Acupoint-syndrome association information processing method and device, terminal equipment and storage medium

Abstract

The application provides an acupoint-syndrome correlation information processing method and device, a terminal device and a storage medium, and the method comprises the following steps: mapping a text to a corresponding node in a graph to obtain structured data; weighting entities in the text according to a pathological degree to obtain a weighting result; obtaining an entity weighting result according to a preset strategy and performing mapping to obtain a syndrome coding result; coding resistance data of acupoints to obtain a coding result and splicing the coding result with the syndrome coding result to obtain a feature vector; initializing a clustering center, iteratively calculating the distance between the feature vector and the clustering center, and assigning the feature vector to a cluster cluster to which the nearest clustering center belongs until a preset requirement is reached, and outputting a syndrome type judgment result of the feature vector. The application standardizes and fuses the expression of medical record texts and acupoint resistance data, constructs a three-dimensional mapping relationship of “symptoms-syndrome type-acupoints”, and can improve the effectiveness and accuracy of cervical spondylosis syndrome type identification.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical technology, and in particular to a method, device, terminal equipment and storage medium for processing acupoint-syndrome correlation information. Background Technology

[0002] Traditional Chinese medicine (TCM) theory posits that "internal conditions inevitably manifest externally," emphasizing that the body's internal pathological state is comprehensively reflected through changes in external symptoms, signs, and meridian acupoints. Syndromes, as a comprehensive summary of the etiology, location, nature, and progression of a disease at a certain stage, are crucial for effective diagnosis and treatment. Modern research further confirms that acupoint sensitization often precedes the appearance of clinical symptoms, more acutely reflecting dynamic changes in organ function. Compared to traditional diagnostic methods relying on tongue and pulse examinations and symptoms, judging organ status based on acupoint reaction characteristics has higher objectivity and early warning value. In the TCM diagnostic system, the core of syndrome differentiation and treatment for cervical spondylosis (referred to as "xiangbi" in TCM) lies in deriving the syndrome type from external manifestations such as symptoms and signs, and then formulating a treatment plan based on meridian and acupoint theory.

[0003] However, traditional data analysis and diagnostic treatment processes for cervical spondylosis face several limitations: First, they are highly subjective, relying excessively on the personal experience of physicians, lacking unified quantitative standards for interpreting symptoms and signs and determining syndrome types, leading to significant differences in treatment outcomes among different physicians. Second, the data dimensions are limited; existing technologies either focus solely on analyzing medical record text data or only assist in acupoint location, failing to establish a complete correlation between "symptoms-syndrome types-acupoints," resulting in a lack of multi-dimensional data support for the diagnostic process and weak early disease identification capabilities. With the rapid development of artificial intelligence technologies such as large-scale clustering algorithms, it has become possible to overcome the bottlenecks of traditional diagnosis and treatment. Based on the holistic relationship theory of "viscera-meridians-acupoints" in Traditional Chinese Medicine, how to combine machine learning with TCM syndrome differentiation to achieve proactive and effective identification and treatment of cervical spondylosis is a pressing technical problem that needs to be solved. Summary of the Invention

[0004] This invention aims to provide a method, device, terminal equipment, and storage medium for processing acupoint-syndrome correlation information to solve the above-mentioned technical problems, and to construct a three-dimensional mapping relationship between "symptom-syndrome-acupoint" to improve the effectiveness and accuracy of cervical spondylosis syndrome identification.

[0005] To address the aforementioned technical problems, this invention provides a method for processing acupoint-syndrome correlation information, including: The preprocessed standardized medical record texts are mapped to the corresponding symptom-sign nodes in the pre-constructed TCM knowledge graph to obtain structured data. The TCM knowledge graph is trained from several TCM classics and textbooks and is used to describe different types of nodes and the semantic relationships between them. The structured data includes syndrome data that are directly related to the nodes. Based on the pathological degree, the symptom and sign entities in the standardized medical record text are initially weighted to obtain the basic weighting results. The basic weighting results are then updated according to the preset weight correction strategy to obtain the entity weighting results. The entity weighting result is mapped to the pre-built syndrome coding system to obtain the syndrome coding result. The acupoint resistance data of the target acupoint is discretely coded to obtain the acupoint resistance coding result. The syndrome coding result and the acupoint resistance coding result are concatenated to obtain the feature vector. Initialize a preset number of cluster centers, iteratively calculate the distance between the feature vector and each cluster center, and assign the feature vector to the cluster to which the nearest cluster center belongs, until the preset requirements are met, and output the pattern judgment result of the feature vector; where each cluster represents a pattern.

[0006] In the aforementioned scheme, the TCM knowledge graph is trained based on several TCM classics and textbooks, integrating traditional TCM theory and practical experience. Processing medical record data through this graph allows classical TCM knowledge to be integrated into modern data processing. Standardized medical record texts are mapped to corresponding symptom-sign nodes in the TCM knowledge graph. By leveraging different types of nodes and their semantic relationships within the knowledge graph, the potential information behind the medical record data can be deeply mined, providing a richer and more accurate data foundation for subsequent syndrome differentiation. Preliminary weighting of symptom and sign entities in the standardized medical record texts, followed by updates to the weighting results according to a preset strategy, comprehensively considers the pathological degree of the disease and other relevant factors. This method more accurately quantifies the importance of syndrome element information in syndrome differentiation, avoiding weight biases caused by single factors or subjective judgments, and improving the accuracy of syndrome differentiation. The entity weighting results are mapped to the syndrome coding system to obtain the syndrome coding results. Simultaneously, acupoint resistance data are discretely coded, and finally, the two are concatenated into a feature vector. This multi-dimensional feature extraction method integrates information from disease symptoms, signs, and acupoint resistance, providing a more comprehensive reflection of the patient's condition and offering richer feature evidence for syndrome differentiation. Using cluster analysis, the distance between the feature vector and the cluster center is iteratively calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs. Each cluster represents a syndrome. This data-driven cluster analysis method avoids the interference of subjective factors and can objectively classify syndromes based on the inherent characteristics of the data, improving the accuracy and reliability of syndrome differentiation.

[0007] In one implementation, the Traditional Chinese Medicine knowledge graph is trained in the following way: Entity recognition models were used to extract core data from several TCM classics and textbooks. The core data included entity keywords related to symptoms, signs, syndromes, internal organs, meridians, and acupoint types. Using symptoms, signs, syndromes, internal organs, meridians, and acupoints as core nodes, each entity keyword is assigned to a core node as a basic node, and the resulting node topology is used as the initial TCM knowledge graph. Define the semantic relationship types between various core nodes; among them, the semantic relationship types include membership, correspondence, dominance, taboo, and association. Assigning attribute values ​​to each basic node yields a TCM knowledge graph; the attributes include the standard TCM terminology name corresponding to the node, the official definition of the node, and the association weight between nodes.

[0008] In the above scheme, core data such as symptoms and signs are extracted from TCM classics through entity recognition models to efficiently obtain key information; entity keywords are divided into core nodes to form basic nodes to construct an initial graph, realizing the structured storage of knowledge; various semantic association types such as affiliation and correspondence are defined to clearly show the internal logic of TCM knowledge; and basic nodes are given attributes such as standard TCM terminology names, official interpretations, and inter-node association weights to promote the digital inheritance of TCM knowledge.

[0009] In one implementation, the preprocessed standardized medical record text is mapped to the corresponding symptom-sign nodes in a pre-constructed TCM knowledge graph to obtain structured data, specifically including: The acquired medical record text data is segmented and entity recognized to extract TCM diagnostic and treatment terms from the medical record text data; the TCM diagnostic and treatment terms include symptoms, signs, syndromes and acupoints; Standardized medical record texts are obtained by mapping TCM diagnostic and treatment terms according to the TCM terminology standardization ... Feature extraction is performed on standardized medical record texts to identify symptom and sign entities within them. A candidate entity set is established by selecting basic nodes in the TCM knowledge graph using a fuzzy matching algorithm, and the semantic similarity between the symptom entity and the sign entity and the description of each entity in the candidate entity set is calculated. Joint reasoning is performed based on the semantic similarity calculation results and the association relationship of the TCM knowledge graph to obtain the symptom-sign nodes corresponding to the symptom entities and sign entities in the TCM knowledge graph and associate them to obtain structured data. The structured data includes: the extracted original identification fields, the corresponding standard TCM terminology fields after mapping, the entity identifier used to describe the node entity ID in the TCM knowledge graph, the encoded weight field used to describe the weight value of the corresponding syndrome differentiation factor of the entity, and the associated syndrome type field used to describe the direct association with the entity.

[0010] The above scheme first performs word segmentation, entity recognition, and standardized mapping on the medical record text, which can accurately extract standardized TCM diagnostic and treatment terminology. Symptom and sign entities are identified through feature extraction, and then a fuzzy matching algorithm is used to establish a candidate entity set and calculate semantic similarity, enabling precise location within the knowledge graph. Combined with semantic similarity and knowledge graph associations, joint reasoning can accurately associate symptom-sign nodes. The resulting structured data contains rich information such as original identification fields, standard terminology fields, entity identifiers, encoded weight fields, and associated syndrome types, providing a systematic, accurate, and standardized data foundation for TCM clinical research and decision-making, and helping to improve the accuracy and efficiency of TCM diagnosis and treatment.

[0011] In one implementation, symptom and sign entities in standardized medical record texts are initially weighted according to the severity of the pathology to obtain a basic weighting result. This basic weighting result is then updated according to a preset weight correction strategy to obtain the entity weighting result, specifically including: The standardized medical record text is parsed to extract symptom entities, sign entities, and detailed descriptive information. Based on the pathological degree, initial weights are assigned to the dialectical factors corresponding to each extracted entity to obtain the basic weighting results; The basic weighting results are updated according to the preset weight correction strategy to obtain the symptom-syndrome factor weight for each syndrome differentiation factor. The symptom-syndrome factor weight is used as the entity weighting result. The preset weight correction strategy is to adjust the weight of the basic weighting results by using the preset weight adjustment rules when there is a time coefficient in the detailed description information, or when the symptom entity and the sign entity constitute a prescribed combination of syndrome features.

[0012] In the above scheme, symptom entities, sign entities, and detailed descriptive information are extracted by parsing the text, providing a comprehensive and accurate data foundation for subsequent weighting. Initial weights are assigned to the diagnostic factors corresponding to each entity based on the severity of the pathology, preliminarily quantifying the importance of different entities in syndrome differentiation and obtaining basic weighting results. Using a preset weight correction strategy, when the detailed descriptive information contains a time coefficient or when symptom and sign entities constitute a combination of prescribed syndrome characteristics, the basic weighting results are adjusted to obtain entity weighting results. This weighting method, which comprehensively considers the severity of the pathology and specific conditions, can more accurately reflect the actual role of symptoms and signs in syndrome differentiation, avoiding the limitations of single-factor weighting, improving the accuracy and reliability of syndrome differentiation, and providing a more scientific basis for TCM clinical diagnosis and treatment.

[0013] In one implementation, after obtaining the entity weighting result, the method further includes calculating the correlation weight between the dialectical factor and the syndrome type based on the entity weighting result. Specifically: Obtain the weights of the symptom-differentiation factors corresponding to all relevant symptom entities and sign entities under the target syndrome type, and calculate the association weight between the differentiation factors and the syndrome type using a weighted average formula; the expression of the weighted average formula is as follows: ; In the formula, Evidence-based diagnostic factors The association weight with the certificate type m; The number of symptom-diagnostic factors. For the proof of type m, the first Each related symptom corresponds to a diagnostic factor. The weights; The total number of all relevant symptom entities under syndrome type m; For all relevant symptoms under syndrome type m, the corresponding syndrome differentiation factors The sum of weights.

[0014] In the above scheme, by obtaining the symptom-differentiation factor weights corresponding to all relevant symptom entities under the target syndrome, and calculating the correlation weights using a specific weighted average formula, the degree of correlation between the differentiation factors and the syndrome can be quantified. This quantification method makes the relationship between differentiation factors and syndromes in TCM diagnosis clearer and more explicit, helping doctors to more accurately grasp the characteristics of syndromes and the role of differentiation factors. It provides a scientific and objective basis for TCM clinical diagnosis, improves the accuracy and reliability of TCM diagnosis, and thus enhances the pertinence and effectiveness of TCM treatment plan formulation.

[0015] In one implementation, the entity weighting result is mapped to a pre-built syndrome coding system to obtain the syndrome coding result. Discrete coding is then performed on the acupoint resistance data of the preset target acupoints to obtain the acupoint resistance coding result, specifically including: A syndrome coding system containing several diagnostic factors is pre-constructed, wherein the coding types of the diagnostic factors include location of disease, etiology, nature of disease, and disease progression. Based on the entity weighting results, the dialectical factors in the syndrome coding system are weighted sequentially to obtain the syndrome coding results; Obtain acupoint resistance data from several target acupoints, and construct an acupoint resistance dataset based on the acupoint resistance data; Calculate the distance between the resistance data of each acupoint and the mean of the acupoint resistance dataset, encode the acupoint resistance data according to the distance, and obtain the acupoint resistance encoding result; The syndrome coding results and acupoint resistance coding results are concatenated, and the concatenated coding results are normalized to obtain the feature vector.

[0016] In the above scheme, the syndrome coding results are obtained by assigning weights to the diagnostic factors in the syndrome coding system based on the entity weighting results. This effectively integrates the entity weight information into the coding, more accurately reflecting disease characteristics. Processing the acupoint resistance data of the target acupoints, constructing a dataset, calculating the distance from the mean, and encoding the acupoint resistance coding results can uncover potential information within the acupoint resistance data. Concatenating and normalizing the syndrome coding results and acupoint resistance coding results yields a feature vector, integrating TCM diagnostic information and acupoint resistance information to form a unified, standardized, and representative feature expression. This provides a multimodal, accurate, and easily processed data foundation for subsequent TCM diagnosis, analysis, and research, helping to improve the accuracy and efficiency of TCM diagnosis and treatment. In one implementation, a preset number of cluster centers are initialized. The distance between the feature vector and each cluster center is iteratively calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs, until a preset requirement is met. The result of the feature vector type judgment is then output, specifically including: A predetermined number of feature vector samples are selected as initial cluster centers in a pre-constructed sample set. The distance from each feature vector to an initial cluster center is calculated, and the feature vector is assigned to the cluster to which the nearest initial cluster center belongs. The sample set includes several common pattern clusters, and each common pattern cluster contains several feature vector samples. Calculate the mean of the feature vectors within each cluster after allocation, and use the mean of the feature vectors as the updated cluster centers of each cluster. Calculate the change from each initial cluster center to the corresponding updated cluster center. Stop iterative training when each change is less than a preset change threshold or the number of iterations reaches a preset number. The evidence type judgment result is generated based on the evidence type corresponding to the clusters assigned by the feature vectors.

[0017] In the above scheme, initial cluster centers are selected from a pre-constructed sample set, utilizing information from existing common syndrome clusters to provide a reasonable starting point for clustering. The distance from the feature vector to the initial cluster center is calculated, and the feature vector is assigned to the nearest cluster, allowing for preliminary classification of the feature vector. During the iteration process, the mean of the feature vectors within each cluster is calculated as the updated cluster centers, and the iteration is stopped based on the change and the number of iterations. This continuously optimizes the clustering results, ensuring accuracy and stability. Finally, syndrome classification results are generated based on the syndromes corresponding to the clusters assigned to the feature vectors, providing an objective and scientific method for syndrome classification, which helps improve the efficiency and accuracy of TCM syndrome classification.

[0018] Secondly, this application also provides an acupoint-syndrome association information processing device, including: a data processing module, a weight assignment module, a vector generation module and a syndrome type judgment module; The data processing module is used to map the pre-processed standardized medical record text to the corresponding symptom-sign nodes in the pre-constructed TCM knowledge graph to obtain structured data. The TCM knowledge graph is trained by several TCM classics and textbooks and is used to describe different types of nodes and the semantic relationships between them. The structured data includes syndrome data that are directly related to the nodes. The weighting module is used to initially assign weights to symptom and sign entities in standardized medical record texts according to the severity of the pathology, to obtain basic weighting results. The basic weighting results are then updated according to a preset weighting correction strategy to obtain entity weighting results. The vector generation module is used to map the entity weighting result to the pre-built syndrome coding system to obtain the syndrome coding result, to perform discrete coding on the acupoint resistance data of the target acupoint to obtain the acupoint resistance coding result, and to concatenate the syndrome coding result and the acupoint resistance coding result to obtain the feature vector. The pattern determination module is used to initialize a preset number of cluster centers, iteratively calculate the distance between the feature vector and each cluster center, and assign the feature vector to the cluster to which the nearest cluster center belongs, until the preset requirements are met, and output the pattern determination result of the feature vector; where each cluster represents a pattern.

[0019] In the aforementioned scheme, the TCM knowledge graph is trained based on several TCM classics and textbooks, integrating traditional TCM theory and practical experience. Processing medical record data through this graph allows classical TCM knowledge to be integrated into modern data processing. Standardized medical record texts are mapped to corresponding symptom-sign nodes in the TCM knowledge graph. By leveraging different types of nodes and their semantic relationships within the knowledge graph, the potential information behind the medical record data can be deeply mined, providing a richer and more accurate data foundation for subsequent syndrome differentiation. Preliminary weighting of symptom and sign entities in the standardized medical record texts, followed by updates to the weighting results according to a preset strategy, comprehensively considers the pathological degree of the disease and other relevant factors. This method more accurately quantifies the importance of syndrome element information in syndrome differentiation, avoiding weight biases caused by single factors or subjective judgments, and improving the accuracy of syndrome differentiation. The entity weighting results are mapped to the syndrome coding system to obtain the syndrome coding results. Simultaneously, acupoint resistance data are discretely coded, and finally, the two are concatenated into a feature vector. This multi-dimensional feature extraction method integrates information from disease symptoms, signs, and acupoint resistance, providing a more comprehensive reflection of the patient's condition and offering richer feature evidence for syndrome differentiation. Using cluster analysis, the distance between the feature vector and the cluster center is iteratively calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs. Each cluster represents a syndrome. This data-driven cluster analysis method avoids the interference of subjective factors and can objectively classify syndromes based on the inherent characteristics of the data, improving the accuracy and reliability of syndrome differentiation.

[0020] In one implementation, the Traditional Chinese Medicine knowledge graph is trained in the following way: Entity recognition models were used to extract core data from several TCM classics. The core data included entity keywords related to symptoms, signs, syndromes, internal organs, meridians, and acupoints. Using symptoms, signs, syndromes, internal organs, meridians, and acupoints as core nodes, each entity keyword is assigned to a core node as a basic node, and the resulting node topology is used as the initial TCM knowledge graph. Define the semantic relationship types between various core nodes; among them, the semantic relationship types include membership, correspondence, dominance, taboo, and association. Assigning attribute values ​​to each basic node yields a TCM knowledge graph; the attributes include the standard TCM terminology name corresponding to the node, the official definition of the node, and the association weight between nodes.

[0021] In the above scheme, core data such as symptoms and signs are extracted from TCM classics through entity recognition models to efficiently obtain key information; entity keywords are divided into core nodes to form basic nodes to construct an initial graph, realizing the structured storage of knowledge; various semantic association types such as affiliation and correspondence are defined to clearly show the internal logic of TCM knowledge; and basic nodes are given attributes such as standard TCM terminology names, official interpretations, and inter-node association weights to promote the digital inheritance of TCM knowledge.

[0022] In one implementation, the data processing module maps the preprocessed standardized medical record text to the corresponding symptom-sign nodes in a pre-constructed TCM knowledge graph to obtain structured data, specifically including: The acquired medical record text data is segmented and entity recognized to extract TCM diagnostic and treatment terms from the medical record text data; the TCM diagnostic and treatment terms include symptoms, signs, syndromes and acupoints; Standardized medical record texts are obtained by mapping TCM diagnostic and treatment terms according to the TCM terminology standardization ... Feature extraction is performed on standardized medical record texts to identify symptom and sign entities within them. A candidate entity set is established by selecting basic nodes in the TCM knowledge graph using a fuzzy matching algorithm, and the semantic similarity between the symptom entity and the sign entity and the description of each entity in the candidate entity set is calculated. Joint reasoning is performed based on the semantic similarity calculation results and the association relationship of the TCM knowledge graph to obtain the symptom-sign nodes corresponding to the symptom entities and sign entities in the TCM knowledge graph and associate them to obtain structured data. The structured data includes: the extracted original identification fields, the corresponding standard TCM terminology fields after mapping, the entity identifier used to describe the node entity ID in the TCM knowledge graph, the encoded weight field used to describe the weight value of the corresponding syndrome differentiation factor of the entity, and the associated syndrome type field used to describe the direct association with the entity.

[0023] The above scheme first performs word segmentation, entity recognition, and standardized mapping on the medical record text, which can accurately extract standardized TCM diagnostic and treatment terminology. Symptom and sign entities are identified through feature extraction, and then a fuzzy matching algorithm is used to establish a candidate entity set and calculate semantic similarity, enabling precise location within the knowledge graph. Combined with semantic similarity and knowledge graph associations, joint reasoning can accurately associate symptom-sign nodes. The resulting structured data contains rich information such as original identification fields, standard terminology fields, entity identifiers, encoded weight fields, and associated syndrome types, providing a systematic, accurate, and standardized data foundation for TCM clinical research and decision-making, and helping to improve the accuracy and efficiency of TCM diagnosis and treatment.

[0024] In one implementation, the weighting module initially assigns weights to symptom and sign entities in the standardized medical record text based on the severity of the pathology, obtaining a basic weighting result. The basic weighting result is then updated according to a preset weight correction strategy to obtain the entity weighting result, specifically including: The standardized medical record text is parsed to extract symptom entities, sign entities, and detailed descriptive information. Based on the pathological degree, initial weights are assigned to the dialectical factors corresponding to each extracted entity to obtain the basic weighting results; The basic weighting results are updated according to the preset weight correction strategy to obtain the symptom-syndrome factor weight for each syndrome differentiation factor. The symptom-syndrome factor weight is used as the entity weighting result. The preset weight correction strategy is to adjust the weight of the basic weighting results by using the preset weight adjustment rules when there is a time coefficient in the detailed description information, or when the symptom entity and the sign entity constitute a prescribed combination of syndrome features.

[0025] In the above scheme, symptom entities, sign entities, and detailed descriptive information are extracted by parsing the text, providing a comprehensive and accurate data foundation for subsequent weighting. Initial weights are assigned to the diagnostic factors corresponding to each entity based on the severity of the pathology, preliminarily quantifying the importance of different entities in syndrome differentiation and obtaining basic weighting results. Using a preset weight correction strategy, when the detailed descriptive information contains a time coefficient or when symptom and sign entities constitute a combination of prescribed syndrome characteristics, the basic weighting results are adjusted to obtain entity weighting results. This weighting method, which comprehensively considers the severity of the pathology and specific conditions, can more accurately reflect the actual role of symptoms and signs in syndrome differentiation, avoiding the limitations of single-factor weighting, improving the accuracy and reliability of syndrome differentiation, and providing a more scientific basis for TCM clinical diagnosis and treatment.

[0026] In one implementation, after obtaining the entity weighting result, the method further includes calculating the correlation weight between the dialectical factor and the syndrome type based on the entity weighting result. Specifically: Obtain the weights of the symptom-differentiation factors corresponding to all relevant symptom entities and sign entities under the target syndrome type, and calculate the association weight between the differentiation factors and the syndrome type using a weighted average formula; the expression of the weighted average formula is as follows: ; In the formula, Evidence-based diagnostic factors The association weight with the certificate type m; The number of symptom-diagnostic factors. For the proof of type m, the first Each related symptom corresponds to a diagnostic factor. The weights; The total number of all relevant symptom entities under syndrome type m; For all relevant symptoms under syndrome type m, the corresponding syndrome differentiation factors The sum of weights.

[0027] In the above scheme, by obtaining the symptom-differentiation factor weights corresponding to all relevant symptom entities under the target syndrome, and calculating the correlation weights using a specific weighted average formula, the degree of correlation between the differentiation factors and the syndrome can be quantified. This quantification method makes the relationship between differentiation factors and syndromes in TCM diagnosis clearer and more explicit, helping doctors to more accurately grasp the characteristics of syndromes and the role of differentiation factors. It provides a scientific and objective basis for TCM clinical diagnosis, improves the accuracy and reliability of TCM diagnosis, and thus enhances the pertinence and effectiveness of TCM treatment plan formulation.

[0028] In one implementation, the vector generation module maps the entity weighting result to a pre-built syndrome coding system to obtain the syndrome coding result. It then performs discrete coding on the acupoint resistance data of the preset target acupoints to obtain the acupoint resistance coding result, specifically including: A syndrome coding system containing several diagnostic factors is pre-constructed, wherein the coding types of the diagnostic factors include location of disease, etiology, nature of disease, and disease progression. Based on the entity weighting results, the dialectical factors in the syndrome coding system are weighted sequentially to obtain the syndrome coding results; Obtain acupoint resistance data from several target acupoints, and construct an acupoint resistance dataset based on the acupoint resistance data; Calculate the distance between the resistance data of each acupoint and the mean of the acupoint resistance dataset, encode the acupoint resistance data according to the distance, and obtain the acupoint resistance encoding result; The syndrome coding results and acupoint resistance coding results are concatenated, and the concatenated coding results are normalized to obtain the feature vector.

[0029] In the above scheme, the syndrome coding results are obtained by assigning weights to the diagnostic factors in the syndrome coding system based on the entity weighting results. This effectively integrates the entity weight information into the coding, more accurately reflecting symptom characteristics. Processing the acupoint resistance data of the target acupoints, constructing a dataset, calculating the distance from the mean, and encoding the acupoint resistance coding results can uncover potential information within the acupoint resistance data. Concatenating and normalizing the syndrome coding results and acupoint resistance coding results yields a feature vector, integrating TCM diagnostic information and acupoint resistance information to form a unified, standardized, and representative feature expression. This provides a multimodal, accurate, and easily processed data foundation for subsequent TCM diagnosis, analysis, and research, helping to improve the accuracy and efficiency of TCM diagnosis and treatment. In one implementation, the pattern recognition module initializes a preset number of cluster centers, iteratively calculates the distance between the feature vector and each cluster center, and assigns the feature vector to the cluster to which the nearest cluster center belongs, until a preset requirement is met, and outputs the pattern recognition result of the feature vector, specifically including: A predetermined number of feature vector samples are selected as initial cluster centers in a pre-constructed sample set. The distance from each feature vector to an initial cluster center is calculated, and the feature vector is assigned to the cluster to which the nearest initial cluster center belongs. The sample set includes several common pattern clusters, and each common pattern cluster contains several feature vector samples. Calculate the mean of the feature vectors within each cluster after allocation, and use the mean of the feature vectors as the updated cluster centers of each cluster. Calculate the change from each initial cluster center to the corresponding updated cluster center. Stop iterative training when each change is less than a preset change threshold or the number of iterations reaches a preset number. The evidence type judgment result is generated based on the evidence type corresponding to the clusters assigned by the feature vectors.

[0030] In the above scheme, initial cluster centers are selected from a pre-constructed sample set, utilizing information from existing common syndrome clusters to provide a reasonable starting point for clustering. The distance from the feature vector to the initial cluster center is calculated, and the feature vector is assigned to the nearest cluster, allowing for preliminary classification of the feature vector. During the iteration process, the mean of the feature vectors within each cluster is calculated as the updated cluster centers, and the iteration is stopped based on the change and the number of iterations. This continuously optimizes the clustering results, ensuring accuracy and stability. Finally, syndrome classification results are generated based on the syndromes corresponding to the clusters assigned to the feature vectors, providing an objective and scientific method for syndrome classification, which helps improve the efficiency and accuracy of TCM syndrome classification.

[0031] Thirdly, this application also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the above-mentioned acupoint-syndrome association information processing method.

[0032] Fourthly, this application also provides a computer-readable storage medium, which includes a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the above-mentioned acupoint-syndrome association information processing method. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating a method for processing acupoint-syndrome correlation information according to an embodiment of the present invention; Figure 2This is a schematic diagram of a module of an acupoint-syndrome association information processing device provided in one embodiment of the present invention. Detailed Implementation

[0034] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0035] The terms "first" and "second," etc., in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0036] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0037] Example 1 See Figure 1 , Figure 1 This is a flowchart illustrating a method for processing acupoint-syndrome correlation information according to an embodiment of the present invention. The embodiment of the present invention provides a method for processing acupoint-syndrome correlation information, including steps 101 to 104, each step being as follows: Step 101: Map the preprocessed standardized medical record text to the corresponding symptom-sign nodes in the pre-constructed TCM knowledge graph to obtain structured data; the TCM knowledge graph is trained from several TCM classics and textbooks and is used to describe different types of nodes and the semantic relationships between them. The structured data includes syndrome data that are directly related to the nodes. In this embodiment of the invention, the medical record text is preprocessed to obtain standardized text, and the standardized medical record text is mapped to the symptom-sign nodes of a traditional Chinese medicine knowledge graph pre-constructed based on several traditional Chinese medicine classics, thereby obtaining structured data, which includes syndrome data directly associated with the nodes.

[0038] In one embodiment, the TCM knowledge graph is trained as follows: Core data from several TCM classics is extracted using an entity recognition model; the core data includes entity keywords related to symptoms, signs, syndromes, internal organs, meridians, and acupoints; symptoms, signs, syndromes, internal organs, meridians, and acupoints are used as core nodes, and each entity keyword is assigned to a subordinate core node as a basic node, with the resulting node topology serving as the initial TCM knowledge graph; semantic association types between various core nodes are defined; these semantic association types include subordinate relationships, correspondence relationships, indication relationships, contraindication relationships, and association relationships; attribute values ​​are assigned to each basic node to obtain the TCM knowledge graph; the attributes include the standard TCM terminology name corresponding to the node, the official definition of the node, and the association weight between nodes.

[0039] In this embodiment of the invention, natural language processing technology is used to perform word segmentation and entity recognition operations on the original medical record text, extracting terms such as symptoms, signs, syndrome types, and acupoints. This is based on the "Standard Edition of Traditional Chinese Medicine Terminology" and the "Guidelines for Clinical Practice of Traditional Chinese Medicine Rehabilitation". The terminology extracted for "Cervical Spondylosis" was standardized and mapped, and expressions that did not conform to TCM norms were removed to ensure the accuracy and consistency of the data. From 39 TCM classics and textbooks such as "Clinical Practice Guidelines for TCM Rehabilitation: Cervical Spondylosis" and "Acupuncture", six types of core data were extracted: "symptoms-signs-syndromes-organs-meridians-acupoints" using manual annotation combined with the BERT entity recognition model. Each type of data corresponds to a core node, and each keyword entity is a basic node. Three attributes were assigned to each entity node: (1) Name: corresponding to standardized TCM terminology to ensure the standardization and uniformity of terminology; (2) Definition: referencing the authoritative interpretation in "TCM Terminology" (standard version) to provide accurate semantic explanation for the node; (3) Association weight: calculated based on the association strength between entities in the data source, and different types of associations have different calculation methods. Then, based on manual experience, the semantic association types between the six types of core nodes were set, including subordinate relationship, correspondence relationship, treatment relationship, contraindication relationship and association relationship. Specifically: (1) Affiliation: Acupoint - (belonging) - Meridian, Syndrome type - (includes) - Symptom; (2) Correspondence: Syndrome type - (corresponding) - Zang-fu / Meridian; (3) Indication: Acupoint - (indication) - Syndrome type / symptom; (4) Contraindication: Syndrome type - (prohibited) - Acupoint; (5) Association: Symptom - (accompanying) - Sign / Syndrome. The standardized medical record text obtained after preprocessing is mapped to the corresponding symptom-sign nodes in the pre-constructed TCM knowledge graph to obtain structured data containing syndrome type data directly associated with the nodes. Furthermore, after the graph is established, the Neo4j graph database is used to store the nodes and relationships to facilitate data management and query.

[0040] For example, suppose there is an original TCM medical record text: "The patient has experienced neck pain, accompanied by dizziness, a purplish-dark tongue, and a choppy pulse for the past week. The diagnosis is Qi stagnation and blood stasis syndrome." Using natural language processing (NLP), the text is segmented to identify the symptoms "neck pain" and "dizziness," the signs "purplish-dark tongue" and "choppy pulse," and the syndrome type "Qi stagnation and blood stasis syndrome." "Neck pain" is standardized to "neck pain," while other terms remain unchanged, and expressions that do not conform to TCM standards are removed. Relevant core data is extracted from 39 TCM classics and textbooks, such as symptom nodes corresponding to "neck pain" and "dizziness," sign nodes corresponding to "purplish-dark tongue" and "choppy pulse," and syndrome type nodes corresponding to "Qi stagnation and blood stasis syndrome." Then, each node is assigned attributes. For example: "Neck pain": the name is "neck pain," the definition is quoted from the "TCM Terminology," and the association weight: the association weight with "Qi stagnation and blood stasis syndrome" is calculated using statistical co-occurrence frequency combined with an information gain algorithm, assumed to be 0.8. Furthermore, semantic relationships are established between nodes: Membership: "Qi stagnation and blood stasis syndrome" - (includes) - "neck pain" and "dizziness"; Association: "Neck pain" - (accompanying) - "purple tongue" and "choppy pulse". Nodes and relationships are stored in the Neo4j graph database. The preprocessed standardized medical record texts "neck pain", "dizziness", "purple tongue", "choppy pulse", and "Qi stagnation and blood stasis syndrome" are mapped to corresponding nodes in the TCM knowledge graph, resulting in structured data containing the relationships and weights between symptoms, signs, and syndrome types, providing a reference for TCM diagnosis and treatment.

[0041] In one embodiment, the preprocessed standardized medical record text is mapped to the corresponding symptom-sign nodes in a pre-constructed TCM knowledge graph to obtain structured data. Specifically, this includes: performing word segmentation and entity recognition on the acquired medical record text data to extract TCM diagnostic and treatment terms from the medical record text data; wherein TCM diagnostic and treatment terms include symptoms, signs, syndrome types, and acupoints; standardizing and mapping the TCM diagnostic and treatment terms according to TCM terminology norms to obtain standardized medical record text; extracting features from the standardized medical record text to identify symptom and sign entities in the standardized medical record text; and selecting basic nodes in the TCM knowledge graph using a fuzzy matching algorithm to establish... A candidate entity set is used to calculate the semantic similarity between the symptom and sign entities and the descriptions of each entity in the candidate entity set. Based on the semantic similarity calculation results and the association relationship of the TCM knowledge graph, joint reasoning is performed to obtain the symptom-sign nodes corresponding to the symptom and sign entities in the TCM knowledge graph and associate them to obtain structured data. The structured data includes: the extracted original identification fields, the corresponding standard TCM terminology fields after mapping, the entity identifier used to describe the node entity ID in the TCM knowledge graph, the encoded weight field used to describe the weight value of the corresponding syndrome differentiation factor of the entity, and the associated syndrome type field used to describe the direct association with the entity.

[0042] In this embodiment of the invention, the acquired medical record text data is segmented and entity recognized to extract traditional Chinese medicine (TCM) diagnostic and treatment terms, including symptoms, signs, syndromes, and acupoints. The extracted TCM diagnostic and treatment terms are standardized and mapped according to TCM terminology norms to obtain standardized medical record text, ensuring the standardization and consistency of terminology. A Transformer-based TCM text BERT model is used to extract features from the standardized medical record text, identifying symptom and sign entities. A candidate entity set is established by selecting basic nodes in the TCM knowledge graph using fuzzy matching algorithms (such as cosine similarity and edit distance) to cover potentially matching entities. The semantic similarity between symptom and sign entities and the descriptions of each entity in the candidate entity set is calculated to provide a basis for subsequent entity disambiguation. Joint reasoning is performed based on the semantic similarity calculation results and the association relationships in the TCM knowledge graph to determine the corresponding symptom-sign nodes in the TCM knowledge graph for symptom and sign entities, and these nodes are then associated. Furthermore, for mentions that cannot be mapped, they can be submitted to an expert feedback system for processing to ensure the accuracy and completeness of the mapping. The generated structured data includes original identification fields, mapped standard TCM terminology fields, entity identifiers, encoded weight fields, and associated syndrome types. The encoded weight field is generated by querying a pre-defined encoding rule mapping table, which defines the correspondence between TCM terminology entities and 48-bit syndrome differentiation factor weight vectors, i.e., the symptom-syndrome differentiation factor weights mentioned later. It should be noted that in this embodiment of the invention, the above mapping process is implemented through entity linking technology. Based on the constructed TCM knowledge graph, non-standardized entities extracted from the text are precisely associated with standardized "symptom-sign" nodes in the knowledge graph, providing standardized mapping targets and data support for entity import.

[0043] For example, suppose there is an original TCM medical record text: "The patient has recently experienced neck stiffness, accompanied by headache, and has a pale red tongue and wiry pulse." Using word segmentation and entity recognition technology, the TCM diagnostic terms "neck stiffness," "headache," "pale red tongue," and "wiry pulse" are extracted. According to TCM terminology standardization, "neck stiffness" is standardized to "neck rigidity," while other terms remain unchanged, resulting in the standardized medical record text "neck rigidity," "headache," "pale red tongue," and "wiry pulse." A Transformer-based BERT model for TCM text is then used to extract features from the standardized medical record text, identifying the symptom entities "neck rigidity" and "headache." The physical signs are defined as "pale red tongue" and "wiry pulse". A fuzzy matching algorithm is used to select basic nodes in the TCM knowledge graph to establish a candidate entity set. For example, for "stiff neck", the candidate entity set might include "stiff neck" and "neck stiffness". The semantic similarity between "stiff neck" and the descriptions of each entity in the candidate entity set is calculated, assuming a semantic similarity of 1 with "stiff neck" and 0.8 with "neck stiffness". Based on the semantic similarity calculation results and the association relationship in the TCM knowledge graph, joint reasoning is performed to determine that the symptom-sign node corresponding to "stiff neck" in the TCM knowledge graph is "stiff neck" (entity ID: 123).

[0044] Similarly, map "headache," "pale red tongue," and "wiry pulse" to the corresponding nodes in the knowledge graph (assuming entity IDs are 456, 789, and 101 respectively). The resulting structured data example is as follows:

[0045] Step 102: Assign preliminary weights to symptom and sign entities in the standardized medical record text based on the pathological degree to obtain basic weighting results. Update the basic weighting results according to the preset weight correction strategy to obtain entity weighting results.

[0046] In this embodiment of the invention, for standardized medical record text, the symptom and sign entities are first preliminarily weighted according to the pathological degree to obtain basic weighting results; then, the basic weighting results are updated according to a preset weight correction strategy to finally obtain entity weighting results. Quantifying and weighting symptoms and signs in traditional Chinese medicine diagnosis makes the diagnostic process more objective and standardized, avoiding potential subjective differences in traditional Chinese medicine diagnosis, thereby improving the reliability and repeatability of diagnostic results.

[0047] In one embodiment, symptom entities and sign entities in standardized medical record text are initially weighted according to the pathological degree to obtain a basic weighting result. The basic weighting result is then updated according to a preset weight correction strategy to obtain an entity weighting result. Specifically, this includes: parsing the standardized medical record text to extract symptom entities, sign entities, and detailed description information; assigning initial weights to the diagnostic factors corresponding to each extracted entity according to the pathological degree to obtain a basic weighting result; updating the basic weighting result according to the preset weight correction strategy to obtain the symptom-diagnosis factor weight for each diagnostic factor, and using the symptom-diagnosis factor weight as the entity weighting result; wherein, the preset weight correction strategy is to adjust the weights of the basic weighting result using preset weight adjustment rules when the detailed description information contains a time coefficient, or when the symptom entities and sign entities constitute a prescribed combination of syndrome characteristics.

[0048] In this embodiment of the invention, symptom and sign entities in standardized medical record texts are assigned appropriate weights to more accurately reflect their importance in disease diagnosis and syndrome differentiation. A preliminary weighting result is obtained, which is then updated according to a preset weight correction strategy to finally obtain the entity weighting result, providing a quantitative basis for subsequent syndrome classification and disease diagnosis. Initial weights are assigned to the syndrome differentiation factors corresponding to each extracted entity based on the pathological degree, thus obtaining the basic weighting result. See the table below, which provides an example of weights assigned initially based on pathological degree.

[0049]

[0050] The basic weight setting rules in the table above are based on the basic theory of traditional Chinese medicine. The scores are divided into multiple levels according to the different pathological degrees, such as severe (7-9 points), moderate to severe (5-6 points), moderate (3-4 points), mild (1-2 points), no effect (0 points), and negative effect (-1, -2 points). The corresponding clinical manifestations, typical symptom examples, and weight correction rules are given for different degrees. The basic weighting results are updated according to the preset weight correction strategy to obtain the symptom-syndrome factor weight of each syndrome differentiation factor, which is used as the entity weighting result. The preset weight correction strategy mainly includes two aspects: time coefficient correction and feature combination additional value: (1) Time coefficient correction: The time coefficient is a dynamic weight adjustment parameter used to quantify the change of symptoms over time. Different symptom persistence characteristics correspond to different time coefficients, such as worsening at night (coefficient 1.5), obvious in the morning (coefficient 1.3), persistent (coefficient 1.2), and intermittent (coefficient 0.8). The weight is corrected by multiplying the basic weight by the time coefficient. (2) Feature Combination Bonus: When a specific symptom and sign combination occurs simultaneously, and the combination includes one core symptom and at least two related signs, which meets the syndrome feature combination specified in the "Standards for Diagnosis and Efficacy of Diseases and Syndromes in Traditional Chinese Medicine", a fixed additional weight of 2 points is triggered. The final weight of the symptom-syndrome differentiation factor = basic weight × time coefficient + feature combination bonus.

[0051] For example, consider a standardized medical record: "The patient experiences stabbing pain in the neck, which worsens at night, accompanied by a purplish-dark tongue and a choppy pulse." Information extraction from this record yields the following: Physical symptoms: stabbing pain in the neck; Physical signs: Dark purple tongue, choppy pulse; Detailed description information: Worsening at night; Then, preliminary weighting was applied based on the severity of the pathology. "Neck stabbing pain" is classified as a severe pathological condition, persistent and significantly affecting daily function; therefore, it is assigned a base weight of 8 points according to the base weighting rules. "Dark purple tongue" and "smooth pulse," as physical signs, are preliminarily judged to be of moderate to severe pathological severity based on their importance in disease diagnosis, and are each assigned a base weight of 6 points. "Neck stabbing pain" exhibits a time characteristic of worsening at night, with a time coefficient of 1.5, resulting in a adjusted weight of 8 × 1.5 = 12 points. "Dark purple tongue" and "smooth pulse" lack clear time-related descriptions; according to the default rules, their time coefficients are 1.0, resulting in adjusted weights of 6 × 1.0 = 6 points respectively. The combination of "stabbing neck pain" (core symptom), "purplish-dark tongue" (related sign), and "hesitant pulse" (related sign) constitutes a characteristic combination of syndromes as defined in the *Standards for Diagnosis and Efficacy of Traditional Chinese Medicine Diseases*, triggering a fixed additional weight of 2 points. The final symptom-syndrome differentiation factor weight for "stabbing neck pain" is 12 + 2 = 14 points. "Purple-dark tongue" and "hesitant pulse," due to their combination with the core symptom, also receive an additional weight of 2 points, resulting in final weights of 6 + 2 = 8 points respectively. Therefore, the final symptom-syndrome differentiation factor weights for each syndrome differentiation factor, i.e., the entity weighting results, are as follows:

[0052] By following the steps above, the symptom and sign entities in the standardized medical record text can be assigned weights.

[0053] In one embodiment, after obtaining the entity weighting result, the method further includes calculating the correlation weight between the dialectical factor and the syndrome type based on the entity weighting result, specifically: Obtain the weights of the symptom-differentiation factors corresponding to all relevant symptom entities and sign entities under the target syndrome type, and calculate the association weight between the differentiation factors and the syndrome type using a weighted average formula; the expression of the weighted average formula is as follows: ; In the formula, Evidence-based diagnostic factors The association weight with the certificate type m; The number of symptom-diagnostic factors. For the proof of type m, the first Each related symptom corresponds to a diagnostic factor. The weights; The total number of all relevant symptom entities under syndrome type m; For all relevant symptoms under syndrome type m, the corresponding syndrome differentiation factors The sum of weights.

[0054] In this embodiment of the invention, after assigning weights to symptom entities and sign entities in standardized medical record texts and obtaining entity weighting results, the correlation weight between syndrome differentiation factors and syndrome types is further calculated. This helps to more accurately determine the patient's syndrome type, provides quantitative evidence for syndrome differentiation and treatment in traditional Chinese medicine, and makes traditional Chinese medicine diagnosis more objective and scientific.

[0055] For example, the target syndrome m is "Qi stagnation and blood stasis syndrome", which has three related symptom entities: "stabbing pain in the chest", "dark purple tongue", and "choppy pulse". The weights of the corresponding symptom-syndrome differentiation factors are as follows: =10 (chest pain) =8 (Dark purple tongue) =8 (Choppy pulse), Total number of all relevant symptom entities under syndrome type m =3. Then, all relevant symptoms under syndrome type m correspond to the syndrome differentiation factors. The sum of weights =10+8+8=26, calculate the correlation weight between the dialectical factor and the syndrome type according to the weighted average formula. =26 3 = 8.666667. That is, the correlation weight between the syndrome differentiation factor and "Qi stagnation and blood stasis syndrome" is approximately 8.67. This correlation weight can be used to further determine whether the patient conforms to this syndrome type. The higher the correlation weight, the stronger the correlation between the patient's symptoms and the syndrome type.

[0056] Step 103: Map the entity weighting result to the pre-built syndrome coding system to obtain the syndrome coding result. Discretize the acupoint resistance data of the target acupoint to obtain the acupoint resistance coding result. Concatenate the syndrome coding result and the acupoint resistance coding result to obtain the feature vector.

[0057] In this embodiment of the invention, the entity weighting result is mapped to a pre-constructed syndrome coding system to obtain the syndrome coding result. The acupoint resistance data of the target acupoint is discretely encoded to obtain the acupoint resistance coding result. The previously obtained syndrome coding result and acupoint resistance coding result are concatenated to form a feature vector. The feature vector simultaneously contains syndrome coding and acupoint resistance coding information, reflecting the body's health status from both traditional Chinese medicine theory and modern detection technology perspectives. A feature vector that comprehensively considers both syndrome and acupoint resistance information can more accurately reflect the patient's condition, reducing potential misdiagnosis and missed diagnosis from a single information source, and improving diagnostic accuracy. Standardized feature vectors can be quickly input into diagnostic models for analysis, shortening diagnostic time and improving diagnostic efficiency, especially suitable for large-scale clinical screening and diagnosis.

[0058] In one embodiment, the entity weighting result is mapped to a pre-constructed syndrome coding system to obtain the syndrome coding result. Discrete coding is then performed on the acupoint resistance data of preset target acupoints to obtain the acupoint resistance coding result. Specifically, this includes: pre-constructing a syndrome coding system containing several diagnostic factors, wherein the coding type of the diagnostic factors includes disease location, etiology, disease nature, and disease progression; sequentially assigning weights to the diagnostic factors in the syndrome coding system according to the entity weighting result to obtain the syndrome coding result; acquiring acupoint resistance data of several target acupoints and constructing an acupoint resistance dataset based on the acupoint resistance data; calculating the distance between each acupoint resistance data and the mean of the acupoint resistance dataset, and coding the acupoint resistance data according to the distance to obtain the acupoint resistance coding result; concatenating the syndrome coding result and the acupoint resistance coding result, and normalizing the concatenated coding result to obtain a feature vector.

[0059] In this embodiment of the invention, a syndrome coding system is pre-constructed, comprising diagnostic factors such as disease location, etiology, nature, and progression. These diagnostic factors cover the multi-dimensional description of diseases in Traditional Chinese Medicine (TCM) diagnosis, providing a basic framework for subsequent coding and analysis. Based on the entity weighting results, each diagnostic factor in the syndrome coding system is weighted. The 48-bit coding element covers multiple aspects of TCM diagnosis. By assigning values ​​from 1 to 9 to represent the influence weight of symptoms or signs, the TCM diagnostic information is further refined and quantified, transforming the symptom and sign information in TCM diagnosis into digital coding results, facilitating computer processing and analysis. Acupoint resistance data for several target acupoints are acquired, and an acupoint resistance dataset is constructed, providing a data foundation for subsequent coding. The distance between each acupoint resistance data point and the mean of the dataset is calculated. The original resistance values ​​are standardized using the Z-score formula, and then encoded into values ​​from 1 to 9 according to a preset discrete coding rule, achieving the discretization and standardization of acupoint resistance data. The syndrome coding results and acupoint resistance coding results are concatenated to form a feature vector containing multiple information, comprehensively reflecting the patient's TCM diagnostic information and acupoint resistance characteristics. The concatenated feature vector is then subjected to Min-Max normalization, mapping all dimensions to the [0,1] interval to eliminate the impact of differences in coding magnitudes on subsequent data analysis and ensure data comparability and consistency.

[0060] It should be noted that the 48-bit coded elements (diagnostic factors) are: heart, liver, spleen, lung, kidney, gallbladder, stomach, intestine, bladder, uterus, seminal vesicle, Ren and Chong meridians, exterior, internal organ excess, imbalance, heat, blood heat, bleeding, deficiency heat, fluid deficiency, accumulation, toxin, pus, cold, summer heat, blood deficiency, qi deficiency, yin deficiency, yang deficiency, irritability, delirium, instability, blood stasis, obstruction of collaterals, qi reversal, qi stagnation, cough and asthma, yang hyperactivity, wind, wind-induced movement, water, dampness, dryness, food, phlegm, parasites, degree, constitution.

[0061] The system includes four 48-bit coding elements: location of the disease (positions 1-12), cause of the disease (positions 13-26), nature of the disease (positions 27-38), and progression of the disease (positions 39-48). This 48-bit coding system is a digital representation of the TCM diagnostic system, comprehensively covering key dimensions of understanding the physiological and pathological states of the human body in TCM theory. By assigning numerical values ​​of 1-9 to these coding elements to represent the influence weight of symptoms or signs, complex TCM diagnostic information can be transformed into standardized, quantifiable data, facilitating computer processing and analysis, and contributing to the intelligent and precise diagnosis of TCM.

[0062] The encoding rules for acupoint resistance data are shown in the table below:

[0063] The resistance value of each acupoint is encoded. First, the resistance of the acupoints is cleaned, and resistance data of 50-9000 ohms are retained. The mean and standard deviation of the resistance of each acupoint are calculated. The improved Z-score algorithm is used to convert the continuous resistance values ​​into discrete codes of 1-9 levels.

[0064] For example, consider a patient with cervical spondylosis. The entity weighting results of their standardized medical record text show: cervical spine-related symptoms have a weight of 8, strain as an etiology has a weight of 7, blood stasis as a pathological nature has a weight of 8, and aggravated condition has a weight of 7. The corresponding syndrome coding system includes differentiation factors: location (cervical spine), etiology (strain), pathological nature (blood stasis), and pathological condition (aggravation). Therefore, the syndrome coding results are: location (cervical spine) code 8, etiology (strain) code 7, pathological nature (blood stasis) code 8, and pathological condition (aggravation) code 7. Three target acupoints are selected: Fengchi (GB20), Tianzhu (BL10), and Dazhui (GV14). The obtained acupoint resistance data are 120Ω, 130Ω, and 140Ω, respectively. The mean μ of the acupoint resistance dataset is calculated to be 130Ω, and the standard deviation σ is calculated to be 5Ω. The Z-score of the resistance data for each acupoint is calculated. Fengchi acupoint: Z1=(120-130) / 5=-2, according to the discrete coding rule, the code is 2.

[0065] Tianzhu acupoint: Z2=(130-130) / 5=0, code is 5.

[0066] The Dazhui acupoint: Z3=(140-130) / 5=2, with the code 8.

[0067] The acupoint resistance coding results are: 2, 5, 8.

[0068] Extending the syndrome coding result to 48 bits yields [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0,0,0,0,0,0,7,8,7]. Concatenating this with the acupoint resistance coding result [2,5,8] results in [0,0,0,0]. The concatenated feature vector is then subjected to Min-Max normalization, mapping all dimensions to the [0,1] interval to obtain the final feature vector. Through these steps, the patient's TCM diagnostic information and acupoint resistance data can be integrated into a unified feature vector for subsequent disease diagnosis and analysis.

[0069] Preferably, this embodiment of the invention further includes a learning module. This learning module uses a three-level processing architecture of data layer, concept layer, and sentence layer to reinforce learning and expert feedback mechanisms, dynamically optimizing algorithm rule weights to achieve accurate processing and structured encoding of TCM medical record text data. The data layer is mainly responsible for extracting symptom entities from the medical record text and performing terminology standardization and feature weight labeling. In terms of terminology standardization: symptom entities are extracted using a Large Language Model (LLM), and terminology is standardized according to the standard version of the "TCM Terminology," such as mapping "numbness in the hands" to "numbness in the upper limbs." In terms of feature weight labeling: features such as time modification and inducing factors of symptoms are weighted, such as separating "stabbing pain" (main symptom), "worsening at night" (time modification), and "worsening with cold" (inducing factor feature) from "neck stabbing pain worsens at night and is aggravated by cold," and determining the encoding influence according to different four diagnostic methods and parsing rules. Furthermore, in order to guide the large language model to complete the TCM syndrome coding task, a detailed prompt word project was constructed: (1) Role anchoring: The large language model was set as a "TCM syndrome coding expert" so that it could perform task processing in a professional role. (2) Task objective definition: It was clarified that symptom extraction, syndrome type judgment and structured coding should be completed based on medical record text data. (3) Rule injection: The logical rules and coding standards of TCM syndromes were injected, and the mapping relationship between specific symptoms and meridians, viscera and syndrome types was listed. (4) Retrieval enhancement: Combined with the retrieval enhancement mechanism, knowledge graphs and historical medical records were dynamically spliced ​​as contextual information to enrich the reference basis for the model to process tasks. (5) Output format constraints: The output structure was formatted and constraints were imposed to clarify the field specifications and data types of the coding results. (6) Semantic restrictions: Semantic restriction clauses were set to limit the model to generate results within the scope of rules. (7) Chain thinking prompts: Chain thinking was used to prompt step by step, and the task was executed according to the logical chain of "entity extraction - rule matching - weight judgment - coding output". (8) Iterative optimization: Through iterative testing and optimization of prompts using multiple case text samples, the accuracy and consistency of the agent's response are improved. The concept layer is used to establish a knowledge base for the mapping of "symptom-syndrome elements" to determine the location of the disease and analyze its nature. In terms of determining the location of the disease: the location of the disease is determined by referring to the meridian circulation rules written in "Ling Shu·Jing Mai" and "Zhen Jiu Xue". For example, numbness in the little finger corresponds to the Hand Taiyang Meridian, which activates the bladder meridian coding position. In terms of analyzing the nature of the disease: a dynamic weight algorithm is used to analyze the nature of the disease, including determining the base value based on symptom characteristics, triggering the superposition rule through feature combination, and introducing correction coefficients based on environmental factors.The statement layer sets up a three-layer verification mechanism through dialectical logic to ensure that the coding conforms to the logic of traditional Chinese medicine theory, including: (1) Syndrome type constraint: setting a lower limit requirement for coding value for specific syndrome types, such as cold (24)≥5 / wind (39)≥3 / dampness (13)≥4; (2) Contradiction handling: handling contradictory situations in the coding, such as when qi deficiency (27)>5, qi stagnation (36)≤3; (3) Dynamic adjustment: adjusting or adding weights to certain features, such as tongue appearance feature weight ×1.5, acute symptoms +1 point.

[0070] Step 104: Initialize a preset number of cluster centers, iteratively calculate the distance between the feature vector and each cluster center, and assign the feature vector to the cluster to which the nearest cluster center belongs, until the preset requirements are met, and output the pattern judgment result of the feature vector; where each cluster represents a pattern.

[0071] In this embodiment of the invention, a preset number of cluster centers are initialized. Then, the distance between the feature vector and each cluster center is iteratively calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs. When a preset requirement is met, the syndrome judgment result corresponding to the feature vector is output. The clustering algorithm automatically classifies based on the distance between the feature vector and the cluster center, avoiding interference from subjective factors in syndrome judgment, reducing errors and uncertainties in human judgment, and making syndrome judgment more objective and accurate.

[0072] In one embodiment, a preset number of cluster centers are initialized, the distance between the feature vector and each cluster center is iteratively calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs, until a preset requirement is met, and the pattern judgment result of the feature vector is output. Specifically, this includes: selecting a preset number of feature vector samples as initial cluster centers in a pre-constructed sample set, calculating the distance from the feature vector to each initial cluster center, and assigning it to the cluster to which the nearest initial cluster center belongs; wherein, the sample set includes several common pattern clusters, and each common pattern cluster contains several feature vector samples; calculating the mean of the feature vectors in each cluster after assignment, and using the mean of the feature vectors as the updated cluster center of each cluster; calculating the change from each initial cluster center to the corresponding updated cluster center in the cluster; stopping iterative training when each change is less than a preset change threshold, or when the number of iterations reaches a preset number; and generating the pattern judgment result of the feature vector based on the pattern corresponding to the cluster to which the feature vector is assigned.

[0073] In this embodiment of the invention, the K-means++ algorithm is used to select a predetermined number (K ​​value) of initial cluster centers from a pre-constructed sample set. The determination of the K value combines the classification of common syndromes of cervical spondylosis in Traditional Chinese Medicine (TCM) theory with data characteristics. First, a preliminary range of K values ​​is set based on TCM classics, then quantitatively evaluated using the "elbow method" and "profile coefficient method," and finally fine-tuned based on expert feedback. The K-means++ algorithm selects initial centers using a probability distribution (proportional to the square of the distance from the sample point to the existing cluster center), avoiding local optima problems caused by random initialization and ensuring cluster stability. The Euclidean distance from the feature vector to each initial cluster center is calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs.

[0074] The mean of the feature vectors within each cluster is calculated and used as the updated cluster center. The change from each initial cluster center to the corresponding updated cluster center is calculated. Iterative training stops when each change is less than a preset threshold or the number of iterations reaches a preset number. Syndrome type determination results are generated based on the syndrome types corresponding to the clusters assigned to the feature vectors. After clustering, the statistical characteristics of the 57-dimensional feature vectors of each category are analyzed (e.g., the mean of the 48-dimensional syndrome code and the distribution of the 9-dimensional acupoint resistance code). These are then correlated with the TCM syndrome type definition to form a three-dimensional mapping relationship of "symptom characteristics - syndrome type category - acupoint resistance pattern," providing a basis for syndrome type determination and acupoint recommendation.

[0075] For example, symptom characteristic data (48-dimensional syndrome coding) and acupoint resistance pattern data (9-dimensional acupoint resistance coding) from 1000 patients with cervical spondylosis were collected to construct a sample set. Data preprocessing was performed, including removing missing values ​​and standardization. This was done according to the *Guidelines for Clinical Practice of Traditional Chinese Medicine Rehabilitation*. Cervical spondylosis is a common disease with five common syndrome types: wind-cold-dampness, phlegm-dampness obstruction, qi stagnation and blood stasis, qi and blood deficiency, and liver and kidney deficiency. The initial K-value range was set at 5. The sum of squared clustering errors (SSE) for different K values ​​(3-10) was calculated. It was found that when K=5, the decrease in SSE slowed significantly, forming an "elbow point." The silhouette coefficient of each sample was calculated, and the K-value with the largest average silhouette coefficient was selected. The results showed that K=5 resulted in the largest average silhouette coefficient. The clustering results were fed back to TCM experts, who believed that the clustering results were consistent with the actual clinical syndrome distribution and that no fine-tuning of the K-value was necessary; therefore, K=5 was finally determined. The K-means++ algorithm was used to select five initial cluster centers from the sample set. The Euclidean distance between the feature vector of each sample and the five centers was calculated, and the samples were assigned to the cluster with the smallest distance. The mean of the 57-dimensional features of all samples within each cluster was calculated to generate new cluster centers. The change between the new center and the previous center is calculated. Iteration stops when the change in all centers is less than 1e-6, for a total of 30 iterations. Based on the cluster assigned to the sample, the syndrome type of each sample is determined. For example, if a sample is assigned to a cluster corresponding to the Qi stagnation and blood stasis type, then the syndrome type of that sample is determined to be Qi stagnation and blood stasis. Furthermore, through the three-dimensional mapping relationship of "symptom-syndrome-acupoint," corresponding treatment acupoints are recommended for each syndrome type. For example, for the Qi stagnation and blood stasis type, acupoints such as Xuehai and Geshu are recommended.

[0076] In this embodiment of the invention, an acupoint-syndrome association information processing device is also provided, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the above-mentioned acupoint-syndrome association information processing method.

[0077] In this embodiment of the invention, a computer-readable storage medium is also provided, which includes a stored computer program, wherein the computer program controls the device where the computer-readable storage medium is located to execute the above-mentioned acupoint-syndrome association information processing method when it is running.

[0078] For example, a computer program can be divided into one or more modules, one or more of which are stored in memory and executed by a processor to complete the present invention. One or more modules can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the acupoint-syndrome association information processing device.

[0079] The acupoint-syndrome association information processing device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The acupoint-syndrome association information processing device may include, but is not limited to, a processor, memory, and display. Those skilled in the art will understand that the above-mentioned components are merely examples of the acupoint-syndrome association information processing device and do not constitute a limitation on the acupoint-syndrome association information processing device. It may include more or fewer components, or a combination of certain components, or different components. For example, the acupoint-syndrome association information processing device may also include input / output devices, network access devices, buses, etc.

[0080] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the acupoint-symptom correlation information processing device, connecting all parts of the device through various interfaces and lines.

[0081] The memory can be used to store computer programs and / or modules. The processor, by running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory, realizes various functions of the acupoint-syndrome correlation information processing device. The memory can mainly include a program storage area and a data storage area. The program storage area can store the operating system, at least one application program required for a function (such as sound playback function, text conversion function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, text message data, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0082] The module based on processing acupoint-syndrome correlation information, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. Those skilled in the art can understand and implement this without any inventive effort.

[0083] This invention provides a method for processing acupoint-syndrome association information. The Traditional Chinese Medicine (TCM) knowledge graph is trained based on several TCM classics and textbooks, integrating traditional TCM theory and practical experience. Processing medical record data through this graph allows the integration of classical TCM knowledge into modern data processing. Standardized medical record texts are mapped to corresponding symptom-sign nodes in the TCM knowledge graph. By leveraging different types of nodes and their semantic relationships within the knowledge graph, the potential information behind the medical record data is deeply mined, providing a richer and more accurate data foundation for subsequent syndrome differentiation. Preliminary weighting is applied to symptom and sign entities in the standardized medical record texts, and the weighting results are updated according to a preset strategy. This approach comprehensively considers the pathological degree of the disease and other relevant factors, more accurately quantifying the importance of syndrome element information in syndrome differentiation, avoiding weight bias caused by a single factor or subjective judgment, and improving the accuracy of syndrome differentiation. The entity weighting results are mapped to the syndrome coding system to obtain the syndrome coding results. Simultaneously, acupoint resistance data are discretely coded, and finally, the two are concatenated into a feature vector. This multi-dimensional feature extraction method integrates information from disease symptoms, signs, and acupoint resistance, providing a more comprehensive reflection of the patient's condition and offering richer feature evidence for syndrome differentiation. Using cluster analysis, the distance between the feature vector and the cluster center is iteratively calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs. Each cluster represents a syndrome. This data-driven cluster analysis method avoids the interference of subjective factors and can objectively classify syndromes based on the inherent characteristics of the data, improving the accuracy and reliability of syndrome differentiation.

[0084] Example 2 See Figure 2 , Figure 2 This is a schematic diagram of a module of an acupoint-syndrome correlation information processing device provided in one embodiment of the present invention. The acupoint-syndrome correlation information processing device provided in this embodiment includes: a data processing module 201, a weight assignment module 202, a vector generation module 203, and a syndrome type judgment module 204; The data processing module 201 is used to map the pre-processed standardized medical record text to the corresponding symptom-sign nodes in the pre-constructed TCM knowledge graph to obtain structured data. The TCM knowledge graph is trained by several TCM classics and textbooks and is used to describe different types of nodes and the semantic relationships between them. The structured data includes syndrome data that are directly related to the nodes. The weight assignment module 202 is used to initially assign weights to symptom entities and sign entities in standardized medical record texts according to the pathological degree, to obtain basic weighting results, and to update the basic weighting results according to the preset weight correction strategy to obtain entity weighting results. The vector generation module 203 is used to map the entity weighting result to the pre-built syndrome coding system to obtain the syndrome coding result, to perform discrete coding on the acupoint resistance data of the target acupoint to obtain the acupoint resistance coding result, and to concatenate the syndrome coding result and the acupoint resistance coding result to obtain the feature vector. The pattern determination module 204 is used to initialize a preset number of cluster centers, iteratively calculate the distance between the feature vector and each cluster center, and assign the feature vector to the cluster to which the nearest cluster center belongs, until the preset requirements are met, and output the pattern determination result of the feature vector; wherein, each cluster represents a pattern.

[0085] In one embodiment, the TCM knowledge graph is trained as follows: Core data from several TCM classics is extracted using an entity recognition model; the core data includes entity keywords related to symptoms, signs, syndromes, internal organs, meridians, and acupoints; symptoms, signs, syndromes, internal organs, meridians, and acupoints are used as core nodes, and each entity keyword is assigned to a subordinate core node as a basic node, with the resulting node topology serving as the initial TCM knowledge graph; semantic association types between various core nodes are defined; these semantic association types include subordinate relationships, correspondence relationships, indication relationships, contraindication relationships, and association relationships; attribute values ​​are assigned to each basic node to obtain the TCM knowledge graph; the attributes include the standard TCM terminology name corresponding to the node, the official definition of the node, and the association weight between nodes.

[0086] In one embodiment, the data processing module 201 is used to map the preprocessed standardized medical record text to the corresponding symptom-sign nodes in a pre-constructed TCM knowledge graph to obtain structured data. Specifically, this includes: performing word segmentation and entity recognition on the acquired medical record text data to extract TCM diagnostic and treatment terms from the medical record text data; wherein TCM diagnostic and treatment terms include symptoms, signs, syndrome types, and acupoints; standardizing and mapping the TCM diagnostic and treatment terms according to TCM terminology standards to obtain standardized medical record text; extracting features from the standardized medical record text to identify symptom and sign entities in the standardized medical record text; and selecting base nodes in the TCM knowledge graph using a fuzzy matching algorithm. A candidate entity set is established at the base node. The semantic similarity between the symptom entity and the sign entity and the description of each entity in the candidate entity set is calculated. Based on the semantic similarity calculation results and the association relationship of the TCM knowledge graph, joint reasoning is performed to obtain the symptom-sign nodes corresponding to the symptom entity and the sign entity in the TCM knowledge graph and associate them to obtain structured data. The structured data includes: the extracted original identification fields, the corresponding standard TCM terminology fields after mapping, the entity identifier used to describe the node entity ID in the TCM knowledge graph, the encoded weight field used to describe the weight value of the corresponding syndrome differentiation factor of the entity, and the associated syndrome type field used to describe the direct association with the entity.

[0087] In one embodiment, the weighting module 202 initially assigns weights to symptom entities and sign entities in the standardized medical record text according to the pathological degree, obtaining a basic weighting result. The basic weighting result is then updated according to a preset weight correction strategy to obtain the entity weighting result. Specifically, this includes: parsing the standardized medical record text to extract symptom entities, sign entities, and detailed description information; assigning initial weights to the diagnostic factors corresponding to each extracted entity according to the pathological degree, obtaining a basic weighting result; updating the basic weighting result according to the preset weight correction strategy to obtain the symptom-diagnostic factor weight for each diagnostic factor, and using the symptom-diagnostic factor weight as the entity weighting result; wherein the preset weight correction strategy is to adjust the weights of the basic weighting result using preset weight adjustment rules when the detailed description information contains a time coefficient, or when the symptom entities and sign entities constitute a prescribed combination of syndrome characteristics.

[0088] In one embodiment, after obtaining the entity weighting result, the method further includes calculating the correlation weight between the diagnostic factor and the syndrome type based on the entity weighting result. Specifically, this involves obtaining the weights of the symptom-diagnostic factors corresponding to all relevant symptom entities and sign entities under the target syndrome type, and calculating the correlation weight between the diagnostic factor and the syndrome type using a weighted average formula. The expression for the weighted average formula is as follows: ; In the formula, Evidence-based diagnostic factors The association weight with the certificate type m; The number of symptom-diagnostic factors. For the proof of type m, the first Each related symptom corresponds to a diagnostic factor. The weights; The total number of all relevant symptom entities under syndrome type m; For all relevant symptoms under syndrome type m, the corresponding syndrome differentiation factors The sum of weights.

[0089] In one embodiment, the vector generation module 203 maps the entity weighting result to a pre-constructed syndrome coding system to obtain the syndrome coding result. It then performs discrete coding on the acupoint resistance data of preset target acupoints to obtain the acupoint resistance coding result. Specifically, this includes: pre-constructing a syndrome coding system containing several diagnostic factors, wherein the coding type of the diagnostic factors includes location, etiology, nature, and severity; sequentially assigning weights to the diagnostic factors in the syndrome coding system according to the entity weighting result to obtain the syndrome coding result; acquiring acupoint resistance data of several target acupoints and constructing an acupoint resistance dataset based on the acupoint resistance data; calculating the distance between each acupoint resistance data and the mean of the acupoint resistance dataset, and coding the acupoint resistance data according to the distance to obtain the acupoint resistance coding result; concatenating the syndrome coding result and the acupoint resistance coding result, and normalizing the concatenated coding result to obtain a feature vector.

[0090] In one embodiment, the pattern determination module 204 is used to initialize a preset number of cluster centers, iteratively calculate the distance between the feature vector and each cluster center, and assign the feature vector to the cluster to which the nearest cluster center belongs, until a preset requirement is met, and output the pattern determination result of the feature vector. Specifically, it includes: selecting a preset number of feature vector samples as initial cluster centers in a pre-constructed sample set, calculating the distance from the feature vector to each initial cluster center, and assigning it to the cluster to which the nearest initial cluster center belongs; wherein, the sample set includes several common pattern clusters, and each common pattern cluster contains several feature vector samples; calculating the mean of the feature vectors in each cluster after assignment, and using the mean of the feature vectors as the updated cluster center of each cluster, calculating the change from each initial cluster center to the updated cluster center in the corresponding cluster, stopping iterative training when each change is less than a preset change threshold, or when the number of iterations reaches a preset number; and generating the pattern determination result of the feature vector according to the pattern corresponding to the cluster to which the feature vector is assigned.

[0091] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0092] This invention provides an acupoint-syndrome association information processing device. The Traditional Chinese Medicine (TCM) knowledge graph is trained based on several TCM classics and textbooks, integrating traditional TCM theory and practical experience. Processing medical record data through this graph allows the integration of classical TCM knowledge into modern data processing. Standardized medical record texts are mapped to corresponding symptom-sign nodes in the TCM knowledge graph. By leveraging different types of nodes and their semantic relationships within the knowledge graph, the potential information behind the medical record data is deeply mined, providing a richer and more accurate data foundation for subsequent syndrome differentiation. Preliminary weighting is applied to symptom and sign entities in the standardized medical record texts, and the weighting results are updated according to a preset strategy. This approach comprehensively considers the pathological degree of the disease and other relevant factors, more accurately quantifying the importance of syndrome element information in syndrome differentiation, avoiding weight bias caused by a single factor or subjective judgment, and improving the accuracy of syndrome differentiation. The entity weighting results are mapped to the syndrome coding system to obtain the syndrome coding results. Simultaneously, acupoint resistance data are discretely coded, and finally, the two are concatenated into a feature vector. This multi-dimensional feature extraction method integrates information from disease symptoms, signs, and acupoint resistance, providing a more comprehensive reflection of the patient's condition and offering richer feature evidence for syndrome differentiation. Using cluster analysis, the distance between the feature vector and the cluster center is iteratively calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs. Each cluster represents a syndrome. This data-driven cluster analysis method avoids the interference of subjective factors and can objectively classify syndromes based on the inherent characteristics of the data, improving the accuracy and reliability of syndrome differentiation.

[0093] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for processing acupoint-syndrome association information, characterized in that, include: The standardized medical record text obtained after preprocessing is mapped to the corresponding symptom-sign nodes in the pre-constructed TCM knowledge graph to obtain structured data. The TCM knowledge graph is trained by several TCM classics and textbooks and is used to describe different types of nodes and the semantic relationships between them. The structured data includes syndrome data that are directly related to the nodes. Based on the pathological degree, the symptom and sign entities in the standardized medical record text are initially weighted to obtain basic weighting results. The basic weighting results are then updated according to a preset weight correction strategy to obtain entity weighting results. The entity weighting result is mapped to a pre-constructed syndrome coding system to obtain the syndrome coding result. The acupoint resistance data of the target acupoint is discretely coded to obtain the acupoint resistance coding result. The syndrome coding result and the acupoint resistance coding result are concatenated to obtain the feature vector. A preset number of cluster centers are initialized, the distance between the feature vector and each cluster center is iteratively calculated, and the feature vector is assigned to the cluster to which the nearest cluster center belongs, until a preset requirement is met, and the pattern judgment result of the feature vector is output; wherein, each cluster represents a pattern.

2. The acupoint-syndrome association information processing method as described in claim 1, characterized in that, The TCM knowledge graph was trained using the following methods: Entity recognition models are used to extract core data from several TCM classics and textbooks; the core data includes entity keywords related to symptoms, signs, syndromes, internal organs, meridians, and acupoint types. Using symptoms, signs, syndromes, internal organs, meridians, and acupoints as core nodes, each entity keyword is assigned to a core node as a basic node, and the resulting node topology is used as the initial TCM knowledge graph. Define the semantic association types between various core nodes; wherein, the semantic association types include membership, correspondence, dominance, taboo, and association; The TCM knowledge graph is obtained by assigning attribute values ​​to each basic node; wherein the attribute includes the standard TCM terminology name corresponding to the node, the official definition of the node, and the association weight between nodes.

3. The acupoint-syndrome association information processing method as described in claim 1, characterized in that, The process of mapping the preprocessed standardized medical record text to the corresponding symptom-sign nodes in a pre-constructed TCM knowledge graph to obtain structured data specifically includes: The acquired medical record text data is segmented and entity recognized to extract traditional Chinese medicine (TCM) diagnostic and treatment terms from the medical record text data; the TCM diagnostic and treatment terms include symptoms, signs, syndromes, and acupoints. The standardized medical record text is obtained by standardizing and mapping the TCM diagnostic and treatment terms according to the TCM terminology standard. Feature extraction is performed on the standardized medical record text to identify symptom entities and sign entities in the standardized medical record text; A candidate entity set is established by selecting basic nodes in the TCM knowledge graph using a fuzzy matching algorithm, and the semantic similarity between the symptom entity and the sign entity and the description of each entity in the candidate entity set is calculated. Joint reasoning is performed based on the semantic similarity calculation results and the association relationship of the TCM knowledge graph to obtain the symptom-sign nodes corresponding to the symptom entity and the sign entity in the TCM knowledge graph and associate them to obtain the structured data; wherein, the structured data includes: the extracted original identification field, the corresponding standard TCM terminology field after mapping, the entity identifier used to describe the node entity ID in the TCM knowledge graph, the encoded weight field used to describe the weight value of the corresponding syndrome differentiation factor of the entity, and the associated syndrome type field used to describe the direct association with the entity.

4. The acupoint-syndrome association information processing method as described in claim 1, characterized in that, The process involves initially assigning weights to symptom and sign entities in the standardized medical record text based on the pathological severity to obtain a basic weighting result. Then, the basic weighting result is updated according to a preset weight correction strategy to obtain the entity weighting result. Specifically, this includes: The standardized medical record text is parsed to extract symptom entities, sign entities, and detailed description information. Based on the pathological degree, initial weights are assigned to the dialectical factors corresponding to each extracted entity to obtain the basic weighting results; The basic weighting result is updated according to a preset weight correction strategy to obtain the symptom-differentiation factor weight for each differentiation factor, and the symptom-differentiation factor weight is used as the entity weighting result; wherein, the preset weight correction strategy is to adjust the weight of the basic weighting result by using a preset weight adjustment rule when there is a time coefficient in the detailed description information, or when the symptom entity and the sign entity constitute a prescribed combination of syndrome features.

5. The acupoint-syndrome association information processing method as described in claim 4, characterized in that, After obtaining the entity weighting results, the process also includes calculating the correlation weight between the dialectical factors and the syndrome types based on the entity weighting results, specifically: Obtain the weights of the symptom-differentiation factors corresponding to all relevant symptom entities and sign entities under the target syndrome type, and calculate the correlation weight between the differentiation factors and the syndrome type using a weighted average formula; wherein, the expression of the weighted average formula is: ； In the formula, Evidence-based diagnostic factors The association weight with the certificate type m; The number of symptom-diagnostic factors. For the proof of type m, the first Each related symptom corresponds to a diagnostic factor. The weights; The total number of all relevant symptom entities under syndrome type m; For all relevant symptoms under syndrome type m, the corresponding syndrome differentiation factors The sum of weights.

6. The acupoint-syndrome association information processing method as described in claim 1, characterized in that, The step of mapping the entity weighting result to a pre-constructed syndrome coding system to obtain a syndrome coding result, and performing discrete coding on the acupoint resistance data of the preset target acupoints to obtain acupoint resistance coding results, specifically includes: A syndrome coding system containing several diagnostic factors is pre-constructed, wherein the coding types of the diagnostic factors include location of disease, etiology, nature of disease, and disease progression. Based on the entity weighting results, the diagnostic factors in the syndrome coding system are weighted sequentially to obtain the syndrome coding results; Obtain acupoint resistance data of several target acupoints, and construct an acupoint resistance dataset based on the acupoint resistance data; Calculate the distance between each acupoint resistance data point and the mean of the acupoint resistance dataset, and encode the acupoint resistance data point based on the distance to obtain the acupoint resistance encoding result; The syndrome encoding result and the acupoint resistance encoding result are concatenated, and the concatenated encoding result is normalized to obtain the feature vector.

7. The acupoint-syndrome association information processing method as described in claim 1, characterized in that, The process involves initializing a preset number of cluster centers, iteratively calculating the distance between the feature vector and each cluster center, assigning the feature vector to the cluster to which the nearest cluster center belongs, and continuing until a preset requirement is met. Finally, the result of the feature vector type judgment is output. Specifically, this includes: A predetermined number of feature vector samples are selected from a pre-constructed sample set as initial cluster centers. The distance from each feature vector to each initial cluster center is calculated, and the feature vector is assigned to the cluster to which the nearest initial cluster center belongs. The sample set includes several common pattern clusters, and each common pattern cluster contains several feature vector samples. Calculate the mean of the feature vectors within each cluster after allocation, and use the mean of the feature vectors as the updated cluster centers of each cluster. Calculate the change from each initial cluster center to the corresponding updated cluster center. Stop iterative training when each change is less than a preset change threshold or the number of iterations reaches a preset number. The evidence type judgment result of the feature vector is generated based on the evidence type corresponding to the cluster assigned by the feature vector.

8. An acupoint-syndrome correlation information processing device, characterized in that, include: The module includes a data processing module, a weight assignment module, a vector generation module, and a proof type judgment module. The data processing module is used to map the pre-processed standardized medical record text to the corresponding symptom-sign nodes in the pre-constructed TCM knowledge graph to obtain structured data. The TCM knowledge graph is trained by several TCM classics and textbooks and is used to describe different types of nodes and the semantic relationships between them. The structured data includes syndrome data that are directly related to the nodes. The weight assignment module is used to initially assign weights to symptom entities and sign entities in the standardized medical record text according to the pathological degree, to obtain basic weight assignment results, and to update the basic weight assignment results according to a preset weight correction strategy to obtain entity weight assignment results. The vector generation module is used to map the entity weighting result to a pre-built syndrome coding system to obtain the syndrome coding result, to perform discrete coding on the acupoint resistance data of the target acupoint to obtain the acupoint resistance coding result, and to concatenate the syndrome coding result and the acupoint resistance coding result to obtain the feature vector. The pattern determination module is used to initialize a preset number of cluster centers, iteratively calculate the distance between the feature vector and each cluster center, and assign the feature vector to the cluster to which the nearest cluster center belongs, until a preset requirement is met, and output the pattern determination result of the feature vector; wherein, each cluster represents a pattern.

9. A terminal device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the acupoint-syndrome association information processing method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the acupoint-syndrome association information processing method as described in any one of claims 1 to 7.

Images