Method and device for mining drug resistance information

Abstract

The invention belongs to the technical field of information mining integration, and particularly relates to a mining method and device for drug resistance information. The method comprises the following steps: S1, respectively acquiring entities and entity categories in medical data of a medicine; s2, constructing a tree map of each medical data according to a preset entity category hierarchical relationship; s3, constructing a standard entity dictionary; s4, standardizing entities in the tree maps according to a standard entity dictionary; s5, combining the standardized tree-shaped maps to obtain a classification system; s6, searching the drug resistance information text of the drug in the multi-source data, and extracting an entity from the obtained drug resistance information text; and after the extracted entity is matched with the classification system of the drug resistance information to obtain a position node, establishing a connection relationship between the position node and a drug resistance information text corresponding to the entity. According to the method, comprehensive and reliable drug resistance information mining is realized, and meanwhile, the implementation efficiency of drug resistance information mining is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of information mining and integration, and in particular relates to a method and device for mining drug resistance information. Background Art

[0002] Drug resistance, also known as drug resistance, refers to the tolerance of microorganisms, parasites and tumor cells to the effects of chemotherapy drugs. Once drug resistance occurs, the chemotherapy effect of the drug is significantly reduced.

[0003] With the development of precision medicine, pharmaceutical companies have increasingly precise research needs on drug efficacy and drug resistance in new drug clinical trials or new drug development, and need to conduct in-depth research in more subdivided therapeutic areas. However, existing drug resistance information often cannot meet this high-granularity, multi-dimensional demand.

[0004] At present, there are widespread problems in the effective mining and utilization of drug resistance information in the global medical and health field, as follows. First, drug resistance information is dispersed and massive. Drug resistance information is widely distributed in data from a variety of different sources, such as clinical guidelines, clinical trial data, and medical literature; and different data sources have different descriptions of drug resistance, classification standards, etc., and lack of unified data standards and specifications, making information integration and comparison difficult. This makes it difficult for clinicians or medical researchers to quickly screen out valuable information from such a huge ocean of information. Second, the description of drug resistance information is broad. The description of drug resistance is diverse and broad, and there is a lack of unified standards or terminology systems, which increases the difficulty of understanding and integrating information. Third, the efficiency of drug resistance information retrieval is low. With the continuous deepening of medical research and the continuous accumulation of data, the amount of data related to drug resistance has exploded, and traditional information retrieval methods are difficult to efficiently process such a large amount of data; and existing information retrieval tools often lack special optimization for drug resistance information, and cannot quickly and accurately extract valuable information from massive data.

[0005] In summary, at present, the effective mining and utilization of drug resistance information has the problems of difficult information integration, low information retrieval efficiency, and mismatch between precision medicine needs and existing information. In order to solve these problems, it is necessary to establish unified data standards and specifications, develop efficient information retrieval tools, and build a complete drug resistance information classification system.

[0006] Therefore, how to achieve comprehensive and reliable drug resistance information mining while effectively improving the efficiency of drug resistance information mining has become a problem that needs to be solved urgently. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for mining drug resistance information, which not only realizes comprehensive and reliable drug resistance information mining, but also effectively improves the efficiency of drug resistance information mining.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solutions: A method for mining drug resistance information comprises the following steps: S1. For each medical data of a drug, entities and corresponding entity categories are obtained respectively; wherein the entity categories include drugs and drug-resistant diseases; S2. According to the preset entity category hierarchical relationship, the entity categories of the entities in the medical data are connected hierarchically to obtain a tree diagram of each medical data; S3, constructing a standard entity dictionary, in which a standard expression method of each entity is set; S4. Standardize the expression of entities in the tree diagram of each medical data according to the standard entity dictionary; S5, merging the standardized dendrograms to obtain a classification system of drug resistance information of the drug; S6. Retrieve the drug resistance information text of the drug in the multi-source data, and extract entities from the obtained drug resistance information text; match the extracted entities with the classification system of drug resistance information, obtain the position node of the entity in the classification system, and establish a connection relationship between the position node and the drug resistance information text corresponding to the entity, so as to form the drug resistance information mining result of the drug; S7. Based on the drug resistance information mining result, analyze and obtain a drug resistance information report of the drug.

[0009] Compared with the prior art, the present invention has the following beneficial effects: 1. Using this method, for centralized drugs, after obtaining the entities and corresponding entity categories in the medical data, the identified entities will be hierarchically connected according to the preset entity category hierarchical relationship to form a tree map. This hierarchical structure helps to understand and represent the complex relationships between entities, and facilitates subsequent standardization and merging operations. After that, this method will construct a standard entity dictionary to unify and standardize the expression of entities in medical data. Through comparison and replacement, the entity expression in the tree map is matched with the standard entity dictionary to eliminate information redundancy and confusion caused by inconsistent expressions. In this way, through the construction of a standard entity dictionary and a hierarchical relationship, the accuracy and consistency of drug resistance information are ensured, and the possibility of information errors and omissions is reduced.

[0010] 2. Construct a standard entity dictionary. This method will standardize the expression of entities in the tree maps of each medical data. After that, the standardized tree maps will be merged to form a complete drug resistance information classification system. This system not only contains the association information between drugs and drug-resistant diseases, but also reflects the hierarchy and logical relationship between them. In this way, integrating multi-source and heterogeneous medical data into a complete drug resistance information classification system can provide comprehensive and systematic reference information for clinicians and medical researchers. By constructing a refined drug resistance information classification system, pharmaceutical companies are supported to conduct research in more subdivided therapeutic areas in new drug clinical trials or new drug development, and promote the development of precision medicine.

[0011] 3. After obtaining the classification system of drug resistance information, this method will retrieve the drug resistance information text of specific drugs in multi-source data, and connect these texts with the classification system of drug resistance information through entity extraction and matching technology. In this way, the rapid positioning and classification of newly emerging drug resistance information is achieved. When the staff queries a node in the classification system, they can also obtain the information expressed by the drug resistance information text that is connected to the node, which facilitates their full mining of the drug resistance information. Afterwards, the generated drug resistance information report can provide strong support for clinicians' medication decisions, reduce the risk of drug resistance, and improve treatment effects.

[0012] In summary, this method not only realizes comprehensive and reliable drug resistance information mining, but also effectively improves the efficiency of drug resistance information mining.

[0013] Preferably, in S1, entities in the medical data and corresponding entity categories are obtained through natural language processing technology.

[0014] Such a setting can greatly shorten the time of mining drug resistance information and improve work efficiency through automated processing and standardized processes.

[0015] Preferably, in S4, the process of standardizing the expression of entities in the tree diagram includes: for any entity in the tree diagram, calculating the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary; then, taking the candidate standard entity with the highest semantic relevance as the standard entity of the entity, replacing the entity, and achieving standardization of the entity.

[0016] In this way, by replacing the entities in the tree map with the corresponding items in the standard entity dictionary, it is ensured that all mentioned entities are expressed in a unified and standardized way. This eliminates the confusion and misunderstanding caused by different data sources using different terms or expressions, and improves the consistency and comparability of the data. In addition, the calculation of semantic relevance is based on a deep understanding of the meaning of the entity, not just a literal match. Therefore, by selecting the candidate standard entity with the highest semantic relevance for replacement, the true meaning of the entity can be more accurately reflected, reducing information errors caused by mismatches or misunderstandings.

[0017] Preferably, the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary is calculated by a preset relevance calculation model.

[0018] This setting can ensure the accuracy and efficiency of drug resistance information mining by calculating the semantic relevance between entities and candidate standard entities in the standard entity dictionary through a preset relevance calculation model. It not only improves the accuracy and robustness of matching, but also supports the identification of complex relationships and optimizes calculation efficiency, promoting the automation and intelligence of the entire mining process.

[0019] Preferably, the entity categories include drugs, drug-resistant diseases, test types, drug-resistant mechanism types, and drug-resistant key molecules.

[0020] Preferably, the hierarchy of each entity category is from high to low: drug, drug-resistant disease, test type, drug-resistant mechanism type, and drug-resistant key molecule.

[0021] Such a setting can form a structured information representation by dividing the drug resistance information into different entity categories and organizing them according to hierarchical relationships. This representation makes the information clearer and more orderly, and easier to understand and analyze. In addition, the clear hierarchical relationship allows more accurate positioning of the required information level when retrieving drug resistance information. Moreover, the hierarchical relationship not only facilitates information retrieval, but also provides a basis for in-depth analysis. By comparing the associations and differences between different levels, the inherent laws and trends of drug resistance information can be revealed, providing strong support for clinical decision-making and drug development.

[0022] Preferably, in S5, the process of merging the standardized tree graphs includes: taking any standardized tree graph as a benchmark, traversing other standardized tree graphs of the same drug, checking whether there are drug-resistant diseases in the traversed tree graphs that are not in the benchmark tree graph, and if so, adding the subtree of the drug-resistant disease to the benchmark tree graph; if not, continuing to check the test type hierarchy under the drug-resistant disease, and so on until all leaf nodes are traversed.

[0023] Such a setting, through the merging process, can ensure that the baseline tree map contains all relevant drug-resistant diseases and their detailed information. Even if some drug-resistant diseases or specific information are missing in an initial tree map, as long as they exist in other tree maps, they can be added to the baseline map, thereby improving the integrity of the information. The merged tree map supports information retrieval across different data sources. Users can find information about the drug resistance of a drug in different diseases and different test conditions in a unified map without having to access multiple data sources separately. This greatly improves the efficiency and convenience of information retrieval.

[0024] The present invention also provides a device for mining drug resistance information, which is used in the above-mentioned method for mining drug resistance information, and comprises an entity acquisition unit, a graph construction unit, a graph standardization unit, a graph merging unit and an information mining unit; The entity acquisition unit is used to acquire entities and entity categories in each medical data; The graph construction unit is used to perform corresponding hierarchical connections on the entity categories of the entities in the medical data based on the acquired entities and their entity categories, as well as the preset hierarchical relationships of the entity categories, to obtain a tree graph of each medical data; The graph standardization unit is used to construct a standard entity dictionary and is also used to standardize each tree graph according to the standard entity dictionary; The map merging unit is used to merge the standardized tree maps to obtain a classification system of drug resistance information; The information mining unit is used for mining drug resistance information based on the classification system of drug resistance information.

[0025] While realizing comprehensive and reliable drug resistance information mining, the device also effectively improves the efficiency of drug resistance information mining.

[0026] Preferably, the process of the graph standardization unit standardizing each tree graph according to the standard entity dictionary includes: for any entity in the tree graph, calculating the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary; then, taking the candidate standard entity with the highest semantic relevance as the standard entity of the entity, replacing the entity, and achieving standardization of the entity.

[0027] Preferably, the process of drug resistance information mining by the information mining unit includes: retrieving the drug resistance information text of the drug in multi-source data, and extracting entities from the obtained drug resistance information text; matching the extracted entities with the classification system of drug resistance information, and after obtaining the position node of the entity in the classification system, establishing a connection relationship between the position node and the drug resistance information text corresponding to the entity, forming the drug resistance information mining result of the drug; and based on the drug resistance information mining result, analyzing to obtain the drug resistance information report of the drug. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to make the purpose, technical solution and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which: Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a logic block diagram of the device in the present invention. DETAILED DESCRIPTION

[0029] The following is a further detailed description through specific implementation methods: Example: like Figure 1 As shown, this embodiment discloses a method for mining drug resistance information, comprising the following steps: S1. For each medical data of a drug, obtain the entities and corresponding entity categories therein respectively; wherein the entity categories include drugs, drug-resistant diseases, test types, drug-resistant mechanism types, and drug-resistant key molecules. In specific implementation, natural language processing technology is used to obtain entities and corresponding entity categories in medical data. In this way, through automated processing and standardized processes, the time for mining drug resistance information can be greatly shortened and work efficiency can be improved.

[0030] The drug is the target drug for medical data research.

[0031] Drug-resistant diseases are diseases that are resistant to the above-mentioned drugs, such as acute myeloid leukemia, colon cancer, kidney cancer, breast cancer, etc.

[0032] There are three types of tests: clinical trial data, in vivo test data, and in vitro test data. Clinical trial data refers to clinical trial data conducted on humans or clinical evidence generated during the use of medical products by patients; in vivo test data refers to pharmacology test data conducted in animals, such as pharmacology test data conducted on mice, rats, rabbits and other animals; in vitro test data refers to data from in vitro cell tests or molecular biology tests.

[0033] There are five types of drug resistance mechanisms: genetic changes, drug structure modifications, changes in drug transport modes, target mutations, and abnormal activation of signal pathways. ① Genetic changes include gene mutations, chromosomal aberrations, changes in DNA methylation patterns, changes in non-coding RNA regulatory patterns, and other genetic changes that lead to drug resistance. ② Drug structure modifications are mutations or abnormal expressions of various drug-metabolizing enzymes that cause the drug structure to be modified and lose activity, leading to drug resistance. ③ Changes in drug transport modes are changes in the absorption, excretion, or transport modes of drugs in the body that lead to reduced concentrations at specific sites, leading to drug resistance. ④ Target mutations are changes in the spatial conformation of drug treatment targets or changes in the amino acids in the composition that lead to decreased affinity for binding to drugs, leading to drug resistance. ⑤ Abnormal activation of signal pathways refers to the activation of certain signal pathways that provide survival advantages for tumor cells or other pathogens, and lead to drug resistance through the disorder of key activities such as proliferation, differentiation, autophagy, or apoptosis.

[0034] Key molecules of drug resistance refer to the key molecules that cause drug resistance, which can be DNA, RNA, protein, enzyme or certain cells, such as EGFR (epidermal growth factor receptor), FLT3 (FMS-like tyrosine kinase 3), CD20 + B cells, etc.

[0035] The preset categories of the above multiple dimensions can classify drug resistance information from different angles. Accordingly, after determining the medical data to be mined, the entities of the above preset categories contained in each medical data can be obtained, and the relationship between the entities of the above preset categories contained in each medical data can be determined based on the representation structure of each medical data.

[0036] The method of obtaining entities contained in medical data can be achieved through entity recognition, rule matching or other methods that can extract entities. The relationship between the above-mentioned preset categories of entities contained in medical data is for each entity contained in a single medical data, and the representation structure of medical data can depend on the data type of medical data. For example, for text-type medical data, its representation structure can be determined based on the outline chapters in the text, and for image or table-type medical data, its representation structure can be determined based on the positional relationship between the text areas in the image or table.

[0037] S2. According to the preset entity category hierarchical relationship, the entity categories of the entities in the medical data are connected hierarchically to obtain a tree diagram of each medical data.

[0038] In specific implementation, the levels of each entity category are drugs, drug-resistant diseases, test types, drug-resistant mechanism types, and drug-resistant key molecules from high to low. That is, drugs are the first level, and drug-resistant diseases, test types, drug-resistant mechanism types, and drug-resistant key molecules are the second, third, fourth, and fifth levels, respectively. The entity categories of each entity contained in the medical data are matched with the hierarchical relationship between the preset categories, so that the hierarchical connection of each entity contained in the medical data is realized according to the hierarchical relationship between the preset categories, and the tree map corresponding to the medical data is obtained.

[0039] In this way, by dividing the drug resistance information into different entity categories and organizing them according to hierarchical relationships, a structured information representation can be formed. This representation makes the information clearer and more orderly, and is easier to understand and analyze. In addition, the clear hierarchical relationship allows more accurate positioning of the required information level when retrieving drug resistance information. Moreover, the hierarchical relationship not only facilitates the retrieval of information, but also provides a basis for in-depth analysis. By comparing the associations and differences between different levels, the inherent laws and trends of drug resistance information can be revealed, providing strong support for clinical decision-making and drug development.

[0040] S3. Construct a standard entity dictionary, in which a standard expression method of each entity is set.

[0041] Build a dictionary of standard drugs and drug-resistant diseases, and then build a standard dictionary of test types, drug-resistant mechanism types, and drug-resistant key molecules corresponding to each drug-resistant disease combination. For example: S4. Standardize the expression of entities in the tree diagram of each medical data according to the standard entity dictionary.

[0042] Among them, the process of standardizing the expression of entities in the tree map includes: for any entity in the tree map, calculating the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary; then, taking the candidate standard entity with the highest semantic relevance as the standard entity of the entity, replacing the entity, and realizing the standardization of the entity. In this way, by replacing the entities in the tree map with the corresponding items in the standard entity dictionary, it is ensured that all mentioned entities are expressed in a unified and standardized way. This eliminates the confusion and misunderstanding caused by different data sources using different terms or expressions, and improves the consistency and comparability of the data. In addition, the calculation of semantic relevance is based on a deep understanding of the meaning of the entity, not just a literal match. Therefore, by selecting the candidate standard entity with the highest semantic relevance for replacement, the true meaning of the entity can be more accurately reflected, reducing information errors caused by mismatching or understanding deviations.

[0043] In specific implementation, the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary is calculated through a preset relevance calculation model. In this way, the accuracy and efficiency of drug resistance information mining can be ensured by calculating the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary through a preset relevance calculation model. It not only improves the accuracy and robustness of matching, but also supports the identification of complex relationships and optimizes the calculation efficiency, promoting the automation and intelligence of the entire mining process.

[0044] Since the same entity may have multiple different expressions (for example, sunitinib, sorafenib, sunitinib, SU011248), it is necessary to match each entity in the tree map with the standard entity words in the standard entity dictionary of the corresponding segmentation dimension to achieve the standardization of the tree map.

[0045] According to the standard entity dictionary matching, the candidate standard entities of each entity in the tree map are obtained. For any entity in any of the tree maps, the semantic relevance between the entity and each candidate standard entity is calculated. The semantic relevance here is used to reflect the degree of relevance of the semantics reflected by the entity and its candidate standard entities. Specifically, the entity and its candidate standard entity can be semantically encoded respectively, and the relevance is calculated based on the semantic features obtained by encoding, or the entity and its candidate standard entity are input into a pre-trained relevance calculation model to obtain the semantic relevance between the two. After obtaining the semantic relevance between any entity in the tree map and each candidate standard entity of the entity, the standard entity of the entity can be determined accordingly. For example, the candidate standard entity with the highest semantic relevance can be selected from each candidate standard entity as the standard entity of the entity, or the candidate standard entity with a semantic relevance greater than a preset threshold can be selected from each candidate standard entity as the standard entity of the entity. Based on the above method, the standard entity of each entity in each tree map can be obtained, and the standardization of each tree map can be achieved by replacing the corresponding entity with the standard entity.

[0046] S5. The standardized dendrograms are combined to obtain a classification system for the drug resistance information of the drug.

[0047] In specific implementation, the process of merging the standardized tree graphs includes: taking any standardized tree graph as a benchmark, traversing other standardized tree graphs of the same drug, checking whether there are drug-resistant diseases in the traversed tree graph that are not in the benchmark tree graph, and if so, adding the subtree of the drug-resistant disease to the benchmark tree graph; if not, continue to check the test type hierarchy under the drug-resistant disease, and so on until all leaf nodes are traversed.

[0048] In this way, through the merging process, it is ensured that the baseline tree map contains all relevant drug-resistant diseases and their detailed information. Even if some drug-resistant diseases or specific information are missing in an initial tree map, as long as they exist in other tree maps, they can be added to the baseline map, thereby improving the integrity of the information. The merged tree map supports information retrieval across different data sources. Users can find information about the drug resistance of a drug in different diseases and different test conditions in a unified map without having to access multiple data sources separately. This greatly improves the efficiency and convenience of information retrieval.

[0049] S6. Retrieve the drug resistance information text of the drug in the multi-source data, and extract entities from the obtained drug resistance information text; match the extracted entities with the classification system of drug resistance information, obtain the position node of the entity in the classification system, and establish a connection relationship between the position node and the drug resistance information text corresponding to the entity, so as to form the drug resistance information mining result of the drug; S7. Based on the drug resistance information mining result, analyze and obtain a drug resistance information report of the drug.

[0050] After completing the construction of a classification system for drug resistance information, drug resistance information mining can be carried out based on the classification system, thereby breaking through the information barriers between data from various sources and realizing the mining and integration of multi-source information.

[0051] For the medical data to be mined, the drug resistance information text can be determined by locating the keywords "drug resistance", "drug resistance", "drug resistance", "drug tolerance", "drug-fast" or synonyms of the above keywords. For example, the paragraph where the above keywords or synonyms of the keywords are located or the adjacent text of the paragraph is used as the drug resistance information text.

[0052] The entities contained in the drug resistance information text are matched with the entities contained in the classification system, and a connection relationship between the drug resistance information text and the classification system is established based on the matching result.

[0053] Specifically, taking any combination of a drug and a drug-resistant disease as an example, the combination may correspond to multi-source data, and the multi-source data here may include data from multiple sources, such as any of clinical trial data, journal literature, clinical guidelines, drug registration information, epidemiological data, and information data. The data from each of the above sources contains text describing drug resistance information, i.e., drug resistance information text. After obtaining the multi-source data, the drug resistance information text of each source data in the multi-source data can be located according to the data structure of each source data in the multi-source data.

[0054] On this basis, entity recognition can be performed on the drug resistance information text in the multi-source data, so as to obtain the entities contained in the drug resistance information text. After obtaining the entities contained in the drug resistance information text, the entities contained in the drug resistance information text can be matched with the entities contained in the classification system corresponding to the drug, so as to locate the position of the entities contained in the drug resistance information text in the classification system, that is, to obtain the entities contained in the drug resistance information text that match the entities in the classification system, and thereby establish a connection relationship between the two matching entities, that is, to establish a connection relationship between the nodes in the classification system and the drug resistance information text in the multi-source data, so that when the user queries a node in the classification system, the information expressed by the drug resistance information text that has a connection relationship with the node can be obtained.

[0055] In order to realize entity recognition for drug resistance information text, an entity recognition model is pre-trained. The drug resistance information text is input into the pre-trained entity recognition model, which performs entity recognition on the treatment plan text and outputs the labeling of each entity in the drug resistance information text. The labeling system for entity recognition here can be BIO, BIOES, etc.

[0056] Using this method, for the centralized drugs, after obtaining the entities and corresponding entity categories in the medical data, the identified entities will be hierarchically connected according to the preset entity category hierarchical relationship to form a tree map. This hierarchical structure helps to understand and represent the complex relationship between entities, and facilitates the subsequent standardization and merging operations. After that, this method will construct a standard entity dictionary to unify and standardize the expression of entities in medical data. Through comparison and replacement, the entity expression in the tree map is matched with the standard entity dictionary to eliminate information redundancy and confusion caused by inconsistent expressions. In this way, the accuracy and consistency of drug resistance information are ensured through the construction of standard entity dictionaries and hierarchical relationships, and the possibility of information errors and omissions is reduced. This method will standardize the expression of entities in the tree maps of each medical data, and then merge the standardized tree maps to form a complete drug resistance information classification system. This system not only contains the association information between drugs and drug-resistant diseases, but also reflects the hierarchy and logical relationship between them. In this way, integrating multi-source, heterogeneous medical data into a complete drug resistance information classification system can provide comprehensive and systematic reference information for clinicians and pharmaceutical R&D personnel. By building a refined drug resistance information classification system, pharmaceutical companies can be supported to conduct research in more subdivided therapeutic areas in new drug clinical trials or new drug development, thus promoting the development of precision medicine.

[0057] After obtaining the classification system of drug resistance information, this method will retrieve the drug resistance information text of specific drugs in multi-source data, and connect these texts with the classification system of drug resistance information through entity extraction and matching technology. In this way, the rapid positioning and classification of newly emerging drug resistance information is achieved. When the staff queries a node in the classification system, they can also obtain the information expressed by the drug resistance information text that is connected to the node, which facilitates their full mining of the drug resistance information. Afterwards, the generated drug resistance information report can provide strong support for clinicians' medication decisions, reduce the risk of drug resistance, and improve treatment effects.

[0058] This method not only realizes comprehensive and reliable drug resistance information mining, but also effectively improves the efficiency of drug resistance information mining.

[0059] like Figure 2 As shown, the present invention also provides a device for mining drug resistance information, which is used in the above-mentioned method for mining drug resistance information, and includes an entity acquisition unit, a graph construction unit, a graph standardization unit, a graph merging unit and an information mining unit; The entity acquisition unit is used to acquire entities and entity categories in each medical data; The graph construction unit is used to perform corresponding hierarchical connections on the entity categories of the entities in the medical data based on the acquired entities and their entity categories, as well as the preset hierarchical relationships of the entity categories, to obtain a tree graph of each medical data; The graph standardization unit is used to construct a standard entity dictionary and is also used to standardize each tree graph according to the standard entity dictionary; The map merging unit is used to merge the standardized tree maps to obtain a classification system of drug resistance information; The information mining unit is used for mining drug resistance information based on the classification system of drug resistance information.

[0060] Among them, the process of the graph standardization unit standardizing each tree graph according to the standard entity dictionary includes: for any entity in the tree graph, calculating the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary; then, taking the candidate standard entity with the highest semantic relevance as the standard entity of the entity, replacing the entity, and realizing the standardization of the entity.

[0061] The process of drug resistance information mining by the information mining unit includes: retrieving the drug resistance information text of the drug in multi-source data, and extracting entities from the obtained drug resistance information text; matching the extracted entities with the classification system of drug resistance information, and after obtaining the position node of the entity in the classification system, establishing a connection relationship between the position node and the drug resistance information text corresponding to the entity, forming the drug resistance information mining result of the drug; and based on the drug resistance information mining result, analyzing to obtain the drug resistance information report of the drug.

[0062] While realizing comprehensive and reliable drug resistance information mining, the device also effectively improves the efficiency of drug resistance information mining.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the technical solution. Those skilled in the art should understand that those modifications or equivalent substitutions of the technical solution of the present invention that do not depart from the purpose and scope of the technical solution should be included in the scope of the claims of the present invention.

Claims

1. A method for mining drug resistance information, characterized in that: The following steps are involved: S1. For each medical data of a drug, entities and corresponding entity categories are obtained respectively; wherein the entity categories include drugs and drug-resistant diseases; S2. According to the preset entity category hierarchical relationship, the entity categories of the entities in the medical data are connected hierarchically to obtain a tree diagram of each medical data; S3, constructing a standard entity dictionary, in which a standard expression method of each entity is set; S4. Standardize the expression of entities in the tree diagram of each medical data according to the standard entity dictionary; S5, merging the standardized dendrograms to obtain a classification system of drug resistance information of the drug; S6. Retrieve the drug resistance information text of the drug in the multi-source data, and extract entities from the obtained drug resistance information text; match the extracted entities with the classification system of drug resistance information, obtain the position node of the entity in the classification system, and establish a connection relationship between the position node and the drug resistance information text corresponding to the entity, so as to form the drug resistance information mining result of the drug; S7. Based on the drug resistance information mining result, analyze and obtain a drug resistance information report of the drug.

2. The method for mining drug resistance information according to claim 1, characterized in that: In S1, natural language processing technology is used to obtain entities in medical data and corresponding entity categories.

3. The method for mining drug resistance information according to claim 1, characterized in that: In S4, the process of standardizing the expression of entities in the tree diagram includes: for any entity in the tree diagram, calculating the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary; then, taking the candidate standard entity with the highest semantic relevance as the standard entity of the entity, replacing the entity, and realizing the standardization of the entity.

4. The method for mining drug resistance information according to claim 3, characterized in that: The semantic relevance between the entity and each candidate standard entity in the standard entity dictionary is calculated through a preset relevance calculation model.

5. The method for mining drug resistance information as claimed in claim 1, characterized in that: The entity categories include drugs, drug-resistant diseases, test types, resistance mechanism types, and key molecules of resistance.

6. The method for mining drug resistance information according to claim 5, characterized in that: The hierarchy of each entity category from high to low is drug, drug-resistant disease, trial type, resistance mechanism type, and key resistance molecules.

7. The method for mining drug resistance information according to claim 6, characterized in that: In S5, the process of merging the standardized tree graphs includes: taking any standardized tree graph as a benchmark, traversing other standardized tree graphs of the same drug, checking whether there is a drug-resistant disease in the traversed tree graph that is not in the benchmark tree graph, and if so, adding the subtree of the drug-resistant disease to the benchmark tree graph; if not, continue to check the test type hierarchy under the drug-resistant disease, and so on until all leaf nodes are traversed.

8. A device for mining drug resistance information, characterized in that: A method for mining the above-mentioned drug resistance information includes an entity acquisition unit, a map construction unit, a map standardization unit, a map merging unit and an information mining unit; The entity acquisition unit is used to acquire entities and entity categories in each medical data; The graph construction unit is used to perform corresponding hierarchical connections on the entity categories of the entities in the medical data based on the acquired entities and their entity categories, as well as the preset hierarchical relationships of the entity categories, to obtain a tree graph of each medical data; The graph standardization unit is used to construct a standard entity dictionary and is also used to standardize each tree graph according to the standard entity dictionary; The map merging unit is used to merge the standardized tree maps to obtain a classification system of drug resistance information; The information mining unit is used for mining drug resistance information based on the classification system of drug resistance information.

9. The device for mining drug resistance information according to claim 8, characterized in that: The process of the graph standardization unit standardizing each tree graph according to the standard entity dictionary includes: for any entity in the tree graph, calculating the semantic relevance between the entity and each candidate standard entity in the standard entity dictionary; then, taking the candidate standard entity with the highest semantic relevance as the standard entity of the entity, replacing the entity, and realizing the standardization of the entity.

10. The drug resistance information mining device according to claim 8, characterized in that: The process of drug resistance information mining by the information mining unit includes: retrieving the drug resistance information text of the drug in multi-source data, and extracting entities from the obtained drug resistance information text; matching the extracted entities with the classification system of drug resistance information, and after obtaining the position node of the entity in the classification system, establishing a connection relationship between the position node and the drug resistance information text corresponding to the entity, forming the drug resistance information mining result of the drug; and based on the drug resistance information mining result, analyzing to obtain the drug resistance information report of the drug.

Images