A method and system for automatically optimizing parameters of mathematical model for individualized medication

Abstract

The present application relates to an automatic optimization method and system for the parameters of a mathematical model for personalized medication. The method includes: obtaining physiological data, medication data, drug concentration monitoring data and test data of a training patient; using a population pharmacokinetic model for processing to obtain a predicted value of blood drug concentration; based on the error information between the predicted value of the blood drug concentration and the actual measured blood drug concentration value, automatically optimizing the model parameters of the population pharmacokinetic model to obtain a trained population pharmacokinetic model; using the trained population pharmacokinetic model for processing to obtain a predicted result, which is a predicted value of the blood drug concentration of the test patient. In this way, the parameters of the population pharmacokinetic model can be adjusted by an error feedback mechanism through automatic iteration, so that the predicted blood drug concentration result is as close as possible to the actual measured value, thereby improving the prediction accuracy of the population pharmacokinetic model.

Description

Technical Field

[0001] The present application relates to the field of personalized medical technology, and in particular, to a method and system for automatically optimizing parameters of a mathematical model for individualized medication. Background Art

[0002] With the advancement of medical technology and biostatistics, personalized medicine is becoming a reality. The goal of precision medicine is to determine the most appropriate type of drug and its dosage based on the individual characteristics of the patient (such as genotype, physiological status, etc.) in order to formulate the most appropriate treatment plan in order to achieve the best treatment effect and reduce the occurrence of side effects. The effectiveness and safety of drug treatment depend to a large extent on the concentration of drugs in the patient's body, and drug concentration is affected by many factors, including the patient's physiological and pathological status, the degree of drug metabolism, and drug interactions. Therefore, accurate prediction of blood drug concentration is crucial to achieve personalized medication.

[0003] Population pharmacokinetic models are statistical tools used to describe the absorption, distribution, metabolism, and excretion (ADME) processes of drugs in a specific population. This model should take into account not only inter-individual differences but also intra-individual variability. It can integrate data from multiple sources, such as clinical trials and laboratory test data, to more accurately predict the pharmacokinetic behavior of drugs in individuals. However, in practical applications, population pharmacokinetic models face challenges, mainly because the core parameters of the model have a large variance, resulting in scattered prediction results, which indicates that there may be large errors for actual patients. In addition, many of the model parameters currently used are calculated based on some actual medical cases in the past, and these data may be affected by factors such as differences in geographical distribution, differences in population characteristics, measurement equipment errors, and insufficient sample size.

[0004] Therefore, it is expected to develop an automatic optimization scheme for the parameters of the mathematical model of personalized medication to enhance the generalization ability of the model and improve the predictive performance of the model. Summary of the invention

[0005] This summary is provided to introduce concepts in a brief form that will be described in detail in the detailed description below. This summary is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0006] In a first aspect, the present application provides a method for automatically optimizing parameters of a mathematical model for personalized medication, the method comprising:

[0007] Obtaining physiological data, medication data, drug concentration monitoring data, and test data of training patient subjects;

[0008] Using a population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the training patient subject to obtain a predicted value of blood drug concentration;

[0009] Based on the error information between the predicted value of the blood drug concentration and the actual measured blood drug concentration value, automatically optimizing the model parameters of the population pharmacokinetic model to obtain a trained population pharmacokinetic model;

[0010] The trained population pharmacokinetic model is used to process the physiological data, medication data, drug concentration monitoring data and test data of the test patient subject to obtain a prediction result, which is the predicted value of the blood drug concentration of the test patient subject.

[0011] Optionally, the physiological data, medication data, drug concentration monitoring data and test data of the training patient objects are processed using a population pharmacokinetic model to obtain a predicted value of blood drug concentration, including: using a physiological data embedding coding matrix to embed the physiological data of the training patient objects to obtain a low-dimensional embedded coding vector for training physiological data; using a medication data embedding coding matrix to embed the medication data of the training patient objects to obtain a low-dimensional embedded coding vector for training medication data; using a drug concentration monitoring data embedding coding matrix to embed the drug concentration monitoring data of the training patient objects to obtain a low-dimensional embedded coding vector for training drug concentration monitoring data; using a test data embedding coding matrix to embed the test data of the training patient objects to obtain a low-dimensional embedded coding vector for training test data; using the population pharmacokinetic model to perform pharmacokinetic encoding on the low-dimensional embedded coding vector for training physiological data, the low-dimensional embedded coding vector for training medication data, the low-dimensional embedded coding vector for training drug concentration monitoring data and the low-dimensional embedded coding vector for training test data to obtain a pharmacokinetic time series coding vector; decoding the pharmacokinetic time series coding vector to obtain the predicted value of blood drug concentration.

[0012] Optionally, the physiological data of the training patient subjects include the underlying diseases, age, gender, weight, height, body surface area, concomitant medications, genotype, liver function indicators, kidney function indicators and clinical test indicators such as blood routine examination of the training patient subjects; the medication data of the training patient subjects include administration time, dosage, administration route, administration frequency and infusion time, etc.; the drug concentration monitoring data of the training patient subjects include monitoring time, blood drug concentration value, detection method and instrument type; the test data include liver and kidney function test data, blood routine examination data, and infection index examination data.

[0013] Optionally, the population pharmacokinetic model is used to perform pharmacokinetic encoding on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain a pharmacokinetic time series coding vector, including: using the population pharmacokinetic model to perform pharmacokinetic encoding based on node potential spatiotemporal information transfer on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain the pharmacokinetic time series coding vector.

[0014] Optionally, the population pharmacokinetic model is used to perform pharmacokinetic encoding based on node potential spatiotemporal information transmission on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain the pharmacokinetic time series coding vector, including: using the training drug concentration monitoring data low-dimensional embedded coding vector as the target data low-dimensional embedded coding vector, and using the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector as the message propagation data low-dimensional embedded coding vector. A sequence of vectors is prepared, and the information transmission dynamic energy spatiotemporal transfer weights of each message propagation data low-dimensional embedded coding vector in the sequence of message propagation data low-dimensional embedded coding vectors relative to the target data low-dimensional embedded coding vector are calculated to obtain a sequence of information transmission dynamic energy spatiotemporal transfer weights; based on the sequence of information transmission dynamic energy spatiotemporal transfer weights, the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training inspection data low-dimensional embedded coding vector are fused, and then the obtained node information fusion representation is passed to the training drug concentration monitoring data low-dimensional embedded coding vector to obtain the pharmacokinetic timing coding vector.

[0015] Optionally, the information transfer dynamic energy spatiotemporal transfer weights of each message propagation data low-dimensional embedded coding vector in the sequence of the message propagation data low-dimensional embedded coding vectors relative to the target data low-dimensional embedded coding vector are calculated to obtain a sequence of information transfer dynamic energy spatiotemporal transfer weights, including: calculating the data low-dimensional embedded coding feature static energy value of each message propagation data low-dimensional embedded coding vector in the sequence of the message propagation data low-dimensional embedded coding vectors; determining the data low-dimensional embedded coding feature energy spatial attenuation factor based on the spatial span coefficient between each message propagation data low-dimensional embedded coding vector and the target data low-dimensional embedded coding vector; determining the data low-dimensional embedded coding feature energy spatial attenuation factor based on the spatial span coefficient between each message propagation data low-dimensional embedded coding vector and the target data low-dimensional embedded coding vector The time span coefficient between the quantities is used to determine the data low-dimensional embedding coding feature energy time decay factor; based on the data low-dimensional embedding coding feature energy space decay factor and the data low-dimensional embedding coding feature energy time decay factor, the data low-dimensional embedding coding feature static energy value of each message propagation data low-dimensional embedding coding vector is temporally modulated to obtain the data low-dimensional embedding coding feature dynamic energy spatiotemporal transfer value of each message propagation data low-dimensional embedding coding vector relative to the target data low-dimensional embedding coding vector; the data low-dimensional embedding coding feature dynamic energy spatiotemporal transfer value of each message propagation data low-dimensional embedding coding vector is input into the energy transfer gating module based on the mask function to obtain the sequence of information transfer dynamic energy spatiotemporal transfer weights.

[0016] Optionally, the data low-dimensional embedded coding feature static energy value of each message propagation data low-dimensional embedded coding vector in the sequence of the message propagation data low-dimensional embedded coding vector is calculated, including: respectively calculating the mean and standard deviation of the message propagation data low-dimensional embedded coding vector to obtain the message propagation data low-dimensional embedded coding feature mean and the message propagation data low-dimensional embedded coding standard deviation; after calculating the fourth power of the positional difference between the message propagation data low-dimensional embedded coding vector and the message propagation data low-dimensional embedded coding feature mean, the expected value of the fourth-power modulated message propagation data low-dimensional embedded coding offset vector is calculated to obtain the data low-dimensional embedded coding feature offset expectation factor; after calculating the division between the data low-dimensional embedded coding feature offset expectation factor and the fourth power of the message propagation data low-dimensional embedded coding standard deviation, the resultant addition is made to the maximum eigenvalue in the message propagation data low-dimensional embedded coding vector to obtain the data low-dimensional embedded coding feature static energy value.

[0017] Optionally, based on the sequence of information transfer dynamic energy spatiotemporal transfer weights, the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector are fused and then the obtained node information fusion representation is passed to the training drug concentration monitoring data low-dimensional embedded coding vector to obtain the pharmacokinetic timing coding vector, including: based on the sequence of information transfer dynamic energy spatiotemporal transfer weights, calculating the weighted sum of the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain the message propagation data low-dimensional embedded coding node feature transfer aggregation modulation vector as the node information fusion representation; calculating the positional sum of the message propagation data low-dimensional embedded coding node feature transfer aggregation modulation vector and the training drug concentration monitoring data low-dimensional embedded coding vector to obtain the pharmacokinetic timing coding vector.

[0018] In a second aspect, the present application provides an automatic optimization system for parameters of a mathematical model of personalized medication, the system comprising:

[0019] A data acquisition module, used to acquire physiological data, medication data, drug concentration monitoring data and test data of training patient subjects;

[0020] A data processing module, for processing the physiological data, medication data, drug concentration monitoring data and test data of the training patient subject using a population pharmacokinetic model to obtain a predicted value of blood drug concentration;

[0021] A model parameter training module, for automatically optimizing the model parameters of the population pharmacokinetic model based on the error information between the predicted value of the blood drug concentration and the actual measured blood drug concentration value to obtain a trained population pharmacokinetic model;

[0022] The prediction result generation module is used to use the trained population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the test patient object to obtain a prediction result, wherein the prediction result is the predicted value of the blood drug concentration of the test patient object.

[0023] Optionally, the data processing module includes: a physiological data embedding coding unit, which is used to embed the physiological data of the training patient object using a physiological data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training physiological data; a medication data embedding coding unit, which is used to embed the medication data of the training patient object using a medication data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training medication data; and a drug concentration monitoring data embedding coding unit, which is used to embed the drug concentration monitoring data of the training patient object using a drug concentration monitoring data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training drug concentration monitoring data. vector; a test data embedding coding unit, used to embed the test data of the training patient object using the test data embedding coding matrix to obtain a training test data low-dimensional embedding coding vector; a pharmacokinetic coding unit, used to use the population pharmacokinetic model to perform pharmacokinetic coding on the training physiological data low-dimensional embedding coding vector, the training medication data low-dimensional embedding coding vector, the training drug concentration monitoring data low-dimensional embedding coding vector and the training test data low-dimensional embedding coding vector to obtain a pharmacokinetic timing coding vector; a vector decoding unit, used to decode the pharmacokinetic timing coding vector to obtain the predicted value of the blood drug concentration.

[0024] The above technical solution is adopted to obtain the physiological data, medication data, drug concentration monitoring data and test data of the training patient object; use the population pharmacokinetic model to process to obtain the predicted value of blood drug concentration; based on the error information between the predicted value of blood drug concentration and the actual measured blood drug concentration value, automatically optimize the model parameters of the population pharmacokinetic model to obtain a trained population pharmacokinetic model; use the trained population pharmacokinetic model to process to obtain the predicted result, which is the predicted value of the blood drug concentration of the test patient object. In this way, the parameters of the population pharmacokinetic model can be adjusted by the error feedback mechanism through automatic iteration, so that the predicted blood drug concentration result is as close as possible to the actual measured value, thereby improving the prediction accuracy of the population pharmacokinetic model.

[0025] Other features and advantages of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The above and other features, advantages and aspects of the embodiments of the present application will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale. In the drawings:

[0027] Figure 1The present invention is a flowchart of a method for automatically optimizing parameters of a mathematical model for individualized medication according to an exemplary embodiment.

[0028] Figure 2 is based on Figure 1 The illustrated embodiment shows a flow chart of step S102 of a method for automatically optimizing parameters of a mathematical model for individualized medication.

[0029] Figure 3 It is a block diagram of an automatic optimization system for parameters of a mathematical model of individualized medication according to an exemplary embodiment.

[0030] Figure 4 It is a block diagram of an electronic device according to an exemplary embodiment.

[0031] Figure 5 It is an application scenario diagram of a method for automatically optimizing parameters of a mathematical model for individualized medication according to an exemplary embodiment. DETAILED DESCRIPTION

[0032] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.

[0033] It should be understood that the various steps described in the method implementation of the present application can be performed in different orders and / or performed in parallel. In addition, the method implementation may include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0034] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0035] It should be noted that the concepts such as "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0036] It should be noted that the modifications of "one" and "plurality" mentioned in the present application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0037] The names of the messages or information exchanged between multiple devices in the embodiments of the present application are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0038] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0039] Figure 1 is a flow chart of a method for automatically optimizing parameters of a mathematical model for personalized medication according to an exemplary embodiment. Figure 1 As shown, the method includes:

[0040] Step S101, obtaining physiological data, medication data, drug concentration monitoring data and test data of a training patient object;

[0041] Step S102, using a population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the training patient subject to obtain a predicted value of blood drug concentration;

[0042] Step S103, based on the error information between the predicted value of the blood drug concentration and the actual measured blood drug concentration value, automatically optimizing the model parameters of the population pharmacokinetic model to obtain a trained population pharmacokinetic model;

[0043] Step S104: using the trained population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the test patient subject to obtain a prediction result, wherein the prediction result is a predicted value of the blood drug concentration of the test patient subject.

[0044] It should be understood that in clinical practice, due to changes in the physiological and pathological characteristics of patients, the kinetic properties of drugs may also be different from those of healthy volunteers. This may lead to poor efficacy in some patients due to too low blood drug levels, or adverse reactions due to too high blood drug concentrations. Therefore, the standard dosing regimen recommended in the instructions or medication guide may not be suitable for all patients. Doctors need to use blood drug concentration monitoring to develop personalized dosing regimens to ensure the effectiveness and safety of drug treatment.

[0045] Population pharmacokinetics, by establishing a mathematical model for personalized medication, allows the physiological differences between individual patients to be taken into account when taking medication, and clearly describes the absorption, distribution, metabolism and excretion process of the drug in the body, thereby achieving personalized prediction of blood drug concentration and providing a scientific basis for personalized medication. However, the existing population pharmacokinetic model has limitations and needs to be further improved to improve its accuracy and reliability in practical applications. In the context of the big data era, more and more data such as physiological parameters, human information, medication information and test data have become available. These big data provide the possibility of building a more accurate personalized medication model. The use of machine learning algorithms and deep learning frameworks can effectively extract useful information from complex data sets and use it to optimize the parameters of the population pharmacokinetic model and learn the source of error information.

[0046] Figure 2 is based on Figure 1 The embodiment shown is a flow chart of step S102 of a method for automatically optimizing parameters of a mathematical model for personalized medication. Figure 2 As shown, step S102, using a population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the training patient object to obtain a predicted value of blood drug concentration, includes: step S1021, using a physiological data embedding coding matrix to embed the physiological data of the training patient object to obtain a low-dimensional embedding coding vector of the training physiological data;

[0047] Step S1022, embedding and encoding the medication data of the training patient object using the medication data embedding coding matrix to obtain a low-dimensional embedding coding vector for the training medication data;

[0048] Step S1023, embedding the drug concentration monitoring data of the training patient object using the drug concentration monitoring data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training drug concentration monitoring data;

[0049] Step S1024, embedding the test data of the training patient object using the test data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training test data;

[0050] Step S1025, using the population pharmacokinetic model to perform pharmacokinetic encoding on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector, and the training test data low-dimensional embedded coding vector to obtain a pharmacokinetic time series coding vector;

[0051] Step S1026, decoding the pharmacokinetic time series coding vector to obtain the predicted value of the blood drug concentration.

[0052] Based on this, in the technical solution of the present application, a method for automatic optimization of parameters of a mathematical model of personalized medication is proposed, which includes: obtaining physiological data, medication data, drug concentration monitoring data and test data of training patient subjects; using a population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the training patient subjects to obtain a predicted value of blood drug concentration; based on the error information between the predicted value of the blood drug concentration and the actual measured blood drug concentration value, automatically optimizing the model parameters of the population pharmacokinetic model to obtain a trained population pharmacokinetic model; obtaining physiological data, medication data, drug concentration monitoring data and test data of a test patient subject; using the trained population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the test patient subject to obtain a prediction result, and the prediction result is the predicted value of the blood drug concentration of the test patient subject.

[0053] In one embodiment of the present application, physiological data reflects the patient's personal health status and physical condition, generally including but not limited to: underlying diseases, age, gender, weight, height, body surface area, combined medication, genotype, liver function indicators (such as ALT, AST, etc.), renal function indicators (such as creatinine clearance), blood routine and other factors that may affect drug metabolism. Medication data refers to the relevant information of the patient's use of drugs, including: administration time point, dosage, route of administration (oral, intravenous injection, etc.), administration frequency, infusion time, other drugs used at the same time and their interactions. Drug concentration monitoring data mainly involves monitoring of drug concentrations in patients, specifically including: actual measured monitoring time, blood drug concentration value, measurement time point, detection method and instrument type. Inspection data: In addition to the above-mentioned data specific to drug treatment, it also includes more extensive clinical examination data, such as: liver and kidney function test data, blood routine test data, infection index test data and marker detection of other specific diseases.

[0054] In particular, in the above-mentioned individualized medication mathematical model parameter optimization scheme, in the process of using the population pharmacokinetic model to generate the predicted value of blood drug concentration, the physiological data, medication data, drug concentration monitoring data and test data of the training patient are input into the data processing and analysis algorithm based on artificial intelligence and deep learning for processing, so as to capture the low-dimensional embedded coding features of physiological data, low-dimensional embedded coding features of medication data, low-dimensional embedded coding features of drug concentration monitoring data and low-dimensional embedded coding features of test data, and extract different parameter data features from these features through the population pharmacokinetic model. The information transmission semantics based on node potential time and space relative to the drug detection embedded coding features is used to reflect the time sequence coding information of pharmacokinetics, so as to predict blood drug concentration and optimize the parameters of the population pharmacokinetic model. In this way, the parameters of the population pharmacokinetic model can be adjusted by using the error feedback mechanism through automatic iteration, and the parameters in the population pharmacokinetic model can be continuously adjusted to better fit the actually observed drug concentration monitoring data, so that the predicted blood drug concentration results are as close as possible to the actual measured value, thereby improving the prediction accuracy of the population pharmacokinetic model. This method not only improves the model's accuracy in predicting blood drug concentrations, but also reduces the need for manual intervention, making the establishment and updating of population pharmacokinetic models more efficient, which helps support doctors in making more scientific and reasonable treatment decisions.

[0055] Specifically, in the process of using the population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the training patient objects to obtain the predicted value of blood drug concentration, since the population pharmacokinetic model can describe the absorption, distribution, metabolism and excretion process of the drug in the body, in order to optimize the parameters of the population pharmacokinetic model to more accurately perform personalized blood drug concentration prediction, in the technical solution of the present application, first, the physiological data of the training patient objects are embedded and encoded using a physiological data embedding coding matrix to obtain a low-dimensional embedded coding vector of training physiological data; the medication data embedding coding matrix is ​​used to embed the medication data of the training patient objects to obtain a low-dimensional embedded coding vector of training medication data; the drug concentration monitoring data embedding coding matrix is ​​used to embed the drug concentration monitoring data of the training patient objects to obtain a low-dimensional embedded coding vector of training drug concentration monitoring data; the test data embedding coding matrix is ​​used to embed the test data of the training patient objects to obtain a low-dimensional embedded coding vector of training test data.

[0056] It should be understood that the original physiological data, medication data, drug concentration monitoring data and test data of the training patient object may be high-dimensional and contain a lot of redundant information. Through embedded coding, high-dimensional data can be mapped to a lower-dimensional space while retaining important feature information of the data. This not only reduces the computational cost, but also avoids the problem of dimensionality disaster, making subsequent model training more efficient. In addition, since different data types may have different units and magnitudes, directly using the original data may cause certain features to dominate the learning process of the model. Through encoding conversion, different types of input data can be standardized, and at the same time, the process of embedded coding is actually a feature extraction process. It can help extract the most relevant features in the data, which is crucial to improving the predictive ability of the model. Especially for unstructured data, embedded coding can convert them into numerical form to facilitate machine learning model processing. Specifically, in the technical solution of the present application, the embedded coding method can capture the low-dimensional embedded coding features of the physiological data of the training patient object, the low-dimensional embedded coding features of the medication data, the low-dimensional embedded coding features of the drug concentration monitoring data, and the low-dimensional embedded coding features of the test data.

[0057] Furthermore, since the training physiological data low-dimensional embedding coding vector, the training medication data low-dimensional embedding coding vector, the training drug concentration monitoring data low-dimensional embedding coding vector and the training test data low-dimensional embedding coding vector respectively contain low-dimensional embedding coding feature representations of the physiological data, medication data, drug concentration monitoring data and test data of the training patient object, and different data embedding representations have different importance for the prediction task of blood drug concentration, this importance is reflected in the information transmission effect of time and space. That is to say, due to the physiological differences between individual patient objects, the absorption, distribution, metabolism and excretion capabilities of drugs in the body are different. Therefore, the blood drug concentration monitoring value and time series information expressed in the training drug concentration monitoring data low-dimensional embedding coding vector will be affected by the physiological indicators, medication parameters and test data parameters of the training patient object. The semantic information carried by these different parameters (including physiological indicators, medication parameters and test data parameters) in the embedding process has different degrees of influence on them. Therefore, in the process of information aggregation, it is necessary to consider this spatiotemporal information transmission to perform the subsequent blood drug concentration prediction task. Based on this, in the technical solution of the present application, the population pharmacokinetics model is used to perform pharmacokinetic encoding based on node potential spatiotemporal information transfer on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain a pharmacokinetic time series coding vector. The pharmacokinetic coding processing based on node potential spatiotemporal information transfer can calculate the node information fusion representation features between different data sources through the dynamic energy spatiotemporal transfer weight sequence, and pass it to the training drug concentration monitoring data low-dimensional embedded coding vector, in order to integrate the information of multiple data sources and improve the prediction accuracy and robustness of the model.

[0058] In one embodiment of the present application, the population pharmacokinetic model is used to perform pharmacokinetic encoding on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain a pharmacokinetic time series coding vector, including: using the population pharmacokinetic model to perform pharmacokinetic encoding based on node potential spatiotemporal information transfer on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain the pharmacokinetic time series coding vector.

[0059] Specifically, the process of pharmacokinetic coding based on node potential spatiotemporal information transmission includes: first, taking the training drug concentration monitoring data low-dimensional embedding coding vector as the target data low-dimensional embedding coding vector, and taking the training physiological data low-dimensional embedding coding vector, the training medication data low-dimensional embedding coding vector and the training inspection data low-dimensional embedding coding vector as a sequence of message propagation data low-dimensional embedding coding vectors, calculating the data low-dimensional embedding coding feature static energy value of each message propagation data low-dimensional embedding coding vector in the sequence of message propagation data low-dimensional embedding coding vectors, the static energy value is related to the maximum eigenvalue, mean and variance of each message propagation data low-dimensional embedding coding vector, the static energy value provides a quantitative measure of the inherent properties of the eigenvector in space, provides a benchmark for the spatiotemporal attenuation of potential energy, and helps the model understand the static importance of the data embedding representation represented by each node in the spatial dimension and the temporal dimension. Next, the data low-dimensional embedded coding feature energy space attenuation factor is determined based on the space span coefficient between the low-dimensional embedded coding vector of each message propagation data and the target data low-dimensional embedded coding vector, and the data low-dimensional embedded coding feature energy time attenuation factor is determined based on the time span coefficient between the low-dimensional embedded coding vector of each message propagation data and the target data low-dimensional embedded coding vector. The space span coefficient and the time span coefficient respectively reflect the spatial distance between nodes and the time distance between the nodes and the test data, and the attenuation factor adjusts the weight of information transmission between nodes based on these coefficients. It should be understood that the energy time attenuation factor and the energy space attenuation factor enable the model to adjust the intensity of information transmission according to the distance between nodes, enhance the model's sensitivity to locality and timeliness, and make these data information transmission and aggregation processes more in line with the actual process. Furthermore, the static energy value of the low-dimensional embedded coding feature of each message propagation data low-dimensional embedded coding vector is spatiotemporally modulated based on the spatial attenuation factor of the data low-dimensional embedded coding feature energy and the temporal attenuation factor of the data low-dimensional embedded coding feature energy, so as to adjust the static energy value by combining the spatial attenuation factor and the temporal attenuation factor through spatiotemporal modulation, so as to reflect the dynamic changes of the node's data parameter embedding semantics and features in time and space. Further, the dynamic energy spatiotemporal transfer value of the low-dimensional embedded coding feature of each message propagation data low-dimensional embedded coding vector is input into the energy transfer gating module based on the mask function, and the mask function is used to screen and strengthen the important spatiotemporal transfer parameter embedding information, and the gating mechanism determines which information is important and which can be ignored through learning. In this way, the accuracy and effectiveness of information transfer are improved through energy transfer gating, and the model can focus more on the features that contribute to the task by adaptively adjusting the information flow, ensuring that only the information that is meaningful to the subsequent blood drug concentration prediction task is transmitted, so as to improve the accuracy and effectiveness of information transfer.Next, based on the sequence of the dynamic energy spatiotemporal transfer weights of the information transfer, the weighted sum of the low-dimensional embedded coding vector of the training physiological data, the low-dimensional embedded coding vector of the training medication data, and the low-dimensional embedded coding vector of the training test data is calculated. That is, the information from the nodes where the embedding semantics of different data parameters are located is integrated by weighted summing of the dynamic energy spatiotemporal transfer values ​​of the low-dimensional embedded coding vector of the message propagation data. The obtained low-dimensional embedded coding node feature transfer aggregation modulation vector of the message propagation data is used as the node information fusion representation, thereby integrating the data embedding semantic information from different nodes, thereby improving the richness and accuracy of the feature representation. Finally, the low-dimensional embedded coding node feature transfer aggregation modulation vector of the message propagation data and the low-dimensional embedded coding vector of the training test data are summed by position to generate a pharmacokinetic time series coding vector that integrates the spatiotemporal information transfer information between the embedded coding semantics of the physiological data, medication data and test data of the training patient object and the embedded coding representation of the drug concentration monitoring data, which provides rich context information for the subsequent blood drug concentration prediction task and enhances the model's ability to handle complex tasks.

[0060] In one embodiment of the present application, the population pharmacokinetic model is used to perform pharmacokinetic encoding based on node potential spatiotemporal information transmission on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain the pharmacokinetic time series coding vector, including: using the training drug concentration monitoring data low-dimensional embedded coding vector as the target data low-dimensional embedded coding vector, and using the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector as the message propagation data low-dimensional Embed a sequence of coding vectors, calculate the information transfer dynamic energy spatiotemporal transfer weights of each message propagation data low-dimensional embedded coding vector in the sequence of message propagation data low-dimensional embedded coding vectors relative to the target data low-dimensional embedded coding vector to obtain a sequence of information transfer dynamic energy spatiotemporal transfer weights; based on the sequence of information transfer dynamic energy spatiotemporal transfer weights, fuse the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training inspection data low-dimensional embedded coding vector, and then pass the obtained node information fusion representation to the training drug concentration monitoring data low-dimensional embedded coding vector to obtain the pharmacokinetic timing coding vector.

[0061] Furthermore, in one embodiment of the present application, the information transfer dynamic energy spatiotemporal transfer weights of each message propagation data low-dimensional embedded coding vector in the sequence of the message propagation data low-dimensional embedded coding vectors relative to the target data low-dimensional embedded coding vector are calculated to obtain a sequence of information transfer dynamic energy spatiotemporal transfer weights, including: calculating the data low-dimensional embedded coding feature static energy value of each message propagation data low-dimensional embedded coding vector in the sequence of the message propagation data low-dimensional embedded coding vectors; determining the data low-dimensional embedded coding feature energy spatial attenuation factor based on the spatial span coefficient between each message propagation data low-dimensional embedded coding vector and the target data low-dimensional embedded coding vector; determining the data low-dimensional embedded coding feature energy spatial attenuation factor based on the spatial span coefficient between each message propagation data low-dimensional embedded coding vector and the target data The time span coefficient between the low-dimensional embedding coding vectors is used to determine the data low-dimensional embedding coding feature energy time decay factor; based on the data low-dimensional embedding coding feature energy space decay factor and the data low-dimensional embedding coding feature energy time decay factor, the data low-dimensional embedding coding feature static energy value of each message propagation data low-dimensional embedding coding vector is temporally modulated to obtain the data low-dimensional embedding coding feature dynamic energy spatiotemporal transfer value of each message propagation data low-dimensional embedding coding vector relative to the target data low-dimensional embedding coding vector; the data low-dimensional embedding coding feature dynamic energy spatiotemporal transfer value of each message propagation data low-dimensional embedding coding vector is input into the energy transfer gating module based on the mask function to obtain the sequence of the information transfer dynamic energy spatiotemporal transfer weights.

[0062] Furthermore, in one embodiment of the present application, the data low-dimensional embedded coding feature static energy value of each message propagation data low-dimensional embedded coding vector in the sequence of the message propagation data low-dimensional embedded coding vector is calculated, including: respectively calculating the mean and standard deviation of the message propagation data low-dimensional embedded coding vector to obtain the message propagation data low-dimensional embedded coding feature mean and the message propagation data low-dimensional embedded coding standard deviation; after calculating the fourth power of the positional difference between the message propagation data low-dimensional embedded coding vector and the message propagation data low-dimensional embedded coding feature mean, the expected value of the fourth-power modulated message propagation data low-dimensional embedded coding offset vector is calculated to obtain the data low-dimensional embedded coding feature offset expectation factor; after calculating the division between the data low-dimensional embedded coding feature offset expectation factor and the fourth power of the message propagation data low-dimensional embedded coding standard deviation, the result is added to the maximum eigenvalue in the message propagation data low-dimensional embedded coding vector to obtain the data low-dimensional embedded coding feature static energy value.

[0063] Furthermore, in one embodiment of the present application, based on the sequence of the information transfer dynamic energy spatiotemporal transfer weights, the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector are fused and then the obtained node information fusion representation is passed to the training drug concentration monitoring data low-dimensional embedded coding vector to obtain the pharmacokinetic timing coding vector, including: based on the sequence of the information transfer dynamic energy spatiotemporal transfer weights, calculating the weighted sum of the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain the message propagation data low-dimensional embedded coding node feature transfer aggregation modulation vector as the node information fusion representation; calculating the positional sum of the message propagation data low-dimensional embedded coding node feature transfer aggregation modulation vector and the training drug concentration monitoring data low-dimensional embedded coding vector to obtain the pharmacokinetic timing coding vector.

[0064] Specifically, the population pharmacokinetic model is used to perform pharmacokinetic encoding based on node potential spatiotemporal information transmission on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector using the following characteristic message propagation formula to obtain the pharmacokinetic time series coding vector;

[0065] Among them, the characteristic message propagation formula is: ; ; ; ; ; ;

[0066] in, is a set of data low-dimensional embedding coding vectors consisting of the training physiological data low-dimensional embedding coding vector, the training medication data low-dimensional embedding coding vector, the training drug concentration monitoring data low-dimensional embedding coding vector and the training inspection data low-dimensional embedding coding vector, are respectively the training physiological data low-dimensional embedding coding vector, the training medication data low-dimensional embedding coding vector, the training inspection data low-dimensional embedding coding vector and the training drug concentration monitoring data low-dimensional embedding coding vector, is the number of vectors in the set of low-dimensional embedding encoding vectors of the data, is the first of the low-dimensional embedding coding vectors for the training physiological data, the low-dimensional embedding coding vectors for the training medication data, and the low-dimensional embedding coding vectors for the training inspection data. data low-dimensional embedding encoding vector, is the first of the low-dimensional embedding coding vectors for the training physiological data, the low-dimensional embedding coding vectors for the training medication data, and the low-dimensional embedding coding vectors for the training inspection data. The low-dimensional embedding encoding vector of the data The eigenvalues ​​at the positions, and They are respectively The feature mean and standard deviation of the low-dimensional embedding encoding vector of each data, is to calculate the expected value, It means to take the maximum value in the vector. The static energy value of the low-dimensional embedding encoding feature is used for data. express and The spatial span coefficient between for and The energy space attenuation factor between and express and timestamp, Indicates a round-down operation. is the descaling parameter that controls the time decay period, is the value of the natural exponential function, for and The energy time decay factor between and is a trainable hyperparameter, for Relative to The dynamic energy space-time transfer value of For mask operation, is the threshold hyperparameter, for function, A vector encoding the pharmacokinetic time sequence.

[0067] Among them, based on the spatial span coefficient between each message propagation data low-dimensional embedded coding vector and the target data low-dimensional embedded coding vector, the data low-dimensional embedded coding feature energy spatial attenuation factor is determined, including: calculating the spatial distance span between the message propagation data low-dimensional embedded coding vector and the target data low-dimensional embedded coding vector to obtain the message propagation data low-dimensional embedded coding space span coefficient; calculating the division between the data low-dimensional embedded coding feature static energy value corresponding to the message propagation data low-dimensional embedded coding vector and the message propagation data low-dimensional embedded coding space span coefficient to obtain the data low-dimensional embedded coding feature energy space attenuation factor.

[0068] Among them, based on the time span coefficient between the low-dimensional embedded coding vectors of each message propagation data and the target data low-dimensional embedded coding vector, the data low-dimensional embedded coding feature energy time decay factor is determined, including: after calculating the subtraction between the timestamp of the target data low-dimensional embedded coding vector and the timestamp of the message propagation data low-dimensional embedded coding vector, the obtained timestamp value is rounded down to obtain the data low-dimensional embedded coding feature time span coefficient; after dividing the data low-dimensional embedded coding feature time span coefficient by the inverse scaling parameter that controls the time decay period, the obtained time span decay representation coefficient is used as the exponential power, and the natural exponential function value with the natural constant e as the base is calculated to obtain the time span decay class support representation coefficient; the division between the data low-dimensional embedded coding feature static energy value and the time span decay class support representation coefficient is calculated to obtain the data low-dimensional embedded coding feature energy time decay factor.

[0069] The method comprises: inputting the dynamic energy spatiotemporal transfer value of the low-dimensional embedding coding feature of each message propagation data low-dimensional embedding coding vector into the energy transfer gating module based on the mask function to obtain the sequence of the dynamic energy spatiotemporal transfer weights of the information transmission, including: in the energy transfer gating module based on the mask function, in response to the dynamic energy spatiotemporal transfer value of the low-dimensional embedding coding feature of the data being greater than the threshold hyperparameter, The function normalizes the dynamic energy spatiotemporal transfer value of the data low-dimensional embedding coding feature that is greater than the threshold hyperparameter to obtain the information transfer dynamic energy spatiotemporal transfer weight.

[0070] Then, the pharmacokinetic time-series coding vector is decoded to obtain the predicted value of the blood drug concentration. In this way, the pharmacokinetic time-series coding information can be used to predict the blood drug concentration and optimize the parameters of the population pharmacokinetic model. In this way, the population pharmacokinetic model parameters can be adjusted by the error feedback mechanism in the subsequent automatic iteration. By continuously adjusting the parameters in the population pharmacokinetic model, it can better fit the actual observed drug concentration monitoring data, so that the predicted blood drug concentration results are as close as possible to the actual measured value, thereby improving the prediction accuracy of the population pharmacokinetic model. This method not only improves the accuracy of the model's blood drug concentration prediction, but also reduces the need for manual intervention, making the establishment and update of the population pharmacokinetic model more efficient, which is conducive to supporting doctors to make more scientific and reasonable treatment decisions.

[0071] Preferably, inputting the pharmacokinetic time sequence coding vector into a decoder-based blood drug concentration prediction module to obtain a predicted value of blood drug concentration comprises:

[0072] Calculating the sum of the absolute values ​​of each eigenvalue of the pharmacokinetic timing coding vector to obtain a first pharmacokinetic timing coding and modulation value, and calculating the square root of the sum of the squares of each eigenvalue of the pharmacokinetic timing coding vector to obtain a second pharmacokinetic timing coding and modulation value;

[0073] After performing dot-wise subtraction on the pharmacokinetic timing coding vector and the second pharmacokinetic timing coding and modulation value, respectively perform dot-wise multiplication on the number of eigenvalues ​​of the pharmacokinetic timing coding vector and the reciprocal of the first pharmacokinetic timing coding and modulation value, and take the reciprocal of each eigenvalue to obtain a first pharmacokinetic timing coding phase conversion vector;

[0074] After point-wise subtraction of the pharmacokinetic timing coding vector and the first pharmacokinetic timing coding and modulation value, point-wise multiplication is performed with the square root of the number of eigenvalues ​​of the pharmacokinetic timing coding vector and the inverse of the second pharmacokinetic timing coding and modulation value, and the inverse of each eigenvalue is taken to obtain the second pharmacokinetic timing coding phase transition vector; point-wise subtraction of the first pharmacokinetic timing coding phase transition vector by the point product vector of the second pharmacokinetic timing coding phase transition vector and the weighted hyperparameter is performed to obtain an optimized pharmacokinetic timing coding vector; the optimized pharmacokinetic timing coding vector is input into a decoder-based blood drug concentration prediction module to obtain a predicted value of blood drug concentration.

[0075] Here, the pharmacokinetic time series encoding vector is recorded as The optimization expression is: ; ; ; ; ;

[0076] in, are the eigenvalues ​​of the pharmacokinetic time series encoding vector, is the first pharmacokinetic timing code and modulation value, is the second pharmacokinetic timing code and modulation value, is the pharmacokinetic temporal coding vector, represents the set of real numbers, is the number of eigenvalues ​​of the pharmacokinetic temporal encoding vector, is the inverse of the first pharmacokinetic timing code and modulation value, is the reciprocal of each eigenvalue, is the first pharmacokinetic timing encoding phase transition vector, is the reciprocal of the second pharmacokinetic timing encoding and modulation value, is the second pharmacokinetic timing encoding phase transition vector, is a weighted hyperparameter, is the optimized pharmacokinetic timing encoding vector, It is subtracted by position point. It is the point product by position.

[0077] More specifically, the decoder-based blood drug concentration prediction module is used to decode and regress the optimized pharmacokinetic time-series coding vector using the following decoding formula to obtain the predicted value of the blood drug concentration, wherein the decoding formula is: ,in, is the optimized pharmacokinetic timing encoding vector, is the predicted value of the blood drug concentration, is the weight matrix of the decoder, Represents matrix multiplication.

[0078] That is, considering that the training physiological data low-dimensional embedding coding vector, the training medication data low-dimensional embedding coding vector, the training drug concentration monitoring data low-dimensional embedding coding vector and the training test data low-dimensional embedding coding vector respectively represent the low-dimensional semantic embedding coding features of physiological data, medication data, drug concentration monitoring data and test data, when they are input into the characteristic message propagation network based on the node potential spatiotemporal transmission characteristics, based on the node potential spatiotemporal transmission differences of each data, the pharmacokinetic time series coding vector will also have global propagation-type aggregation differences. Therefore, it is expected to improve the detailed semantic aggregation expression effect of the pharmacokinetic time series coding vector based on the local feature propagation-type aggregation differences of each sample in the global sample domain.

[0079] Based on this, the present application uses the difference of the characteristic value of the pharmacokinetic time-series coding vector relative to the vector of the pharmacokinetic time-series coding vector as the semantic change intensity information, and performs a phase-like conversion corresponding to the position-based intensity modulation through different and modulation representation forms, so that the aggregation enhancement of the semantic change phase perception can improve the axial aggregation receptive field along the feature aggregation direction by performing a spatial translation operation based on alternating stacking under the vector set scale balance of the pharmacokinetic time-series coding vector, thereby improving the perception effect of the aggregation semantics of the pharmacokinetic time-series coding vector on the detail semantic change, so as to improve the expression effect of the pharmacokinetic time-series coding vector, and improve the accuracy of the predicted value of the blood drug concentration obtained by inputting the blood drug concentration prediction module based on the decoder. In this way, the error feedback mechanism is used to adjust the parameters of the population pharmacokinetic model through automatic iteration, and the parameters in the population pharmacokinetic model are continuously adjusted to make it better fit the actually observed drug concentration monitoring data, so that the predicted blood drug concentration result is as close as possible to the actual measured value, thereby improving the prediction accuracy of the population pharmacokinetic model.

[0080] In summary, in the process of using the above scheme to generate the predicted value of blood drug concentration using the population pharmacokinetic model, the physiological data, medication data, drug concentration monitoring data and test data of the training patient are input into the data processing and analysis algorithm based on artificial intelligence and deep learning for processing, so as to capture the low-dimensional embedded coding features of physiological data, low-dimensional embedded coding features of medication data, low-dimensional embedded coding features of drug concentration monitoring data and low-dimensional embedded coding features of test data, and extract different parameter data features from these features through the population pharmacokinetic model. The information transmission semantics based on node potential time and space between the embedded coding features of drug detection is used to reflect the time sequence coding information of pharmacokinetics, so as to predict blood drug concentration and optimize the parameters of the population pharmacokinetic model. In this way, the parameters of the population pharmacokinetic model can be adjusted by using the error feedback mechanism through automatic iteration, and the parameters in the population pharmacokinetic model can be adjusted to better fit the actually observed drug concentration monitoring data, so that the predicted blood drug concentration results are as close as possible to the actual measured values, thereby improving the prediction accuracy of the population pharmacokinetic model.

[0081] Figure 3 FIG. 1 is a block diagram of an automatic optimization system for parameters of a mathematical model for personalized medication according to an exemplary embodiment. Figure 3 As shown, the system 200 includes:

[0082] A data acquisition module 201 is used to acquire physiological data, medication data, drug concentration monitoring data and test data of a training patient object;

[0083] A data processing module 202, for processing the physiological data, medication data, drug concentration monitoring data and test data of the training patient subject using a population pharmacokinetic model to obtain a predicted value of blood drug concentration;

[0084] A model parameter training module 203, for automatically optimizing the model parameters of the population pharmacokinetic model based on the error information between the predicted value of the blood drug concentration and the actual measured blood drug concentration value to obtain a trained population pharmacokinetic model;

[0085] The prediction result generation module 204 is used to use the trained population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the test patient subject to obtain a prediction result, wherein the prediction result is a predicted value of the blood drug concentration of the test patient subject.

[0086] In one embodiment of the present application, the data processing module includes: a physiological data embedding coding unit, which is used to embed the physiological data of the training patient object using a physiological data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training physiological data; a medication data embedding coding unit, which is used to embed the medication data of the training patient object using a medication data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training medication data; and a drug concentration monitoring data embedding coding unit, which is used to embed the drug concentration monitoring data of the training patient object using a drug concentration monitoring data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training drug concentration monitoring data. Embedded coding vector; a test data embedded coding unit, used to use the test data embedded coding matrix to embed the test data of the training patient object to obtain a training test data low-dimensional embedded coding vector; a pharmacokinetic coding unit, used to use the population pharmacokinetic model to perform pharmacokinetic coding on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector and the training test data low-dimensional embedded coding vector to obtain a pharmacokinetic timing coding vector; a vector decoding unit, used to decode the pharmacokinetic timing coding vector to obtain the predicted value of the blood drug concentration.

[0087] In another embodiment of the present application, a computer software system for automatic optimization of parameters of a mathematical model for personalized medication is provided, the system comprising:

[0088] (1) Acquisition unit, used to collect patient physiological data, medication data, drug concentration monitoring data and test data;

[0089] (2) A prediction unit, which predicts blood drug concentration based on the existing population pharmacokinetic model and using the data obtained by the acquisition unit;

[0090] (3) A learning unit, based on a machine learning algorithm, identifies the source of prediction errors by learning the difference between the predicted blood drug concentration and the actual measured blood drug concentration; or automatically optimizes the parameters of the population pharmacokinetic model to correct the model prediction accuracy;

[0091] (4) The diagnosis unit automatically provides the first medication regimen, the adjustment medication regimen, the medication regimen for similar medical records, etc. based on the patient data obtained by the acquisition unit and the prediction results of the prediction unit;

[0092] (5) Database unit, which maintains relevant data and provides relevant data queries.

[0093] The population pharmacokinetic model used is a one-compartment model, a two-compartment model, or a one-compartment model and a two-compartment model. The machine learning algorithm used is a Bayesian maximum a posteriori estimation algorithm (MAP), a reinforcement learning algorithm, or a fusion algorithm of maximum a posteriori estimation and reinforcement learning.

[0094] The acquisition unit may be a medical record input interface in the software, and also have a data access interface for the hospital HIS system and LIS system, for automatically acquiring treatment information from the HIS system and LIS system.

[0095] The model used in the prediction unit uses different model parameters for different patient populations; multiple sets of model parameters are also provided for the same patient population; the system can automatically evaluate the effects of multiple sets of model parameters.

[0096] Reference below Figure 4 , which shows a schematic diagram of the structure of an electronic device 600 suitable for implementing the embodiment of the present application. The terminal device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0097] like Figure 4 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a ROM (read-only memory) 602 or a program loaded from a storage device 608 into a RAM (random access memory) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An I / O (input / output) interface 605 is also connected to the bus 604.

[0098] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 4The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0099] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present application are executed.

[0100] It should be noted that the computer-readable medium mentioned above in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0101] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0102] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0103] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0104] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0105] The modules involved in the embodiments described in this application can be implemented by software or hardware. The name of the module does not limit the module itself in some cases. For example, the test parameter acquisition module can also be described as a "module for acquiring device test parameters corresponding to the target device".

[0106] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0107] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0108] Figure 5 FIG. 1 is an application scenario diagram of an automatic optimization method for parameters of a mathematical model for personalized medication according to an exemplary embodiment. Figure 5 As shown, in this application scenario, first, the physiological data of the training patient object is obtained (for example, Figure 5 C1 as indicated in the figure), medication data (e.g. Figure 5 C2 as shown in the figure), drug concentration monitoring data (e.g. Figure 5 C3 as shown in the figure) and test data (e.g. Figure 5 Then, the acquired physiological data, medication data, drug concentration monitoring data and test data are input into a server (for example, Figure 5In S) as shown in , the server can process the physiological data, medication data, drug concentration monitoring data and test data based on the automatic optimization algorithm of the individualized medication mathematical model parameters to obtain a prediction result, which is the predicted value of the blood drug concentration of the test patient subject.

[0109] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

[0110] In addition, although each operation is described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or to be performed in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the application. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0111] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely exemplary forms of implementation. With respect to the apparatus in the above-mentioned embodiments, the specific manner in which each module performs the operation has been described in detail in the embodiments related to the method, and will not be described in detail here.

Claims

1. A method for automatically optimizing parameters of a mathematical model for personalized medication, characterized in that: include: Obtaining physiological data, medication data, drug concentration monitoring data, and test data of training patient subjects; Using a population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the training patient subject to obtain a predicted value of blood drug concentration; Based on the error information between the predicted value of the blood drug concentration and the actual measured blood drug concentration value, automatically optimizing the model parameters of the population pharmacokinetic model to obtain a trained population pharmacokinetic model; Using the trained population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the test patient subject to obtain a prediction result, wherein the prediction result is a predicted value of the blood drug concentration of the test patient subject; The method of using a population pharmacokinetic model to process the physiological data, medication data, drug concentration monitoring data and test data of the training patient subjects to obtain a predicted value of blood drug concentration includes: Embedding the physiological data of the training patient object using a physiological data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training physiological data; Using a medication data embedding coding matrix to embed the medication data of the training patient object to obtain a low-dimensional embedding coding vector for the training medication data; Embedding the drug concentration monitoring data of the training patient object using a drug concentration monitoring data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training drug concentration monitoring data; Embedding the test data of the training patient object using a test data embedding coding matrix to obtain a low-dimensional embedding coding vector for the training test data; Using the population pharmacokinetic model, the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector, and the training test data low-dimensional embedded coding vector are pharmacokinetic encoded to obtain a pharmacokinetic time series coding vector; Decoding the pharmacokinetic time series coding vector to obtain a predicted value of the blood drug concentration; The step of inputting the pharmacokinetic time sequence coding vector into a decoder-based blood drug concentration prediction module to obtain a predicted value of blood drug concentration includes: Calculating the sum of the absolute values ​​of each eigenvalue of the pharmacokinetic timing coding vector to obtain a first pharmacokinetic timing coding and modulation value, and calculating the square root of the sum of the squares of each eigenvalue of the pharmacokinetic timing coding vector to obtain a second pharmacokinetic timing coding and modulation value; After performing dot-wise subtraction on the pharmacokinetic timing coding vector and the second pharmacokinetic timing coding and modulation value, respectively perform dot-wise multiplication on the number of eigenvalues ​​of the pharmacokinetic timing coding vector and the reciprocal of the first pharmacokinetic timing coding and modulation value, and take the reciprocal of each eigenvalue to obtain a first pharmacokinetic timing coding phase conversion vector; After point-wise subtraction of the pharmacokinetic timing coding vector and the first pharmacokinetic timing coding and modulation value, point-wise multiplication is performed with the square root of the number of eigenvalues ​​of the pharmacokinetic timing coding vector and the inverse of the second pharmacokinetic timing coding and modulation value, and the inverse of each eigenvalue is taken to obtain the second pharmacokinetic timing coding phase transition vector; point-wise subtraction of the first pharmacokinetic timing coding phase transition vector by the point product vector of the second pharmacokinetic timing coding phase transition vector and the weighted hyperparameter is performed to obtain an optimized pharmacokinetic timing coding vector; the optimized pharmacokinetic timing coding vector is input into a decoder-based blood drug concentration prediction module to obtain a predicted value of blood drug concentration.

2. The method for automatic optimization of parameters of a mathematical model for personalized medication according to claim 1, characterized in that: The physiological data of the training patient subjects include the basic diseases, age, gender, weight, height, body surface area, concomitant medications, genotype, liver function indicators, kidney function indicators and blood routine tests of the training patient subjects; the medication data of the training patient subjects include administration time, dosage, administration route, administration frequency and infusion time; the drug concentration monitoring data of the training patient subjects include monitoring time, blood drug concentration value, detection method and instrument type; the test data include liver and kidney function test data, blood routine test data, and infection index test data.

3. The method for automatic optimization of parameters of a mathematical model for personalized medication according to claim 2, characterized in that: The population pharmacokinetic model is used to perform pharmacokinetic encoding on the low-dimensional embedded coding vector of the training physiological data, the low-dimensional embedded coding vector of the training medication data, the low-dimensional embedded coding vector of the training drug concentration monitoring data and the low-dimensional embedded coding vector of the training test data to obtain a pharmacokinetic time series coding vector, including: using the population pharmacokinetic model to perform pharmacokinetic encoding based on node potential spatiotemporal information transfer on the low-dimensional embedded coding vector of the training physiological data, the low-dimensional embedded coding vector of the training medication data, the low-dimensional embedded coding vector of the training drug concentration monitoring data and the low-dimensional embedded coding vector of the training test data to obtain the pharmacokinetic time series coding vector.

4. The method for automatic optimization of parameters of a mathematical model for personalized medication according to claim 3, characterized in that: Using the population pharmacokinetic model, the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector, and the training test data low-dimensional embedded coding vector are subjected to pharmacokinetic coding based on node potential spatiotemporal information transfer to obtain the pharmacokinetic time series coding vector, including: The training drug concentration monitoring data low-dimensional embedding coding vector is used as the target data low-dimensional embedding coding vector, and the training physiological data low-dimensional embedding coding vector, the training medication data low-dimensional embedding coding vector and the training inspection data low-dimensional embedding coding vector are used as a sequence of message propagation data low-dimensional embedding coding vectors, and the information transmission dynamic energy spatiotemporal transfer weight of each message propagation data low-dimensional embedding coding vector in the sequence of message propagation data low-dimensional embedding coding vectors relative to the target data low-dimensional embedding coding vector is calculated to obtain a sequence of information transmission dynamic energy spatiotemporal transfer weights; Based on the sequence of information transfer dynamic energy spatiotemporal transfer weights, the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training inspection data low-dimensional embedded coding vector are fused, and then the obtained node information fusion representation is passed to the training drug concentration monitoring data low-dimensional embedded coding vector to obtain the pharmacokinetic timing coding vector.

5. The method for automatic optimization of parameters of a mathematical model for personalized medication according to claim 4, characterized in that: Calculating the information transfer dynamic energy spatiotemporal transfer weight of each message propagation data low-dimensional embedding coding vector in the sequence of the message propagation data low-dimensional embedding coding vector relative to the target data low-dimensional embedding coding vector to obtain a sequence of information transfer dynamic energy spatiotemporal transfer weights, including: Calculating a data low-dimensional embedding coding feature static energy value of each message propagation data low-dimensional embedding coding vector in the sequence of message propagation data low-dimensional embedding coding vectors; Determining a data low-dimensional embedding coding feature energy spatial attenuation factor based on a spatial span coefficient between the low-dimensional embedding coding vectors of the respective message propagation data and the low-dimensional embedding coding vectors of the target data; Determine a data low-dimensional embedding coding feature energy time decay factor based on a time span coefficient between the low-dimensional embedding coding vector of each message propagation data and the low-dimensional embedding coding vector of the target data; Based on the data low-dimensional embedding coding feature energy space decay factor and the data low-dimensional embedding coding feature energy time decay factor, the data low-dimensional embedding coding feature static energy value of each message propagation data low-dimensional embedding coding vector is temporally modulated to obtain the data low-dimensional embedding coding feature dynamic energy space-time transfer value of each message propagation data low-dimensional embedding coding vector relative to the target data low-dimensional embedding coding vector; The data low-dimensional embedding coding feature dynamic energy spatiotemporal transfer value of each message propagation data low-dimensional embedding coding vector is input into the energy transfer gating module based on the mask function to obtain the sequence of the information transfer dynamic energy spatiotemporal transfer weights.

6. The method for automatic optimization of parameters of a mathematical model for personalized medication according to claim 5, characterized in that: Calculating the data low-dimensional embedding coding feature static energy value of each message propagation data low-dimensional embedding coding vector in the sequence of message propagation data low-dimensional embedding coding vectors, including: Respectively calculating the mean and standard deviation of the low-dimensional embedding coding vector of the message propagation data to obtain the mean of the low-dimensional embedding coding feature of the message propagation data and the standard deviation of the low-dimensional embedding coding of the message propagation data; After calculating the fourth power of the positional difference between the low-dimensional embedding coding vector of the message propagation data and the mean value of the low-dimensional embedding coding feature of the message propagation data, the expected value of the fourth-power modulated low-dimensional embedding coding offset vector of the message propagation data is calculated to obtain the expected factor of the data low-dimensional embedding coding feature offset; After calculating the division between the expected factor of the data low-dimensional embedding coding feature offset and the fourth power of the standard deviation of the message propagation data low-dimensional embedding coding, it is added to the maximum eigenvalue in the message propagation data low-dimensional embedding coding vector to obtain the data low-dimensional embedding coding feature static energy value.

7. The method for automatic optimization of parameters of a mathematical model for personalized medication according to claim 6, characterized in that: Based on the sequence of the information transfer dynamic energy spatiotemporal transfer weights, the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector and the training inspection data low-dimensional embedded coding vector are fused, and then the obtained node information fusion representation is transmitted to the training drug concentration monitoring data low-dimensional embedded coding vector to obtain the pharmacokinetic time series coding vector, including: Based on the sequence of the information transfer dynamic energy spatiotemporal transfer weights, a weighted sum of the training physiological data low-dimensional embedding coding vector, the training medication data low-dimensional embedding coding vector and the training inspection data low-dimensional embedding coding vector is calculated to obtain a message propagation data low-dimensional embedding coding node feature transfer aggregation modulation vector as the node information fusion representation; The position-wise sum of the message propagation data low-dimensional embedded coding node feature transfer aggregation modulation vector and the training drug concentration monitoring data low-dimensional embedded coding vector is calculated to obtain the pharmacokinetic time series coding vector.

8. An automatic optimization system for parameters of a mathematical model for personalized medication, characterized in that: include: A data acquisition module, used to acquire physiological data, medication data, drug concentration monitoring data and test data of training patient subjects; A data processing module, for processing the physiological data, medication data, drug concentration monitoring data and test data of the training patient subject using a population pharmacokinetic model to obtain a predicted value of blood drug concentration; A model parameter training module, for automatically optimizing the model parameters of the population pharmacokinetic model based on the error information between the predicted value of the blood drug concentration and the actual measured blood drug concentration value to obtain a trained population pharmacokinetic model; A prediction result generating module, used to process the physiological data, medication data, drug concentration monitoring data and test data of the test patient subject using the trained population pharmacokinetic model to obtain a prediction result, wherein the prediction result is a predicted value of the blood drug concentration of the test patient subject; Wherein, the data processing module includes: A physiological data embedding coding unit, configured to embed the physiological data of the training patient object using a physiological data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training physiological data; A medication data embedding coding unit, configured to embed the medication data of the training patient object using a medication data embedding coding matrix to obtain a low-dimensional embedding coding vector for the training medication data; A drug concentration monitoring data embedding coding unit, used to embed the drug concentration monitoring data of the training patient object using a drug concentration monitoring data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training drug concentration monitoring data; A test data embedding coding unit, used for embedding coding the test data of the training patient object using a test data embedding coding matrix to obtain a low-dimensional embedding coding vector of the training test data; a pharmacokinetic encoding unit, configured to use the population pharmacokinetic model to perform pharmacokinetic encoding on the training physiological data low-dimensional embedded coding vector, the training medication data low-dimensional embedded coding vector, the training drug concentration monitoring data low-dimensional embedded coding vector, and the training test data low-dimensional embedded coding vector to obtain a pharmacokinetic time series coding vector; A vector decoding unit, used for decoding the pharmacokinetic time-series coding vector to obtain a predicted value of the blood drug concentration; The step of inputting the pharmacokinetic time sequence coding vector into a decoder-based blood drug concentration prediction module to obtain a predicted value of blood drug concentration includes: Calculating the sum of the absolute values ​​of each eigenvalue of the pharmacokinetic timing coding vector to obtain a first pharmacokinetic timing coding and modulation value, and calculating the square root of the sum of the squares of each eigenvalue of the pharmacokinetic timing coding vector to obtain a second pharmacokinetic timing coding and modulation value; After performing dot-wise subtraction on the pharmacokinetic timing coding vector and the second pharmacokinetic timing coding and modulation value, respectively perform dot-wise multiplication on the number of eigenvalues ​​of the pharmacokinetic timing coding vector and the reciprocal of the first pharmacokinetic timing coding and modulation value, and take the reciprocal of each eigenvalue to obtain a first pharmacokinetic timing coding phase conversion vector; After point-wise subtraction of the pharmacokinetic timing coding vector and the first pharmacokinetic timing coding and modulation value, point-wise multiplication is performed with the square root of the number of eigenvalues ​​of the pharmacokinetic timing coding vector and the inverse of the second pharmacokinetic timing coding and modulation value, and the inverse of each eigenvalue is taken to obtain the second pharmacokinetic timing coding phase transition vector; point-wise subtraction of the first pharmacokinetic timing coding phase transition vector by the point product vector of the second pharmacokinetic timing coding phase transition vector and the weighted hyperparameter is performed to obtain an optimized pharmacokinetic timing coding vector; the optimized pharmacokinetic timing coding vector is input into a decoder-based blood drug concentration prediction module to obtain a predicted value of blood drug concentration.

Images