CN115775635A - Medicine risk identification method and device based on deep learning model and terminal equipment - Google Patents
Medicine risk identification method and device based on deep learning model and terminal equipment Download PDFInfo
- Publication number
- CN115775635A CN115775635A CN202211465161.9A CN202211465161A CN115775635A CN 115775635 A CN115775635 A CN 115775635A CN 202211465161 A CN202211465161 A CN 202211465161A CN 115775635 A CN115775635 A CN 115775635A
- Authority
- CN
- China
- Prior art keywords
- adverse reaction
- training data
- data
- medicine
- determining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The application is applicable to the technical field of neural network models, and provides a medicine risk identification method, a medicine risk identification device and terminal equipment based on a deep learning model, wherein the method comprises the following steps: acquiring real world data of a medicine to be processed, inputting the real world data of the medicine to be processed into a pre-trained key variable relation extraction model for processing to obtain a variable relation, determining a target variable corresponding to the target variable relation, determining a corresponding adverse reaction name according to the target variable, matching the adverse reaction name with a pre-stored adverse reaction name, and determining the risk type of the real world data of the medicine to be processed based on a matching result. According to the method and the device, the variable relation of the real-world data of the medicine is accurately identified based on the pre-trained key variable relation extraction model, so that the risk type of the real-world data of the medicine to be processed is obtained based on the key variable relation, the medicine risk identification efficiency and precision are improved, and the accuracy and the reliability of judgment of the correlation between the risk and the medicine are improved.
Description
Technical Field
The application belongs to the technical field of neural network models, and particularly relates to a medicine risk identification method and device based on a deep learning model, and a terminal device.
Background
In the related research process aiming at the use condition, potential benefit and risk of the medicine, a large amount of real world data (derived from various data related to the health condition and/or diagnosis and treatment and health care of patients) such as hospital information system data, medical insurance payment data, disease registration data, public health monitoring data (such as medicine safety monitoring, death information registration and out-of-hospital health monitoring), natural crowd queue data, individual health monitoring data from mobile equipment and the like needs to be acquired based on health medicine big data.
Disclosure of Invention
The embodiment of the application provides a medicine risk identification method and device based on a deep learning model, and can solve the problems of low efficiency and low accuracy of a related medicine risk identification method.
In a first aspect, an embodiment of the present application provides a drug risk identification method based on a deep learning model, including:
acquiring real world data of a medicine to be processed;
inputting the real world data of the medicine to be processed into a pre-trained key variable relation extraction model for processing to obtain a variable relation;
determining a target variable corresponding to the target variable relation, and determining a corresponding adverse reaction name according to the target variable;
and matching the adverse reaction name with a pre-stored adverse reaction name, and determining the risk type of the real world data of the medicine to be processed based on the matching result.
In one embodiment, before acquiring the real world data of the drug to be processed, the method includes:
acquiring a plurality of original training data;
preprocessing the original training data, constructing a training data set based on the preprocessed original training data, inputting the training data set into a key variable relation extraction model for pre-training, and obtaining the pre-trained key variable relation extraction model.
In one embodiment, the key variable relation extraction model is formed by sequentially connecting a sequence annotation model and a seasonal autoregressive moving average model;
the preprocessing the original training data, constructing a training data set based on the preprocessed original training data, inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model, and the method comprises the following steps:
preprocessing the original training data by a preset data processing method to obtain preprocessed original training data; the preset data processing method comprises at least one of the processes of de-marking and de-duplicating;
when the preprocessed original training data are detected to be structured data, constructing a training data set based on the preprocessed original training data;
when the preprocessed original training data are detected to be unstructured data, adding corresponding variable labels to the preprocessed original training data, and constructing a training data set based on the preprocessed original training data carrying the variable labels;
and inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
In one embodiment, the determining a target variable corresponding to a target variable relationship and determining a corresponding adverse reaction name according to the target variable includes:
determining a target variable relation which meets a preset condition;
determining a target variable corresponding to the target variable relation, and performing code conversion on the target variable through a preset code conversion method to obtain an adverse reaction name corresponding to the target variable; the preset code conversion method comprises a MedDRA code conversion method, a WHOART code conversion method, a WHO-DD code conversion method or an ICD-10 code conversion method.
In one embodiment, the matching the adverse reaction name with a pre-stored adverse reaction name, and determining the risk type of the real-world data of the drug to be processed based on the matching result includes:
matching the adverse reaction name with a pre-stored adverse reaction name;
and when a pre-stored adverse reaction name matched with the adverse reaction name is detected, determining the risk type of the real world data of the medicine to be processed as the identified risk.
In one embodiment, after the matching the adverse reaction name with a pre-stored adverse reaction name, the method further includes:
and when the pre-stored adverse reaction name matched with the adverse reaction name is not detected, determining the risk type of the real world data of the medicine to be processed as a potential risk.
In a second aspect, an embodiment of the present application provides a drug risk identification device based on a deep learning model, including:
the original data acquisition module is used for acquiring real world data of the medicine to be processed;
the model processing module is used for inputting the real world data of the medicine to be processed into a pre-trained key variable relation extraction model for processing to obtain a variable relation;
the adverse reaction determining module is used for determining a target variable corresponding to the target variable relation and determining a corresponding adverse reaction name according to the target variable;
and the risk type determining module is used for matching the adverse reaction name with a pre-stored adverse reaction name and determining the risk type of the real world data of the medicine to be processed based on the matching result.
In one embodiment, the apparatus further comprises:
the training data acquisition module is used for acquiring a plurality of original training data;
and the pre-training module is used for preprocessing the original training data, constructing a training data set based on the preprocessed original training data, and inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
In one embodiment, the key variable relation extraction model is formed by sequentially connecting a sequence annotation model and a seasonal autoregressive moving average model;
in one embodiment, the pre-training module comprises:
the preprocessing unit is used for preprocessing the original training data through a preset data processing method to obtain preprocessed original training data; the preset data processing method comprises at least one of a de-marking process and a de-duplication process;
a first training set constructing unit, configured to construct a training data set based on the preprocessed original training data when it is detected that the preprocessed original training data is structured data;
a second training set constructing unit, configured to add a corresponding variable label to the preprocessed original training data when it is detected that the preprocessed original training data is unstructured data, and construct a training data set based on the preprocessed original training data carrying the variable label;
and the pre-training unit is used for inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
In one embodiment, the adverse reaction determination module comprises:
the category determining unit is used for determining a target variable relation meeting a preset condition;
the variable conversion unit is used for determining a target variable corresponding to the target variable relation, and performing code conversion on the target variable through a preset code conversion method to obtain an adverse reaction name corresponding to the target variable; the preset code conversion method comprises a MedDRA code conversion method, a WHOART code conversion method, a WHO-DD code conversion method or an ICD-10 code conversion method.
In one embodiment, the risk type determination module includes:
the matching unit is used for matching the adverse reaction name with a pre-stored adverse reaction name;
and the first risk identification unit is used for determining the risk type of the real world data of the medicine to be processed as the identified risk when a pre-stored adverse reaction name matched with the adverse reaction name is detected.
In one embodiment, the adverse reaction determination module further comprises:
and the second risk identification unit is used for determining that the risk type of the real world data of the medicine to be processed is a potential risk when the pre-stored adverse reaction name matched with the adverse reaction name is not detected.
In a third aspect, an embodiment of the present application provides a terminal device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor, when executing the computer program, implements the deep learning model-based drug risk identification method according to any one of the above first aspects.
In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the computer program implements the deep learning model-based drug risk identification method according to any one of the above first aspects.
In a fifth aspect, the present application provides a computer program product, which, when run on a terminal device, causes the terminal device to execute the deep learning model-based drug risk identification method according to any one of the first aspect.
Compared with the prior art, the embodiment of the application has the advantages that: identifying real world data of a drug to be processed by a pre-trained key variable relation extraction model to obtain a variable relation, determining a corresponding adverse reaction name according to a target variable corresponding to the target variable relation, matching the adverse reaction name with a pre-stored adverse reaction name, thereby determining a risk type of the real world data of the drug to be processed, realizing accurate identification of the variable relation of the drug data based on a neural network model, obtaining a risk type of the corresponding real world data of the drug to be processed based on the key variable relation, improving the efficiency and precision of drug risk identification, and further improving the accuracy and reliability of risk and drug correlation judgment.
It is to be understood that, for the beneficial effects of the second aspect to the fifth aspect, reference may be made to the relevant description in the first aspect, and details are not described herein again.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a schematic flow chart of a drug risk identification method based on a deep learning model according to an embodiment of the present application;
FIG. 2 is another schematic flow chart of a drug risk identification method based on a deep learning model according to an embodiment of the present application;
FIG. 3 is a schematic flowchart of step S202 of a deep learning model-based drug risk identification method according to an embodiment of the present application;
FIG. 4 is a schematic structural diagram of a medicine risk identification device based on a deep learning model according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a terminal device according to an embodiment of the present application.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.
It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It should also be understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to" determining "or" in response to detecting ". Similarly, the phrase "if it is determined" or "if a [ described condition or event ] is detected" may be interpreted contextually to mean "upon determining" or "in response to determining" or "upon detecting [ described condition or event ]" or "in response to detecting [ described condition or event ]".
Furthermore, in the description of the present application and the appended claims, the terms "first," "second," "third," and the like are used for distinguishing between descriptions and not necessarily for describing or implying relative importance.
Reference throughout this specification to "one embodiment" or "some embodiments," or the like, means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," or the like, in various places throughout this specification are not necessarily all referring to the same embodiment, but rather "one or more but not all embodiments" unless specifically stated otherwise. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless expressly specified otherwise.
The drug risk identification method based on the deep learning model provided by the embodiment of the application can be applied to terminal devices such as mobile phones, tablet computers, notebook computers, ultra-mobile personal computers (UMPCs) and the like, and the embodiment of the application does not limit the specific types of the terminal devices.
Fig. 1 shows a schematic flow chart of the drug risk identification method based on deep learning model provided in the present application, which can be applied to the above-mentioned notebook computer by way of example and not limitation.
S101, acquiring real world data of the medicine to be processed.
Specifically, the relevant data of the drug to be treated, which needs risk identification, is used as the real-world data of the drug to be treated. Real World Studies (RWS), i.e. collecting Data (RWD) related to a patient in a Real World environment, and obtaining clinical evidence (RWE) of the use value and potential benefit or risk of a medical product through analysis, the main type of Study is observational Study, and also clinical trials. Real world data includes, but is not limited to, hospital Information System (HIS) data acquired based on HIS, related medical insurance reimbursement data for the drug, and on-board device data related to the drug (e.g., aerosol data related to asthma drugs, etc.).
It is to be understood that the above-mentioned real world data of the pharmaceutical product to be processed may be structured data or unstructured data. Wherein the structured data is data with a high degree of organization and regular formatting. Unstructured data is data other than structured data.
S102, inputting the real world data of the medicine to be processed into a pre-trained key variable relation extraction model for processing to obtain a variable relation.
Specifically, the real world data of the medicine to be processed is input into a pre-trained key variable relation extraction model for processing, so that the pre-trained key variable relation extraction model identifies variables in the real world data of the medicine to be processed, and a corresponding variable relation is determined based on the variables. The variables are attribute data of the drug, including but not limited to disease (specifically referring to a name of a certain disease), examination (specifically referring to related medical examination items and corresponding examination data for the disease), treatment (specifically referring to related drug treatment methods and means for the disease), drug (specifically referring to a name of a drug used by a user when the disease is treated), symptom/diagnosis (specifically referring to a diagnosis basis of the disease, vital sign data of the user, and a body state representation), and outcome (specifically referring to a certain treatment method, vital sign data of the user after the drug is treated, and a body state representation). Variable relationships are used to represent causal relationships between two variables, including but not limited to TrWS (treatment worsens symptoms), trWD (treatment worsens disease), trNAS (treatment is not taken because of symptoms), trNAD (treatment is not taken because of disease), trCS (treatment causes symptoms), trCD (treatment causes disease), trs (treatment improves symptoms), trID (treatment improves disease), trAS (treatment is applied to symptoms), trAD (treatment is applied to disease), teRS (examination confirms symptoms), teRD (examination confirms disease), teCD (examination takes examination to confirm disease), teAS (collection of treatment because of symptoms), SID (symptoms indicate disease), DCS (disease causes symptoms).
S103, determining a target variable corresponding to the target variable relation, and determining a corresponding adverse reaction name according to the target variable.
Specifically, based on different influence degrees of different causal relationships on drug risk identification, one or more variable relationships that are most important for drug risk identification need to be selected from variable relationships among multiple drug variables as target variable relationships, target variables corresponding to output target variable relationships are determined, and adverse reaction names corresponding to drugs to be treated are determined by processing the target variables.
For example, setting the target variable relationships includes TrWS (treatment aggravates symptoms), trWD (treatment aggravates disease), trCS (treatment causes symptoms), trCD (treatment causes disease).
S104, matching the adverse reaction name with a pre-stored adverse reaction name, and determining the risk type of the real world data of the medicine to be processed based on the matching result.
Specifically, when it is detected that the relevant data such as the drug specification of the drug to be treated and the like contain relevant one or more adverse reaction data, the adverse reaction data is converted into an adverse reaction name and is pre-stored, so as to obtain a pre-stored adverse reaction name.
Specifically, the adverse reaction name related to the target variable is matched with pre-stored adverse reaction names to obtain a corresponding matching result (the matching result comprises a first matching result that the adverse reaction name is the same as a certain pre-stored adverse reaction name or a second matching result that the adverse reaction name is different from all the pre-stored adverse reaction names), and the risk type of the real world data of the drug to be processed is determined according to the matching result of the adverse reaction name and the pre-stored adverse reaction names.
As shown in fig. 2, in an embodiment, before acquiring the real-world data of the drug to be processed, the method includes:
s201, acquiring a plurality of original training data;
s202, preprocessing the original training data, constructing a training data set based on the preprocessed original training data, inputting the training data set into a key variable relation extraction model for pre-training, and obtaining the pre-trained key variable relation extraction model.
Specifically, a plurality of original training data are obtained through an authorized related medical information database, the original training data are preprocessed, data which are difficult to identify variables and are repetitive are screened out, preprocessed original training data are obtained, a training data set is constructed on the basis of the preprocessed original training data, the training data set is input into a key variable relation extraction model to be pre-trained, and a pre-trained key variable relation extraction model is obtained.
In one embodiment, the key variable relation extraction model is formed by connecting a sequence annotation model and a seasonal autoregressive moving average model in sequence.
Specifically, the key variable relation extraction model is formed by sequentially connecting a sequence annotation model (Bi-LSTM-CRF) and an applied seasonal Autoregressive Moving Average model (ARIMA). Wherein, the sequence labeling (Bi-LSTM-CRF) model comprises an input characteristic layer, a Bi-LSTM intermediate layer and a CRF output layer; the input feature layer divides words of unstructured data through a Stanford Parser tool, defines features according to the characteristics of entities in the medical field, constructs feature vectors, obtains the feature vectors of the entities as input feature data, and inputs the input feature data into the Bi-LSTM intermediate layer; the Bi-LSTM middle layer is used for extracting context characteristics of input sequence information by utilizing a bidirectional LSTM recurrent neural network, splicing LSTM characteristic results in two directions and inputting the splicing results into a CRF output layer; the CRF output layer is used for outputting different variable labels, namely sequence labeling results of unstructured data (wherein the structured data can be automatically labeled based on a semantic relation network formed by the structured data). And determining the variable relation corresponding to the variable label according to the sequence labeling result carried by the variable label. And the ARIMA model corrects the recognition result of the sequence labeling (Bi-LSTM-CRF) model to improve the classification precision of the variable relation and obtain a pre-trained key variable relation extraction model.
As shown in fig. 3, in an embodiment, the step S202 of preprocessing the original training data, constructing a training data set based on the preprocessed original training data, and inputting the training data set into a key variable relationship extraction model for pre-training to obtain the pre-trained key variable relationship extraction model includes:
s2021, preprocessing the original training data by a preset data processing method to obtain preprocessed original training data; the preset data processing method comprises at least one of a de-marking process and a de-duplication process;
s2022, when the preprocessed original training data are detected to be structured data, constructing a training data set based on the preprocessed original training data;
s2023, when the preprocessed original training data are detected to be unstructured data, adding corresponding variable labels to the preprocessed original training data, and constructing a training data set based on the preprocessed original training data carrying the variable labels;
s2024, inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
Specifically, the raw training data set is preprocessed by a preset data processing method (the preset data processing method includes, but is not limited to, at least one of a de-labeling process and a de-duplication process), so that non-repetitive preprocessed raw training data capable of conveniently identifying variables is obtained. When the preprocessed original training data are detected to be structured data, the structured data can identify and obtain corresponding variables based on the constructed semantic relation network, so that a training data set can be directly constructed based on the preprocessed original training data. When it is detected that the preprocessed original training data is unstructured data, corresponding variable labels need to be added to the preprocessed original training data due to the characteristics of the unstructured data (corresponding variables are difficult to identify due to no high organization and regular formatting), and a training data set is constructed based on the preprocessed original training data carrying the variable labels. And inputting the training data set into the key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
Specifically, the process of de-marking is to replace the private information such as the name of the patient, the home address of the patient, and the like in the real world data of the medicine to be processed with distinguishable and ordered numbers and hide the identity information. The duplication elimination processing refers to eliminating repeated items of information of the same case number in the database, and comprises the following steps: filtering the data by using Bloom filter to remove repeated data; firstly, mapping data into a bit array by using N hash functions, calculating N hash values, and if the calculated hash values exist in the bit array, indicating that the data already exists, and filtering the data.
In one embodiment, the determining a target variable corresponding to a target variable relationship and determining a corresponding adverse reaction name according to the target variable includes:
determining a target variable relation which meets a preset condition;
determining a target variable corresponding to the target variable relation, and performing code conversion on the target variable through a preset code conversion method to obtain an adverse reaction name corresponding to the target variable; the preset code conversion method comprises a MedDRA code conversion method, a WHOART code conversion method, a WHO-DD code conversion method or an ICD-10 code conversion method.
Specifically, the preset conditions are set as variable relationships (for example, including TrWS (treatment worsened symptoms), trWD (treatment worsened diseases), trCS (treatment worsened symptoms), trCD (treatment caused diseases)) in which the influence between the treatment and the disease or symptom variables is negative, corresponding target variable relationships meeting the preset conditions are screened from variable relationships output by a pre-trained key variable relationship extraction model, target variables corresponding to the target variable relationships are determined, and the target variables are subjected to coding conversion through a preset coding method to obtain adverse reaction names corresponding to the target variables; the preset transcoding method includes, but is not limited to, a MedDRA transcoding method, a WHOART transcoding method, a WHO-DD transcoding method, or an ICD-10 transcoding method. MedDRA is a set of international medical terms created under the initiative of ICH. The MedDRA is used for the administrative management of the whole research and development and application period of medical products, and is used for classifying, retrieving, reporting and exchanging medical information. The MedDRA is a medical dictionary for new drug registration, and is suitable for safety reports of all medical and diagnostic products under government registration jurisdiction. In the new drug registration link, the product information of MedDRA, such as clinical research, spontaneous report of adverse reaction, registration report and government registration management, is required. MedDRA is used for adverse reaction monitoring after the medicine is on the market, and is used for adverse reaction report and data analysis of the medicine, and the like. The MedDRA code conversion method is an adverse reaction code conversion method realized based on a MedDRA dictionary.
WHOART is specifically a highly accurate set of terms used to encode and encode clinical information during drug therapy, primarily for free use in member countries participating in the WHO drug monitoring program, as well as in pharmaceutical enterprises and clinical research institutions throughout the world. The WHOART term set has been the basis for rational coding of adverse reaction terms since 30 years of development. WHOART covers almost all of the medical terms required in adverse event reports, but is still small and delicate and can be printed in the form of a line list. Because new drugs and new indications will generate new adverse reaction terms, the structure of the term set is flexible and variable, allowing new terms to be incorporated on the basis of preserving the structure of the term set, while the relationship between previous terms may not be lost. The WHOART code conversion method is an adverse reaction code conversion method realized by a base WHOART term set.
WHO-DD is specifically the world health organization drug dictionary (WHO drug dictionary). The WHO-DD coding conversion method is an adverse reaction coding conversion method realized based on a WHO-DD dictionary.
The International national Classification of Diseases and Related Health issues 10th resolution (ICD-10), the tenth edition of The International disease injury and death cause Classification Standard, is a system which classifies Diseases according to certain characteristics of Diseases and rules by The World Health Organization (WHO) and is expressed by a coding method. The ICD-10 code conversion method is an adverse reaction code conversion method realized based on an ICD-10 dictionary.
For example, the name of a drug for treating a disease a is a, the a drug carries a treatment variable a1 and a disease variable a2, and when the variable relationship of the a drug is detected to be TrCD, the a1 and the a2 are converted by a MedDRA code conversion method to obtain an adverse reaction name corresponding to a target variable (for example, the adverse reaction name is an adverse reaction name which may cause a disease a2 to appear in a user by a treatment means a1: taking the drug a).
In one embodiment, the matching the adverse reaction name with a pre-stored adverse reaction name, and determining the risk type of the real-world data of the drug to be processed based on the matching result includes:
matching the adverse reaction name with a pre-stored adverse reaction name;
and when a pre-stored adverse reaction name matched with the adverse reaction name is detected, determining the risk type of the real world data of the medicine to be processed as the identified risk.
Specifically, the adverse reaction name of the drug to be processed obtained based on code conversion is matched with a pre-stored adverse reaction name of the drug to be processed, when a certain pre-stored adverse reaction name is detected to be the same as the adverse reaction name of the drug to be processed, the existence of the pre-stored adverse reaction name matched with the adverse reaction name is judged, and the risk type of the real world data of the drug to be processed is determined to be an identified risk, namely the drug to be processed does not have other risks. Correspondingly, a first risk warning notice is generated, wherein the first risk warning notice comprises the identified risk of the risk type of the real world data of the medicine to be processed and the corresponding adverse reaction name.
In one embodiment, the matching the adverse reaction name with a pre-stored adverse reaction name, and determining the risk type of the real-world data of the drug to be processed based on the matching result further includes:
and when the pre-stored adverse reaction name matched with the adverse reaction name is not detected, determining the risk type of the real world data of the medicine to be processed as a potential risk.
Specifically, when any pre-stored adverse reaction name is detected to be different from the aforementioned adverse reaction name, judging that the adverse reaction name is not matched with the pre-stored adverse reaction name, and determining the risk type of the real world data of the drug to be processed as a potential risk. Correspondingly, a second risk warning notification is generated, wherein the second risk warning notification includes that the risk type of the real world data of the medicine to be processed is a potential risk, and the adverse reaction name which is not matched with the pre-stored adverse reaction name.
The risk types of the medicines to be processed are identified through the neural network model, and the obtained identification results are used as the basis for formulating a medicine risk management plan and the important basis for considering the medicine benefit risk, so that the efficiency and the precision of medicine risk identification are improved, and the inspection efficiency for evaluating the effectiveness and the safety of medical products is improved.
In one embodiment, when it is detected that a certain drug to be treated contains at least two adverse reaction names, a matching operation with a pre-stored adverse reaction name needs to be performed for each adverse reaction name, and when it is detected that any one adverse reaction name is not matched with the corresponding pre-stored adverse reaction name, it is determined that the risk type of the real world data of the drug to be treated is a potential risk.
In this embodiment, a pre-trained key variable relationship extraction model is used to identify real world data of a drug to be processed to obtain a variable relationship, a corresponding adverse reaction name is determined according to a target variable corresponding to the target variable relationship, and the adverse reaction name is matched with a pre-stored adverse reaction name, so as to determine a risk type of the real world data of the drug to be processed, and accurately identify a variable relationship of the drug data based on a neural network model, thereby obtaining a risk type of the corresponding real world data of the drug to be processed based on the key variable relationship, improving the efficiency and accuracy of drug risk identification, and further improving the accuracy and reliability of risk and drug correlation determination.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.
Corresponding to the method for identifying drug risk based on deep learning model described in the above embodiments, fig. 4 shows a block diagram of a device for identifying drug risk based on deep learning model provided in the embodiments of the present application, and for convenience of description, only the relevant parts of the embodiments of the present application are shown.
Referring to fig. 4, the drug risk identification device 100 based on the deep learning model includes:
the original data acquisition module 101 is used for acquiring real world data of the medicine to be processed;
the model processing module 102 is configured to input the real world data of the drug to be processed into a pre-trained key variable relationship extraction model for processing to obtain a variable relationship;
the adverse reaction determining module 103 is used for determining a target variable corresponding to the target variable relationship and determining a corresponding adverse reaction name according to the target variable;
and a risk type determining module 104, configured to match the adverse reaction name with a pre-stored adverse reaction name, and determine a risk type of the real-world data of the drug to be processed based on the matching result.
In one embodiment, the apparatus further comprises:
the training data acquisition module is used for acquiring a plurality of original training data;
and the pre-training module is used for preprocessing the original training data, constructing a training data set based on the preprocessed original training data, and inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
In one embodiment, the key variable relation extraction model is formed by sequentially connecting a sequence annotation model and a seasonal autoregressive moving average model;
in one embodiment, the pre-training module comprises:
the preprocessing unit is used for preprocessing the original training data by a preset data processing method to obtain preprocessed original training data; the preset data processing method comprises at least one of a de-marking process and a de-duplication process;
a first training set constructing unit, configured to construct a training data set based on the preprocessed original training data when it is detected that the preprocessed original training data is structured data;
a second training set constructing unit, configured to add a corresponding variable label to the preprocessed original training data when it is detected that the preprocessed original training data is unstructured data, and construct a training data set based on the preprocessed original training data carrying the variable label;
and the pre-training unit is used for inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
In one embodiment, the adverse reaction determination module comprises:
the category determining unit is used for determining a target variable relation meeting a preset condition;
the variable conversion unit is used for determining a target variable corresponding to the target variable relation, and performing code conversion on the target variable through a preset code conversion method to obtain an adverse reaction name corresponding to the target variable; the preset code conversion method comprises a MedDRA code conversion method, a WHOART code conversion method, a WHO-DD code conversion method or an ICD-10 code conversion method.
In one embodiment, the risk type determination module includes:
the matching unit is used for matching the adverse reaction name with a pre-stored adverse reaction name;
and the first risk identification unit is used for determining that the risk type of the real world data of the medicine to be processed is identified risk when a pre-stored adverse reaction name matched with the adverse reaction name is detected.
In one embodiment, the adverse reaction determination module further comprises:
and the second risk identification unit is used for determining the risk type of the real world data of the medicine to be processed as a potential risk when the pre-stored adverse reaction name matched with the adverse reaction name is not detected.
In the embodiment, the variable relation is obtained by identifying the real world data of the drug to be processed through the pre-trained key variable relation extraction model, the corresponding adverse reaction name is determined according to the target variable corresponding to the target variable relation, and the adverse reaction name is matched with the pre-stored adverse reaction name, so that the risk type of the real world data of the drug to be processed is determined, the variable relation of the drug data is accurately identified based on the neural network model, the risk type of the real world data of the drug to be processed is obtained based on the key variable relation, the drug risk identification efficiency and accuracy are improved, and the accuracy and reliability of risk and drug correlation judgment are improved.
It should be noted that, for the information interaction, execution process, and other contents between the above devices/units, the specific functions and technical effects thereof based on the same concept as those of the method embodiment of the present application can be specifically referred to the method embodiment portion, and are not described herein again.
Fig. 5 is a schematic structural diagram of the terminal device provided in this embodiment. As shown in fig. 5, the terminal device 5 of this embodiment includes: at least one processor 50 (only one shown in fig. 5), a memory 51, and a computer program 52 stored in the memory 51 and executable on the at least one processor 50, wherein the processor 50 executes the computer program 52 to implement the steps of any of the above-mentioned various deep learning model-based drug risk identification method embodiments.
The terminal device 5 may be a desktop computer, a notebook, a palm computer, a cloud server, or other computing devices. The terminal device may include, but is not limited to, a processor 50, a memory 51. Those skilled in the art will appreciate that fig. 5 is only an example of the terminal device 5, and does not constitute a limitation to the terminal device 5, and may include more or less components than those shown, or combine some components, or different components, such as an input-output device, a network access device, and the like.
The Processor 50 may be a Central Processing Unit (CPU), and the Processor 50 may be other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 51 may in some embodiments be an internal storage unit of the terminal device 5, such as a hard disk or a memory of the terminal device 5. The memory 51 may also be an external storage device of the terminal device 5 in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, which are provided on the terminal device 5. Further, the memory 51 may also include both an internal storage unit and an external storage device of the terminal device 5. The memory 51 is used for storing an operating system, an application program, a BootLoader (BootLoader), data, and other programs, such as program codes of the computer program. The memory 51 may also be used to temporarily store data that has been output or is to be output.
It should be clear to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional units and modules is only used for illustration, and in practical applications, the above function distribution may be performed by different functional units and modules as needed, that is, the internal structure of the apparatus may be divided into different functional units or modules to perform all or part of the above described functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. For the specific working processes of the units and modules in the system, reference may be made to the corresponding processes in the foregoing method embodiments, which are not described herein again.
An embodiment of the present application further provides a network device, where the network device includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, the processor implementing the steps of any of the various method embodiments described above when executing the computer program.
The embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps in the above-mentioned method embodiments.
The embodiments of the present application provide a computer program product, which when running on a mobile terminal, enables the mobile terminal to implement the steps in the above method embodiments when executed.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium and can implement the steps of the embodiments of the methods described above when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include at least: any entity or device capable of carrying computer program code to a photographing apparatus/terminal apparatus, a recording medium, computer Memory, read-Only Memory (ROM), random Access Memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc. In some jurisdictions, computer-readable media may not be an electrical carrier signal or a telecommunications signal in accordance with legislative and proprietary practices.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus/network device and method may be implemented in other ways. For example, the above-described apparatus/network device embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implementing, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the embodiments of the present application, and they should be construed as being included in the present application.
Claims (10)
1. A drug risk identification method based on a deep learning model is characterized by comprising the following steps:
acquiring real world data of a medicine to be processed;
inputting the real world data of the medicine to be processed into a pre-trained key variable relation extraction model for processing to obtain a variable relation;
determining a target variable corresponding to the target variable relation, and determining a corresponding adverse reaction name according to the target variable;
and matching the adverse reaction name with a pre-stored adverse reaction name, and determining the risk type of the real world data of the medicine to be processed based on the matching result.
2. The drug risk identification method based on the deep learning model as claimed in claim 1, wherein before the acquiring real world data of the drug to be processed, the method comprises:
acquiring a plurality of original training data;
preprocessing the original training data, constructing a training data set based on the preprocessed original training data, inputting the training data set into a key variable relation extraction model for pre-training, and obtaining the pre-trained key variable relation extraction model.
3. The drug risk identification method based on the deep learning model as claimed in claim 2, wherein the key variable relation extraction model is formed by connecting a sequence labeling model and a seasonal autoregressive moving average model in sequence;
the preprocessing the original training data, constructing a training data set based on the preprocessed original training data, inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model, and the method comprises the following steps:
preprocessing the original training data by a preset data processing method to obtain preprocessed original training data; the preset data processing method comprises at least one of a de-marking process and a de-duplication process;
when the preprocessed original training data are detected to be structured data, constructing a training data set based on the preprocessed original training data;
when the preprocessed original training data are detected to be unstructured data, adding corresponding variable labels to the preprocessed original training data, and constructing a training data set based on the preprocessed original training data carrying the variable labels;
and inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
4. The drug risk identification method based on the deep learning model as claimed in claim 1, wherein the determining the target variable corresponding to the target variable relationship and determining the corresponding adverse reaction name according to the target variable comprises:
determining a target variable relation which meets a preset condition;
determining a target variable corresponding to the target variable relation, and performing code conversion on the target variable through a preset code conversion method to obtain an adverse reaction name corresponding to the target variable; the preset code conversion method comprises a MedDRA code conversion method, a WHOART code conversion method, a WHO-DD code conversion method or an ICD-10 code conversion method.
5. The drug risk identification method based on the deep learning model as claimed in claim 1, wherein the matching the adverse reaction name with a pre-stored adverse reaction name, and the determining the risk type of the real world data of the drug to be processed based on the matching result comprises:
matching the adverse reaction name with a pre-stored adverse reaction name;
and when a pre-stored adverse reaction name matched with the adverse reaction name is detected, determining the risk type of the real world data of the medicine to be processed as the identified risk.
6. The drug risk identification method based on the deep learning model according to any one of claims 1 to 5, wherein after matching the adverse reaction name with a pre-stored adverse reaction name, further comprising:
and when the pre-stored adverse reaction name matched with the adverse reaction name is not detected, determining the risk type of the real world data of the medicine to be processed as a potential risk.
7. A medicine risk identification device based on a deep learning model is characterized by comprising:
the original data acquisition module is used for acquiring real world data of the medicine to be processed;
the model processing module is used for inputting the real world data of the medicine to be processed into a pre-trained key variable relation extraction model for processing to obtain a variable relation;
the adverse reaction determining module is used for determining a target variable corresponding to the target variable relation and determining a corresponding adverse reaction name according to the target variable;
and the risk type determining module is used for matching the adverse reaction name with a pre-stored adverse reaction name and determining the risk type of the real world data of the medicine to be processed based on the matching result.
8. The deep learning model-based drug risk identification apparatus of claim 7, wherein the apparatus further comprises:
the training data acquisition module is used for acquiring a plurality of original training data;
and the pre-training module is used for preprocessing the original training data, constructing a training data set based on the preprocessed original training data, and inputting the training data set into a key variable relation extraction model for pre-training to obtain the pre-trained key variable relation extraction model.
9. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the method according to any of claims 1 to 6 when executing the computer program.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211465161.9A CN115775635A (en) | 2022-11-22 | 2022-11-22 | Medicine risk identification method and device based on deep learning model and terminal equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211465161.9A CN115775635A (en) | 2022-11-22 | 2022-11-22 | Medicine risk identification method and device based on deep learning model and terminal equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115775635A true CN115775635A (en) | 2023-03-10 |
Family
ID=85389689
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211465161.9A Pending CN115775635A (en) | 2022-11-22 | 2022-11-22 | Medicine risk identification method and device based on deep learning model and terminal equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115775635A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117334347A (en) * | 2023-12-01 | 2024-01-02 | 北京大学 | Method, device, equipment and storage medium for evaluating treatment effect |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190228864A1 (en) * | 2018-01-24 | 2019-07-25 | International Business Machines Corporation | Evaluating Drug-Adverse Event Causality Based on an Integration of Heterogeneous Drug Safety Causality Models |
| CN111415757A (en) * | 2020-03-18 | 2020-07-14 | 南京邮电大学 | Method for mining medicine risk characteristics based on concept generalization |
| CN112669991A (en) * | 2020-12-28 | 2021-04-16 | 山东健康医疗大数据有限公司 | Method for detecting adverse drug reaction signals |
| CN114298484A (en) * | 2021-12-03 | 2022-04-08 | 阿里健康科技(杭州)有限公司 | Drug risk identification method, device, equipment and storage medium |
| CN114492386A (en) * | 2022-01-29 | 2022-05-13 | 清华大学 | Combined detection method for drug name and adverse drug reaction in web text |
| CN114743692A (en) * | 2022-03-21 | 2022-07-12 | 上海工程技术大学 | System and method for predicting adverse drug reactions based on knowledge graph |
| CN114822872A (en) * | 2022-04-14 | 2022-07-29 | 北京左医科技有限公司 | Training method and device of risk signal recognition model and drug risk signal mining method and device |
| CN114898895A (en) * | 2022-04-29 | 2022-08-12 | 国网新疆电力有限公司信息通信公司 | Xinjiang local adverse drug reaction identification method and related device |
| CN115132374A (en) * | 2022-06-23 | 2022-09-30 | 长沙砝码柯数据科技有限责任公司 | System, method, equipment and medium for high-throughput identification of potential risks of drugs |
-
2022
- 2022-11-22 CN CN202211465161.9A patent/CN115775635A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190228864A1 (en) * | 2018-01-24 | 2019-07-25 | International Business Machines Corporation | Evaluating Drug-Adverse Event Causality Based on an Integration of Heterogeneous Drug Safety Causality Models |
| CN111415757A (en) * | 2020-03-18 | 2020-07-14 | 南京邮电大学 | Method for mining medicine risk characteristics based on concept generalization |
| CN112669991A (en) * | 2020-12-28 | 2021-04-16 | 山东健康医疗大数据有限公司 | Method for detecting adverse drug reaction signals |
| CN114298484A (en) * | 2021-12-03 | 2022-04-08 | 阿里健康科技(杭州)有限公司 | Drug risk identification method, device, equipment and storage medium |
| CN114492386A (en) * | 2022-01-29 | 2022-05-13 | 清华大学 | Combined detection method for drug name and adverse drug reaction in web text |
| CN114743692A (en) * | 2022-03-21 | 2022-07-12 | 上海工程技术大学 | System and method for predicting adverse drug reactions based on knowledge graph |
| CN114822872A (en) * | 2022-04-14 | 2022-07-29 | 北京左医科技有限公司 | Training method and device of risk signal recognition model and drug risk signal mining method and device |
| CN114898895A (en) * | 2022-04-29 | 2022-08-12 | 国网新疆电力有限公司信息通信公司 | Xinjiang local adverse drug reaction identification method and related device |
| CN115132374A (en) * | 2022-06-23 | 2022-09-30 | 长沙砝码柯数据科技有限责任公司 | System, method, equipment and medium for high-throughput identification of potential risks of drugs |
Non-Patent Citations (1)
| Title |
|---|
| 朱笑笑等: "基于 Bi-LSTM 和 CRF 的药品不良反应抽取模型构建", 数据分析与知识发现, vol. 3, no. 2, pages 90 - 97 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117334347A (en) * | 2023-12-01 | 2024-01-02 | 北京大学 | Method, device, equipment and storage medium for evaluating treatment effect |
| CN117334347B (en) * | 2023-12-01 | 2024-03-22 | 北京大学 | Method, device, equipment and storage medium for evaluating treatment effect |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110910976A (en) | Medical record detection method, device, equipment and storage medium | |
| CN109670173B (en) | Checking method of reimbursement data, identification server and storage medium | |
| CN111145844B (en) | Comprehensive medical supervision platform | |
| CN113724848A (en) | Medical resource recommendation method, device, server and medium based on artificial intelligence | |
| CN111710420A (en) | Complication morbidity risk prediction method, system, terminal and storage medium based on electronic medical record big data | |
| CN109584980B (en) | Data verification method and device, electronic equipment and storage medium | |
| CN112635011A (en) | Disease diagnosis method, disease diagnosis system, and readable storage medium | |
| CN109671476A (en) | Recognition methods, device, terminal and the computer readable storage medium of unrelated medication | |
| CN109636650A (en) | Recognition methods, device, terminal and the readable storage medium storing program for executing of therapeutic regimen exception | |
| CN113657548A (en) | Medical insurance abnormity detection method and device, computer equipment and storage medium | |
| CN110752027B (en) | Electronic medical record data pushing method, device, computer equipment and storage medium | |
| CN115775635A (en) | Medicine risk identification method and device based on deep learning model and terminal equipment | |
| CN111785383A (en) | Data processing method and related equipment | |
| CN111383726B (en) | Electronic medical record data processing method and device, electronic equipment and readable medium | |
| CN114334175A (en) | Hospital epidemic situation monitoring method and device, computer equipment and storage medium | |
| CN114328968A (en) | Construction method and device of medical knowledge graph, electronic equipment and medium | |
| CN113724830A (en) | Medicine taking risk detection method based on artificial intelligence and related equipment | |
| Roosan et al. | Artificial intelligent context-aware machine-learning tool to detect adverse drug events from social media platforms | |
| CN113157864A (en) | Key information extraction method and device, electronic equipment and medium | |
| CN107680636B (en) | Clinical information processing system and method based on metadata | |
| CN112530602A (en) | Method and device for analyzing side effect of medicine, electronic equipment and computer storage medium | |
| CN111863173A (en) | Medical record quality evaluation method and computing device | |
| CN116884612A (en) | Intelligent analysis method, device, equipment and storage medium for disease risk level | |
| CN116502261A (en) | Data desensitization method and device for retaining data characteristics | |
| US20230377697A1 (en) | System and a way to automatically monitor clinical trials - virtual monitor (vm) and a way to record medical history |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |