A method for mining traditional Chinese medical records data

Abstract

The present invention discloses a method for data mining of TCM medical records. On the basis of collecting a large number of TCM medical records, a TCM medical record knowledge base based on granular computing is constructed. The research uses association rule mining algorithms and clustering algorithms to efficiently mine TCM medical records. The method solves the problems of fuzzy and complicated terminology expressions in massive TCM medical records, as well as the low efficiency and limited application of data mining algorithms caused by the fuzzy and complicated terminology expressions in massive TCM medical records. The method broadens the application of data mining technology in the field of TCM and provides theoretical and technical support for promoting the development of related industries.

Description

Technical Field

[0001] The present invention relates to the field of traditional Chinese medicine, and in particular to a method for mining traditional Chinese medicine medical case data. Background Art

[0002] TCM medical records are also called case records, which are continuous records of doctors’ dialectical treatment, legislation, and prescription of drugs when treating diseases. Medical records record the process of TCM clinical doctors trying to treat diseases based on dialectical treatment, reflecting the doctors’ thinking activities, academic views, and rich clinical experience, and provide valuable research materials for the innovation and development of TCM theory. Studying medical records and learning relevant knowledge from them can not only enrich and deepen theoretical knowledge, but also improve clinical diagnosis and treatment level, broaden horizons, and inspire ideas. Therefore, medical records have great guiding significance in TCM clinical research and practice for those engaged in TCM clinical, teaching, and scientific research.

[0003] However, there are some vague terms, complicated titles and unclear meanings in TCM medical records, which makes TCM medical record data mining inefficient and not widely used. Therefore, establishing an efficient and scientific TCM medical record database and using it to mine massive medical record data is a problem to be solved. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention aims to provide a method for mining TCM medical records data, which can broaden the application of data mining technology in the field of TCM and provide support for the development of related industries.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for mining TCM medical records data, comprising the following steps:

[0007] S1: Collect massive patient medical records;

[0008] S2: Construct a TCM medical records knowledge base based on particle computing;

[0009] S3: using an information extraction algorithm to normalize the patient symptom information recorded in the TCM medical records knowledge base document;

[0010] S4: Cluster analysis of standardized patient symptom information of the same type of symptoms was performed using a text clustering algorithm;

[0011] S5: Through the association rule mining algorithm, the multiple associations between symptoms and prescriptions, symptoms and syndromes, syndromes and prescriptions, and Chinese medicine combinations in TCM medical records are mined.

[0012] Furthermore, the specific steps of constructing a TCM medical records knowledge base based on granular computing described in step S2 include:

[0013] S21: Constructing the concept set of TCM medical records;

[0014] S22: combining the concept set, using the Chinese lexical analysis system ICTCLAS word segmentation system to segment the TCM medical records;

[0015] S23: appropriately granulate the TCM medical record data, and synthesize and decompose the particles;

[0016] S24: Extracting ontology concepts and relationships through concept acquisition algorithms and relationship extraction algorithms based on granular computing;

[0017] S25: The extracted ontology concepts are organized into a knowledge base.

[0018] Furthermore, the specific steps of constructing the concept set of TCM medical records in step S21 include:

[0019] S211: The related concepts generated by the concept are defined as concept set, concept, term set, and term, and the relationship between them is shown in formula (1), which is used to represent the set of the same term for different concepts:

[0020] It represents m different terms t,

[0021]

[0022]

[0023] in, represents a set of l words with the same meaning but different terms, and C is the corpus;

[0024] S212: Using formula (2) to set the threshold, and by optimizing the threshold, construct a concept set of TCM medical records. Formula (2) is expressed as:

[0025]

[0026] Among them, DMD t，k The term t t，k For D k The domain membership, CF t，k D k Domain feature words, 0≤α≤1, where α is the control and The random factor is set to the optimal value of α according to different background expectations. is the standardized domain relevance, are standardized domain consistency, and their definitions are as follows:

[0027]

[0028] Among them, t is a term, k is a domain identifier, and nom represents balanced processing.

[0029] Furthermore, the information extraction algorithm described in step S3 is processed based on semantic feature analysis, and its specific steps include:

[0030] S31: Combine semantic dictionary and TCM medical records knowledge base to perform semantic disambiguation and synonym merging;

[0031] S32: Using a redundancy elimination module to eliminate redundancy in sentences in TCM medical records.

[0032] Furthermore, the redundancy elimination module is performed based on sentence similarity, and the calculation formula (2) for iteratively updating the sentence similarity is:

[0033]

[0034] Among them, R(a,b) is a, b is the similarity of the two compared objects, C is a damping coefficient, which is a constant, C∈[0,1]; if a≠b, and a or b has no neighbors, then R(a,b)=0.

[0035] Furthermore, the text clustering algorithm in step S4 adopts text clustering technology, based on semantic text similarity measurement, combined with the ontology TCM medical records knowledge base, and analyzes the information of patients with the same symptoms according to the text similarity formula. The specific steps include:

[0036] S41: Divide the TCM case document collection in the TCM case knowledge base into clusters;

[0037] S42: Grouping normalized patient symptom information of similar symptoms into a cluster;

[0038] S43: Use vector space model to represent the symptom information of each patient;

[0039] S44: Construct a feature matrix based on symptom information;

[0040] S45: Use cosine distance to measure the similarity between texts, modify the cluster center through the improved k-means algorithm, reduce the number of iterations, and output the optimal clustering result.

[0041] Furthermore, the association rule mining algorithm described in step S5 mines multiple associations among symptoms, prescriptions and medicines, and syndrome types in TCM medical records based on the ontology TCM medical record knowledge base.

[0042] The beneficial effects of the present invention are:

[0043] 1. Granular computing is a new concept and computing paradigm for information processing. In this invention, the characteristics of TCM medical case data, such as vague expressions, complicated names, and unclear meanings, are taken into consideration. A TCM medical case knowledge base is constructed by using a concept acquisition algorithm based on granular computing and extracting multi-layer associations between concepts, thus laying a foundation for subsequent data mining.

[0044] 2. The present invention proposes the concept of information extraction when normalizing patient symptom information. Traditional information extraction uses sentence extraction algorithms based on shallow feature analysis and semantic feature analysis to extract abstract sentences. However, due to the data characteristics of TCM medical records, this traditional method is prone to ambiguity in the results. Therefore, the present invention uses semantic feature analysis for processing, combines the TCM medical record knowledge base, and proposes semantic disambiguation and synonym merging strategies, which effectively improves the standardization and accuracy of patient symptom information processing;

[0045] 3. Due to the complexity of the data structure of massive TCM medical records, the present invention studies the latest trends of association rules, applies swarm intelligence optimization algorithm and granular computing to data mining of TCM medication rules, establishes the corresponding relationship between swarm intelligence optimization algorithm and association rule mining, and proposes a perturbation strategy guided by small-probability normative knowledge to solve the problem that swarm intelligence algorithm is prone to fall into local optimality, enhances the ability of the algorithm to jump out of local optimality, and improves the operation efficiency of the algorithm. Therefore, the use of the mining algorithm in the present invention can effectively reduce redundancy and improve the timeliness of algorithm completion;

[0046] In summary, the present invention builds a TCM medical case knowledge base based on granular computing on the basis of collecting a large number of medical cases, studies the efficient data mining of TCM medical cases through association rule mining algorithms and cluster analysis, solves the problems of fuzzy and complicated expressions in massive TCM medical cases, and the low efficiency and limited application of data mining algorithms. It broadens the application of data mining technology in the field of traditional Chinese medicine and provides theoretical and technical support for promoting the development of related industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A flowchart of a method for mining TCM medical case data provided by the present invention. DETAILED DESCRIPTION

[0048] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with embodiments.

[0049] Reference Figure 1 It can be seen that a method for mining TCM medical records data includes the following steps:

[0050] S1: Taking cardiovascular medical records as an example, 10,000 medical records of cardiovascular patients were collected through manual sorting and online platforms. At the same time, a dictionary of Chinese medical records terms, a dictionary of Chinese medicine synonyms, a dictionary of symptom synonyms, a dictionary of syndrome synonyms, a dictionary of stop words, and a dictionary of Chinese medicine properties and flavors;

[0051] S2: Construct a TCM medical records knowledge base based on particle computing;

[0052] S3: using an information extraction algorithm to normalize the patient symptom information recorded in the TCM medical records knowledge base document, so as to improve efficiency while improving the accuracy of normalization;

[0053] S4: Analyze patients with the same symptoms through text clustering algorithm to study the dialectical law between symptoms and syndrome types;

[0054] S5: Through the association rule mining algorithm, the multiple associations between symptoms and prescriptions, symptoms and syndromes, syndromes and prescriptions, and Chinese medicine combinations in TCM medical records are mined.

[0055] Furthermore, the specific steps of constructing a TCM medical records knowledge base based on granular computing described in step S2 include:

[0056] S21: Construct the concept set of TCM medical records. The concept set is the basis for constructing the knowledge base and determines whether the ontology after construction can fully describe the conceptual model of the TCM medical records.

[0057] S22: combining the concept set, using the Chinese lexical analysis system ICTCLAS word segmentation system to segment the TCM medical records;

[0058] S23: appropriately granulate the TCM medical record data, and synthesize and decompose the particles;

[0059] S24: Extracting ontology concepts and relationships through concept acquisition algorithms and relationship extraction algorithms based on granular computing;

[0060] S25: The extracted ontology concepts are organized into a knowledge base.

[0061] Furthermore, the specific steps of constructing the concept set of TCM medical records in step S21 include:

[0062] S211: The related concepts generated by the concept are defined as concept set, concept, term set, and term, and the relationship between them is shown in formula (1), which is used to represent the set of the same term for different concepts:

[0063] represents m different terms.

[0064]

[0065] in, represents a set of l words with the same meaning but different terms, and C is the corpus;

[0066] S212: Using formula (2) to set the threshold, and by optimizing the threshold, construct a concept set of TCM medical records, wherein the setting of the threshold is the key to constructing the concept set, and formula (2) is expressed as:

[0067]

[0068] Among them, DMD t，k The term t t，k For D k The domain membership, CF t，k D k Domain feature words, 0≤α≤1, where α is the control and The random factor is set to the optimal value of α according to different background expectations. is the standardized domain relevance, are standardized domain consistency, and their definitions are as follows:

[0069]

[0070] Among them, t is a term, k is a domain identifier, and nom represents balanced processing.

[0071] That is, the TCM medical case text set is used as the foreground corpus set, and the non-TCM medical case text set is used as the background corpus set. By optimizing the threshold of the threshold formula (2), the concept set of TCM medical cases is constructed.

[0072] Furthermore, the traditional information extraction uses a sentence extraction algorithm based on shallow feature analysis and semantic feature analysis to extract abstract sentences, which is prone to ambiguity. The information extraction algorithm described in step S3 of the present invention uses a processing based on semantic feature analysis, and its specific steps include:

[0073] S31: Combine semantic dictionary and TCM medical records knowledge base to perform semantic disambiguation and synonym merging on TCM medical records;

[0074] S32: Using a redundancy elimination module to eliminate redundancy in sentences in TCM medical records.

[0075] Furthermore, the redundancy elimination module is performed based on sentence similarity, and the calculation formula (2) for iteratively updating the sentence similarity is:

[0076]

[0077] Among them, R(a, b) is a, b is the similarity of the two compared objects, C is a damping coefficient, which is a constant, C∈[0,1]; if a≠b, and a or b has no neighbors, then R(a, b)=0.

[0078] Text clustering technology refers to dividing a document collection into clusters so that the documents in the same cluster are as similar as possible, while the similarity between clusters is as small as possible. Cluster analysis is performed on standardized patient symptom information, and similar patient symptom information is divided into a cluster to analyze the dialectical law between symptoms and syndrome types.

[0079] When studying the dialectical law between symptoms and syndromes through text clustering technology, the design of text similarity measurement method is the key technology in this link. Most traditional text similarity measurement methods use the TF-IDF method to model text as a word frequency vector, and use methods such as cosine similarity measurement to calculate the similarity between texts. These methods ignore the semantic information of the terms in the text. This study uses a semantic-based text similarity measurement method, with the help of an ontology-based Chinese medicine case knowledge base, to expand semantically similar terms in the traditional word frequency vector, further increase the dimension of the symptom text representation vector, and finally use text similarity to perform cluster analysis on symptom information.

[0080] Furthermore, the text clustering algorithm in step S4 adopts text clustering technology, based on semantic text similarity measurement, combined with the ontology TCM medical records knowledge base, and analyzes the information of patients with the same symptoms according to the text similarity formula. The specific steps include:

[0081] S41: Divide the TCM case document collection in the TCM case knowledge base into clusters;

[0082] S42: Grouping normalized patient symptom information of similar symptoms into a cluster;

[0083] S43: Vector Space Model (VSM) is used to represent the symptom information of each patient;

[0084] S44: Construct a feature matrix based on symptom information;

[0085] S45: Use cosine distance to measure the similarity between texts, modify the cluster center through the improved k-means algorithm, reduce the number of iterations, and output the optimal clustering result.

[0086] Furthermore, the association rule mining algorithm described in step S5 mines multiple associations among symptoms, prescriptions and medicines, and syndrome types in TCM medical records based on the ontology TCM medical record knowledge base.

[0087] When mining the relationship between symptoms, syndromes, and prescriptions in patients' medical records through association rule mining algorithms, improving time efficiency and space efficiency and reducing redundancy are the key issues solved in the present invention. By studying the latest trends of association rules, group intelligence optimization algorithms and granular computing are used for data mining of traditional Chinese medicine rules;

[0088] Example:

[0089] First, 10,000 medical records of cardiovascular patients from famous traditional Chinese medicine practitioners were collected and sorted out from the outpatient medical record system;

[0090] Secondly, the collected medical records of cardiovascular patients are used to build a TCM medical record knowledge base based on granular computing, and then the medical records in the TCM medical record knowledge base are standardized using information extraction algorithms to obtain standardized patient medical record information;

[0091] Secondly, the text clustering algorithm was used to cluster the normalized patient information with the same cardiovascular symptoms, and the medication rules and experience of famous traditional Chinese medicine practitioners were analyzed;

[0092] Finally, the association rule mining algorithm is used to mine the multiple associations between symptoms and prescriptions, symptoms and syndromes, syndromes and prescriptions, and Chinese medicine compatibility. Through the derived associations, effective prescriptions are obtained to treat patients and provide guidance for the formulation of Chinese patent medicines.

[0093] In summary, the present invention uses a TCM medical case data mining algorithm, from the collection, organization, and standardized processing of medical cases of famous old TCM doctors to data mining and experience rule mining. On the basis of fully utilizing the standardized outpatient medical case text data, it deeply analyzes the medication rules and experience of famous old TCM doctors, obtains effective empirical prescriptions, provides guidance for the formulation of Chinese patent medicines, develops Chinese patent medicines and in-hospital preparations, and creates economic benefits.

[0094] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for mining TCM medical records data, characterized in that: The following steps are involved: S1: Collect massive patient medical records; S2: Construct a TCM medical records knowledge base based on particle computing; S3: using an information extraction algorithm to normalize the patient symptom information recorded in the TCM medical records knowledge base document; S4: Cluster analysis of standardized patient symptom information of the same type of symptoms was performed using a text clustering algorithm; S5: Use association rule mining algorithms to mine multiple associations between symptoms and prescriptions, symptoms and syndromes, syndromes and prescriptions, and Chinese medicine combinations in TCM medical records; The specific steps of constructing the TCM medical records knowledge base based on particle computing described in step S2 include: S21: Constructing the concept set of TCM medical records; S22: combining the concept set, using the Chinese lexical analysis system ICTCLAS word segmentation system to segment the TCM medical records; S23: appropriately granulate the TCM medical record data, and synthesize and decompose the particles; S24: Extracting ontology concepts and relationships through concept acquisition algorithms and relationship extraction algorithms based on granular computing; S25: The extracted ontology concepts are combined into a knowledge base; The specific steps of constructing the concept set of TCM medical records in step S21 include: S211: The related concepts generated by the concept are defined as concept sets, and the relationship between them is shown in formula (1), which is used to represent the set of the same term for different concepts: , , which represents m different terms t, ； （1）， in, Indicates that A set with the same meaning but different terminology, For corpus; S212: Formula (2) is used to set the threshold, and by optimizing the threshold, a concept set of TCM medical records is constructed. Formula (2) is expressed as: （2）， in, For term for The domain affiliation of for Domain feature words, ,in For control and The random factors are set according to different background expectations. Optimal value, is the standardized domain relevance, are standardized domain consistency, and their definitions are as follows: in For terminology, To identify a certain field, Indicates balanced processing; The information extraction algorithm described in step S3 is based on semantic feature analysis, and its specific steps include: S31: Combine semantic dictionary and TCM medical records knowledge base to perform semantic disambiguation and synonym merging; S32: using a redundancy elimination module to eliminate redundancy in sentences in TCM medical records; The redundancy elimination module is based on sentence similarity, and the calculation formula for iterative update of sentence similarity is: ； in, for is the similarity between the two compared objects, is a damping coefficient, which is a constant. ;if ,and or If there are no neighbors, ; The text clustering algorithm in step S4 uses text clustering technology, semantic-based text similarity measurement, combined with the ontology TCM medical case knowledge base, and analyzes the information of patients with the same symptoms according to the text similarity formula. The specific steps include: S41: Divide the TCM case document collection in the TCM case knowledge base into clusters; S42: Grouping normalized patient symptom information of similar symptoms into a cluster; S43: Use vector space model to represent the symptom information of each patient; S44: Construct a feature matrix based on symptom information; S45: Use cosine distance to measure the similarity between texts, and improve The algorithm corrects the cluster center, reduces the number of iterations, and outputs the optimal clustering result.

2. A method for mining TCM medical records data according to claim 1, characterized in that: The association rule mining algorithm described in step S5 mines multiple associations among symptoms, prescriptions and medicines, and syndrome types in TCM medical records based on the ontology TCM medical records knowledge base.

Images