Artificial intelligence-based antibacterial drug purpose judgment method and automatic judgment system

Abstract

The application discloses an artificial intelligence-based antibacterial drug purpose judgment method and automatic judgment system, which comprises the following steps: after preprocessing the obtained patient medical data, antibacterial drug clinical application guidelines and literature, and antibacterial drug clinical application guiding principles, constructing a patient historical medical data set, randomly dividing the patient historical medical data set into a training set, a verification set and a test set according to a set proportion for training, converting the trained antibacterial drug purpose judgment model into antibacterial drug purpose judgment rules, receiving an antibacterial drug order opening request, calling the antibacterial drug purpose judgment rules, generating a matched antibacterial drug purpose according to collected antibacterial drug order opening information and current medical data of the corresponding patient, realizing automatic judgment of the antibacterial drug purpose, and selecting corresponding antibacterial drug control rules and paths according to the matched purpose.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical data processing, and in particular to an artificial intelligence-based judgment method and automatic judgment system for the purpose of using antibacterial drugs. BACKGROUND

[0002] Currently, the purposes of clinical use of antibacterial drugs can be divided into surgical prophylactic use (including preoperative prevention, intraoperative prevention, postoperative prevention, etc.), non-surgical prevention, therapeutic use and discharge medication. Rational use of antibacterial drugs is very important for improving efficacy and safety and reducing the risk of drug resistance. In order to curb the development of bacterial resistance, it is necessary to carry out scientific management of the clinical application of antibacterial drugs. The first step in the management process is to accurately determine the purpose of using antibacterial drugs, and the determination result directly determines the subsequent management process.

[0003] However, the current determination of the purpose of using antibacterial drugs is a subjective judgment of the doctor. This judgment process should be based on a comprehensive evaluation of the patient's genetic background, underlying disease, surgical details, immune status, complications, infection status, past use of antibacterial drugs, pathogenic test results, and inflammation indicators. However, due to the lack of infection diagnosis and treatment ability of the doctors themselves and the large differences in clinical experience, it is difficult to make consistent judgments. In addition, in order to evade supervision and control by management departments, some doctors deliberately take the wrong clinical path, such as choosing therapeutic use as prophylactic use to avoid microbiological specimen sampling, or changing prophylactic use to therapeutic use to extend the time of postoperative prophylactic use or use antibacterial drug varieties beyond the control of surgical / non-surgical prophylaxis. The error of the purpose of using antibacterial drugs will directly lead to changes in the subsequent management path of antibacterial drugs, thereby losing reasonable supervision of the doctor's antibacterial drug orders, and further causing a decline in patient medical quality, overuse of antibacterial drugs, development of bacterial resistance, and waste of medical resources. SUMMARY

[0004] The present application discloses an artificial intelligence-based judgment method for the purpose of using antibacterial drugs, which comprises the following steps:

[0005] S1, preprocessing the obtained patient medical data, antibacterial drug clinical application guidelines and literature, and antibacterial drug clinical application guidelines, constructing a patient historical medical data set, the patient medical data including patient basic information, disease diagnosis information, surgical application information, medical record, surgical record, pathogenic test results and other laboratory test results, historical antibacterial drug use and treatment / prevention effect;

[0006] S2, randomly divide the patient historical medical data set into training set, validation set and test set according to a set proportion, use the training set to train the initial antibacterial drug use purpose judgment model, then use the validation set to adjust parameters and optimize, use the test set to evaluate and screen to obtain the trained antibacterial drug use purpose judgment model, and convert it into antibacterial drug use purpose judgment rule;

[0007] S3, after receiving the antibacterial drug order request, calling the antibacterial drug use purpose judgment rule, generating the matched antibacterial drug use purpose according to the collected antibacterial drug order information and the current medical data of the corresponding patient, and selecting the corresponding control rule and path according to the matched antibacterial drug use purpose.

[0008] Preferably, the step S2 further comprises: before using the patient historical medical data set to train the initial antibacterial drug use purpose judgment model, preprocessing the collected patient historical medical data set, the preprocessing includes but is not limited to data cleaning, data labeling, missing value filling, numerical standardization and classification data coding.

[0009] Preferably, the step S1 specifically comprises:

[0010] S11, obtaining and recording the basic information B of the patient, the basic information includes but is not limited to identity information, date of birth, gender, medical record number, admission date and discharge date:

[0011] ;

[0012] Wherein, represents the basic information set of the i th patient, J is the number of basic information attributes, is the j th basic information attribute, is the data source of the extracted information, is the function of extracting information from the data source ; represents the extracted basic information set of the i th patient;

[0013] S12, obtaining the medical record M of the corresponding patient, the medical record contains the clinical symptoms, historical medical conditions, surgical information, diagnosed diseases, treatment process records and complication records of the patient:

[0014] ;

[0015] Wherein, represents the weighted total sum of the medical record information of the i th patient, K is the total number of medical records, is the numerical representation of the k th medical record item, is the weight of the medical record item;

[0016] S13, collect the results of the etiology test and other laboratory tests L of the corresponding patient, which include the recorded pathogen species, detection time and antibiotic sensitivity level:

[0017] ;

[0018] ;

[0019] ;

[0020] wherein, is the comprehensive laboratory test result of the ith patient, and are the weights of bacterial culture and drug sensitivity test, respectively, is the comprehensive result of bacterial culture, is the comprehensive result of drug sensitivity test, and are the relevant pathogen and drug factors, respectively, is the time decay factor, is the detection time; is the sensitivity coefficient of the mth factor of the jth specific test result, is the reaction value of the nth drug or chemical substance related to drug sensitivity test k;

[0021] S14, summarize the historical use of antibacterial drugs and treatment / prevention effect D of the corresponding patient, which include drug name, dosage, duration of use, start and end date and corresponding treatment / prevention effect:

[0022] ;

[0023] wherein, is the weighted sum of the historical drug use of the ith patient, is the code of the drug name, is the dosage, is the duration of use, , and are adjustment coefficients, L is the drug list containing all considered antibacterial drugs, and l is one drug in the drug list;

[0024] S15, determine the integrated weight of each patient according to the antibacterial drug clinical application guidelines and literature, and the antibacterial drug clinical application guiding principles , which reflects the degree of compliance of the patient's historical medical process with the guidelines and literature on the clinical application of antibacterial drugs and the principles of the clinical application of antibacterial drugs; and data of all collected research samples are combined to form a patient historical medical data set H:

[0025] ;

[0026] wherein is the integrated weight of the i-th patient, N is the number of all patients, represents a basic information set of the i-th patient, represents a weighted sum of medical record information of the i-th patient, is the i-th patient's pathogenic examination result and other laboratory test results, is the i-th patient's historical antibacterial drug use and treatment / prevention effect weighted sum.

[0027] Preferably, the antibacterial drug purpose judgment model adopts a multi-classification model formed by a plurality of target support vector machines, wherein step S2 specifically comprises:

[0028] S21, receiving the patient historical medical data set H, setting the input variable set X of the machine learning algorithm of the antibacterial drug purpose judgment model, which is composed of the basic information B, medical record M, pathogenic examination result and other laboratory test results L, and historical antibacterial drug use and treatment / prevention effect D of each patient, and defining the target variable Y to represent the antibacterial drug purpose of the patient;

[0029] S22, using the training set data to train each target support vector machine; by finding a maximum interval hyperplane to achieve the goal of minimizing the empirical risk and structural risk while maximizing the classification interval, which is expressed as:

[0030] ;

[0031] wherein is a feature vector, is a class label, w is a weight vector, is the norm of the vector w, b is a bias term, and m is the number of training data points;

[0032] The constraint condition ensures that each sample in the training data is located outside the margin of its own class, and the objective function is to maximize the classification interval; introduce the Lagrange multiplier to construct the Lagrange function, and dualize it:

[0033] ;

[0034] The form of the dual problem is:

[0035] where is the objective function of the dual problem, and are the feature vectors of the i-th and j-th data points, and are the class labels of the i-th and j-th data points, is the vector representation of the Lagrange multipliers, is the Lagrange multiplier associated with the i-th data point and satisfies the Karush-Kuhn-Tucker condition; the optimal hyperplane is determined after training, which is the preliminary multi-classification model;

[0036] S23, using the validation set data to adjust and optimize the preliminary multi-classification model; using grid search or random search method to determine the best parameter combination in the predetermined range through cross-validation;

[0037] S24, using the test set data to evaluate the performance of the optimized multi-classification model, selecting the best model as the trained model among the optimized models, and converting the trained multi-classification model into the judgment rule for the purpose of using antibacterial drugs.

[0038] Preferably, the step S22 further comprises: using a set kernel function to map the data to a higher dimensional space, and introducing a slack variable and a penalty parameter to control the cost of misclassification, so that the optimization problem becomes:

[0039] ;

[0040] After training, the optimal hyperplane is determined as the trained preliminary multi-classification model.

[0041] Preferably, the step S24 specifically comprises:

[0042] S241, using a rule extraction algorithm to analyze the trained multi-classification model, and formalizing each condition and decision output into a judgment rule for the purpose of using antibacterial drugs;

[0043] S242, formalizing each judgment rule R for the purpose of using antibacterial drugs into a conditional expression:

[0044] ;

[0045] where is a condition derived from basic information B, medical records M, etiology test results and other laboratory test results L, and historical antibacterial drug use and treatment / prevention effect D of each patient, is the purpose of the antibacterial drug in the current medical order;

[0046] S243, using the trained multi-classification model to analyze the newly acquired case data, updated antibacterial drug clinical application guidelines and literature to update the antibacterial drug purpose judgment rule:

[0047] ; wherein R is the antibacterial drug purpose judgment rule before updating, is the antibacterial drug purpose judgment rule after updating, is the learning rate, is the observed purpose of the medication, is the purpose of the medication predicted by the current antibacterial drug purpose judgment rule R; is the partial derivative of the antibacterial drug purpose judgment rule R with respect to the condition , used to represent the sensitivity of the predicted result when changes, and n is the number of real-time acquired antibacterial drug order items.

[0048] Preferably, the purpose of the medication includes but is not limited to preoperative prophylactic medication, intraoperative prophylactic medication, postoperative prophylactic medication, non-surgical prophylactic medication, therapeutic medication and discharge medication.

[0049] The application also discloses an automatic antibacterial drug purpose judgment system, comprising: a data construction module for preprocessing acquired patient medical data, antibacterial drug clinical application guidelines and literature, and antibacterial drug clinical application guidelines, and constructing a patient historical medical data set, wherein the patient medical data includes patient basic information, disease diagnosis information, surgery application information, medical records, surgery records, etiology test results and other laboratory test results, and historical antibacterial drug use and treatment / prevention effect. A model training module is used to randomly divide the patient historical medical data set into a training set, a validation set and a test set according to a set proportion, train an initial antibacterial drug purpose judgment model using the training set, then use the validation set to adjust parameters and optimize, use the test set to evaluate and screen to obtain a trained antibacterial drug purpose judgment model, and convert it into an antibacterial drug purpose judgment rule. The rule application module is used to call the antibacterial drug purpose judgment rule after receiving the antibacterial drug order request, generate a matched antibacterial drug purpose according to the collected antibacterial drug order information and the current medical data of the corresponding patient, and select the corresponding control rule and path according to the matched antibacterial drug purpose.

[0050] The application also discloses a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of the above when executing the computer program.

[0051] The application also discloses a computer program product comprising computer programs / instructions, wherein the computer programs / instructions implement the steps of the method according to any one of the above when executed by a processor.

[0052] The application discloses an artificial intelligence-based antibacterial drug purpose judgment method and automatic judgment system, which automatically judges the purpose of using antibacterial drugs through data processing and machine learning technology, overcomes the disadvantages of relying on subjective judgment of doctors in traditional medical processes, and realizes intelligent and refined management of the use of antibacterial drugs. By collecting basic information of a patient, disease diagnosis information, surgery application information, medical records, surgery records, pathogenic examination results and other laboratory test results, historical antibacterial drug use and treatment / prevention effect, and antibacterial drug clinical application guidelines and literature, an accurate patient historical medical data set is formed, an initial machine learning model is trained, and an accurate antibacterial drug purpose judgment model is constructed and converted into antibacterial drug purpose judgment rules. The application provides a premise and basis for realizing standard control and management of surgical antibacterial drug prevention and sampling control of microorganisms before antibacterial drug treatment. Therefore, after receiving an antibacterial drug order opening request, the antibacterial drug purpose can be automatically judged according to the collected antibacterial drug order opening information and the current medical data of the patient.

[0053] Additional aspects and advantages of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following and / or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0054] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an implementation of the application and together with the description, serve to explain the application.

[0055] Figure 1 A step schematic diagram of the artificial intelligence-based antibacterial drug purpose judgment method disclosed by an embodiment of the application.

[0056] Figure 2 A structure block diagram of the automatic antibacterial drug purpose judgment system disclosed by another embodiment of the application. DETAILED DESCRIPTION

[0057] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will be combined with the drawings of the embodiments of the present application to make a clear and complete description of the technical solutions of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without any creative effort belong to the scope of protection of the present application.

[0058] Unless otherwise defined, technical terms or scientific terms used herein should be understood as having the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terms "first", "second", and similar terms used in the specification and claims of the present patent application do not denote any order, quantity, or importance, but are used to distinguish different components. Similarly, the terms "one" or "a" or similar terms do not denote a quantity limitation, but mean at least one.

[0059] The purposes of the use of antibacterial drugs in clinical practice are diverse, which can be divided into surgical prophylactic use, non-surgical prophylactic use, therapeutic use, and discharge medication, etc., wherein the surgical prophylactic use includes preoperative prophylactic use, intraoperative prophylactic use, intraoperative additional use, postoperative prophylactic use, and postoperative prolonged prophylactic use. Rational use of antibacterial drugs is very important for improving efficacy and safety, and reducing the risk of drug resistance. Therefore, it is necessary to carry out scientific management of the clinical application of antibacterial drugs, and the basis of the management process is to accurately determine the purpose of the use of antibacterial drugs, and the determination result directly determines the subsequent management process. However, the current determination of the purpose of the use of antibacterial drugs is a subjective judgment of the doctor, and this judgment process should be based on the comprehensive evaluation of the genetic background, underlying disease, surgical details, immune status, complications, infection status, past use of antibacterial drugs, pathogenic examination results, and inflammation indicators of the patient, but due to the lack of infection diagnosis and treatment ability of the doctor himself and the great difference in clinical experience, it is difficult to make consistent judgments. In addition, in order to evade the supervision and control of the management department, some doctors deliberately take the wrong clinical path, such as choosing the therapeutic purpose as the prophylactic purpose in order to avoid microbial specimen sampling, changing the prophylactic purpose to the therapeutic purpose in order to prolong the postoperative prophylactic time or use antibacterial drug varieties beyond the surgical / non-surgical prophylactic control, etc.

[0060] Therefore, in order to overcome the above problems, the embodiment discloses a method for determining the purpose of the use of antibacterial drugs based on artificial intelligence, as shown in the following steps. Figure 1

[0061] ​Step S1, preprocessing the acquired patient medical data, antibacterial drug clinical application guidelines and literature, and antibacterial drug clinical application guidelines to construct a patient historical medical data set, wherein the patient medical data includes patient basic information, disease diagnosis information, surgery application information, medical record, surgery record, pathogenic examination result and other laboratory test results, historical antibacterial drug use and treatment / prevention effect.

[0062] Specifically, in a hospital working environment, doctors face a large number of patients who need to use antibacterial drugs every day. In the process of handling cases, doctors need to minimize the irrational use of antibacterial drugs while ensuring efficacy to avoid the development of drug resistance. In addition, the specific conditions of the patient, such as surgery information, infection type, pathogen test result and drug sensitivity test result, antibacterial drug use details, clinical treatment response, adverse reactions, etc. should be considered to make individualized drug purpose judgment. The antibacterial drug use purpose judgment method disclosed in this embodiment can automatically collect detailed historical medical data of the patient. After data cleaning and screening based on antibacterial drug clinical application guidelines and literature, and antibacterial drug clinical application guidelines, the data is used to train a machine learning model and generate accurate drug purpose judgment rules, which provides strong assistance for doctors when prescribing antibacterial drugs. Not only does it reduce the probability of irrational decision-making by doctors, ensure the individualization and optimization of treatment, but also promotes the rational allocation of medical resources.

[0063] In this embodiment, step S1 can specifically include the following contents.

[0064] Step S11, acquiring and recording the basic information B of the patient, including but not limited to identification information, date of birth, gender, medical record number, admission date and discharge date:

[0065] ;

[0066] wherein, represents the basic information set of the i-th patient, J is the number of basic information attributes, is the j-th basic information attribute, is the data source of the extracted information, is a function for extracting information from the data source , represents the extracted set of all basic information of the i-th patient.

[0067] Step S12, acquiring the medical record M corresponding to the patient, which contains the patient's clinical symptoms, medical history, surgery information, diagnosed diseases, treatment process record and complication record:

[0068] ;

[0069] wherein, represents the weighted sum of the medical record information of the i-th patient, is the numerical representation of the k-th medical record entry, is the weight of the medical record entry.

[0070] Step S13, collect the etiology test results and other laboratory test results L of the corresponding patient, which include the recorded pathogen species, detection time and antibiotic sensitivity level:

[0071] ;

[0072] ;

[0073] ;

[0074] wherein, is the comprehensive laboratory test result of the i-th patient, and are the weights of bacterial culture and drug sensitivity test, respectively, is the comprehensive result of bacterial culture, is the comprehensive result of drug sensitivity test, and are the relevant pathogen and drug factors, respectively, is the time decay factor, is the detection time; is the sensitivity coefficient of the m-th factor of the j-th specific test result, is the reaction value of the n-th drug or chemical substance related to drug sensitivity test k.

[0075] Step S14, summarize the historical antibacterial drug use and treatment / prevention effect D of the corresponding patient, which includes drug name, use dose, medication duration, start and end date, and corresponding treatment / prevention effect:

[0076] ;

[0077] wherein, is the weighted sum of the historical drug use of the i-th patient, is the code of the drug name, is the dose, is the medication duration, , and are adjustment coefficients, L is the drug list containing all considered antibacterial drugs, and l is one drug in the drug list.

[0078] Step S15, according to the clinical application of antibacterial drugs and literature, and the clinical application of antibacterial drugs guidelines to determine the weight of each patient, the coefficient reflects the degree of conformity of the patient's medical history and the clinical application of antibacterial drugs and literature, and the clinical application of antibacterial drugs guidelines; all the data collected from the research sample are combined to construct a patient medical history data set H:

[0079]

[0080] Wherein is the integrated weight of the ith patient, N is the number of all patients, represents the basic information set of the ith patient, indicates the weighted sum of the medical record information of the ith patient, is the pathogenic examination result and other laboratory test results of the ith patient, is the weighted sum of the history of antibacterial drug use and treatment / prevention effect of the ith patient.

[0081] Step S2, the patient medical history data set is randomly divided into training set, validation set and test set according to the set proportion, the initial antibacterial drug use purpose judgment model is trained using the training set, then the validation set is used for parameter adjustment and optimization, the test set is used for evaluation and screening to obtain the trained antibacterial drug use purpose judgment model, and is converted into antibacterial drug use purpose judgment rule.

[0082] Wherein, the purpose of medication includes but is not limited to preoperative prophylactic medication, intraoperative prophylactic medication, postoperative prophylactic medication, non-surgical prophylactic medication, therapeutic medication and discharge medication. Among them, the surgical prophylactic medication can include preoperative prophylactic medication, intraoperative prophylactic medication, intraoperative additional medication, postoperative prophylactic medication, postoperative prolonged prophylactic medication, etc.

[0083] Further, in order to ensure the quality and effectiveness of the training data, before using the patient medical history data set to train the initial antibacterial drug use purpose judgment model, the collected patient medical history data set is pretreated, and the pretreatment includes but is not limited to data cleaning, data labeling, missing value filling, numerical standardization and classification data coding. Data preprocessing process can significantly improve the quality and availability of patient medical history data set, and provide strong support for subsequent antibacterial drug use purpose judgment model training and selection. In the specific implementation process of this embodiment, more than 50000 patients' medical data are used, which are pretreated by using advanced data preprocessing technology and machine learning algorithm, so as to improve the data processing efficiency and further improve the accuracy of the constructed antibacterial drug use purpose judgment model. ​

[0084] Specifically, the patient historical medical data set can be randomly divided into a training set, a validation set and a test set in a ratio of 6:2:2. After training the initial antibacterial drug use purpose judgment model using the training set, the validation set is used for parameter adjustment and optimization. Finally, the test set is used to evaluate the optimized model, and the model with the best comprehensive performance is selected and converted into a drug use purpose judgment rule.

[0085] The evaluation criteria include but are not limited to accuracy, precision, recall, harmonic mean F value, and area under ROC curve AUC. The machine learning algorithms for antibacterial drug use purpose judgment model include but are not limited to decision tree algorithm, logistic regression algorithm, target support vector machine algorithm, naive Bayes algorithm, K-nearest neighbor algorithm, neural network algorithm, Bayesian network algorithm, linear discriminant analysis algorithm, maximum entropy model, random forest algorithm, adaptive boosting algorithm, gradient boosting decision tree algorithm GBDT, extreme gradient boosting algorithm XGBoost, etc.

[0086] In this embodiment, the antibacterial drug use purpose judgment model adopts a multi-classification model formed by a plurality of target support vector machines. The input layer of the multi-classification model is composed of a plurality of target support vector machine models SVM which are the same as the number of drug use purpose categories to be judged and identified, and each target support vector machine model is used to identify a corresponding category of drug use purpose. The output layer of the multi-classification model receives the prediction probability of the drug use purpose category output by each target support vector machine model, and selects the drug use purpose category with the highest prediction probability as the final output classification result. The training of the multi-classification model can include the following contents.

[0087] Step S21, receiving a patient historical medical data set H, setting the input variable set X of the machine learning algorithm of the antibacterial drug use purpose judgment model, which is composed of the basic information B, medical record M, pathogenic examination result and other laboratory test result L of each patient, and historical antibacterial drug use and treatment / prevention effect D, and defining the target variable Y to represent the antibacterial drug use purpose of the patient.

[0088] Step S22, using the training set data to train each target support vector machine model; by finding a maximum interval hyperplane to achieve the goal of minimizing the empirical risk and structural risk while maximizing the classification interval, which is expressed as:

[0089] ;

[0090] where is the feature vector, is the class label, w is the weight vector, is the norm of vector w, b is the bias term, and m is the number of training data points;

[0091] The constraints ensure that each sample in the training data lies outside the margin of its own class, while the objective function is to maximize the margin between classes; introduce Lagrange multipliers Construct the Lagrangian function, dualize it:

[0092] ;

[0093] The dual problem takes the form:

[0094] where is the objective function of the dual problem, and are the feature vectors of the i-th and j-th data points, and are the class labels of the i-th and j-th data points, is the vector representation of the Lagrange multipliers, is the Lagrange multiplier associated with the i-th data point and satisfies the Karush-Kuhn-Tucker conditions, and > 0; after training, the optimal hyperplane is determined as the trained preliminary multi-classification model.

[0095] In this embodiment, the step S22 can further include: using a set kernel function to map the data to a higher dimensional space, and by introducing a slack variable and a penalty parameter to control the cost of misclassification, the optimization problem becomes:

[0096] ;

[0097] After training, the optimal hyperplane is determined as the trained preliminary multi-classification model.

[0098] Specifically, in other embodiments, the kernel function used can include but is not limited to Linear Kernel, Polynomial Kernel, Radial Basis Function, Exponential Kernel, Laplacian Kernel, ANOVA Kernel, Rational Quadratic Kernel, Multiquadric Kernel, Inverse Multiquadric Kernel, etc.

[0099] Step S23, using the validation set data to adjust and optimize the preliminary multi-classification model, using grid search or random search method to determine the best parameter combination in the predetermined range.

[0100] Specifically, the adjusted parameters include but are not limited to the penalty coefficient C of the objective function, the selection of kernel function, the highest power degree of the polynomial kernel function, the coefficient gamma of the kernel function, the independent term coef0 in the kernel function, whether to use the probability estimation probablity, whether to perform heuristic shrinking, the accuracy of the support vector machine model end standard, the memory required for training, the weight of each category of the specified sample, the maximum number of iterations, the classification decision, the seed of the pseudo-random number generator for data rearrangement for probability estimation, etc.

[0101] Step S24, using the test set data to evaluate the performance of the optimized multi-classification model, selecting the best model from the optimized models as the trained model, and converting the trained multi-classification model into the antibacterial drug use purpose judgment rule.

[0102] Specifically, the evaluation methods include but are not limited to accuracy, precision, recall, harmonic mean F value, and area under the ROC curve, etc. Through continuous fine-tuning of the model and repeating the above steps, until a satisfactory performance is obtained.

[0103] In this embodiment, the step S24 specifically includes the following contents.

[0104] Step S241, using a rule extraction algorithm to analyze the trained multi-classification model, and formalizing each condition and decision output into an antibacterial drug use purpose judgment rule.

[0105] Step S242, formalizing each antibacterial drug use purpose judgment rule R into a conditional expression:

[0106] ; wherein is a condition derived based on the basic information B of each patient, medical record M, pathogenic examination results and other laboratory test results L, and historical antibacterial drug use and treatment / prevention effect D, is the purpose of the use of the antibacterial drug.

[0107] Step S243, using the trained multi-classification model to analyze the real-time acquired new case data, updated antibacterial drug clinical application guidelines and literature to update the antibacterial drug use purpose judgment rule:

[0108] ; where R is the updated antibacterial drug purpose judgment rule, is the updated antibacterial drug purpose judgment rule, is the learning rate, is the observed purpose of medication, is the predicted purpose of medication by the current antibacterial drug purpose judgment rule R. is the partial derivative of the antibacterial drug purpose judgment rule R with respect to the condition , which represents the sensitivity of the prediction result when changes, and n is the number of real-time acquired antibacterial drug order items.

[0109] Specifically, by introducing a machine learning algorithm for dynamic adjustment, the adaptability and individualization level of the rule are further improved. The real-time updated antibacterial drug purpose judgment rule can automatically judge the purpose of antibacterial drug use in the electronic medical record system and enter the corresponding management path by constructing a rule engine and integrating the rule engine into the hospital information system. The electronic system can automatically judge the purpose of antibacterial drug use according to the current health status and historical medical data of the patient, and activate the corresponding management path for targeted management.

[0110] Step S3, after receiving the antibacterial drug order request, calling the antibacterial drug purpose judgment rule, generating the matched antibacterial drug purpose according to the collected antibacterial drug order information and the current medical data of the corresponding patient, and selecting the corresponding management rule and path according to the matched antibacterial drug purpose.

[0111] In specific embodiments, the hospital information system employing the above-mentioned artificial intelligence-based antibacterial drug purpose judgment method is deployed, which can automatically collect detailed historical medical data of patients, including basic information, disease diagnosis, surgery application, medical record, surgery record, laboratory test results such as bacterial culture and drug sensitivity test results, and historical antibacterial drug use. After these data go through the pretreatment process, they can be used to train the antibacterial drug purpose judgment model to generate accurate antibacterial drug purpose judgment rules. When a doctor needs to prescribe antibacterial drugs, the system can quickly judge the purpose of drug use by virtue of the integrated rule engine, and automatically match the applicable management rules and processes, effectively eliminating the interference of subjective judgment and simplifying the process of the clinical doctors prescribing antibacterial drug orders.

[0112] In the first six months of system implementation, pre- and post-implementation medication data were collected and analyzed. The results showed that before the system was deployed, the proportion of pathogenic microorganism specimen sampling before antimicrobial therapy by doctors in the hospital was only 33.45%; after the system was implemented, the proportion jumped to 81.81%. In the first year of system implementation, the number of patients who used antimicrobial therapy without infection indications decreased by 82.34%. At the same time, the number of patients who used antimicrobial drugs for surgical or non-surgical prevention without preventive indications also decreased by 79.98%. Due to the significant increase in pathogen detection rate, the number of cases of precision antimicrobial therapy for specific pathogenic bacteria increased in clinical practice, and the use of unnecessary high-level and broad-spectrum antimicrobial drugs was reduced. In addition, surgical and non-surgical prophylactic drug use is more standardized, and the hospital has observed a reduction in the average length of stay for patients, which directly promotes the more efficient allocation of medical resources and the reduction of overall medical costs.

[0113] The above-mentioned embodiments disclose an artificial intelligence-based judgment method for the purpose of antimicrobial drug use. Through data processing and machine learning technology, the purpose of antimicrobial drug use is automatically judged, overcoming the drawbacks of relying on subjective judgment by doctors in traditional medical processes, and realizing intelligent and refined management of antimicrobial drug use. By collecting the patient's basic information, disease diagnosis information, surgical application information, medical records, surgical records, pathogen test results and other laboratory test results, historical antimicrobial drug use and treatment / prevention effects, and antimicrobial drug clinical application guidelines and literature, an accurate patient historical medical data set is formed, an initial machine learning model is trained, and a precise antimicrobial drug use purpose judgment model is constructed and converted into an antimicrobial drug use purpose judgment rule. This provides a prerequisite and foundation for realizing the standardized control of surgical antimicrobial prophylaxis and the control of microbial specimen sampling before antimicrobial therapy. Thus, upon receiving an antimicrobial drug prescription request, the antimicrobial drug use purpose can be automatically judged based on the collected antimicrobial drug prescription information and the current patient's medical data.

[0114] In another embodiment, an automatic judgment system for the purpose of antimicrobial drug use is also disclosed, such as Figure 2As shown, it comprises a data construction module 1, a model training module 2 and a rule application module 3. Among them, the data construction module 1. Among them, the data construction module 1 is used to preprocess the obtained patient medical data, antibacterial drug clinical application guidelines and literature, and antibacterial drug clinical application guidelines, construct a patient historical medical data set, and the patient medical data includes the patient's basic information, disease diagnosis information, operation application information, medical record, operation record, pathogenic examination result and other laboratory test results, historical antibacterial drug use and treatment / prevention effect. The model training module 2 is used to randomly divide the patient historical medical data set into a training set, a validation set and a test set according to a set proportion, train the initial antibacterial drug use purpose judgment model using the training set, then use the validation set to adjust the parameters and optimize, use the test set to evaluate and screen to obtain the trained antibacterial drug use purpose judgment model, and convert it into antibacterial drug use purpose judgment rule. The rule application module 3 receives the antibacterial drug order opening request, calls the antibacterial drug use purpose judgment rule, generates the matched antibacterial drug use purpose according to the collected antibacterial drug order opening information and the current medical data of the corresponding patient, and selects the corresponding control rule and path according to the matched antibacterial drug use purpose.

[0115] It should be noted that the embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The similar parts of each embodiment can be referred to each other. For the antibacterial drug use purpose judgment system based on artificial intelligence disclosed in the embodiments, since it corresponds to the antibacterial drug use purpose judgment method based on artificial intelligence disclosed in the embodiments, the description is relatively simple, and the relevant parts are described in the foregoing method part.

[0116] In some other embodiments, a computer device is also provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to realize each step of the antibacterial drug use purpose judgment method based on artificial intelligence described in the above embodiments. The server can include but is not limited to a processor and a memory. Those skilled in the art can understand that the schematic diagram is only an example of the server and does not constitute a limitation on the server, which can include more or fewer components than the diagram, or combine certain components, or different components.

[0117] If the artificial intelligence-based antibacterial drug medication purpose judgment system is realized in the form of a software function unit and sold or used as a stand-alone product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the processor executes the computer program, the steps of each of the above-mentioned artificial intelligence-based antibacterial drug medication purpose judgment methods can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer-readable medium can include any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0118] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

[0119] In summary, the above is only a preferred embodiment of the present application, and any equivalent changes and modifications made within the scope of the patent application of the present application shall be included in the scope of the present application.

Claims

1. A method for determining the purpose of antimicrobial drug use based on artificial intelligence, characterized in that, Includes the following steps: S1. Preprocess the acquired patient medical data, clinical application guidelines and literature for antimicrobial drugs, and clinical application guidelines for antimicrobial drugs to construct a patient historical medical dataset. The patient medical data includes the patient's basic information, disease diagnosis information, surgical application information, medical records, surgical records, etiological test results and other laboratory test results, and historical antimicrobial drug use and treatment / prevention effects. S2, the patient's historical medical dataset is randomly divided into a training set, a validation set and a test set according to a set ratio. The initial antimicrobial drug use purpose judgment model is trained using the training set, then the validation set is used for parameter tuning and optimization, and the test set is used for evaluation and screening to obtain the trained antimicrobial drug use purpose judgment model, which is then converted into antimicrobial drug use purpose judgment rules. S3, after receiving the request to prescribe an antimicrobial drug, invoke the antimicrobial drug use purpose determination rule, generate a matching antimicrobial drug use purpose based on the collected antimicrobial drug prescription information and the corresponding patient's current medical data, and select the corresponding control rules and path based on the matched antimicrobial drug use purpose; The antibacterial drug use purpose determination model adopts a multi-classification model formed by combining multiple target support vector machines, wherein step S2 specifically includes: S21, Receive patient's historical medical dataset H, Set the input variable set X of the machine learning algorithm for the antimicrobial drug use purpose judgment model, which consists of each patient's basic information B, medical record M, etiological test results and other laboratory test results L, and historical antimicrobial drug use and treatment / prevention effects D, and Define the target variable Y to represent the patient's antimicrobial drug use purpose; S22, using training set data (x i ,y i Train the support vector machines for each target; find a maximum margin hyperplane w T The goal of x+b=0 is to minimize empirical risk and structural risk while maximizing the classification margin, expressed as: Where, x i ∈R n It is an eigenvector, y i Let ∈{-1,+1} be the class labels, w be the weight vector, ||w|| be the norm of vector w, b be the bias term, and m be the number of training data points; the constraints ensure that each sample in the training data lies outside the boundary of its class, while the objective function... To maximize the classification margin, we introduce Lagrange multipliers α to construct the Lagrange function and then dualize it: The dual problem takes the form of: Where W(α) is the objective function of the dual problem, x i and x j These are the feature vectors of the i-th and j-th data points, respectively. i and y j These are the class labels for the i-th and j-th data points, respectively, and α is the vector representation of the Lagrange multipliers. i It is the Lagrange multiplier associated with the i-th data point and satisfies the Carlow-Kun-Tucker condition; after training, the optimal hyperplane is determined, which is the preliminary multi-classification model; S23, Use validation set data to tune and optimize the initial multi-class classification model; Use grid search or random search methods to cross-validate within a predetermined range to determine the optimal parameter combination; S24. Use test set data to evaluate the performance of the optimized multi-classification model, select the best model among the optimized models as the trained model, and transform the trained multi-classification model into a rule for judging the purpose of antibacterial drug use. Step S22 further includes: Using a defined kernel function, the data is mapped to a higher-dimensional space by introducing a slack variable ξ. i The penalty parameter C is used to control the cost of misclassification, transforming the optimization problem into: After training, the optimal hyperplane is determined as the initial multi-class classification model.

2. The method for determining the purpose of antibacterial drug use based on artificial intelligence according to claim 1, characterized in that, Step S2 further includes: Before using the patient history medical dataset to train the initial antimicrobial drug use purpose judgment model, the collected patient history medical dataset is preprocessed, including data cleaning, data labeling, missing value imputation, numerical standardization, and encoding of categorical data.

3. The method for determining the purpose of antibacterial drug use based on artificial intelligence according to claim 2, characterized in that, Step S1 specifically includes: S11, Obtain and record the patient's basic information B, which includes identification information, date of birth, gender, medical record number, admission date, and discharge date: Among them, B i Let J represent the set of basic information for the i-th patient, where J is the number of basic information attributes, and b is the number of attributes. ij It is the j-th basic information attribute, data j It is the data source for extracting information, f j From data source j Functions for extracting information This represents the set of all basic information extracted from the i-th patient. S12, obtain the corresponding patient's medical record M, which includes the patient's clinical symptoms, medical history, surgical information, diagnosed diseases, treatment process records, and complication records: Among them, M i Let m represent the weighted sum of the medical record information for the i-th patient, where K is the total number of medical record entries. ik w is the numerical representation of the k-th medical record entry. k This is the weight of the medical record entry; S13, collect the etiological test results and other laboratory test results L for the corresponding patient, including the recorded pathogen type, test time, and antibiotic sensitivity level: Among them, L i This is the comprehensive laboratory test result of the i-th patient, α j and β k These are the weights of bacterial culture and drug susceptibility testing, respectively, L icj This is a comprehensive result of bacterial culture, L idk It is a comprehensive result of drug sensitivity testing, γ pj and δ dk These are the relevant pathogens and drug factors, λ tj It is the time decay factor, t j It is the detection time; s jm r is the sensitivity coefficient of the m-th factor for the j-th specific detection result. kn The response value of the nth drug or chemical substance associated with drug sensitivity test k; S14, Summarize the historical antimicrobial drug use and treatment / prevention effects D for the corresponding patient. The historical antimicrobial drug use and treatment / prevention effects include drug name, dosage, duration of use, start and end dates, and corresponding treatment / prevention effects: Among them, D i It is the weighted sum of the historical medication use of the i-th patient, d l It is the code for the drug name, v l It is the dosage, τ l Duration of medication, ξ l ψ l and κ l This is an adjustment factor, where L is the drug list containing all considered antimicrobial drugs, and l is a single drug in the drug list; S15, determine the integrated weight η for each patient based on guidelines and literature on the clinical application of antimicrobial drugs, as well as the guiding principles for the clinical application of antimicrobial drugs. i This coefficient reflects the degree to which the patient's historical medical history conforms to the aforementioned guidelines and literature on the clinical application of antimicrobial drugs, as well as the guiding principles for the clinical application of antimicrobial drugs; data from all collected research samples are combined to construct a patient historical medical history dataset H: Where η i B is the integrated weight of the i-th patient, N is the total number of patients, and B is the weight of the i-th patient. i M represents the set of basic information for the i-th patient. i L represents the weighted sum of the medical record information of the i-th patient. i The etiological test results and other laboratory test results of the i-th patient, D i It is the weighted sum of the historical antimicrobial drug use and treatment / prevention effects of the i-th patient.

4. The method for determining the purpose of antimicrobial drug use based on artificial intelligence according to claim 1, characterized in that, Step S24 specifically includes: S241, use a rule extraction algorithm to parse the trained multi-classification model, and formalize each condition and decision output into a rule for judging the purpose of antibacterial drug use; S242, formalize each rule R for determining the purpose of antimicrobial drug use into a conditional expression: R:if(c1∧c2∧…∧c n then(d1,d2,…,d m ); where c i It is a condition derived from each patient's basic information (B), medical records (M), etiological test results and other laboratory test results (L), and historical antimicrobial drug use and treatment / prevention effects (D). j This is the intended use of the antibacterial drug; S243, using a pre-trained multi-classification model to automatically and continuously analyze real-time acquired new case data, updated clinical application guidelines and literature for antimicrobial drugs, to update the rules for determining the purpose of antimicrobial drug use: Where R represents the previous rule for determining the purpose of antimicrobial drug use. new This is the updated rule for determining the purpose of antimicrobial drug use; β is the learning rate, and y is the antimicrobial drug dosage. i It is the observed purpose of medication, The intended use of antibiotics is predicted by rule R, which determines the intended use of antibiotics. The rule for determining the purpose of antimicrobial drug use, R, is based on condition c. i The partial derivative of c is used to represent the partial derivative of c when c is zero. i When changes occur, the prediction results The sensitivity is denoted by n, which is the number of antimicrobial drug prescription entries obtained in real time.

5. The method for determining the purpose of antimicrobial drug use based on artificial intelligence according to any one of claims 1-4, characterized in that: The purposes of medication include preoperative prophylaxis, intraoperative prophylaxis, postoperative prophylaxis, non-surgical prophylaxis, therapeutic medication, and discharge medication.

6. An automatic system for determining the intended use of antimicrobial drugs, used to implement the artificial intelligence-based method for determining the intended use of antimicrobial drugs as described in any one of claims 1-5, characterized in that, include: The data construction module is used to preprocess the acquired patient medical data, clinical application guidelines and literature for antimicrobial drugs, and clinical application guidelines for antimicrobial drugs to construct a patient historical medical dataset. The patient medical data includes the patient's basic information, disease diagnosis information, surgical application information, medical records, surgical records, etiological test results and other laboratory test results, and historical antimicrobial drug use and treatment / prevention effects. The model training module is used to randomly divide the patient's historical medical dataset into a training set, a validation set, and a test set according to a set ratio. After training the initial antimicrobial drug use purpose judgment model using the training set, the validation set is used for parameter tuning and optimization. After evaluation and screening using the test set, the trained antimicrobial drug use purpose judgment model is obtained and converted into antimicrobial drug use purpose judgment rules. The rule application module, upon receiving a request to prescribe an antimicrobial drug, invokes the antimicrobial drug use purpose determination rule, generates a matching antimicrobial drug use purpose based on the collected antimicrobial drug prescription information and the corresponding patient's current medical data, and selects the appropriate control rules and paths based on the matched antimicrobial drug use purpose.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-5.

8. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-5.

Images