ensure the legality of the preliminary TCM diagnosis and treatment knowledge graph structure;

[0175] After the preliminary TCM diagnosis and treatment knowledge map is constructed, the diagnosis and treatment logic based on the "image-feature-medicine-effect" path can be reconstructed, and the thinking structure of famous Chinese medicine practitioners can be restored.

[0176] Specifically, the nonlinear association relationship between node entities in the preliminary TCM diagnosis and treatment knowledge graph is subjected to non-discretization enhancement processing to form an enhanced knowledge graph, including:

[0177] Construct a node entity association matrix based on the node entities and relationship edges in the preliminary TCM diagnosis and treatment knowledge graph;

[0178] Specifically, a node entity refers to a structured diagnosis and treatment unit with clear meaning at the diagnosis and treatment semantic level, i.e., diagnosis and treatment elements, such as a thin and rapid pulse, white and greasy tongue coating, and worsening symptoms at night;

[0179] A relationship edge refers to a semantic path connecting two node entities, which is used to express pairing, deduction, or co-occurrence relationships in diagnosis and treatment logic, such as prompts, guidance, and pointing.

[0180] The build process is as follows:

[0181] After numbering all node entities, they are used as row labels and column labels of the matrix in turn;

[0182] If there is a direct relationship edge between two node entities, the initial association value is assigned to the corresponding matrix unit position, otherwise it is assigned to zero;

[0183] The initially constructed matrix is ​​a binary matrix, which represents the existence connection relationship between nodes rather than the strength relationship.

[0184] Perform continuous weight assignment processing on the association relationships in the node entity association matrix to form a node entity continuous association matrix;

[0185] Specifically, the co-occurrence frequency between each node entity in the preliminary knowledge graph is analyzed, that is, the number of times two node entities appear together in the same image feature pattern;

[0186] Analyze the semantic label attributes of relationship edges. For example, the strength of “compatibility” edges is usually higher than that of “reference” edges.

[0187] Based on the co-occurrence frequency and semantic label level, a weight calculation function is established to assign the initial binary association relationship to a continuous real value between zero and one;

[0188] Generate a node-entity continuous association matrix, where each cell value in the entity continuous association matrix represents the semantic coupling strength between two node entities.

[0189] For example, if red tongue with little coating and Ophiopogon japonicus appear repeatedly together in multiple characteristic patterns and there is a compatibility edge relationship between them, then their values ​​in the continuous association matrix are higher.

[0190] Establish a node-entity multidimensional relationship network based on the node-entity continuous association matrix;

[0191] Specifically, the node entities are considered as nodes in the network graph structure;

[0192] The non-zero values ​​in the continuous incidence matrix are regarded as the weights of the edges;

[0193] A weighted undirected graph is established, where the edge weight represents the comprehensive diagnosis and treatment semantic strength between node entities;

[0194] During the construction process, multiple edge weight attributes between each node are retained, such as symptom co-occurrence weight, drug compatibility weight, constitution orientation weight, etc., to form a multi-dimensional edge attribute structure;

[0195] The constructed node-entity multidimensional relationship network can be used to represent the interaction logic of diagnosis and treatment characteristics in multiple semantic dimensions.

[0196] The strength of the association relationship in the multidimensional relationship network of node entities is weighted and optimized to form a non-discretized enhanced network structure of node entities;

[0197] Specifically, basic weight factors are set for various semantic edges. For example, the weight factor of drug compatibility is higher than the tongue image prompt factor.

[0198] All edges are recalculated with weights, with the optimization goal of improving the local aggregation and path traversability of the graph structure;

[0199] Execute network structure sparsity constraint operations to remove low-weight redundant edges and avoid introducing weak semantic interference paths;

[0200] In the generated enhanced network structure, the connection between each node entity has higher diagnosis and treatment logic density, and the structure has higher clinical semantic consistency.

[0201] The enhanced network structure has formally deviated from the discrete node connection mode of the original graph and is manifested as a continuous value semantic expression network. It is the core carrier for realizing image thinking graph reasoning.

[0202] Based on the node entity non-discretization enhanced network structure, the continuity weights of node entities and relationship edges in the preliminary TCM diagnosis and treatment knowledge graph are updated to generate an enhanced knowledge graph;

[0203] Specifically, the updating process includes: extracting the final weighted value between each pair of node entities in the node entity non-discretization enhanced network structure;

[0204] Write the final weighted value into the corresponding relationship edge in the preliminary knowledge graph, replacing the original fixed edge label;

[0205] For node pairs that originally did not have a connection relationship in the preliminary knowledge graph, but whose edge weights in the enhanced network structure reach the set threshold, new relationship edges are established and the enhanced edge attributes are marked;

[0206] Perform structural integrity verification on the entire preliminary knowledge graph to ensure there are no isolated nodes and no logical loop conflicts;

[0207] The final output enhanced knowledge graph has stronger semantic continuous expression and reasoning support capabilities.

[0208] Specifically, we conduct a mapping distortion analysis on the enhanced knowledge graph and the image feature pattern library, and build an iterative adjustment model, including:

[0209] Generate enhanced knowledge graph structure data based on the enhanced knowledge graph, and generate image feature pattern benchmark data based on the image feature pattern library;

[0210] Specifically, for the enhanced knowledge graph: extract all node entities and their connected relationship edges in the enhanced knowledge graph; extract the continuous weight information of each node entity and relationship edge to form a node entity and relationship edge weight table.

[0211] The node entity and relationship edge weight table is called enhanced knowledge graph structure data, which represents the actual status of the current enhanced knowledge graph nodes and edges.

[0212] For object feature pattern libraries:

[0213] Extracting data of all the image feature patterns in the image feature pattern library, including the diagnostic and treatment elements of the image feature patterns, the boundary information between the image feature patterns, and the membership index of the image feature patterns;

[0214] A standard data structure of the image feature pattern is constructed, called the image feature pattern benchmark data, which is used to represent the standard reference state of real diagnosis and treatment experience data.

[0215] For example, the node entities extracted from the enhanced knowledge graph may include red tongue with little coating, thin and rapid pulse, and worsening at night, and record continuous weights respectively; the data extracted from the image feature pattern library contains aggregated affiliation indicators and boundary data corresponding to the same diagnostic and treatment elements.

[0216] Calculate the mapping error values ​​of node entities and relationship edges based on the enhanced knowledge graph structure data and the image feature pattern benchmark data;

[0217] Specifically, the mapping error of the node entity is calculated by numerically comparing the continuous weight of each node entity in the enhanced knowledge graph with the corresponding diagnosis and treatment element aggregation membership index in the image feature pattern benchmark data;

[0218] After subtracting the corresponding feature pattern aggregation membership index from the enhanced knowledge graph node entity weight, the absolute value is calculated to obtain the error value of the node entity.

[0219] The mapping error calculation method of the relationship edge is:

[0220] Compare the continuous weight of each relationship edge in the enhanced knowledge graph with the standard weight of the corresponding diagnosis and treatment feature connection structure in the feature pattern library;

[0221] The error value of the relationship edge is obtained by subtracting the absolute value of the difference between the relationship edge weight in the enhanced knowledge graph and the connection structure weight of the corresponding feature pattern.

[0222] Determine the overall mapping distortion of the enhanced knowledge graph based on the mapping error value between node entities and relationship edges;

[0223] Specifically, the overall mapping distortion of the enhanced knowledge graph is determined by comprehensively evaluating the mapping error values ​​of all node entities and relationship edges:

[0224] Accumulate the error values ​​of all node entities and take the average value as the node entity average error;

[0225] Accumulate the error values ​​of all relationship edges and take the average value as the average error of the relationship edge;

[0226] The average node entity error and the average relationship edge error are summed to obtain the overall mapping distortion index.

[0227] The mapping error feedback function is established using the overall mapping distortion of the enhanced knowledge graph;

[0228] Specifically, the mapping error feedback function takes the overall mapping distortion as the core input variable;

[0229] The output of the error feedback function is used to guide the adjustment direction and adjustment range of node entities and relationship edges during the optimization process;

[0230] The error feedback function is designed such that the larger the overall mapping distortion value is, the larger the output adjustment amplitude is.

[0231] Based on the mapping error feedback function, iterative adjustment rules for node entities and relationship edges are constructed to form an iterative adjustment model;

[0232] Specifically, the node entity adjustment rule stipulates that when the node entity mapping error value is large, the node entity weight is increased or decreased to gradually approach the standard membership index in the image feature pattern benchmark data;

[0233] The relationship edge adjustment rule stipulates that when the relationship edge error value is large, the relationship edge weight is increased or decreased to gradually approach the standard weight in the image feature pattern benchmark data;

[0234] The adjustment amplitude and direction are determined by the adjustment amplitude output by the error feedback function;

[0235] The above rule combination constitutes an iterative adjustment model, which ensures that the gap between the enhanced knowledge graph and the image feature pattern library gradually narrows during the iteration process until the convergence criteria are met.

[0236] Specifically, the ontology nodes and relationship edge parameters of the knowledge graph are optimized and enhanced according to the iterative adjustment model, and the TCM data management knowledge graph is output, including:

[0237] According to the iterative adjustment rules of the node entities in the iterative adjustment model, the definition attributes and hierarchical structure of the ontology nodes in the enhanced knowledge graph are modified;

[0238] Specifically, the iterative adjustment rule set for the node entity in the iterative adjustment model is read. Each ontology node in the enhanced knowledge graph is traversed and the existing defined attributes of the ontology node are compared with the target attributes in the iterative adjustment rule. If the concept name, semantic label, attribute range, or hierarchical position of the ontology node is inconsistent with the target attribute, the following correction operations are performed:

[0239] Revise concept names to standard expressions that conform to the unified coding dictionary.

[0240] Supplement or replace semantic tags with standard tags that are consistent with the image feature pattern library.

[0241] Expand or shrink the attribute range to keep it consistent with the classification boundaries of diagnosis and treatment factors.

[0242] Adjust the position of the ontology hierarchy tree to ensure that the relationship between parent nodes and child nodes conforms to the concept hierarchy arrangement rules.

[0243] For example, if the ontology node "tongue red category" in the enhanced knowledge graph is incorrectly placed in the "tongue coating category" branch, the iterative adjustment rule indicates that the ontology node should be moved to the "tongue quality category" branch, and the attribute "tongue quality color" should be updated synchronously. After correction, the ontology hierarchical structure can restore the correct semantics.

[0244] According to the iterative adjustment rules of the relationship edges in the iterative adjustment model, the weight parameters and association directions of the relationship edges in the enhanced knowledge graph are modified;

[0245] Specifically, read the set of weight and direction adjustment rules for the relationship edge in the iterative adjustment model. Traverse each relationship edge in the enhanced knowledge graph and extract the current continuous weight value and direction label. Based on the adjustment range output by the mapping error feedback function, calculate the new weight parameters, specifically:

[0246] If the mapping error feedback function indicates that the weight is too low, the weight is increased by the “original weight plus increment” method.

[0247] If the mapping error feedback function indicates that the weight is too high, the weight is reduced by using the “original weight reduction” method.

[0248] The size of the increment or decrement is proportional to the mapping error.

[0249] If the iterative adjustment rule indicates that the association direction does not conform to the diagnosis and treatment logic, the starting node and the ending node are swapped and the direction is re-labeled.

[0250] For example, if the weight of the relationship edge "red tongue with little coating guides the compatibility of Ophiopogon japonicus" is too low and leads to the missing of the reasoning path, the rule is adjusted to give an increment. After the weight is increased, the reasoning link can be connected.

[0251] Perform consistency check on the revised ontology nodes and relationship edges to generate a structured updated dataset that conforms to semantic logic;

[0252] Specifically, after completing the node entity and relationship edge correction, the correction results are fully Figure 1 Consistency check. The check contents include:

[0253] There are no circular references in the concept hierarchy.

[0254] Ontology node names are unique and non-overlapping.

[0255] The direction of the relationship edges matches the semantic labels and there is no conflict.

[0256] Continuous weights lie within a preset range of values.

[0257] If the verification passes, all correction records will be written into the structure update data set; if the verification fails, the corresponding nodes or relationships will be traced back and readjusted until all verifications pass.

[0258] Apply the structure-updated dataset to the enhanced knowledge graph to form a knowledge graph for TCM data management;

[0259] Specifically, start the enhanced knowledge graph update program and load the structure update dataset. Perform add, delete, and modify operations on the enhanced knowledge graph in the order of records:

[0260] Replace the ontology node attributes and hierarchical information.

[0261] Update the relationship edge weight and direction attributes.

[0262] Add necessary nodes or relationships to fill semantic gaps.

[0263] Remove redundant nodes or relationships to streamline the structure.

[0264] Specifically, based on the TCM data management knowledge graph, the system performs real-time mapping and reasoning on newly input patient diagnosis and treatment data, generating diagnosis and treatment decision recommendations that match the thinking logic of famous TCM practitioners, including:

[0265] Extract diagnosis and treatment elements from newly input patient diagnosis and treatment data to obtain diagnosis and treatment feature items;

[0266] Specifically, the newly input patient diagnosis and treatment data comes from electronic medical record texts and tongue and pulse images, including the patient's chief complaint, current medical history and physician observation records.

[0267] Natural language segmentation and term matching were performed according to the unified coding dictionary to extract symptom manifestation items, tongue description items, pulse description items and constitution type items.

[0268] Each entry is mapped to a diagnosis and treatment element coding label to form a set of four types of diagnosis and treatment feature items.

[0269] For example, when the text contains "cough worsens at night, tongue is red with little coating, pulse is thin and rapid, and constitution tends to be yin deficiency", the corresponding extraction result is "symptoms worsen at night, tongue is red with little coating, pulse is thin and rapid, and constitution tends to be yin deficiency".

[0270] Semantically match the diagnosis and treatment feature items with the ontology node set in the TCM data management knowledge graph to determine the semantic relevance with each ontology node;

[0271] Specifically, the set of diagnosis and treatment feature items is vectorized and encoded, and the vector dimension corresponds to all standard labels in the ontology node set.

[0272] Traverse each dimension of the vector, multiply the value of the new input vector in that dimension with the value of the knowledge graph node vector in that dimension, and then sum them to obtain the semantic association score.

[0273] After calculating the semantic relevance scores for all ontology nodes, normalization is performed to convert them into semantic relevance between zero and 1. The closer the semantic relevance is to 1, the closer the diagnosis and treatment feature item is to the semantics of the ontology node.

[0274] Based on the semantic correlation, a multi-hop correlation path set is constructed;

[0275] Specifically, taking the ontology node with a semantic relevance greater than a preset threshold as the starting point, the relationship edge of no more than three hops is expanded outward in the knowledge graph of traditional Chinese medicine data management.

[0276] Perform breadth search along the priority order of continuous weights from high to low, record all possible node sequences, and form a multi-hop associated path set.

[0277] Each associated path saves the cumulative continuous weight, path length and node semantic label for sorting.

[0278] For example, if the cumulative weight of the path "red tongue with little coating - hint - heat syndrome - direction - heat-clearing drugs" is higher than other paths, then the path is ranked higher.

[0279] Generate a diagnosis and treatment decision reasoning path sequence based on the multi-hop association path set, and extract the diagnosis recommendation nodes and medication recommendation nodes in the diagnosis and treatment decision reasoning path sequence;

[0280] Specifically, the multi-hop associated path set is sorted from high to low according to the cumulative continuous weight, and the top several paths are taken to form the reasoning path sequence.

[0281] In each reasoning path, the first appearing disease conclusion type node is identified as a diagnosis suggestion node; the first appearing drug or prescription type node is identified as a medication suggestion node.

[0282] If there are multiple drug nodes in a path, they are sorted according to the edge weights and the top two are retained as the medication recommendation node group.

[0283] For example, the path "red tongue with little coating - hint - heat syndrome - direction - heat-clearing drugs - compatibility - scutellaria baicalensis" outputs the diagnosis suggestion node "heat syndrome" and the medication suggestion node "heat-clearing drugs, scutellaria baicalensis".

[0284] The diagnosis suggestion node and the medication suggestion node corresponding to the diagnosis suggestion node and the medication suggestion node are combined to form a diagnosis and treatment decision recommendation;

[0285] Specifically, all diagnosis suggestion nodes generated by the reasoning path sequence are deduplicated and sorted from high to low according to the frequency of occurrence, and the most frequent diagnosis suggestion is selected as the main diagnosis.

[0286] Conflict detection is performed on the medication suggestion node group. If the medicinal properties are opposite, they are removed or replaced according to the medicinal property coordination rules in the knowledge graph.

[0287] Output diagnosis and treatment decision recommendation text, the format of which includes main diagnosis recommendation, auxiliary diagnosis alternatives, recommended prescriptions and dosage ranges, and adjustment prompts. Example

[0288] The difference between Example 2 of the present invention and Example 1 is that this example introduces a traditional Chinese medicine data record management platform that uses a knowledge graph.

[0289] Figure 2 The present invention provides a schematic structural diagram of a TCM data record management platform using a knowledge graph, which includes a data construction module, a pattern clustering module, an ontology construction module, a graph enhancement module, an iterative adjustment module, a graph optimization module, and an intelligent decision-making module.

[0290] The data construction module collects and structures the diagnosis and treatment experience data from multiple clinical cases of famous Chinese medicine practitioners to construct a dataset of Chinese medicine image thinking characteristics;

[0291] The pattern clustering module analyzes the continuity characteristics of the Xiang concept based on the TCM Xiang thinking feature dataset, uses the fuzzy clustering algorithm to determine the continuity characteristic pattern of the Xiang feature in the diagnosis and treatment experience data, and outputs the Xiang feature pattern library;

[0292] The ontology construction module object feature pattern library is symbolically mapped, and the ontology construction method is used to form a preliminary TCM diagnosis and treatment knowledge map;

[0293] The graph enhancement module performs non-discretization enhancement processing on the nonlinear association relationship between node entities in the preliminary TCM diagnosis and treatment knowledge graph to form an enhanced knowledge graph;

[0294] The iterative adjustment module performs mapping distortion analysis on the enhanced knowledge graph and the image feature pattern library to build an iterative adjustment model;

[0295] The graph optimization module optimizes the ontology nodes and relationship edge parameters of the enhanced knowledge graph based on the iterative adjustment model, and outputs the TCM data management knowledge graph;

[0296] The intelligent decision-making module is based on the knowledge graph of traditional Chinese medicine data management, and performs real-time mapping and reasoning on newly input patient diagnosis and treatment data to generate diagnosis and treatment decision recommendations that match the thinking logic of famous Chinese medicine practitioners.

[0297] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0298] The above embodiments can be implemented in whole or in part via software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in the embodiments of this application are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0299] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0300] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0301] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0302] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0303] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0304] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0305] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0306] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for managing traditional Chinese medicine data records using a knowledge graph, characterized in that: The steps include: We collected and structured the diagnosis and treatment experience data from multiple famous TCM clinical cases to construct a TCM image thinking feature dataset, specifically: Collect diagnostic description information and treatment medication information from multiple clinical cases of famous Chinese medicine practitioners; Classify and mark the diagnostic description information according to symptoms, tongue description, pulse description, and disease change characteristics; Structurally mark the therapeutic drug information according to drug type, drug dosage, preparation method, and drug compatibility relationship; Establish a clinical case database of famous Chinese medicine practitioners based on the classified and labeled diagnostic description information and the structured and labeled therapeutic medication information; Extract the associated data of diagnosis and treatment elements based on the clinical case database of famous Chinese medicine practitioners to form a dataset of TCM image thinking characteristics; Based on the TCM Xiang thinking feature dataset, we conduct a continuous feature analysis of Xiang concepts, use fuzzy clustering algorithms to determine the continuous feature patterns of Xiang features in the diagnosis and treatment experience data, and output the Xiang feature pattern library. The object feature pattern library is symbolically mapped and the initial TCM diagnosis and treatment knowledge map is formed using the ontology construction method; The nonlinear association relationship between node entities in the preliminary TCM diagnosis and treatment knowledge graph is processed by non-discretization enhancement to form an enhanced knowledge graph, specifically: Construct a node entity association matrix based on the node entities and relationship edges in the preliminary TCM diagnosis and treatment knowledge graph; Perform continuous weight assignment processing on the association relationships in the node entity association matrix to form a node entity continuous association matrix; Establish a node-entity multidimensional relationship network based on the node-entity continuous association matrix; The strength of the association relationship in the multidimensional relationship network of node entities is weighted and optimized to form a non-discretized enhanced network structure of node entities; Based on the non-discretized enhanced network structure of node entities, the continuity weights of node entities and relationship edges in the preliminary TCM diagnosis and treatment knowledge graph are updated to generate an enhanced knowledge graph. Perform mapping distortion analysis on the enhanced knowledge graph and the image feature pattern library, and build an iterative adjustment model, specifically: Generate enhanced knowledge graph structure data based on the enhanced knowledge graph, and generate image feature pattern benchmark data based on the image feature pattern library; Calculate the mapping error values ​​of node entities and relationship edges based on the enhanced knowledge graph structure data and the image feature pattern benchmark data; Determine the overall mapping distortion of the enhanced knowledge graph based on the mapping error value between node entities and relationship edges; The mapping error feedback function is established using the overall mapping distortion of the enhanced knowledge graph; Based on the mapping error feedback function, iterative adjustment rules for node entities and relationship edges are constructed to form an iterative adjustment model; Based on the iterative adjustment model, the ontology node and relationship edge parameters of the enhanced knowledge graph are optimized and output as a knowledge graph for TCM data management. Based on the knowledge graph of traditional Chinese medicine data management, the newly input patient diagnosis and treatment data are mapped and inferred in real time to generate diagnosis and treatment decision recommendations that match the thinking logic of famous Chinese medicine practitioners.

2. A method for managing traditional Chinese medicine data records using a knowledge graph according to claim 1, characterized in that: Based on the TCM Xiang thinking feature dataset, we conduct a continuous feature analysis of Xiang concepts, use fuzzy clustering algorithm to determine the continuous feature pattern of Xiang features in the diagnosis and treatment experience data, and output the Xiang feature pattern library, specifically: The diagnosis and treatment element association data in the TCM Xiang Si thinking feature dataset is vectorized and coded to construct a diagnosis and treatment feature vector set; The fuzzy clustering algorithm based on membership calculation is used to perform feature pattern recognition on the diagnosis and treatment feature vector set to generate multiple initial clusters of image features; Calculate the aggregate membership index of each diagnosis and treatment feature vector in each initial cluster of image features; Determine the continuity boundary between image features based on the aggregation membership index; The continuity boundary is subjected to fuzzy constraint optimization processing, and an image feature pattern set is output to form an image feature pattern library.

3. A method for managing traditional Chinese medicine data records using a knowledge graph according to claim 2, characterized in that: The object feature pattern library is symbolically mapped and the ontology construction method is used to form a preliminary TCM diagnosis and treatment knowledge map, specifically: Assign a unique symbol identifier to each image feature pattern in the image feature pattern library to generate an image feature symbol table; Determine the arrangement rules of the image concept node set and the image concept hierarchy relationship according to the image feature symbol table; Determine the node entity set and relationship edge set based on the diagnosis and treatment element association data; Adopting the knowledge domain ontology construction rules, the ontology structure framework is established according to the hierarchical relationship arrangement rules of image concepts, node entity sets, and relationship edge sets; The concept node set, node entity set and relationship edge set are written into the ontology structure framework to generate a preliminary TCM diagnosis and treatment knowledge graph.

4. A method for managing traditional Chinese medicine data records using a knowledge graph according to claim 3, characterized in that: The ontology node and relationship edge parameters of the knowledge graph are optimized and enhanced according to the iterative adjustment model, and the TCM data management knowledge graph is output. Specifically: According to the iterative adjustment rules of the node entities in the iterative adjustment model, the definition attributes and hierarchical structure of the ontology nodes in the enhanced knowledge graph are modified; According to the iterative adjustment rules of the relationship edges in the iterative adjustment model, the weight parameters and association directions of the relationship edges in the enhanced knowledge graph are modified; Perform consistency check on the revised ontology nodes and relationship edges to generate a structured updated dataset that conforms to semantic logic; The structure-updated dataset is applied to the enhanced knowledge graph to form a knowledge graph for traditional Chinese medicine data management.

5. A method for managing traditional Chinese medicine data records using a knowledge graph according to claim 4, characterized in that: Based on the TCM data management knowledge graph, the newly input patient diagnosis and treatment data is mapped and reasoned in real time to generate diagnosis and treatment decision suggestions that match the thinking logic of famous TCM practitioners. Specifically: Extract diagnosis and treatment elements from newly input patient diagnosis and treatment data to obtain diagnosis and treatment feature items; Semantically match the diagnosis and treatment feature items with the ontology node set in the TCM data management knowledge graph to determine the semantic relevance with each ontology node; Based on the semantic correlation, a multi-hop correlation path set is constructed; Generate a diagnosis and treatment decision reasoning path sequence based on the multi-hop association path set, and extract the diagnosis recommendation nodes and medication recommendation nodes in the diagnosis and treatment decision reasoning path sequence; The diagnosis and treatment recommendations corresponding to the diagnosis and treatment recommendation nodes are combined to form treatment decision recommendations.

6. A TCM data record management platform using a knowledge graph, used to implement a TCM data record management method using a knowledge graph as described in any one of claims 1 to 5, characterized in that: It includes data construction module, pattern clustering module, ontology construction module, graph enhancement module, iterative adjustment module, graph optimization module and intelligent decision-making module; The data construction module collects and structures the diagnosis and treatment experience data from multiple clinical cases of famous Chinese medicine practitioners to construct a dataset of Chinese medicine image thinking characteristics; The pattern clustering module analyzes the continuity characteristics of the Xiang concept based on the TCM Xiang thinking feature dataset, uses the fuzzy clustering algorithm to determine the continuity characteristic pattern of the Xiang feature in the diagnosis and treatment experience data, and outputs the Xiang feature pattern library; The ontology construction module object feature pattern library is symbolically mapped, and the ontology construction method is used to form a preliminary TCM diagnosis and treatment knowledge map; The graph enhancement module performs non-discretization enhancement processing on the nonlinear association relationship between node entities in the preliminary TCM diagnosis and treatment knowledge graph to form an enhanced knowledge graph; The iterative adjustment module performs mapping distortion analysis on the enhanced knowledge graph and the image feature pattern library to build an iterative adjustment model; The graph optimization module optimizes the ontology nodes and relationship edge parameters of the enhanced knowledge graph based on the iterative adjustment model, and outputs the TCM data management knowledge graph; The intelligent decision-making module is based on the knowledge graph of traditional Chinese medicine data management, and performs real-time mapping and reasoning on newly input patient diagnosis and treatment data to generate diagnosis and treatment decision recommendations that match the thinking logic of famous Chinese medicine practitioners.

Images