Precision drug treatment management system and method for patients with severe infection

Abstract

The present invention relates to the field of medical information processing technology, and in particular to a precise drug treatment management system and method for patients with severe infections. The content includes: collecting multi-source medical data of patients and preprocessing them, performing multi-source fusion processing on the preprocessed medical data to obtain the fused comprehensive patient data; generating a preliminary personalized drug treatment plan based on the fused comprehensive patient data; based on the preliminary personalized drug treatment plan, quantifying the risk of the drug combination in the preliminary personalized drug treatment plan through a multi-dimensional interaction field and a dynamic entropy weight risk assessment algorithm to obtain a risk score; based on the risk score, optimizing the preliminary personalized drug treatment plan to generate a safe personalized drug treatment plan. It solves the technical problems of inaccurate processing of multi-source heterogeneous patient medical data and inaccurate and unreasonable treatment plans in the implementation of precise drug treatment management for patients with severe infections in existing drug treatment plans.

Description

Technical Field

[0001] The present invention relates to the field of medical information processing technology, and in particular to a precise drug treatment management system and method for patients with severe infections. Background Art

[0002] With the continuous advancement of medical technology and the increasing abundance of medical data, precision medicine has gradually become an important development direction in the treatment of patients with severe infections. Patients with severe infections usually have complex and changeable conditions, and there are significant individual differences in their response to drug treatment. Therefore, it is necessary to accurately consider the individual characteristics of patients in drug selection and dosage adjustment. However, existing drug treatment plans mostly rely on empirical medicine and traditional guideline recommendations, lack in-depth analysis of individual characteristics of patients, and it is difficult to effectively respond to the complex changes in the condition and individual differences of patients with severe infections.

[0003] In summary, the existing drug treatment plans have the following technical problems: in the implementation of precise drug treatment management for patients with severe infections, the processing of multi-source heterogeneous patient medical data is inaccurate and the treatment plans are inaccurate and unreasonable. Summary of the invention

[0004] The present invention provides a precise drug treatment management system and method for patients with severe infections, so as to solve the technical problems that the existing drug treatment schemes have inaccurate processing of multi-source heterogeneous patient medical data and inaccurate and unreasonable treatment schemes in the implementation of precise drug treatment management for patients with severe infections.

[0005] The precise drug treatment management system and method for patients with severe infections of the present invention specifically include the following technical solutions:

[0006] The precise drug treatment management method for patients with severe infection includes the following steps:

[0007] S1. Collect multi-source medical data of patients and perform preprocessing to obtain preprocessed medical data; perform multi-source fusion processing on the preprocessed medical data to obtain fused comprehensive patient data;

[0008] S2. Generate a preliminary personalized drug treatment plan based on the integrated patient data; based on the preliminary personalized drug treatment plan, quantify the risk of the drug combination in the preliminary personalized drug treatment plan through a multi-dimensional interaction field and a dynamic entropy weight risk assessment algorithm to obtain a risk score; based on the risk score, optimize the preliminary personalized drug treatment plan to generate a safe personalized drug treatment plan.

[0009] Preferably, the S1 specifically includes:

[0010] The specific steps of the multi-source fusion processing are as follows: first, the pre-processed medical data is merged to obtain merged medical data; then, the merged medical data is feature extracted to obtain preliminary multi-dimensional feature data of the patient's medical data; finally, the preliminary multi-dimensional feature data of the patient's medical data is processed by a feature fusion algorithm based on multi-dimensional similarity embedding and nonlinear perturbation to obtain fused comprehensive patient data.

[0011] Preferably, the S1 specifically includes:

[0012] In the process of implementing the feature fusion algorithm based on multidimensional similarity embedding and nonlinear perturbation, the preliminary multidimensional feature data of the patient medical data is standardized and transformed, and a standardized feature matrix is ​​generated based on the standardized preliminary multidimensional feature data of the patient medical data; based on the standardized feature matrix, a similarity measurement matrix is ​​calculated; based on the similarity measurement matrix, the standardized feature matrix is ​​spatially mapped to obtain a multidimensional similarity embedding matrix.

[0013] Preferably, the S1 specifically includes:

[0014] A nonlinear perturbation function is introduced to perturb the multidimensional similarity embedding matrix to generate a nonlinear perturbation matrix; any element in the nonlinear perturbation matrix is ​​expressed as:

[0015] ;

[0016] in, is the nonlinear perturbation matrix Line Column elements; is the first dimension of the multidimensional similarity embedding matrix Line Column elements; is a nonlinear perturbation function; is the scaling factor that controls the hyperbolic tangent function; is the translation amount that controls the hyperbolic tangent function; is the perturbation strength that controls the influence of the sine function; , It is the frequency and phase shift of the controlled sine function; is the hyperbolic tangent function.

[0017] Preferably, the S1 specifically includes:

[0018] Based on the nonlinear perturbation matrix, an adaptive multiple embedding aggregation mechanism is introduced to calculate the contribution weight, and a comprehensive feature generation function is introduced to obtain the fused patient comprehensive data; the specific formula of the comprehensive feature generation function is:

[0019] ;

[0020] in, It is the comprehensive patient data after fusion; is the comprehensive feature generation function; and Respectively represent the row index and column index in the nonlinear perturbation matrix; represents the total number of rows of the nonlinear perturbation matrix; represents the total number of columns of the nonlinear perturbation matrix; is the contribution weight; is the nonlinear perturbation matrix Line Column elements; is the sinusoidal function perturbation intensity coefficient in the comprehensive feature generating function; is the frequency coefficient of the sine function in the comprehensive feature generating function; is the activation function.

[0021] Preferably, the S2 specifically includes:

[0022] In the process of implementing the multidimensional interaction field and dynamic entropy weight risk assessment algorithm, a drug interaction matrix is ​​constructed based on the preliminary personalized drug treatment plan; based on the drug interaction matrix, the interaction intensity of the interaction dimension is obtained; the interaction intensity on all interaction dimensions is integrated to obtain the multidimensional interaction field intensity; the calculation formula of the interaction intensity of the interaction dimension is as follows:

[0023] ;

[0024] in, In each interaction dimension Drugs in the initial personalized drug treatment plan and medications The strength of the interaction; It is a drug interaction information; It's medicine and medications In the The weight coefficient of the dimensional interaction; Indicates drug and medications In the The interaction factors of the dimensions; and Respectively represent drugs and Specific dosage; and Respectively represent drugs and medications In the The duration of the dimension's action; It is the adjustment coefficient for controlling the duration of drug action.

[0025] Preferably, the S2 specifically includes:

[0026] A dynamic entropy weight adjustment mechanism is introduced to obtain the entropy value of the interaction dimension based on the interaction intensity of the interaction dimension; based on the entropy value of the interaction dimension, the weight is dynamically adjusted to calculate the dynamic weight of each interaction dimension; the dynamic weight of each interaction dimension is applied to the multidimensional interaction field intensity to obtain the dynamic weighted interaction intensity.

[0027] Preferably, the S2 specifically includes:

[0028] Based on the dynamic weighted interaction intensity, the risk entropy weight is obtained; based on the risk entropy weight, the risk of each pair of drug combinations in the preliminary personalized drug treatment plan is evaluated to obtain a risk score; the risk score is compared with the preset risk score threshold, and when the risk score is greater than the risk score threshold, the current drug combination in the preliminary personalized drug treatment plan is optimized to generate a safe personalized drug treatment plan.

[0029] The precise drug treatment management system for patients with severe infection includes the following parts:

[0030] Data acquisition and preprocessing module, multi-source data fusion module, drug treatment database, individualized treatment decision module, drug interaction database, drug interaction detection module, dynamic monitoring and feedback regulation module;

[0031] The data acquisition and preprocessing module collects the multi-source medical data of the patient and performs preprocessing to obtain the preprocessed medical data; the preprocessed medical data is sent to the multi-source data fusion module;

[0032] The multi-source data fusion module performs multi-source fusion processing on the pre-processed medical data to obtain the fused comprehensive patient data; the fused comprehensive patient data is sent to the individualized treatment decision module;

[0033] A drug therapy database, storing drug therapy data and sending the drug therapy data to an individualized treatment decision module;

[0034] The individualized treatment decision module generates a preliminary personalized drug treatment plan based on the integrated patient data and the drug treatment data provided by the drug treatment database; and sends the preliminary personalized drug treatment plan to the drug interaction detection module;

[0035] a drug interaction database, storing drug interaction data and sending the drug interaction data to a drug interaction detection module;

[0036] The drug interaction detection module combines drug interaction data and uses a multidimensional interaction field and a dynamic entropy weight risk assessment algorithm to quantify the preliminary personalized drug treatment plan, evaluate the risk of each pair of drug combinations in the preliminary personalized drug treatment plan, and obtain a risk score; compare the risk score with a preset risk score threshold, and when the risk score is greater than the risk score threshold, optimize the current drug combination in the preliminary personalized drug treatment plan to generate a safe personalized drug treatment plan; send the safe personalized drug treatment plan to the dynamic monitoring and feedback adjustment module;

[0037] The dynamic monitoring and feedback adjustment module treats patients through a safe and personalized drug treatment plan. During the treatment process, the patient's condition change trend is tracked in real time and the safe and personalized drug treatment plan is evaluated. When changes in the patient's condition are detected, the changes are fed back to the data acquisition and preprocessing module to form a closed loop until the patient's condition stabilizes.

[0038] The beneficial effects of the technical solution of the present invention are:

[0039] 1. By introducing multidimensional similarity embedding, the preliminary multidimensional feature data of patient medical data is mapped into the similarity embedding space, and combined with the nonlinear perturbation mechanism, the similarity relationship between different features can be more accurately characterized and expressed; multidimensional similarity embedding can effectively capture the global similarity relationship between features, and the nonlinear perturbation function further introduces nonlinear changes, making the generation of fused patient comprehensive data more flexible and diversified, thereby improving the quality and expression ability of feature fusion. The generated fused patient comprehensive data can more comprehensively reflect the patient's condition and provide a solid data foundation for the formulation of personalized drug treatment plans.

[0040] 2. By calculating the contribution weight of each feature in the nonlinear perturbation matrix, the contribution of each feature in the final feature fusion is determined, and then all the features of the nonlinear perturbation matrix are weighted and aggregated; the introduction of contribution weights enables the precision drug treatment management system to dynamically adjust the importance of different features in the fused patient comprehensive data, improving the accuracy and robustness of feature fusion. The fused patient comprehensive data can better reflect the individual characteristics of the patient and provide reliable data support for subsequent personalized treatment.

[0041] 3. The construction of multi-dimensional interaction fields can carefully capture the complex interaction behaviors between drugs in the preliminary personalized drug treatment plan. The introduction of dynamic weights can adaptively adjust the importance of each interaction dimension according to the characteristics and medication conditions of different patients, thereby more accurately quantifying the interaction risk of drug combinations. Through optimization strategies (such as dose adjustment, time separation, etc.), the risk of drug interactions can be effectively reduced and a safe personalized drug treatment plan can be generated. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a structural diagram of the precise drug treatment management system for patients with severe infections described in the present invention;

[0043] Figure 2 This is a flow chart of the precise drug treatment management method for patients with severe infections described in the present invention. DETAILED DESCRIPTION

[0044] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0046] The specific scheme of the precise drug treatment management system and method for patients with severe infection provided by the present invention is described in detail below with reference to the accompanying drawings.

[0047] See attached Figure 1 , which shows a structural diagram of a precise drug treatment management system for severe infection patients provided by an embodiment of the present invention, the system includes the following parts:

[0048] Data acquisition and preprocessing module, multi-source data fusion module, drug treatment database, individualized treatment decision module, drug interaction database, drug interaction detection module, dynamic monitoring and feedback regulation module;

[0049] The data collection and preprocessing module collects the multi-source medical data of the patient from various data sources, and preprocesses the collected multi-source medical data to obtain preprocessed medical data; and sends the preprocessed medical data to the multi-source data fusion module;

[0050] The data sources include drug databases, disease knowledge bases, electronic health record systems, real-time monitoring equipment, laboratory test data, and patient self-filled data;

[0051] The medical data include the patient's genomic data, past medical history and allergy history, real-time physiological index data (such as heart rate, blood oxygen saturation, body temperature, blood pressure, etc.), laboratory test data (such as white blood cell count, C-reactive protein, blood culture results, etc.), the patient's living habits, medication history and other information data;

[0052] The multi-source data fusion module performs multi-source fusion processing on the pre-processed medical data to obtain the fused comprehensive patient data; the fused comprehensive patient data is sent to the individualized treatment decision module;

[0053] A drug therapy database is used to store drug therapy data such as drug indications, contraindications, and pharmacokinetic parameters, and send the drug therapy data to the individualized treatment decision module;

[0054] The individualized treatment decision module generates a preliminary personalized drug treatment plan based on the integrated patient data and the drug treatment data provided by the drug treatment database; and sends the preliminary personalized drug treatment plan to the drug interaction detection module;

[0055] A drug interaction database is used to store interaction data, potential risks, interaction mechanisms between drugs, and send the drug interaction data to the drug interaction detection module;

[0056] The drug interaction detection module combines drug interaction data and uses a multidimensional interaction field and a dynamic entropy weight risk assessment algorithm to quantify the preliminary personalized drug treatment plan, evaluate the risk of each pair of drug combinations in the preliminary personalized drug treatment plan, and obtain a risk score; compare the risk score with a preset risk score threshold, and when the risk score is greater than the risk score threshold, optimize the current drug combination in the preliminary personalized drug treatment plan to obtain a safe personalized drug treatment plan; send the safe personalized drug treatment plan to the dynamic monitoring and feedback adjustment module;

[0057] The dynamic monitoring and feedback adjustment module treats patients through a safe and personalized drug treatment plan. During the treatment process, the patient's condition change trend is tracked in real time to evaluate the effectiveness of the safe and personalized drug treatment plan. When changes in the patient's condition are detected, the changes are fed back to the data acquisition and preprocessing module to form a closed loop until the patient's condition stabilizes.

[0058] See attached Figure 2 , which shows a flow chart of a method for precise drug treatment management of severe infection patients provided by an embodiment of the present invention, the method comprising the following steps:

[0059] S1. Collect multi-source medical data of patients and perform preprocessing to obtain preprocessed medical data; perform multi-source fusion processing on the preprocessed medical data to obtain fused comprehensive patient data;

[0060] Collect multi-source medical data from various data sources, such as drug databases, disease knowledge bases, electronic health record systems, real-time monitoring equipment, laboratory test data, and patient self-filled data; perform preprocessing such as data cleaning, consistency checking, standardization, and classification data encoding on the collected multi-source medical data to obtain preprocessed medical data; the classification data encoding is to facilitate subsequent data processing, and the medical data is classified by existing technologies, such as one-hot encoding, which is used to convert categorical variables (such as gender, disease type) into binary vector form; label encoding is used to convert ordered categorical variables (such as disease level) into ordered integer labels. Data cleaning, consistency checking, and standardization are all methods well known to those skilled in the art and will not be elaborated here.

[0061] The pre-processed medical data is subjected to multi-source fusion processing, and the specific steps of the multi-source fusion processing are as follows:

[0062] First, the pre-processed medical data is merged to obtain merged medical data; the data merging includes vertical merging and horizontal merging; vertical merging refers to merging data of the same data type at different time points or data from different sources by row, such as vertical merging of medical data collected from the same patient at different time points, for example, merging physiological indicators (such as blood pressure, body temperature, etc.) at different time points into a time series data set; for the same type of data (such as multiple blood test results) from different data sources (such as different hospitals, laboratories), they are merged into a complete data set after time alignment; horizontal merging is performed after the vertical merging is completed, which refers to merging different types of feature data (such as genetic information, laboratory data, real-time physiological indicators, etc.) by column, such as horizontally merging the time series data after vertical merging with the patient's static information (such as genetic information, medical history, etc.), each row represents a different time point or different source of the same patient, and different columns represent different feature types of the same patient;

[0063] Then, the merged medical data is subjected to feature extraction using existing feature engineering techniques to obtain preliminary multi-dimensional feature data of the patient's medical data;

[0064] Finally, the preliminary multidimensional feature data of the patient's medical data is processed using a feature fusion algorithm based on multidimensional similarity embedding and nonlinear perturbation to obtain the fused patient comprehensive data.

[0065] The feature fusion algorithm based on multidimensional similarity embedding and nonlinear perturbation is to introduce multidimensional similarity embedding to spatially map the preliminary multidimensional feature data of the patient's medical data, and combine it with the nonlinear perturbation mechanism to generate new fusion features in the multidimensional feature space, that is, the fused patient comprehensive data. The specific implementation process is as follows:

[0066] First, the preliminary multidimensional feature data of the patient's medical data is recorded as , and convert it into a standardized form to ensure that different features are comparable, and obtain the preliminary multidimensional feature data of the standardized patient medical data; based on the preliminary multidimensional feature data of the standardized patient medical data, a standardized feature matrix is ​​constructed ;

[0067] Furthermore, in order to effectively capture the similarity between the preliminary multidimensional feature data of the patient's medical data, the standardized feature matrix is ​​calculated The similarity measure matrix between each feature in . Similarity measure matrix Each element of is calculated by Pearson's correlation coefficient;

[0068] Furthermore, in order to map the similarity information of the multidimensional feature space into the embedding space composed of eigenvalues ​​and eigenvectors and lay the foundation for the subsequent multidimensional similarity embedding, the similarity measurement matrix Perform eigendecomposition to obtain the eigenvalue matrix and the eigenvector matrix :

[0069] ;

[0070] in, is the similarity measure matrix The eigenvalue matrix of , the diagonal elements represent the variance of each principal component; is the similarity measure matrix The eigenvector matrix of , each column is a principal component direction vector; represents transpose;

[0071] After the eigendecomposition is completed, the multidimensional similarity embedding matrix is ​​calculated , which is used to map the multidimensional features of each patient into a similarity embedding space to better capture the global similarity relationship between features:

[0072] ;

[0073] Multidimensional Similarity Embedding Matrix The calculation process is as follows: through the standardized feature matrix Left multiply the eigenvector matrix , map the preliminary multidimensional feature data of the standardized patient medical data into the principal component space; then multiply by , that is, the eigenvalue matrix The purpose is to normalize the principal components and eliminate the influence of inconsistent feature dimensions on the embedding results; finally, multiply the right , restoring the embedding results to the original multidimensional feature space, thereby retaining the similarity structure of the preliminary multidimensional feature data of the standardized patient medical data;

[0074] Furthermore, in order to introduce more nonlinear expression capabilities based on multidimensional similarity embedding, a nonlinear perturbation function is introduced. For multi-dimensional similarity embedding matrix Make a perturbation:

[0075] The specific formula of the nonlinear perturbation function is:

[0076] ;

[0077] in, , is the linear modulation that controls the hyperbolic tangent function; It is the scaling factor of the hyperbolic tangent function in the nonlinear perturbation function, which adjusts the amplification or reduction of the input variable; It is the translation amount of the hyperbolic tangent function in the nonlinear perturbation function, which adjusts the offset of the input variable; is the perturbation intensity, which is used to control the influence of the sine function; , It controls the frequency and phase shift of the sine function, which are used to adjust the frequency amplification and phase shift of the input variable respectively; represents the input variable of the nonlinear perturbation function; is the hyperbolic tangent function;

[0078] Furthermore, the nonlinear perturbation function Application to multi-dimensional similarity embedding matrix Each element in generates a nonlinear perturbation matrix , the nonlinear perturbation matrix Any element in is represented as:

[0079] ;

[0080] in, is the nonlinear perturbation matrix No. Line Column elements represent the patient's characteristic values ​​after nonlinear perturbation; is the multidimensional similarity embedding matrix No. Line Column elements;

[0081] Further, we get the nonlinear perturbation matrix Finally, the adaptive multiple embedding aggregation mechanism is introduced. First, the nonlinear perturbation matrix is ​​calculated The contribution weight of each feature in the final feature fusion , contribution weight By the nonlinear perturbation matrix For multi-dimensional similarity embedding matrix The partial derivatives of are quantified as:

[0082] ;

[0083] in, represents the nonlinear perturbation matrix The total number of rows; represents the nonlinear perturbation matrix Total number of columns; and They represent the nonlinear perturbation matrices row index and column index; represents the nonlinear perturbation matrix No. Line Column elements; is the multidimensional similarity embedding matrix No. Line Column elements; ;

[0084] Get contribution weight Finally, the comprehensive feature generation function is introduced , and obtain the comprehensive feature vector, that is, the fused comprehensive data of the patient , the formula is as follows:

[0085] ;

[0086] in, It is the sinusoidal function perturbation intensity coefficient in the comprehensive feature generation function, which controls the influence of the sinusoidal perturbation term; It is the frequency coefficient of the sine function in the comprehensive feature generation function, which adjusts the frequency amplification of the sine disturbance term; is the activation function.

[0087] S2. Generate a preliminary personalized drug treatment plan based on the integrated patient data; based on the preliminary personalized drug treatment plan, quantify the risk of the drug combination in the preliminary personalized drug treatment plan through a multi-dimensional interaction field and a dynamic entropy weight risk assessment algorithm to obtain a risk score; based on the risk score, optimize the preliminary personalized drug treatment plan to generate a safe personalized drug treatment plan.

[0088] Based on the integrated patient data The drug treatment data provided by the drug treatment database is used to generate a preliminary personalized drug treatment plan using the existing multi-task learning model. The multi-task learning model is input with the fused patient comprehensive data and drug treatment data, select appropriate drug combinations and generate specific doses and medication frequencies according to the probability distribution and dosage recommendations output by the multi-task learning model. The multi-task learning model includes an input layer, multiple shared hidden layers (the specific number is determined by the usage scenario and will not be elaborated here), and task-specific layers (different task-specific output layers are designed for different tasks, such as drug selection and dosage recommendations).

[0089] Furthermore, in complex drug treatment plans, there are multiple interactions between different drugs (pharmacokinetics, pharmacodynamics, etc.), and the interactions will show dynamic changes under the influence of different time, dosage and patient characteristics. For this reason, a multidimensional interaction field and a dynamic entropy weight risk assessment algorithm are introduced. The multidimensional interaction field and a dynamic entropy weight risk assessment algorithm analyze the complex dynamic behavior of drug interactions from a multidimensional perspective, and dynamically adjust the weights of each interaction dimension in combination with the entropy weight theory to obtain more accurate risk quantification results, and obtain safe and personalized drug treatment plans through optimization strategies. The specific implementation process is as follows:

[0090] First, a drug interaction matrix was constructed; the initial personalized drug treatment plan included Drugs, denoted as (in ). Define the drug interaction matrix , whose elements Indicates the drugs in the initial personalized drug treatment plan and medications Drug interaction matrix The construction process is as follows:

[0091] The first step is data acquisition. The interaction data between each pair of drugs are extracted from the drug interaction database, such as pharmacokinetic interaction, pharmacodynamic interaction, metabolic pathway competition, etc. Each type of interaction is represented by different indicators.

[0092] The second step is matrix filling; according to the drug The drug interaction matrix Corresponding position in Fill in the corresponding interaction strength, which can be expressed as a standardized score (such as a relative strength value in the range of 0-1). The drug interaction matrix The diagonal elements of are 0;

[0093] The drug interaction matrix Provide basic data structure for subsequent multi-dimensional interaction field modeling, ensuring that the interaction information of each pair of drugs can be systematically expressed;

[0094] Secondly, in order to decompose the complex interaction information between drugs into different dimensions, a multidimensional interaction field vector is introduced , Indicates drug The multidimensional interaction field vector, each component of which Indicates drug In the The interaction strength on the interaction dimension, each component The calculation process is as follows:

[0095] Determine the interaction dimension; classify all possible drug interaction factors into dimensions (such as metabolic pathway competition, receptor affinity, drug metabolizing enzyme inhibition, etc.), represents the total number of interaction dimensions, each of which represents an independent interaction type;

[0096] Calculate the interaction strength of the interaction dimension; in each interaction dimension On, calculate the drug and medications The interaction strength , the formula is as follows:

[0097] ;

[0098] in, It's medicine and medications In the The weight coefficient of the dimension interaction reflects the effect of this dimension on the drug and medications the importance of overall interactions; Indicates drug and medications In the Dimensional interaction influencing factors, such as receptor affinity (pKi value), metabolic enzyme competition coefficient, etc. and Respectively represent drugs and The specific dose is a dimensionless value after dimensionless processing; and Respectively represent drugs and medications In the The duration of action of the dimension was obtained from the drug treatment database; is the adjustment coefficient, which controls the degree of influence of drug action time. The above formula comprehensively considers the influence of drug dose, interaction influencing factor and drug action time on the interaction intensity, ensuring the multi-dimensional expression ability of multi-dimensional interaction field vector;

[0099] Furthermore, all drug interactions and medications The interaction strength of the drug and medications The multidimensional interaction field strength between , and its calculation formula is:

[0100] ;

[0101] Indicates drug and medications The composite strength in all interaction dimensions reflects the overall interaction strength between the two drugs. By integrating the interaction strengths of each interaction dimension, an overall interaction field strength index is formed, laying the foundation for subsequent dynamic weight adjustment and risk assessment. represents the total number of interaction dimensions;

[0102] Furthermore, a dynamic entropy weight adjustment mechanism is introduced; in multidimensional interactions, the importance of different interaction dimensions will change dynamically according to the different characteristics of the patients. The entropy weight method is used to determine the weight of each interaction dimension to dynamically adjust the risk assessment. The entropy value The calculation steps are as follows:

[0103] First, normalization is performed; The interaction strength on each interaction dimension Normalized to the interval [0,1], we get Normalized interaction strength on each interaction dimension ;

[0104] Further, the probability distribution of each drug pair is calculated , for each drug pair , calculate its The proportion of normalized interaction strength on the interaction dimension:

[0105] ;

[0106] Finally, the entropy value of the interaction dimension is calculated; according to the probability distribution of each drug pair , calculate the The entropy value of the interaction dimension :

[0107] ;

[0108] in, is the total number of drug pairs; The entropy value of the interaction dimension To measure the effect of drugs on The degree of discreteness of the interaction intensity on each interaction dimension. The higher the entropy value, the more dispersed the information in this dimension is and the lower its importance is.

[0109] Furthermore, based on the entropy value of the interaction dimension obtained , dynamically adjust the weight according to the importance of each interaction dimension, and calculate the dynamic weight of each interaction dimension , which provides a basis for the subsequent calculation of dynamic weighted interaction intensity:

[0110] ;

[0111] in, Indicates The entropy value of the interaction dimension; represents the total number of interaction dimensions;

[0112] Dynamic Weight The calculation process is as follows: First, The entropy value of the interaction dimension Perform reverse mapping to obtain the reverse entropy value of the interaction dimension; The entropy value of the interaction dimension Converted into a negative indicator of weight, that is , the smaller the entropy value, the more concentrated the information in this dimension is and the higher its importance is; then the dynamic weights are normalized; the reverse entropy values ​​of all interaction dimensions are normalized to ensure that the sum of all dynamic weights is 1;

[0113] Furthermore, the dynamic weight Applied to multidimensional interaction field strength, dynamic weighted interaction strength is obtained :

[0114] ;

[0115] The above formula combines the dynamic weight and the interaction strength of each interaction dimension to express the drug and medications The total interaction strength under dynamic weighting reflects the importance of each interaction dimension under different patient conditions, and dynamically adjusts the total interaction strength accordingly; represents the total number of interaction dimensions;

[0116] Furthermore, to quantify the risk of the entire drug combination, the dynamic weighted interaction strength of all drug pairs is calculated as Perform standardization to obtain the risk entropy weight ; Based on risk entropy weight , assessing the risk score for each drug pair in the initial personalized drug regimen :

[0117] ;

[0118] The above formula represents the risk entropy weighted score of each pair of drug combinations in the preliminary personalized drug treatment plan. The higher the risk score, the more concentrated the risk of drug interaction, and the lower the risk score, the more dispersed the risk of drug interaction.

[0119] Finally, the risk score threshold is set according to the expert experience method ,like , the current drug combination in the preliminary personalized drug treatment plan is considered safe; otherwise, the current drug combination is judged to be high-risk and needs to be optimized and adjusted through optimization strategies.

[0120] The optimization strategies include dose adjustment and time separation; among them, dose adjustment: for high-risk drug combinations, priority is given to reducing their specific doses. or , thereby reducing the dynamic weighted interaction strength (The influence of each drug is quantified by the partial derivative of the dynamic weighted interaction intensity with respect to the specific dose, and the influence is compared to further adjust the drugs with large influence); Time separation: high-risk drugs are used at different times to reduce their overlap in the multidimensional interaction field, thereby reducing the risk entropy weighted score.

[0121] Through the above optimization strategy, a safe personalized drug treatment plan is generated. The safe personalized drug treatment plan is based on multiple strategies such as dosage adjustment and time separation, and is repeatedly optimized to ensure the safety of the personalized drug treatment plan.

[0122] The patient is treated with a safe and personalized drug therapy plan. During the treatment process, the patient's condition change trend is tracked in real time to evaluate the effectiveness of the safe and personalized drug therapy plan. When changes in the patient's condition are detected, the changes are fed back to the data acquisition and preprocessing module to form a closed loop until the patient's condition stabilizes.

[0123] In summary, the precise drug treatment management system and method for patients with severe infections has been completed.

[0124] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0125] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0126] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for the precise drug treatment management of patients with severe infection, characterized in that: The following steps are involved: S1. Collect multi-source medical data of patients and pre-process them to obtain pre-processed medical data; Perform multi-source fusion processing on the pre-processed medical data to obtain fused comprehensive patient data; S2. Generate a preliminary personalized drug treatment plan based on the integrated patient data; introduce a multidimensional interaction field and a dynamic entropy weight risk assessment algorithm, and construct a drug interaction matrix based on the preliminary personalized drug treatment plan; based on the drug interaction matrix, obtain the interaction strength of the interaction dimension; the calculation formula for the interaction strength of the interaction dimension is as follows: in, In each interaction dimension Drugs in the initial personalized drug treatment plan and medications The strength of the interaction; It is a drug interaction information; It's medicine and medications In the The weight coefficient of the dimensional interaction; Indicates drug and medications In the The interaction factors of the dimensions; and Respectively represent drugs and Specific dosage; and Respectively represent drugs and medications In the The duration of the dimension's action; It is the adjustment coefficient of the degree of influence of controlling the duration of drug action; The interaction strengths on all interaction dimensions are combined to obtain the multidimensional interaction field strength; Introduce a dynamic entropy weight adjustment mechanism, obtain the entropy value of the interaction dimension based on the interaction intensity of the interaction dimension; dynamically adjust the weight based on the entropy value of the interaction dimension, and calculate the dynamic weight of each interaction dimension; apply the dynamic weight of each interaction dimension to the multi-dimensional interaction field intensity to obtain the dynamic weighted interaction intensity; Based on the dynamic weighted interaction intensity, the risk entropy weight is obtained; based on the risk entropy weight, the risk of the drug combination in the preliminary personalized drug treatment plan is quantified to obtain a risk score; based on the risk score, the preliminary personalized drug treatment plan is optimized to generate a safe personalized drug treatment plan.

2. The method for precise drug treatment management of patients with severe infection according to claim 1, characterized in that: The S1 specifically includes: The specific steps of the multi-source fusion processing are as follows: first, the pre-processed medical data is merged to obtain merged medical data; then, the merged medical data is feature extracted to obtain preliminary multi-dimensional feature data of the patient's medical data; finally, the preliminary multi-dimensional feature data of the patient's medical data is processed by a feature fusion algorithm based on multi-dimensional similarity embedding and nonlinear perturbation to obtain fused comprehensive patient data.

3. The method for precise drug treatment management of patients with severe infection according to claim 2, characterized in that: In the S1, it also includes: In the process of implementing the feature fusion algorithm based on multidimensional similarity embedding and nonlinear perturbation, the preliminary multidimensional feature data of the patient medical data is standardized and transformed, and a standardized feature matrix is ​​generated based on the standardized preliminary multidimensional feature data of the patient medical data; based on the standardized feature matrix, a similarity measurement matrix is ​​calculated; based on the similarity measurement matrix, the standardized feature matrix is ​​spatially mapped to obtain a multidimensional similarity embedding matrix.

4. The method for precise drug treatment management of patients with severe infection according to claim 3, characterized in that: In the S1, it also includes: A nonlinear perturbation function is introduced to perturb the multidimensional similarity embedding matrix to generate a nonlinear perturbation matrix; any element in the nonlinear perturbation matrix is ​​expressed as: in, is the nonlinear perturbation matrix Line Column elements; is the first element of the multidimensional similarity embedding matrix Line Column elements; is a nonlinear perturbation function; is the scaling factor that controls the hyperbolic tangent function; is the translation amount that controls the hyperbolic tangent function; is the perturbation strength that controls the influence of the sine function; , It is the frequency and phase shift of the controlled sine function; is the hyperbolic tangent function.

5. The method for precise drug treatment management of patients with severe infection according to claim 4, characterized in that: In the S1, it also includes: Based on the nonlinear perturbation matrix, an adaptive multiple embedding aggregation mechanism is introduced to calculate the contribution weight, and a comprehensive feature generation function is introduced to obtain the fused patient comprehensive data; the specific formula of the comprehensive feature generation function is: in, It is the comprehensive patient data after fusion; is the comprehensive feature generating function; and Respectively represent the row index and column index in the nonlinear perturbation matrix; represents the total number of rows of the nonlinear perturbation matrix; represents the total number of columns of the nonlinear perturbation matrix; is the contribution weight; is the nonlinear perturbation matrix Line Column elements; is the sinusoidal function perturbation intensity coefficient in the comprehensive feature generating function; is the frequency coefficient of the sine function in the comprehensive feature generating function; is the activation function.

6. The method for precise drug treatment management of patients with severe infection according to claim 1, characterized in that: The S2 specifically includes: The risk score is compared with a preset risk score threshold. When the risk score is greater than the risk score threshold, the current drug combination in the preliminary personalized drug treatment plan is optimized to generate a safe personalized drug treatment plan.

7. A precise drug treatment management system for patients with severe infections, applied to the precise drug treatment management method for patients with severe infections as claimed in claim 1, characterized in that: Includes the following parts: Data acquisition and preprocessing module, multi-source data fusion module, drug treatment database, individualized treatment decision module, drug interaction database, drug interaction detection module, dynamic monitoring and feedback regulation module; The data collection and preprocessing module collects and preprocesses the multi-source medical data of patients to obtain preprocessed medical data; Send the preprocessed medical data to the multi-source data fusion module; The multi-source data fusion module performs multi-source fusion processing on the pre-processed medical data to obtain the fused comprehensive patient data; the fused comprehensive patient data is sent to the individualized treatment decision module; A drug therapy database, storing drug therapy data and sending the drug therapy data to an individualized treatment decision module; The individualized treatment decision module generates a preliminary personalized drug treatment plan based on the integrated patient data and the drug treatment data provided by the drug treatment database; sending the preliminary personalized drug treatment plan to the drug interaction detection module; a drug interaction database, storing drug interaction data and sending the drug interaction data to a drug interaction detection module; The drug interaction detection module combines drug interaction data and uses a multidimensional interaction field and a dynamic entropy weight risk assessment algorithm to quantify the preliminary personalized drug treatment plan, assess the risk of each pair of drug combinations in the preliminary personalized drug treatment plan, and obtain a risk score; The risk score is compared with a preset risk score threshold, and when the risk score is greater than the risk score threshold, the current drug combination in the preliminary personalized drug treatment plan is optimized to generate a safe personalized drug treatment plan; Send safe personalized drug treatment plans to the dynamic monitoring and feedback regulation module; Dynamic monitoring and feedback regulation module, patients are treated with safe personalized drug treatment plans. During the treatment process, the patient's condition change trend is tracked in real time to evaluate the safe personalized drug treatment plan; When changes in the patient's condition are detected, the changes are fed back to the data acquisition and preprocessing module to form a closed loop until the patient's condition stabilizes.

Images