Risk prediction method and system for related adverse reactions caused by combination of clinical traditional Chinese medicine and western medicine

Abstract

The invention provides a risk prediction method and system for related adverse reactions of clinical Chinese and western medicine combination, and the method comprises the steps: collecting prior data, historical clinical data and real electronic medical record data, and constructing a drug-adverse reaction mapping relation database based on the prior data and the historical clinical data; performing knowledge representation learning and knowledge graph modeling of the drug-adverse reaction according to the mapping relation database; according to the real electronic medical record data, a self-supervised model is adopted to acquire target features of doctor's advice information and potential adverse reactions on the time dynamic graph; and constructing a prediction model, analyzing and predicting the relationship between the doctor's advice information input into the model and the adverse reaction, and visually outputting the relevance between the drug and the adverse reaction, the potential risk and the risk basis. Clinical doctors and pharmaceutical personnel are assisted to dynamically analyze and monitor the clinical medication condition in real time according to the relevance, potential risks and risk basis of the adverse reactions, the clinical medication safety is improved, and the occurrence rate of the adverse reactions of the drugs is reduced.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for predicting the risk of adverse reactions associated with the clinical use of Chinese and Western medicines. Background Art

[0002] With the development of biomedical science in recent years, drugs, disease mechanisms and the biological pathways between them have been widely studied.

[0003] In existing clinical practice, the combined use of Chinese and Western medicine is becoming more and more common, but this combination of medicines may cause adverse reactions, which will greatly reduce the therapeutic effect and safety of the drugs. At present, due to the complexity of the combined use of Chinese and Western medicines, the detection and identification of adverse reactions of the combined use of Chinese and Western medicines is of vital importance to clinical medicine.

[0004] Therefore, there is an urgent need for a risk prediction method for adverse reactions related to the combination of Chinese and Western medicines in clinical practice to solve the risk prediction problem of the combination of Chinese and Western medicines. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present invention proposes a method and system for predicting the risk of adverse reactions associated with the clinical combination of Chinese and Western medicines.

[0006] The first aspect of the present invention discloses a method for predicting the risk of adverse reactions related to the clinical combination of Chinese and Western medicines, the method comprising:

[0007] Collecting prior data, historical clinical data and actual electronic medical record data, and building a drug-adverse reaction mapping database based on the prior data and historical clinical data;

[0008] Performing knowledge representation learning and knowledge graph modeling of the drug-adverse reaction according to the mapping relationship database;

[0009] According to the actual electronic medical record data, a self-supervised model is used to obtain the target features of the medical advice information and potential adverse reactions on a time dynamic graph;

[0010] Construct a prediction model, analyze and predict the relationship between the doctor's order information input into the prediction model and the adverse reaction, and visualize the correlation between the drug and the adverse reaction, the potential risk and the basis of the risk.

[0011] As an optional implementation, in the first aspect of the present invention, collecting prior data, historical clinical data and current electronic medical record data specifically includes:

[0012] Collect prior data and historical clinical data from public data sources, including prior knowledge bases, laboratory databases, bioinformatics bases, Chinese and Western medicine knowledge bases, and adverse reaction reports. The prior data and historical clinical data are knowledge and experience data that describe the relationship and knowledge information between drugs and adverse reactions, and use unified coding to integrate the relationship between drugs and adverse reactions;

[0013] Collect actual electronic medical record data in the medical information system and the electronic medical record system. The actual electronic medical record data is medical information related to adverse reactions in the actual medication process. The medical information includes clinical medical orders, patient medication, drug prescriptions in drug trials, medication symptoms, and adverse reaction symptoms.

[0014] As an optional embodiment, in the first aspect of the present invention, a drug-adverse reaction mapping database is constructed based on the prior data and historical clinical data, specifically comprising:

[0015] Entity recognition and relationship extraction are used to obtain the prior relationship between drug-adverse reactions in the prior data and historical clinical data, and to predict the actual relationship between drug-adverse reactions in the actual electronic medical record data;

[0016] Modify the unified coding of drug-adverse reaction relationships based on the a priori and actual relationships;

[0017] Based on the unified coding, entity relationships are constructed through semantic networks, and the entity relationships include drug names, drug numbers, drug ingredients, drug applicable disease types, drug side effects, and relationships between targets.

[0018] As an optional implementation, in the first aspect of the present invention, the drug-adverse reaction knowledge representation learning is performed according to the mapping relationship database, specifically comprising:

[0019] Obtaining original feature signals in the adverse reaction dataset;

[0020] Constructing an initial adverse reaction graph according to the entity relationship information and the original feature signal in the mapping relationship database;

[0021] The nodes in the initial adverse reaction graph are mapped to a low-dimensional vector space to obtain node vectors and relationship node vectors, wherein the node vectors represent entities and the relationship node vectors represent associated knowledge of entities.

[0022] As an optional implementation, in the first aspect of the present invention, the knowledge graph modeling of the drug-adverse reaction is performed according to the mapping relationship database, specifically including:

[0023] Obtain an adverse reaction dynamic graph based on the historical adverse reaction data set to form a knowledge graph of drugs and adverse reactions, wherein the adverse reaction dynamic graph includes a historical node time series set of drugs and adverse reactions in historical clinical data;

[0024] A self-supervised learning model is constructed based on a time-series graph network and an attention mechanism, and the historical node time-series set, node vector, and relationship node vector are used as inputs of the self-supervised learning model.

[0025] As an optional embodiment, in the first aspect of the present invention, the target features of the doctor's advice information and potential adverse reactions on the time dynamic graph are obtained using a self-supervised model based on the historical clinical data, specifically including:

[0026] The local features of the self-supervised learning model are fused by an attention mechanism to obtain target features in the adverse reaction dynamic graph that represent changes in drugs and adverse reactions over time, wherein the target features are entity relationships expressed in nodes in the adverse reaction dynamic graph;

[0027] Feature ablation is used to evaluate the impact of the target feature on predicting the association between the drug and the adverse reaction.

[0028] As an optional implementation, in the first aspect of the present invention, a prediction model is constructed, the relationship between the doctor's order information input into the prediction model and the adverse reaction is analyzed and predicted, and the correlation between the drug and the adverse reaction, the potential risk and the risk basis are visually output, specifically including:

[0029] A prediction model is constructed using a deep neural network, the target features pre-trained by the self-supervised learning model are integrated according to the attention mechanism, the cross entropy of the prediction model is selected based on binary label classification, and the prediction model is optimized by continuously updating the network weights using the adam optimizer;

[0030] Obtaining node time series data in the medical order information, and inputting the original feature signal, node vector, relationship node vector, target feature and influence value as prediction features together with the node time series data into the prediction model for prediction analysis;

[0031] The output of the prediction model is set as a new link classification task according to the binary label classification, and the prediction result is output according to the classification task to obtain the correlation between the drug and the adverse reaction of the patient within the predicted course of disease according to the doctor's advice information, and the influence degree of the drug and the adverse reaction is determined according to the influence value;

[0032] The correlation and impact degree are visualized as a target adverse reaction graph, and the potential risk of the medical order information is identified according to the target adverse reaction graph.

[0033] The second aspect of the present invention discloses a risk prediction system for adverse reactions related to the clinical combination of Chinese and Western medicines, the system comprising:

[0034] A data acquisition module is used to collect prior data, historical clinical data and actual electronic medical record data, and to construct a drug-adverse reaction mapping database based on the prior data and historical clinical data;

[0035] A knowledge modeling module, used for performing knowledge representation learning and knowledge graph modeling of the drug-adverse reaction according to the mapping relationship database;

[0036] A prediction training module, used for obtaining target features of medical advice information and potential adverse reactions on a time dynamic graph using a self-supervised model based on the actual electronic medical record data;

[0037] The risk identification module is used to build a prediction model, analyze and predict the relationship between the doctor's order information input into the prediction model and the adverse reaction, and visualize the correlation between the drug and the adverse reaction, the potential risk and the risk basis.

[0038] As an optional implementation, in the second aspect of the present invention, the data acquisition module collects prior data, historical clinical data and current electronic medical record data, specifically including:

[0039] Collect prior data and historical clinical data from public data sources, including prior knowledge bases, laboratory databases, bioinformatics bases, Chinese and Western medicine knowledge bases, and adverse reaction reports. The prior data and historical clinical data are knowledge and experience data that describe the relationship and knowledge information between drugs and adverse reactions, and use unified coding to integrate the relationship between drugs and adverse reactions;

[0040] Collect actual electronic medical record data in the medical information system and the electronic medical record system. The actual electronic medical record data is medical information related to adverse reactions in the actual medication process. The medical information includes clinical medical orders, patient medication, drug prescriptions in drug trials, medication symptoms, and adverse reaction symptoms.

[0041] As an optional implementation, in the second aspect of the present invention, the data acquisition module constructs a drug-adverse reaction mapping database based on the prior data and historical clinical data, specifically including:

[0042] Entity recognition and relationship extraction are used to obtain the prior relationship between drug-adverse reactions in the prior data and historical clinical data, and to predict the actual relationship between drug-adverse reactions in the actual electronic medical record data;

[0043] Modify the unified coding of drug-adverse reaction relationships based on the a priori and actual relationships;

[0044] Based on the unified coding, entity relationships are constructed through semantic networks, and the entity relationships include drug names, drug numbers, drug ingredients, drug applicable disease types, drug side effects, and relationships between targets.

[0045] As an optional implementation, in the second aspect of the present invention, the knowledge modeling module performs the knowledge representation learning of the drug-adverse reaction according to the mapping relationship database, specifically including:

[0046] Obtaining original feature signals in the adverse reaction dataset;

[0047] Constructing an initial adverse reaction graph according to the entity relationship information and the original feature signal in the mapping relationship database;

[0048] The nodes in the initial adverse reaction graph are mapped to a low-dimensional vector space to obtain node vectors and relationship node vectors, wherein the node vectors represent entities and the relationship node vectors represent associated knowledge of entities.

[0049] As an optional implementation, in the second aspect of the present invention, the knowledge modeling module performs knowledge graph modeling of the drug-adverse reaction according to the mapping relationship database, specifically comprising:

[0050] Obtain an adverse reaction dynamic graph based on the historical adverse reaction data set to form a knowledge graph of drugs and adverse reactions, wherein the adverse reaction dynamic graph includes a historical node time series set of drugs and adverse reactions in historical clinical data;

[0051] A self-supervised learning model is constructed based on a time-series graph network and an attention mechanism, and the historical node time-series set, node vector, and relationship node vector are used as inputs of the self-supervised learning model.

[0052] As an optional embodiment, in the second aspect of the present invention, the prediction training module uses a self-supervised model to obtain the target features of the doctor's advice information and potential adverse reactions on the time dynamic graph based on the historical clinical data, specifically including:

[0053] The local features of the self-supervised learning model are fused by an attention mechanism to obtain target features in the adverse reaction dynamic graph that represent changes in drugs and adverse reactions over time, wherein the target features are entity relationships expressed in nodes in the adverse reaction dynamic graph;

[0054] Feature ablation is used to evaluate the impact of the target feature on predicting the association between the drug and the adverse reaction.

[0055] As an optional implementation, in the second aspect of the present invention, the risk identification module constructs a prediction model, analyzes and predicts the relationship between the doctor's order information input into the prediction model and the adverse reaction, and visualizes the correlation between the drug and the adverse reaction, the potential risk and the risk basis, specifically including:

[0056] A prediction model is constructed using a deep neural network, the target features pre-trained by the self-supervised learning model are integrated according to the attention mechanism, the cross entropy of the prediction model is selected based on binary label classification, and the prediction model is optimized by continuously updating the network weights using the adam optimizer;

[0057] Obtaining node time series data in the medical order information, and inputting the original feature signal, node vector, relationship node vector, target feature and influence value as prediction features together with the node time series data into the prediction model for prediction analysis;

[0058] The output of the prediction model is set as a new link classification task according to the binary label classification, and the prediction result is output according to the classification task to obtain the correlation between the drug and the adverse reaction of the patient within the predicted course of disease according to the doctor's advice information, and the influence degree of the drug and the adverse reaction is determined according to the influence value;

[0059] The correlation and impact degree are visualized as a target adverse reaction graph, and the potential risk of the medical order information is identified according to the target adverse reaction graph.

[0060] The third aspect of the present invention discloses a risk prediction device for adverse reactions related to the clinical combination of Chinese and Western medicines, comprising:

[0061] at least one processor, and

[0062] a memory communicatively connected to the at least one processor; wherein,

[0063] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the risk prediction method for adverse reactions related to the combination of Chinese and Western medicines in clinical practice as described in any one of the items disclosed in the first aspect of the present invention.

[0064] The fourth aspect of the present invention discloses a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the risk prediction method for adverse reactions related to the combination of Chinese and Western medicines in clinical practice as described in any one of the above-mentioned aspects.

[0065] Compared with the prior art, the present invention has the following advantages:

[0066] The present invention collects prior data, historical clinical data and actual electronic medical record data, and constructs a drug-adverse reaction mapping relationship database based on the prior data and historical clinical data; performs knowledge representation learning and knowledge graph modeling of the drug-adverse reaction according to the mapping relationship database; obtains the target features of the doctor's order information and potential adverse reactions on the time dynamic graph according to the data set of knowledge representation learning and knowledge graph modeling; constructs a prediction model, analyzes the doctor's order information of the prediction input model, and can visualize the output of the correlation, potential risks and risk basis between the drug and the adverse reaction, so as to assist clinicians in detecting the potential adverse reactions in the combined use of Chinese medicine formula granules and western medicines at an early stage, and monitor and evaluate them subsequently. The framework can not only accurately predict the actual adverse drug reactions, but also provide a complex visualized adverse reaction graph analysis. Through visualized analysis, the present framework enhances the interpretability of the model, so that users can understand the logic and reasons behind the prediction results, and can be widely used in the screening of clinical drug combinations, formulation of personalized medical plans and drug safety monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0068] Figure 1 This is a flow chart of the method for predicting the risk of adverse reactions related to the clinical combination of Chinese and Western medicines of the present invention;

[0069] Figure 2 It is a schematic diagram of the risk prediction system for adverse reactions related to the clinical combination of Chinese and Western medicines of the present invention. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0071] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0072] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0073] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0074] Example 1

[0075] See also Figure 1 The present invention discloses a method for predicting the risk of adverse reactions associated with the clinical use of Chinese and Western medicines, the method comprising:

[0076] 101. Collect prior data, historical clinical data and actual electronic medical record data, and build a drug-adverse reaction mapping database based on the prior data and historical clinical data;

[0077] 102. Conduct knowledge representation learning and knowledge graph modeling of drug-adverse reactions based on the mapping relationship database;

[0078] 103. Based on the actual electronic medical record data, a self-supervised model is used to obtain the target features of medical advice information and potential adverse reactions on the time dynamic graph;

[0079] 104. Build a prediction model, analyze the relationship between the medical advice information input into the prediction model and adverse reactions, and visualize the correlation between drugs and adverse reactions, potential risks, and risk basis.

[0080] As an optional implementation, in the first aspect of the present invention, collecting prior data, historical clinical data and current electronic medical record data specifically includes:

[0081] Collect prior data and historical clinical data from public data sources, including prior knowledge bases, laboratory databases, bioinformatics databases, Chinese and Western medicine knowledge bases, and adverse reaction reports. Prior data and historical clinical data are knowledge and experience data that describe the relationship and knowledge information between drugs and adverse reactions. Use unified coding to integrate the relationship between drugs and adverse reactions.

[0082] The actual electronic medical record data in the medical information system and the electronic medical record system are collected. The actual electronic medical record data is the medical information related to the adverse reactions in the actual medication process. The medical information includes clinical medical instructions, patient medication, drug prescriptions in drug trials, medication symptoms and adverse reaction symptoms.

[0083] As an optional implementation, in the first aspect of the present invention, a drug-adverse reaction mapping database is constructed based on prior data and historical clinical data, specifically including:

[0084] Entity recognition and relationship extraction are used to obtain the prior relationship between drugs and adverse reactions in prior data and historical clinical data, and to predict the actual relationship between drugs and adverse reactions in real electronic medical record data;

[0085] Modify the unified coding of drug-adverse reaction relationships based on a priori and actual relationships;

[0086] Entity relationships are constructed through semantic networks based on unified coding. Entity relationships include drug name, drug number, drug ingredient, drug applicable disease type, drug side effects, and the relationship between targets.

[0087] As an optional implementation, in the first aspect of the present invention, the knowledge representation learning of drug-adverse reactions is performed according to the mapping relationship database, specifically including:

[0088] Obtaining original feature signals in the adverse reaction dataset;

[0089] Constructing an initial adverse reaction graph based on entity relationship information and original feature signals in a mapping relationship database;

[0090] The nodes in the initial adverse reaction graph are mapped to a low-dimensional vector space to obtain node vectors and relationship node vectors. The node vectors represent entities, and the relationship node vectors represent the associated knowledge of entities.

[0091] As an optional implementation, in the first aspect of the present invention, the knowledge graph modeling of drug-adverse reactions is performed according to the mapping relationship database, specifically including:

[0092] Obtain an adverse reaction dynamic graph based on the historical adverse reaction data set to form a knowledge graph of drugs and adverse reactions. The adverse reaction dynamic graph includes a historical node time series set of drugs and adverse reactions in historical clinical data;

[0093] A self-supervised learning model is constructed based on the temporal graph network and attention mechanism, and the historical node temporal set, node vector and relationship node vector are used as the input of the self-supervised learning model.

[0094] As an optional embodiment, in the first aspect of the present invention, a self-supervised model is used to obtain target features of medical advice information and potential adverse reactions on a time dynamic graph based on historical clinical data, specifically including:

[0095] The local features of the input self-supervised learning model are fused through the attention mechanism to obtain the target features in the adverse reaction dynamic graph that represent the changes of drugs and adverse reactions over time. The target features are the entity relationships expressed in the nodes in the adverse reaction dynamic graph.

[0096] Feature ablation was used to evaluate the impact of target features in predicting the association between drugs and adverse reactions.

[0097] As an optional embodiment, in the first aspect of the present invention, a prediction model is constructed, the relationship between the doctor's advice information input into the prediction model and the adverse reaction is analyzed and predicted, and the correlation between the drug and the adverse reaction, the potential risk and the risk basis are visualized and output, specifically including:

[0098] A deep neural network is used to build a prediction model. The target features pre-trained by the self-supervised learning model are integrated according to the attention mechanism. The cross entropy of the prediction model is selected based on binary label classification. The Adam optimizer is used to continuously update the network weights to optimize the prediction model.

[0099] Obtaining node time series data in the medical order information, and inputting the original feature signal, node vector, relationship node vector, target feature and influence value as prediction features into the prediction model together with the node time series data for prediction analysis;

[0100] The output of the prediction model is set as a new link classification task according to the binary label classification, and the prediction result is output according to the classification task to obtain the correlation between the drug and the adverse reaction of the patient according to the doctor's advice information within the predicted course of the disease, and the influence degree of the drug and the adverse reaction is determined according to the influence value;

[0101] The relationship and impact degree are visualized as a target adverse reaction graph, and the potential risks of medical order information are identified based on the target adverse reaction graph.

[0102] The present invention collects prior data, historical clinical data and real electronic medical record data, and constructs a drug-adverse reaction mapping database based on prior data and historical clinical data; performs knowledge representation learning and knowledge graph modeling of drug-adverse reactions according to the mapping relationship database; obtains the target features of doctor's order information and potential adverse reactions on a time dynamic graph according to the data set of knowledge representation learning and knowledge graph modeling; constructs a prediction model, analyzes the doctor's order information of the prediction input model, and can visualize the correlation, potential risks and risk basis of the output drug and adverse reactions, so as to assist clinicians in detecting potential adverse reactions in the combined use of Chinese medicine formula granules and Western medicines in the early stage, and monitor and evaluate them subsequently. The framework can not only accurately predict actual adverse drug reactions, but also provide a complex visualization adverse reaction graph analysis. Through visualization analysis, the framework enhances the interpretability of the model, so that users can understand the logic and reasons behind the prediction results, and can be widely used in the screening of clinical drug combinations, the formulation of personalized medical plans and the field of drug safety monitoring.

[0103] like Figure 2 As shown, the second aspect of the present invention discloses a risk prediction system for clinical adverse reactions related to the combination of Chinese and Western medicines, the system comprising:

[0104] The data acquisition module is used to collect prior data, historical clinical data and actual electronic medical record data, and build a drug-adverse reaction mapping database based on the prior data and historical clinical data;

[0105] The knowledge modeling module is used to learn the knowledge representation of drug-adverse reactions and model the knowledge graph based on the mapping relationship database;

[0106] The prediction training module is used to obtain the target features of medical advice information and potential adverse reactions on a time-dynamic graph using a self-supervised model based on real electronic medical record data;

[0107] The risk identification module is used to build a prediction model, analyze the relationship between the doctor's advice information input into the prediction model and adverse reactions, and visualize the correlation between drugs and adverse reactions, potential risks, and risk basis.

[0108] As an optional implementation, in the second aspect of the present invention, the data acquisition module collects prior data, historical clinical data, and current electronic medical record data, specifically including:

[0109] Collect prior data and historical clinical data from public data sources, including prior knowledge bases, laboratory databases, bioinformatics databases, Chinese and Western medicine knowledge bases, and adverse reaction reports. Prior data and historical clinical data are knowledge and experience data that describe the relationship and knowledge information between drugs and adverse reactions. Use unified coding to integrate the relationship between drugs and adverse reactions.

[0110] The actual electronic medical record data in the medical information system and the electronic medical record system are collected. The actual electronic medical record data is the medical information related to the adverse reactions in the actual medication process. The medical information includes clinical medical instructions, patient medication, drug prescriptions in drug trials, medication symptoms and adverse reaction symptoms.

[0111] As an optional implementation, in the second aspect of the present invention, the data acquisition module constructs a drug-adverse reaction mapping database based on prior data and historical clinical data, specifically including:

[0112] Entity recognition and relationship extraction are used to obtain the prior relationship between drugs and adverse reactions in prior data and historical clinical data, and to predict the actual relationship between drugs and adverse reactions in real electronic medical record data;

[0113] Modify the unified coding of drug-adverse reaction relationships based on a priori and actual relationships;

[0114] Entity relationships are constructed through semantic networks based on unified coding. Entity relationships include drug name, drug number, drug ingredient, drug applicable disease type, drug side effects, and the relationship between targets.

[0115] As an optional implementation, in the second aspect of the present invention, the knowledge modeling module performs drug-adverse reaction knowledge representation learning according to the mapping relationship database, specifically including:

[0116] Obtaining original feature signals in the adverse reaction dataset;

[0117] Constructing an initial adverse reaction graph based on entity relationship information and original feature signals in a mapping relationship database;

[0118] The nodes in the initial adverse reaction graph are mapped to a low-dimensional vector space to obtain node vectors and relationship node vectors. The node vectors represent entities, and the relationship node vectors represent the associated knowledge of entities.

[0119] As an optional implementation, in the second aspect of the present invention, the knowledge modeling module performs drug-adverse reaction knowledge graph modeling according to the mapping relationship database, specifically including:

[0120] Obtain an adverse reaction dynamic graph based on the historical adverse reaction data set to form a knowledge graph of drugs and adverse reactions. The adverse reaction dynamic graph includes a historical node time series set of drugs and adverse reactions in historical clinical data;

[0121] A self-supervised learning model is constructed based on the temporal graph network and attention mechanism, and the historical node temporal set, node vector and relationship node vector are used as the input of the self-supervised learning model.

[0122] As an optional embodiment, in the second aspect of the present invention, the prediction training module uses a self-supervised model to obtain the target features of the doctor's advice information and potential adverse reactions on the time dynamic graph based on the historical clinical data, specifically including:

[0123] The local features of the input self-supervised learning model are fused through the attention mechanism to obtain the target features in the adverse reaction dynamic graph that represent the changes of drugs and adverse reactions over time. The target features are the entity relationships expressed in the nodes in the adverse reaction dynamic graph.

[0124] Feature ablation was used to evaluate the impact of target features in predicting the association between drugs and adverse reactions.

[0125] As an optional implementation, in the second aspect of the present invention, a prediction model is constructed, the relationship between the doctor's advice information input into the prediction model and the adverse reaction is analyzed, and the correlation between the drug and the adverse reaction, the potential risk and the risk basis are visualized and output, specifically including:

[0126] A deep neural network is used to build a prediction model. The target features pre-trained by the self-supervised learning model are integrated according to the attention mechanism. The cross entropy of the prediction model is selected based on binary label classification. The Adam optimizer is used to continuously update the network weights to optimize the prediction model.

[0127] Obtaining node time series data in the medical order information, and inputting the original feature signal, node vector, relationship node vector, target feature and influence value as prediction features into the prediction model together with the node time series data for prediction analysis;

[0128] The output of the prediction model is set as a new link classification task according to the binary label classification, and the prediction result is output according to the classification task to obtain the correlation between the drug and the adverse reaction of the patient according to the doctor's advice information within the predicted course of the disease, and the influence degree of the drug and the adverse reaction is determined according to the influence value;

[0129] The relationship and impact degree are visualized as a target adverse reaction graph, and the potential risks of medical order information are identified based on the target adverse reaction graph.

[0130] The third aspect of the present invention discloses a risk prediction device for adverse reactions related to the clinical combination of Chinese and Western medicines, comprising:

[0131] at least one processor, and

[0132] a memory communicatively connected to at least one processor; wherein,

[0133] The memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that the at least one processor can execute any of the risk prediction methods for clinical adverse reactions related to the combination of Chinese and Western medicines disclosed in the first aspect of the present invention.

[0134] The computer device may be a terminal, and the computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a risk prediction method for adverse reactions related to the combination of Chinese and Western medicines in clinical practice is implemented. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0135] The fourth aspect of the present invention discloses a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute any of the risk prediction methods for adverse reactions related to the combination of Chinese and Western medicines in clinical practice disclosed in the first aspect of the present invention.

[0136] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the risk prediction method for the adverse reactions related to the clinical combination of Chinese and Western medicines as described above. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0137] Alternatively, if the above modules of the present invention are implemented in the form of software function modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the relevant technology. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to execute all or part of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, RAM, ROM, magnetic disks or optical disks.

[0138] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0139] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0140] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0141] Finally, it should be noted that the method and system for predicting the risk of adverse reactions related to the clinical combination of Chinese and Western medicines disclosed in the embodiments of the present invention only disclose the preferred embodiments of the present invention, which are only used to illustrate the technical scheme of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical schemes described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical schemes from the spirit and scope of the technical schemes of the embodiments of the present invention.

Claims

1. A method for predicting the risk of adverse reactions related to the combination of Chinese and Western medicines in clinical practice, applied to a risk prediction system, characterized in that: The method comprises: Collecting prior data, historical clinical data and actual electronic medical record data, and building a drug-adverse reaction mapping database based on the prior data and historical clinical data; Performing knowledge representation learning and knowledge graph modeling of the drug-adverse reaction according to the mapping relationship database; According to the actual electronic medical record data, a self-supervised model is used to obtain the target features of the medical advice information and potential adverse reactions on a time dynamic graph; Construct a prediction model, analyze and predict the relationship between the doctor's order information input into the prediction model and the adverse reaction, and visualize the correlation between the drug and the adverse reaction, the potential risk and the basis of the risk.

2. The method for predicting the risk of adverse reactions associated with the clinical use of Chinese and Western medicines according to claim 1, characterized in that: Collect prior data, historical clinical data, and real-world electronic medical record data, including: Collect prior data and historical clinical data from public data sources, including prior knowledge bases, laboratory databases, bioinformatics bases, Chinese and Western medicine knowledge bases, and adverse reaction reports. The prior data and historical clinical data are knowledge and experience data that describe the relationship and knowledge information between drugs and adverse reactions, and use unified coding to integrate the relationship between drugs and adverse reactions; Collect actual electronic medical record data in the medical information system and the electronic medical record system. The actual electronic medical record data is medical information related to adverse reactions in the actual medication process. The medical information includes clinical medical orders, patient medication, drug prescriptions in drug trials, medication symptoms, and adverse reaction symptoms.

3. The method for predicting the risk of adverse reactions associated with the clinical use of Chinese and Western medicines according to claim 2, characterized in that: A drug-adverse reaction mapping database is constructed based on the prior data and historical clinical data, specifically including: Entity recognition and relationship extraction are used to obtain the prior relationship between drug-adverse reactions in the prior data and historical clinical data, and to predict the actual relationship between drug-adverse reactions in the actual electronic medical record data; Modify the unified coding of drug-adverse reaction relationships based on the a priori and actual relationships; Based on the unified coding, entity relationships are constructed through semantic networks, and the entity relationships include drug names, drug numbers, drug ingredients, drug applicable disease types, drug side effects, and relationships between targets.

4. The method for predicting the risk of adverse reactions associated with the clinical use of Chinese and Western medicines according to claim 3, characterized in that: The knowledge representation learning of the drug-adverse reaction is performed according to the mapping relationship database, specifically including: Obtaining original feature signals in the adverse reaction dataset; Constructing an initial adverse reaction graph according to the entity relationship information and the original feature signal in the mapping relationship database; The nodes in the initial adverse reaction graph are mapped to a low-dimensional vector space to obtain node vectors and relationship node vectors, wherein the node vectors represent entities and the relationship node vectors represent associated knowledge of entities.

5. The method for predicting the risk of adverse reactions associated with the clinical combination of Chinese and Western medicines according to claim 4, characterized in that: The knowledge graph modeling of the drug-adverse reaction is performed according to the mapping relationship database, specifically including: Obtain an adverse reaction dynamic graph based on the historical adverse reaction data set to form a knowledge graph of drugs and adverse reactions, wherein the adverse reaction dynamic graph includes a historical node time series set of drugs and adverse reactions in historical clinical data; A self-supervised learning model is constructed based on a time-series graph network and an attention mechanism, and the historical node time-series set, node vector, and relationship node vector are used as inputs of the self-supervised learning model.

6. The method for predicting the risk of adverse reactions associated with the clinical combination of Chinese and Western medicines according to claim 5, characterized in that: According to the historical clinical data, a self-supervised model is used to obtain the target features of the doctor's advice information and potential adverse reactions on the time dynamic graph, including: The local features of the self-supervised learning model are fused by an attention mechanism to obtain target features in the adverse reaction dynamic graph that represent changes in drugs and adverse reactions over time, wherein the target features are entity relationships expressed in nodes in the adverse reaction dynamic graph; Feature ablation is used to evaluate the impact of the target feature on predicting the association between the drug and the adverse reaction.

7. The method for predicting the risk of adverse reactions associated with the clinical use of Chinese and Western medicines according to claim 1, characterized in that: Construct a prediction model, analyze and predict the relationship between the doctor's advice information input into the prediction model and the adverse reactions, and visualize the correlation between the drug and the adverse reactions, potential risks, and risk basis, including: A prediction model is constructed using a deep neural network, the target features pre-trained by the self-supervised learning model are integrated according to the attention mechanism, the cross entropy of the prediction model is selected based on binary label classification, and the prediction model is optimized by continuously updating the network weights using the adam optimizer; Obtaining node time series data in the medical order information, and inputting the original feature signal, node vector, relationship node vector, target feature and influence value as prediction features together with the node time series data into the prediction model for prediction analysis; The output of the prediction model is set as a new link classification task according to the binary label classification, and the prediction result is output according to the classification task to obtain the correlation between the drug and the adverse reaction of the patient within the predicted course of disease according to the doctor's advice information, and the influence degree of the drug and the adverse reaction is determined according to the influence value; The correlation and impact degree are visualized as a target adverse reaction graph, and the potential risk of the medical order information is identified according to the target adverse reaction graph.

8. A risk prediction system for adverse reactions related to the combination of Chinese and Western medicines in clinical practice, characterized in that: The system comprises: A data acquisition module is used to collect prior data, historical clinical data and actual electronic medical record data, and to construct a drug-adverse reaction mapping database based on the prior data and historical clinical data; A knowledge modeling module, used for performing knowledge representation learning and knowledge graph modeling of the drug-adverse reaction according to the mapping relationship database; A prediction training module, used for obtaining target features of medical advice information and potential adverse reactions on a time dynamic graph using a self-supervised model based on the actual electronic medical record data; The risk identification module is used to build a prediction model, analyze and predict the relationship between the doctor's order information input into the prediction model and the adverse reaction, and visualize the correlation between the drug and the adverse reaction, the potential risk and the risk basis.

9. A clinical risk prediction device for adverse reactions related to the combination of Chinese and Western medicines, characterized in that: include: at least one processor, and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the risk prediction method for adverse reactions related to the clinical combination of Chinese and Western medicines as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the risk prediction method for adverse reactions related to the clinical combination of Chinese and Western medicines as described in any one of claims 1 to 7.

Images