Pharmacokinetic model construction method and device, equipment and storage medium

Abstract

The invention discloses a pharmacokinetic model construction method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining multi-modal physiological parameters of a plurality of patients; wherein the multi-modal physiological parameters comprise dynamic reaction data and physical sign change data of each patient under different drug metabolism; constructing a multi-modal neural network model based on the deep neural network model; training a multi-modal neural network model according to the multi-modal physiological parameters to obtain a pre-training model; and training the pre-training model according to the multi-modal physiological parameters of the target patient in the target field to obtain a pharmacokinetic model. According to the method, the pre-training model is trained by using the multi-modal physiological parameters of the target patient with a small sample size, the ability of the pre-training model can be migrated to different target fields, the pharmacokinetic model is obtained, and the generalization ability is improved; meanwhile, the sample requirement for a target patient with a small sample size is reduced, and the cost for constructing the pharmacokinetic model is reduced.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and storage medium for constructing a pharmacokinetic model. Background Art

[0002] Pharmacokinetics (PK) studies the absorption, distribution, metabolism and excretion of drugs in the human body. Its core goal is to quantitatively describe the change of drug concentration over time through mathematical models, and provide a scientific basis for personalized medication. With the rapid development of precision medicine and smart medicine, the optimization of personalized drug dosage, drug efficacy monitoring and prediction of adverse drug reactions have become important research directions in pharmacokinetics. However, the pharmacokinetic models of existing technologies still have problems such as high construction cost and poor generalization ability. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose a method, device, equipment and storage medium for constructing a pharmacokinetic model to reduce the cost of constructing a pharmacokinetic model and improve the generalization ability of the pharmacokinetic model.

[0004] To achieve the above purpose, one aspect of the embodiments of the present application proposes a method for constructing a pharmacokinetic model, the method comprising the following steps:

[0005] Acquiring multimodal physiological parameters of multiple patients; wherein the multimodal physiological parameters include dynamic response data and vital sign change data of each of the patients under different drug metabolism;

[0006] Build a multimodal neural network model based on the deep neural network model;

[0007] Training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model;

[0008] The pre-trained model is trained according to the multimodal physiological parameters of the target patients in the target field to obtain a pharmacokinetic model.

[0009] In some embodiments, the step of obtaining multimodal physiological parameters of multiple patients comprises the following steps:

[0010] The multimodal physiological parameters of the plurality of patients are acquired through a multi-sensor acquisition system.

[0011] In some embodiments, constructing a multimodal neural network model based on a deep neural network model comprises the following steps:

[0012] Selecting a convolutional neural network or a recurrent neural network as the deep neural network model;

[0013] The multimodal neural network model is constructed based on the convolutional neural network or the recurrent neural network.

[0014] In some embodiments, the step of training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model comprises the following steps:

[0015] The multimodal neural network model is trained on a classification task, a regression task or a multi-task joint training according to the multimodal physiological parameters, so that the multimodal neural network model learns the characteristics between different drug metabolisms and the multimodal physiological parameters, thereby obtaining the pre-trained model.

[0016] In some embodiments, training the pre-trained model according to the multimodal physiological parameters of the target patient in the target field to obtain a pharmacokinetic model comprises the following steps:

[0017] Part of the model parameters of the pre-trained model are frozen, and then the pre-trained model is trained according to the multimodal physiological parameters of the target patient in the target field to obtain the pharmacokinetic model.

[0018] In some embodiments, the method further comprises the following steps:

[0019] Acquiring dynamic physiological feedback parameters of the target patient;

[0020] The model parameters of the pharmacokinetic model are adjusted according to the dynamic physiological feedback parameters to reduce the deviation between the expected pharmacodynamic value predicted by the pharmacokinetic model and the actual pharmacodynamic value.

[0021] In some embodiments, the method further comprises the following steps:

[0022] Obtaining the deviation value between the expected pharmacokinetic value predicted by the pharmacokinetic model and the actual pharmacokinetic value;

[0023] generating an error feedback signal according to the deviation value;

[0024] Dynamically adjust the value of the drug injection rate or the value of the drug dosage concentration.

[0025] To achieve the above purpose, another aspect of the embodiment of the present application provides a pharmacokinetic model construction device, the device comprising:

[0026] A parameter acquisition unit, used to acquire multimodal physiological parameters of multiple patients; wherein the multimodal physiological parameters include dynamic response data and physical sign change data of each of the patients under different drug metabolism;

[0027] A model building unit, used for building a multimodal neural network model based on the deep neural network model;

[0028] A model pre-training unit, used for training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model;

[0029] The model training unit is used to train the pre-trained model according to the multimodal physiological parameters of the target patients in the target field to obtain a pharmacokinetic model.

[0030] To achieve the above objective, another aspect of an embodiment of the present application provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the above method when executing the computer program.

[0031] To achieve the above objective, another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0032] The embodiments of the present application include at least the following beneficial effects:

[0033] The present application can obtain multimodal physiological parameters of multiple patients; wherein, the multimodal physiological parameters include dynamic response data and physical sign change data of each patient under different drug metabolism; construct a multimodal neural network model based on a deep neural network model; train the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model; train the pre-trained model according to the multimodal physiological parameters of target patients in the target field to obtain a pharmacokinetic model. The present application first obtains a pre-trained model by training a multimodal neural network model, and then trains the pre-trained model using the multimodal physiological parameters of target patients with a smaller sample size, which can transfer the capabilities of the pre-trained model to different target fields, obtain a pharmacokinetic model, and improve the generalization ability; at the same time, it reduces the sample requirements for target patients with a smaller sample size, and reduces the cost of building a pharmacokinetic model. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 A schematic diagram of a method for constructing a pharmacokinetic model provided in an embodiment of the present application;

[0036] Figure 2An exemplary flow chart of a method for constructing a pharmacokinetic model provided in an embodiment of the present application;

[0037] Figure 3 An example flow chart of a pre-trained multimodal neural network model provided in an embodiment of the present application;

[0038] Figure 4 A schematic diagram of the structure of a pharmacokinetic model building device provided in an embodiment of the present application;

[0039] Figure 5 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0041] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0042] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0044] Before describing the embodiments of the present application in detail, some related technologies involved in the embodiments of the present application are first described as follows:

[0045] At present, the research methods of pharmacokinetics mainly include the following categories:

[0046] 1. Traditional statistical model.

[0047] Common pharmacokinetic models, such as one-compartment model, multi-compartment model and physiological pharmacokinetic model, are based on classical differential equations and biological laws, and fit the kinetic parameters of drugs in vivo (such as absorption rate, distribution volume, clearance rate, etc.) through experimental data. These models are simple and intuitive, and easy to explain, but they also have the following problems:

[0048] (1) Insufficient adaptability to individual differences: Traditional models assume that parameters between groups are consistent, which makes it difficult to accurately reflect the physiological differences between individuals.

[0049] (2) Large data requirements: Model fitting requires a large amount of high-quality experimental data, especially in multi-drug combination therapy or special populations (such as children, the elderly, and those with abnormal liver and kidney function), where data are difficult to obtain.

[0050] (3) Difficulty in processing complex data: For nonlinear and multivariate drug metabolism processes, traditional models have limitations in prediction accuracy and generalization ability.

[0051] 2. Pharmacokinetic model based on artificial intelligence.

[0052] With the rapid development of deep learning and artificial intelligence, data-driven pharmacokinetic modeling methods have gradually emerged. This type of method learns features from large-scale data through models such as neural networks, overcomes the traditional model's reliance on prior assumptions, and demonstrates higher predictive capabilities. Commonly used techniques include:

[0053] (1) Machine learning methods: such as support vector machine (SVM) and random forest (RF), which are used to predict drug metabolism-related parameters.

[0054] (2) Deep learning methods: Methods such as convolutional neural networks (CNN) and recurrent neural networks (RNN) have shown high accuracy in drug concentration prediction and personalized dose optimization.

[0055] However, the above related technologies also have the following problems:

[0056] (1) High dependence on data volume: The training of deep learning models requires a large amount of labeled data, while experimental data in the field of pharmacokinetics are often scarce and expensive.

[0057] (2) Insufficient transferability: Since drug metabolism characteristics vary greatly among different populations, the model often needs to be trained separately for each type of patient, which limits its application in actual scenarios.

[0058] Therefore, the embodiments of the present application provide a method, device, equipment and storage medium for constructing a pharmacokinetic model. The technical solution of the present application includes: obtaining multimodal physiological parameters of multiple patients; wherein the multimodal physiological parameters include dynamic response data and physical sign change data of each patient under different drug metabolism; constructing a multimodal neural network model based on a deep neural network model; training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model; training the pre-trained model according to the multimodal physiological parameters of target patients in the target field to obtain a pharmacokinetic model. The present application first obtains a pre-trained model by training a multimodal neural network model, and then trains the pre-trained model using the multimodal physiological parameters of target patients with a small sample size, so as to transfer the capabilities of the pre-trained model to different target fields, obtain a pharmacokinetic model, and improve the generalization ability; at the same time, it reduces the sample requirements for target patients with a small sample size, and reduces the cost of constructing a pharmacokinetic model.

[0059] The embodiments of the present application provide a method, device, equipment and storage medium for constructing a pharmacokinetic model, and relate to the field of artificial intelligence technology. The method, device, equipment and storage medium for constructing a pharmacokinetic model provided in the embodiments of the present application can be applied to a terminal, a server, or a software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and a car terminal, etc., but are not limited to this; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application that implements a knowledge extraction method, etc., but is not limited to the above forms.

[0060] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0061] Reference Figure 1 , the present application embodiment provides a method for constructing a pharmacokinetic model, which may include but is not limited to S100 to S130, as follows:

[0062] S100: Acquire multimodal physiological parameters of a plurality of patients; wherein the multimodal physiological parameters include dynamic response data and vital sign change data of each of the patients under different drug metabolism.

[0063] Further, S100 may include S101:

[0064] S101: Acquire the multimodal physiological parameters of the plurality of patients through a multi-sensor acquisition system.

[0065] Specifically, the multimodal physiological parameters of multiple patients are acquired in real time through a multi-sensor acquisition system. The multimodal physiological parameters may include: EEG index (reflecting the state of the central nervous system), heart rate and blood pressure (monitoring cardiovascular status), end-tidal carbon dioxide partial pressure and oxygen saturation (reflecting respiratory system function), and drug concentration in blood samples (used to quantify drug metabolism). The multimodal physiological parameters cover the dynamic response data and physical sign change data of patients under different drug metabolism, forming a multi-dimensional data set. The collected multimodal physiological parameters undergo routine cleaning, including removing outliers, filling missing data, and standardization to ensure data consistency and reliability. The high-quality multimodal physiological parameters after cleaning will serve as the basis for model training and verification.

[0066] S110: Construct a multimodal neural network model based on the deep neural network model.

[0067] Furthermore, S110 may include S111-S112:

[0068] S111: Select a convolutional neural network or a recurrent neural network as the deep neural network model;

[0069] S112: Constructing the multimodal neural network model based on the convolutional neural network or the recurrent neural network.

[0070] Specifically, this embodiment can select an appropriate deep neural network model (such as a convolutional neural network CNN or a recurrent neural network RNN) and build a multimodal neural network model.

[0071] S120: Training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model.

[0072] Further, S120 may include S121:

[0073] S121: Performing classification tasks, regression tasks or multi-task joint training on the multimodal neural network model according to the multimodal physiological parameters, so that the multimodal neural network model learns the characteristics between different drug metabolism and the multimodal physiological parameters, and then obtains the pre-trained model.

[0074] Specifically, the multimodal neural network model is pre-trained on large-scale source domain data (such as different patients, drugs, and treatment scenarios) to learn the metabolic characteristics and efficacy patterns of drugs under different signs. In the pre-training stage, classification tasks, regression tasks, or multi-task joint training are used to extract the features between drug metabolism and multimodal physiological parameters and build a pre-trained model with strong generalization ability. Pre-training enables the multimodal neural network model to form a set of general feature extractors, provide a feature representation basis for the target task, and reduce the demand for target domain data.

[0075] S130: Training the pre-trained model according to the multimodal physiological parameters of the target patients in the target field to obtain a pharmacokinetic model.

[0076] Further, S130 may include S131:

[0077] S131: Freeze some model parameters of the pre-trained model, and then train the pre-trained model according to the multimodal physiological parameters of the target patient in the target field to obtain the pharmacokinetic model.

[0078] Specifically, the pre-trained model is transferred to the target domain, combined with the individual characteristics of the target patients and a small amount of labeled data, and rapid adaptation to the target task is achieved through transfer learning strategies. For small sample patients, a lightweight fine-tuning method is used, such as freezing the parameters of some pre-trained models and only fine-tuning the last few layers to obtain a pharmacokinetic model. Furthermore, the pharmacokinetic model can not only make full use of the knowledge in the source domain, but also adjust the drug dosage prediction scheme according to the data of individual patients, and realize accurate individualized medication parameter adjustment.

[0079] To further improve the accuracy of the pharmacokinetic model, the embodiment of the present application may further include steps S141 to S142:

[0080] S141: Acquire dynamic physiological feedback parameters of the target patient;

[0081] S142: Adjusting the model parameters of the pharmacokinetic model according to the dynamic physiological feedback parameters to reduce the deviation between the expected pharmacodynamic value predicted by the pharmacokinetic model and the actual pharmacodynamic value.

[0082] Specifically, this embodiment can be based on the dynamic physiological feedback parameters of the target patient (such as EEG index, blood pressure, heart rate, etc.), through an adaptive adjustment mechanism dynamic pharmacokinetic model. The pharmacokinetic model is used to compare the deviation between the expected pharmacodynamic value (such as drug concentration or physiological index target value) and the actual pharmacokinetic value in real time.

[0083] Furthermore, the embodiment of the present application may further include S151 to S153:

[0084] S151: Obtaining the deviation value between the expected pharmacokinetic value predicted by the pharmacokinetic model and the actual pharmacokinetic value;

[0085] S152: generating an error feedback signal according to the deviation value;

[0086] S153: Dynamically adjust the value of the drug injection rate or the value of the drug dosage concentration.

[0087] Specifically, this embodiment can generate an error feedback signal according to the deviation between the expected pharmacokinetic value predicted by the pharmacokinetic model and the actual pharmacokinetic value, and then use the error feedback signal to dynamically adjust the value of the drug injection rate or the value of the drug dosage concentration. By using an efficient online learning algorithm, the pharmacokinetic model can quickly update the medication parameters (such as the value of the drug injection rate or the value of the drug dosage concentration) when the state of the target patient changes, respond to the fluctuation of the real-time monitoring data, so that the feedback control system always maintains efficient and stable operation, and achieves the goal of individualized and real-time drug dosage optimization.

[0088] The embodiment of the present application is based on a pre-trained model of deep learning, and uses transfer learning to solve the problem of poor generalization ability of the pharmacokinetic model in the case of small samples, which can effectively deal with individual differences in patients and provide a more accurate drug efficacy monitoring and dosage value determination scheme. The scheme of the embodiment of the present application has adaptive learning and real-time adjustment functions, and can dynamically adjust the medication parameters in scenarios such as anesthesia and intensive care.

[0089] Next, the solution of the embodiment of the present application will be introduced and explained in detail with reference to specific application examples.

[0090] This embodiment relates to the field of smart medicine, especially the drug efficacy evaluation method of multi-source signals, drug efficacy prediction and optimization of personalized pharmacokinetic models, and pre-training methods based on deep learning. This embodiment can be applied to the field of personalized medical parameter determination, pharmacokinetic analysis, and physiological data monitoring and evaluation.

[0091] This embodiment aims to overcome the shortcomings of existing pharmacokinetic modeling technology and solve technical problems such as insufficient personalized support, weak modeling capabilities for multi-drug combination therapy, poor dynamic response capabilities, scarcity of small sample data, and lack of interpretability of the model. This embodiment introduces pre-training models, transfer learning, multi-task learning, and real-time feedback control mechanisms to achieve accurate capture of individualized physiological characteristics of patients and dynamic adjustment of drug dosages, effectively support multi-drug combination therapy scenarios, improve the generalization ability of pharmacokinetic models under small sample data, and provide strong support for personalized medication parameter determination and real-time dosage parameter optimization.

[0092] This embodiment provides a method for constructing a pharmacokinetic model based on a pre-trained deep learning model. The pharmacokinetic model obtains a wide range of pharmacodynamic metabolic characteristics by learning from source domain data and migrating them to individualized small sample data to achieve pharmacodynamic evaluation and dynamic dosage parameter adjustment when multiple drugs are used in combination.

[0093] Reference Figure 2 , this embodiment includes the following technical solutions:

[0094] 1. Data collection.

[0095] The patient's name, gender, age, height, weight and medical history were recorded to determine the drugs needed by the patient, such as a variety of intravenous drugs (such as propofol in anesthetic drugs, remifentanil in analgesics, and atracurium in muscle relaxants) as the subjects of the combined drug test. Based on the pharmacological effects, EEG index, blood pressure, and heart rate were selected as the main clinical efficacy indicators. The central nervous system was monitored by EEG detection instruments, and the results were synchronized and digitized; the respiratory system was monitored by end-tidal carbon dioxide partial pressure and oxygen saturation; the cardiovascular system was monitored by invasive or non-invasive blood pressure and electrocardiogram. In clinical practice, patients were given medication, and blood was drawn from patients at different time points to detect and record drug concentrations. Data collection was completed through the above methods, and routine cleaning, including elimination of outliers, filling in missing data, and standardization, was performed to improve data consistency and reliability, providing data support for the construction of pharmacodynamic models.

[0096] 2. Construction of pre-trained model.

[0097] The data sources in the pharmacokinetic model are mainly divided into patient clinical static data, physiological indicator signals and medical image data. In order to effectively combine different types of data, this embodiment designs a multimodal architecture to uniformly extract and fuse the features of different data sources, such as Figure 3 shown.

[0098] The data types of clinical static data are basic patient information (such as age, gender, height, weight, medical history, etc.) and drug types (drug name, dosage, injection rate). The applicable architecture is a fully connected network (FCN). Therefore, this embodiment uses a multilayer perceptron (MLP) to extract a high-dimensional representation of static data. The input form is to encode each static feature, for example, categorical variables (such as gender, drug name) are represented by one-hot encoding (One-Hot Encoding) or embedding layer (Embedding Layer), and numerical variables are normalized (Normalization).

[0099] The data types of physiological indicator signal data are physiological indicators (such as EEG index, heart rate, blood pressure, oxygen saturation, end-tidal carbon dioxide partial pressure) and drug concentration changes over time. Applicable architectures are recurrent neural networks (RNN), long short-term memory networks (LSTM), and gated recurrent units (GRU). Therefore, this embodiment uses multi-layer LSTM to capture short-term and long-term dependencies of time series data. The input form is to expand the time series data by time steps, and the features of each time step are used as input vectors.

[0100] The data type of medical image data is a time-related signal generated by a device such as an electroencephalogram or electrocardiogram. The applicable architecture is a convolutional neural network (CNN). Therefore, this embodiment uses the classic CNN architecture ResNet to extract the spatial features of image data.

[0101] Since different features are highly correlated, this embodiment uses a cross-attention mechanism to weight the fusion features to capture the most important information.

[0102] Based on the multimodal features, this embodiment designs downstream task branches, including classification tasks and regression tasks.

[0103] The task goal of the classification task branch is to predict whether the patient will have a specific physiological reaction (such as respiratory depression, adverse reactions, etc.). The structure is designed to build a classifier based on a shared feature vector using a fully connected network. The last layer uses a Softmax or Sigmoid activation function to output the category probability. The loss function is the cross-entropy loss (Cross-EntropyLoss).

[0104] The task goal of the regression task branch is to predict the changes in future physiological indicators and drug concentrations. The structure is designed to use a fully connected network for continuous value prediction based on a shared feature vector, and the last layer uses a linear activation function to output physiological indicators and drug concentration values. The loss function is the mean square error (MSE).

[0105] This example uses joint training: using a multi-task learning framework, sharing underlying features, and optimizing classification and regression tasks separately. Define the joint loss function:

[0106] L total =α·L classification +β·L regression ;

[0107] Among them, α and β are the weight parameters of the pharmacokinetic model, which are used to balance the classification and regression tasks.

[0108] 3. Model migration and individualized adjustment.

[0109] The pre-trained model has learned general drug metabolism characteristics on large-scale data, but the physiological state and drug response characteristics of each patient are different, and the model needs to be adapted to individualized scenarios through transfer learning. Due to the small amount of individual patient data, this embodiment uses a lightweight fine-tuning mechanism to freeze most of the layers of the pre-trained model and only fine-tune the last few layers (such as feature fusion layers or classification / regression branches). The data of individual patients (such as basic features, real-time physiological monitoring data) are used as input to fine-tune the pre-trained model. The model weights are adjusted using small-scale individualized data through an online learning mechanism to improve the adaptability to specific patients. By continuously receiving the patient's real-time monitoring data and dynamically adjusting the model parameters, the model can quickly adapt to changes in the patient's status. Real-time monitoring data of patients (such as EEG index, heart rate, blood pressure, oxygen saturation, etc.) is obtained through multiple sensors and input into the model in an appropriate form. Use real-time data to update the weights of the model in small steps without retraining the entire pre-trained model:

[0110]

[0111] Among them, θ t are the parameters of the model at time t, x t ,y t are the input features and target values ​​at the current time t, η is the learning rate, and L is the loss function (such as mean squared error MSE or cross entropy loss).

[0112] 4. Real-time feedback and model optimization.

[0113] Real-time feedback compares the actual pharmacodynamic value (such as drug concentration or EEG index) with the predicted pharmacokinetic value predicted by the pharmacokinetic model, calculates the deviation value and dynamically adjusts the output or weight of the pharmacokinetic model.

[0114] Input the real monitoring data of the patient's current state (such as drug concentration or EEG index), compare the actual efficacy value with the predicted efficacy value, and calculate the deviation value:

[0115] e(t)=y real -y pred ;

[0116] where y real is the actual efficacy value, y pred is the predicted efficacy value predicted by the pharmacokinetic model, and e(t) is the deviation value at time t.

[0117] The deviation value is used for training the pharmacokinetic model and optimizing the weight parameters:

[0118]

[0119] Among them, θ tare the parameters of the pharmacokinetic model at time t, e(t) is the deviation value at time t, η is the learning rate, and L is the loss function (such as mean squared error MSE or cross entropy loss).

[0120] The deviation value is mapped to the adjustment amount of the input drug dose (or other input variables), so that the patient's monitoring indicators gradually approach the target value. Comprehensive proportional, integral and differential control (PID controller) balances response speed, steady-state error and dynamic adjustment:

[0121]

[0122] Where K p Determine the sensitivity of the input adjustment to the current error, K i Accumulate historical errors to correct steady-state deviations, K d By predicting the error change trend, we can reduce overshoot or undershoot. e(t) is the deviation value at time t. is the input adjustment amount.

[0123] First, the pharmacokinetic model is updated in real time according to the latest input data, that is, the parameters are adjusted through the online learning mechanism to adapt the pharmacokinetic model to the new data. After the prediction output is generated, the error feedback signal is used to optimize the model parameters of the pharmacokinetic model, and the input is further adjusted based on the error feedback to correct the future output of the pharmacokinetic model.

[0124] In summary, the technical solution of this embodiment includes:

[0125] 1. Build a pharmacokinetic model based on the pre-trained model.

[0126] By designing a pharmacokinetic model based on a multimodal pre-trained model, pre-training it on a large scale on data from different patients, and fine-tuning it for specific patients, we can achieve rapid migration and individualized adjustment for target patients in the target area.

[0127] 2. Fast transfer learning and real-time PID feedback.

[0128] By designing fast transfer learning, the weights of the pharmacokinetic model are adjusted using small-scale individualized data through an online learning mechanism, so that the pharmacokinetic model can quickly adapt to changes in patient status. In addition, by combining real-time PID feedback, the pharmacokinetic model can accurately predict changes in patient status.

[0129] Reference Figure 4 The present application also provides a pharmacokinetic model construction device, which can implement the above-mentioned pharmacokinetic model construction method, and the device includes:

[0130] A parameter acquisition unit, used to acquire multimodal physiological parameters of multiple patients; wherein the multimodal physiological parameters include dynamic response data and physical sign change data of each of the patients under different drug metabolism;

[0131] A model building unit, used for building a multimodal neural network model based on the deep neural network model;

[0132] A model pre-training unit, used for training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model;

[0133] The model training unit is used to train the pre-trained model according to the multimodal physiological parameters of the target patients in the target field to obtain a pharmacokinetic model.

[0134] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0135] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned method for constructing a pharmacokinetic model when executing the computer program. The electronic device can be any smart terminal including a tablet computer, a car computer, etc.

[0136] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0137] See also Figure 5 , Figure 5 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0138] The processor 501 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0139] The memory 502 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 502 can store an operating system and other application programs. When the technical solution provided in the embodiments of this specification is implemented by software or firmware, the relevant program code is stored in the memory 502, and the processor 501 calls and executes a pharmacokinetic model construction method of the embodiment of the present application;

[0140] Input / output interface 503, used to implement information input and output;

[0141] Communication interface 504, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0142] A bus 505 that transmits information between the various components of the device (e.g., the processor 501, the memory 502, the input / output interface 503, and the communication interface 504);

[0143] The processor 501 , the memory 502 , the input / output interface 503 and the communication interface 504 are connected to each other in communication within the device via the bus 505 .

[0144] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned pharmacokinetic model construction method is implemented.

[0145] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0146] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0147] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0148] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0149] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0150] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0151] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0152] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0153] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0154] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0155] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0156] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0157] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for constructing a pharmacokinetic model, characterized in that: The method comprises the following steps: Acquiring multimodal physiological parameters of multiple patients; wherein the multimodal physiological parameters include dynamic response data and vital sign change data of each of the patients under different drug metabolism; Build a multimodal neural network model based on the deep neural network model; Training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model; The pre-trained model is trained according to the multimodal physiological parameters of the target patients in the target field to obtain a pharmacokinetic model.

2. A method for constructing a pharmacokinetic model according to claim 1, characterized in that: The method of obtaining multimodal physiological parameters of multiple patients comprises the following steps: The multimodal physiological parameters of the plurality of patients are acquired through a multi-sensor acquisition system.

3. A method for constructing a pharmacokinetic model according to claim 1, characterized in that: The multimodal neural network model is constructed based on the deep neural network model, comprising the following steps: Selecting a convolutional neural network or a recurrent neural network as the deep neural network model; The multimodal neural network model is constructed based on the convolutional neural network or the recurrent neural network.

4. A method for constructing a pharmacokinetic model according to claim 1, characterized in that: The step of training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model comprises the following steps: The multimodal neural network model is trained on a classification task, a regression task or a multi-task joint training according to the multimodal physiological parameters, so that the multimodal neural network model learns the characteristics between different drug metabolisms and the multimodal physiological parameters, thereby obtaining the pre-trained model.

5. A method for constructing a pharmacokinetic model according to claim 1, characterized in that: The step of training the pre-trained model according to the multimodal physiological parameters of the target patient in the target field to obtain a pharmacokinetic model comprises the following steps: Part of the model parameters of the pre-trained model are frozen, and then the pre-trained model is trained according to the multimodal physiological parameters of the target patient in the target field to obtain the pharmacokinetic model.

6. A method for constructing a pharmacokinetic model according to claim 1, characterized in that: The method further comprises the following steps: Acquiring dynamic physiological feedback parameters of the target patient; The model parameters of the pharmacokinetic model are adjusted according to the dynamic physiological feedback parameters to reduce the deviation between the expected pharmacodynamic value predicted by the pharmacokinetic model and the actual pharmacodynamic value.

7. A method for constructing a pharmacokinetic model according to claim 6, characterized in that: The method further comprises the following steps: Obtaining the deviation value between the expected pharmacokinetic value predicted by the pharmacokinetic model and the actual pharmacokinetic value; generating an error feedback signal according to the deviation value; Dynamically adjust the value of the drug injection rate or the value of the drug dosage concentration.

8. A pharmacokinetic model building device, characterized in that: The device comprises: A parameter acquisition unit, used to acquire multimodal physiological parameters of multiple patients; wherein the multimodal physiological parameters include dynamic response data and physical sign change data of each of the patients under different drug metabolism; A model building unit, used for building a multimodal neural network model based on the deep neural network model; A model pre-training unit, used for training the multimodal neural network model according to the multimodal physiological parameters to obtain a pre-trained model; The model training unit is used to train the pre-trained model according to the multimodal physiological parameters of the target patients in the target field to obtain a pharmacokinetic model.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images