A sensor-based method and system for monitoring pharmacokinetics

Abstract

The application discloses a kind of based on sensor's drug metabolism dynamic monitoring method and system.The method is by deploying multimodal sensor node to collect the biochemical response data, physiological state data and environmental behavior associated data in the drug metabolism process of target monitoring object, constructs drug metabolism state feature model based on multidimensional state space.Further, the application establishes state transition relationship model according to historical metabolism data, obtains state transition parameter by analyzing the state evolution law between continuous time nodes, and dynamically adjusts the fusion weight of different data sources in combination with sensor data credibility evaluation mechanism, constructs dynamic metabolism combined feature vector, and adaptively optimizes model parameter.The application improves the accuracy and stability of drug metabolism state analysis under complex individualized environment, and provides data support for precise medication and drug safety management.

Description

Technical Field

[0001] This invention relates to the field of drug metabolism monitoring, and in particular to a sensor-based method and system for dynamic monitoring of drug metabolism. Background Technology

[0002] Biosensor-based drug metabolism monitoring technology is increasingly being applied in the field of clinical drug management. By acquiring real-time data on changes in drug concentration, physiological responses, and related metabolic parameters in the human body, it provides data support for drug dosage adjustment and efficacy evaluation. For example, CN119517450A discloses an integrated dynamic drug metabolism monitoring and early warning system. This technical solution uses optical sensors, electrophysiological sensors, and chemical sensors to acquire biological signals and achieve dynamic monitoring of drug metabolism.

[0003] SD rats are widely used in the analysis of drug absorption, distribution, metabolism, and excretion processes due to their stable genetic background, suitable body size, well-defined metabolic characteristics, and high experimental reproducibility. When conducting pharmacokinetic experiments based on candidate drugs, it is often necessary to determine the dosage based on factors such as the drug's physicochemical properties and expected therapeutic concentrations, analyze drug concentration change curves, and then obtain key pharmacokinetic parameters. In existing technologies, SD rat pharmacokinetic analysis methods typically rely on timed sampling and offline detection. This not only requires multiple animal sample collections, increasing the complexity of experimental procedures, but also, due to the limited sampling time intervals, makes it difficult to continuously reflect the dynamic changes in drug absorption, metabolic transformation, and clearance. For example, the technical solution disclosed in CN119517450A offers some convenience compared to traditional laboratory detection methods. However, this method mainly analyzes drug concentrations at a single time point, lacking the ability to model continuous dynamic metabolic processes and failing to effectively characterize the state changes during drug absorption, transformation, and clearance.

[0004] Furthermore, some existing drug metabolism monitoring systems combine physiological parameters such as heart rate, blood oxygen, and body temperature to assist in judging drug efficacy. These technologies typically use fixed thresholds or simple data correlation methods for analysis, failing to consider the temporal correlations between different sensor data and individual metabolic differences. In the scenario of continuous pharmacokinetic monitoring in SD rats, traditional methods require blood sampling at preset time points, making it impossible to obtain real-time trends in drug concentration changes within the animal. During actual monitoring, differences in sampling frequency, data type, and signal stability among different sensors can easily lead to data drift, abnormal fluctuations, and information loss, resulting in biased analysis results. Therefore, how to utilize sensors to acquire multi-source dynamic data, effectively fuse and process the correlations between different data types, and establish a data model that reflects individualized changes in drug metabolism status has become a pressing technical problem in the field of intelligent drug monitoring. Summary of the Invention

[0005] To overcome the problems of single data acquisition dimensions, insufficient dynamic change characterization, and weak adaptability to individual differences in existing drug metabolism monitoring technologies, this invention provides a sensor-based dynamic drug metabolism monitoring method. This method collects biochemical response data, physiological state data, and environmental behavior correlation data during drug metabolism through multimodal sensor nodes. It performs time synchronization and feature encoding on different types of sensor data to construct a drug metabolism state feature space. Based on the migration relationship of drug metabolism states, it performs credibility assessment and fusion processing on multi-source data to establish a dynamic metabolism joint feature vector. A trend prediction model is then used to dynamically analyze the trends of drug concentration changes, metabolic rate changes, and abnormal metabolic risks. Furthermore, this invention also provides a monitoring system for implementing the sensor-based dynamic drug metabolism monitoring method.

[0006] To achieve the above objectives, the present invention provides the following technical solution, comprising the following steps: Step 1: Collect continuous dynamic monitoring data of drug metabolism process by deploying multimodal sensor nodes on the target monitoring object, classify and process the data according to the data type to obtain the first biochemical response data, the second physiological state data and the third environmental behavior correlation data; Step 2: Perform time synchronization processing and feature encoding processing on the first biochemical response data, the second physiological state data, and the third environmental behavior correlation data respectively to obtain the corresponding first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C ; Step 3: Based on the first metabolic feature sequence X A Second physiological characteristic sequence X B and the third associated feature sequence X C A multidimensional drug metabolism state feature space is constructed, and sensor data corresponding to different time points are mapped to a unified state representation space to obtain the current drug metabolism state feature X. S ; Step 4: Construct a state transition relationship model based on historically collected dynamic drug metabolism data, and analyze the current drug metabolism state characteristics X. S Perform time-series correlation analysis to obtain the drug metabolism state transition parameter M; Step 5: Based on the state transition parameter M, process the first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X CA credibility assessment is performed, and the data weight coefficients corresponding to different sensor data sources are calculated respectively. Based on the weight coefficients, multi-source data fusion processing is performed to obtain the dynamic metabolic joint feature vector Z. Step 6: Input the dynamic metabolic joint feature vector Z into the pre-constructed drug metabolism trend prediction model, and perform nonlinear mapping analysis on the current drug metabolism state based on the drug metabolism trend prediction model to obtain the predicted drug concentration change trend Y and metabolic rate change parameter V within the future time window. Step 7: Obtain actual sensor feedback data, and calculate the error between the actual feedback data and the drug concentration change trend prediction result Y obtained in Step 6 to obtain the prediction deviation parameter E; Step 8: Determine whether the prediction deviation parameter E is greater than the preset error threshold T; if E > T, adjust the state transition parameter M and data weight coefficient in the state transition relationship model according to the prediction deviation parameter E; if E ≤ T, maintain the current model parameters. Step 9: Re-execute the drug metabolism state prediction process based on the adjusted state transition relationship model, and output the current drug metabolism state information of the target object, the drug concentration change curve, and the prediction results of future metabolism trends; Step 10: Generate dynamic monitoring results based on the drug metabolism status information and future metabolism trend prediction results, and mark abnormal drug metabolism statuses for risk.

[0007] In this invention, the first biochemical response data includes drug concentration response signals, drug metabolite concentration change signals, and biochemical reaction parameters related to the drug metabolism process; the second physiological state data includes heart rate change parameters, blood oxygen change parameters, body temperature change parameters, and skin conductance response parameters; the third environmental behavior correlation data includes the target monitoring object's motion state parameters, drug administration time parameters, environmental temperature and humidity parameters, and behavioral activity parameters.

[0008] In this invention, the first metabolic feature sequence X A At least including drug concentration change trend characteristics, metabolite change rate characteristics, and drug clearance characteristics; the second physiological characteristic sequence X B It includes at least the characteristics of vital sign fluctuations and the intensity of human response; the third associated feature sequence X C It includes at least the characteristics associated with drug administration time, the characteristics associated with behavioral state, and the characteristics associated with environmental disturbances.

[0009] In this invention, the drug metabolic state characteristic X S These are comprehensive state characterization parameters constructed based on multi-source sensor data, used to describe the current drug metabolism state changes of the target monitored object; the drug metabolism state feature X SIt should include at least the characteristics of drug concentration changes, metabolic rate, physiological response, environmental influence, and individual metabolic deviation.

[0010] In this invention, the state transition relationship model is used to describe the changing relationship of drug metabolic state between adjacent time points, and it is expressed as follows: ,in, This indicates the characteristics of the drug metabolism state at the current time point. Indicates the predicted state at the next time point. This refers to external input parameters that affect changes in the metabolic state of a drug.

[0011] In this invention, in step 5, the data weighting coefficient is obtained through sensor data reliability calculation, and the calculation method is as follows: ,in, This represents the data fusion weight corresponding to the first-order sensor data at the current time point. This represents the reliability evaluation parameter for the type I sensor data.

[0012] In this invention, the dynamic metabolic joint feature vector Z is represented as: Among them, X A X represents the first metabolic characteristic sequence. B The second physiological characteristic sequence, X C The third associated feature sequence is given, and M represents the state transition parameter. , , , Let represent the corresponding feature fusion weight coefficients, and satisfy the following: .

[0013] In this invention, the drug metabolism trend prediction model in step 6 is represented as follows: ,in, Represents the dynamic metabolic joint feature vector. Represents the set of parameters for the prediction model. Represents a nonlinear mapping function; the prediction model is achieved through: Obtain the probability distribution of future drug metabolism states, where W i Let b represent the weight matrix of the i-th layer network. i This indicates the corresponding bias parameter. This represents a non-linear activation function.

[0014] In this invention, in step 7, the prediction deviation parameter daily is expressed as: Among them, Y r Y represents the drug metabolism state parameter obtained from actual sensor feedback. p This represents the prediction parameters output by the drug metabolism trend prediction model.

[0015] In this invention, the monitoring system of the sensor-based drug metabolism dynamic monitoring method includes a multimodal sensing acquisition module, a data synchronization encoding module, a metabolic state modeling module, a state transition analysis module, a multi-source data fusion module, a trend prediction module, an adaptive optimization module, and a result output module. The multimodal sensing acquisition module is used to collect the first biochemical response data, the second physiological state data, and the third environmental behavior correlation data of the target monitored object during drug metabolism. The data synchronization encoding module is used to perform time synchronization processing on sensor data from different sources and with different sampling periods, and to generate a first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C ; The metabolic state modeling module is used to construct a drug metabolic state feature space based on multi-source sensing features and output drug metabolic state features X. S ; The state transition analysis module is used to establish a state transition relationship model based on historical drug metabolism data and to calculate the drug metabolism state transition parameter M. The multi-source data fusion module is used to evaluate the credibility of different sensor data based on the state transition parameter M, and to construct a dynamic metabolic joint feature vector Z. The trend prediction module is used to input the dynamic metabolic joint feature vector Z into the drug metabolism trend prediction model and output the future drug concentration change trend prediction result Y and the metabolic rate change parameter V. The adaptive optimization module is used to adjust the model parameters according to the prediction deviation parameter E between the prediction results and the actual feedback data. The result output module is used to output information on the drug metabolism status of the target monitoring object, concentration change curves, and abnormal risk labeling results.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention achieves a transformation from single-index detection to multi-dimensional state perception and dynamic analysis by constructing a dynamic monitoring method for drug metabolism based on multimodal sensor data. By simultaneously collecting, feature-coding, and fusing biochemical response data, physiological state data, and environmental behavior-related data, the completeness and accuracy of drug metabolism information acquisition are improved. Simultaneously, this invention establishes a dynamic state characterization model based on the migration relationship of drug metabolism states, which can effectively analyze the impact of changes in drug concentration, metabolic rate, and individual differences on monitoring results, reducing misjudgments caused by individual differences in traditional detection processes. Furthermore, by introducing a predictive feedback optimization mechanism, the model parameters are dynamically corrected based on real-time detection data, improving the stability and adaptability during long-term continuous monitoring. This invention not only enables real-time monitoring of drug metabolism states but also predicts future trends and abnormal metabolic risks, providing reliable data support for precision medicine, personalized treatment, and drug safety management. Attached Figure Description

[0017] Figure 1 This is a flowchart of the sensor-based dynamic monitoring method for drug metabolism of the present invention; Figure 2 This is a flowchart of the construction of a drug metabolism state transition relationship model in the sensor-based drug metabolism dynamic monitoring method of the present invention; Figure 3 This is a flowchart of the drug metabolism trend prediction process in the sensor-based dynamic monitoring method for drug metabolism according to the present invention; Figure 4 This is a hardware diagram of the monitoring system for the sensor-based dynamic monitoring method for drug metabolism of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1

[0019] In Example 1, addressing the challenges of complex data types from multiple sensors, difficulty in effectively characterizing the relationships between different data points, and difficulty in accurately predicting the dynamic changes in individualized metabolic states during drug metabolism, this example provides a sensor-based method for dynamic monitoring of drug metabolism. By collecting, encoding, modeling, predicting trends, and optimizing the multidimensional dynamic data of the target monitoring object, continuous analysis of the entire drug metabolism process is achieved. (Refer to...) Figure 1 The specific implementation process includes the following steps: Step 1: Continuous dynamic monitoring data of the drug metabolism process is collected by multimodal sensing nodes deployed on the target monitoring object. This data is then classified and processed according to data type to obtain first biochemical response data, second physiological state data, and third environmental behavior correlation data. The first biochemical response data includes drug concentration response signals, drug metabolite concentration change signals, and biochemical reaction parameters related to the drug metabolism process. The second physiological state data includes heart rate change parameters, blood oxygen change parameters, body temperature change parameters, and skin conductance response parameters. The third environmental behavior correlation data includes the target monitoring object's motion state parameters, drug administration time parameters, environmental temperature and humidity parameters, and behavioral activity parameters.

[0020] Step 2: Perform time synchronization processing and feature encoding processing on the first biochemical response data, the second physiological state data, and the third environmental behavior correlation data respectively to obtain the corresponding first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C In this embodiment, the first metabolic feature sequence X A At least including drug concentration change trend characteristics, metabolite change rate characteristics, and drug clearance characteristics; the second physiological characteristic sequence X B It includes at least the characteristics of vital sign fluctuations and the intensity of human response; the third associated feature sequence X C It includes at least the characteristics associated with drug administration time, the characteristics associated with behavioral state, and the characteristics associated with environmental disturbances.

[0021] Step 3: Based on the first metabolic feature sequence X A Second physiological characteristic sequence X B and the third associated feature sequence X C A multidimensional drug metabolism state feature space is constructed, and sensor data corresponding to different time points are mapped to a unified state representation space to obtain the current drug metabolism state feature X. S In this embodiment, the drug metabolic state characteristic X SThese are comprehensive state characterization parameters constructed based on multi-source sensor data, used to describe the current drug metabolism state changes of the target monitored object; the drug metabolism state feature X S It should include at least the characteristics of drug concentration changes, metabolic rate, physiological response, environmental influence, and individual metabolic deviation.

[0022] Step 4: Construct a state transition relationship model based on historically collected dynamic drug metabolism data, and analyze the current drug metabolism state characteristics X. S A time-series correlation analysis is performed to obtain the drug metabolism state transition parameter M. In this embodiment, the state transition relationship model is used to describe the change relationship of drug metabolism state between adjacent time points, and it is expressed as follows: ,in, This indicates the characteristics of the drug metabolism state at the current time point. Indicates the predicted state at the next time point. This refers to external input parameters that affect changes in the metabolic state of a drug.

[0023] Step 5: Based on the state transition parameter M, process the first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C A reliability assessment is performed by calculating data weight coefficients for different sensor data sources, and then multi-source data fusion processing is conducted based on these weight coefficients to obtain the dynamic metabolic joint feature vector Z. Specifically, the data weight coefficients are obtained through sensor data reliability calculation, and the calculation method is as follows: ,in, This represents the data fusion weight corresponding to the first-order sensor data at the current time point. This represents the reliability evaluation parameter for the type I sensor data.

[0024] Step 6: Input the dynamic metabolic joint feature vector Z into the pre-constructed drug metabolism trend prediction model, and perform nonlinear mapping analysis on the current drug metabolism state based on the drug metabolism trend prediction model to obtain the predicted drug concentration change trend Y and metabolic rate change parameter V within the future time window.

[0025] In this embodiment, the dynamic metabolic joint feature vector Z is represented as: Among them, X A X represents the first metabolic characteristic sequence. B The second physiological characteristic sequence, X C The third associated feature sequence is given, and M represents the state transition parameter. , , , Let represent the corresponding feature fusion weight coefficients, and satisfy the following: .

[0026] In this embodiment, the drug metabolism trend prediction model is represented as follows: ,in, This represents the dynamic metabolic joint feature vector. Represents the set of parameters for the prediction model. Represents a nonlinear mapping function; the prediction model is achieved through: Obtain the probability distribution of future drug metabolism states, where W i Let b represent the weight matrix of the i-th layer network. i This represents the corresponding bias parameter. This represents a non-linear activation function.

[0027] Step 7: Obtain actual sensor feedback data, and calculate the error between the actual feedback data and the drug concentration change trend prediction result Y obtained in Step 6 to obtain the prediction deviation parameter E. The prediction deviation parameter E is expressed as: Among them, Y r Y represents the drug metabolism state parameter obtained from actual sensor feedback. p This represents the prediction parameters output by the drug metabolism trend prediction model.

[0028] Step 8: Determine whether the prediction deviation parameter E is greater than the preset error threshold T; if E > T, adjust the state transition parameter M and data weight coefficients in the state transition relationship model according to the prediction deviation parameter E; if E ≤ T, maintain the current model parameters. In this embodiment, after completing the drug metabolism trend prediction, the prediction result output by the model is compared with the actual metabolic state data collected by the sensor in real time, and the prediction deviation parameter E between the two is calculated. When the prediction deviation parameter E exceeds the preset error threshold T, it indicates that the current state transition relationship model cannot accurately reflect the actual drug metabolism change law of the target monitoring object. At this time, the state transition parameter M is adjusted according to the degree of change of the deviation parameter E, and the data weight coefficients corresponding to different sensor data sources are reallocated.

[0029] Preferably, when the deviation between drug concentration sensing data and prediction results is large, the weight of drug concentration-related features in the fusion process is increased; when physiological state changes are inconsistent with the predicted trend, the influence parameters corresponding to physiological response features are adjusted to improve the model's adaptability to the current metabolic state. When the prediction deviation parameter E does not exceed the preset error threshold T, the current model is considered to meet the monitoring requirements, and the original state transition parameters and data fusion weights remain unchanged.

[0030] Step 9: Re-execute the drug metabolism state prediction process based on the adjusted state transition relationship model, outputting the current drug metabolism state information of the target object, the drug concentration change curve, and the prediction results of future metabolic trends. In this embodiment, after the state transition relationship model completes parameter updates, the updated state transition parameters M and the adjusted data weight coefficients are re-inputted into the drug metabolism state prediction process to re-integrate and analyze the multi-source sensor data collected at the current time node. The updated model dynamically extrapolates the drug metabolism state of the target monitoring object, obtaining current drug metabolism state information, including drug concentration levels, metabolic rate changes, and physiological response states. Simultaneously, a drug concentration change curve is generated based on the prediction results at continuous time nodes to represent the trend of drug changes in vivo over time; further, the drug metabolism trend within future time windows is predicted based on the state transition relationship, providing data support for subsequent abnormal state judgment and medication adjustment.

[0031] Step 10: Generate dynamic monitoring results based on the drug metabolism status information and future metabolism trend prediction results, and mark abnormal drug metabolism statuses for risk.

[0032] In this embodiment, based on the current drug metabolism status information output in step 9 and the future metabolic trend prediction results, dynamic monitoring results corresponding to the target monitoring object are generated. The dynamic monitoring results include at least the current drug metabolism level, drug concentration change trend, expected metabolic cycle, and abnormal changes. When the rate of change in drug concentration exceeds a preset range, abnormal fluctuations in metabolic rate are detected, or the prediction results indicate a risk of drug accumulation, the corresponding drug metabolism status is risk-marked, and abnormal status alert information is output. The risk marking information includes the time of abnormality occurrence, the type of abnormality, and the corresponding degree of impact, to facilitate subsequent individualized medication adjustments and safety risk assessments. Example 2

[0033] This embodiment details the process of constructing a drug metabolism state transition relationship model for implementing a sensor-based dynamic drug metabolism monitoring method, referring to... Figure 2 The process of constructing the state transition relationship model includes the following steps: Step 1: Obtain the characteristic sequence of drug metabolism status continuously collected within a preset time window, and construct a historical state change sequence in chronological order: ,in, This represents the drug metabolism state characteristics corresponding to the nth time point.

[0034] In this embodiment, by sorting the drug metabolism states at consecutive time points, a historical state sequence reflecting the dynamic changes in drug metabolism is formed, providing a data foundation for state transition relationship analysis.

[0035] Step 2: Calculate the drug metabolism state transition characteristics based on the state change relationship between adjacent time points: ,in, This represents the change in drug metabolism state between adjacent time points. Furthermore, a state transition matrix is ​​constructed based on the state changes within multiple consecutive time windows: , of which M ij Indicates the current state X i Next state X j The probability of transition, P represents the state transition probability function.

[0036] In this embodiment, a state transition matrix is ​​used to describe the relationship between different stages of drug metabolism, enabling the model to characterize the continuous changes in drug absorption, transformation, and clearance processes.

[0037] Step 3: Perform a matching analysis on the current drug metabolism state based on the state transition matrix to obtain the state transition parameters corresponding to the current state. ,in, M represents the current characteristics of drug metabolism. ij Indicates the historical state migration relationship, Here, F represents the external influence parameter corresponding to the current time point, and F represents the state fusion calculation function. In this embodiment, by fusing current state information, historical migration patterns, and external influence factors, a state transition parameter M is generated to describe the current trend of drug metabolism change.

[0038] Step 4: Predict the direction of drug metabolism state changes at future time points based on the obtained state transition parameters M: ,in, This represents the predicted drug metabolism state at the next time point. In this embodiment, by using the state transition parameter M to extrapolate the current drug metabolism state, the model can predict future state change trends based on historical metabolic change patterns, providing data support for subsequent dynamic monitoring of drug metabolism. Example 3

[0039] This embodiment details a drug metabolism trend prediction process in a sensor-based dynamic drug metabolism monitoring method, referring to... Figure 3 The drug metabolism trend prediction process includes the following steps: Step 1: Obtain the dynamic metabolic joint feature vector Z, and perform feature normalization on the dynamic metabolic joint feature vector Z to map feature parameters of different dimensions to a unified numerical range. The normalization process is expressed as follows: Among them, Z i Z represents the i-th feature parameter in the dynamic metabolic joint feature vector Z.max With Z min These represent the maximum and minimum values ​​in the corresponding feature dimensions, respectively. This represents the normalized feature parameters. In this embodiment, by standardizing the data input to the prediction model, the impact of dimensional differences between different types of sensor data on the prediction results is reduced.

[0040] Step 2: Input the normalized dynamic metabolic joint feature vector Z' into the drug metabolism trend prediction model, and extract the nonlinear correlation features in the drug metabolism state through a multi-layer feature mapping structure.

[0041] The feature mapping process is represented as follows: , where H i W represents the output feature of the i-th hidden layer. i Let b represent the weight parameters of the i-th layer of the network. i Indicates the bias parameter. H represents a nonlinear activation function, and H0 represents the input dynamic metabolic joint feature vector. In this embodiment, a multi-layer nonlinear mapping method is used to extract the potential correlation between changes in drug concentration, changes in physiological response, and changes in metabolic state.

[0042] Step 3: Perform regression analysis on the hidden layer features based on the output layer of the prediction model to obtain the predicted trend Y of drug concentration change within the future time window. The prediction result... Represented as , where H n W represents the output feature of the last hidden layer. y b represents the output layer weight parameters. y represents the output layer bias parameter, and Y represents the predicted drug concentration within the future time window.

[0043] In this embodiment, by jointly analyzing the historical drug metabolism state changes and the current dynamic characteristics, the future trend of drug concentration changes can be predicted.

[0044] Step 4: Based on the prediction results within the continuous time window Calculate parameters of drug metabolism rate change This is used to characterize the dynamic changes in drug concentration over time. The metabolic rate change parameter is expressed as: Where, represents the predicted drug concentration at the current time point. This represents the predicted drug concentration at a future time point. This indicates a time interval. In this embodiment, by calculating the rate of change of drug concentration between different time points, the prediction results are further converted into metabolic rate change information to reflect the dynamic changes in drug absorption, transformation, and clearance processes.

[0045] Step 5: Generate drug metabolism trend prediction information within the future time window based on the drug concentration change trend prediction result Y and the metabolic rate change parameter V. The prediction information includes at least: the future direction of drug concentration change, the trend of peak drug concentration change, and the metabolic rate change. In this embodiment, by combining the drug concentration prediction result Y and the metabolic rate change parameter V, trend analysis of the drug metabolism process is achieved, providing a basis for subsequent abnormal state judgment and dynamic monitoring result output. Example 4

[0046] In this invention, a monitoring system for implementing a sensor-based dynamic monitoring method for drug metabolism is described, with reference to... Figure 4 It includes a multimodal sensing acquisition module, a data synchronization and encoding module, a metabolic state modeling module, a state transition analysis module, a multi-source data fusion module, a trend prediction module, an adaptive optimization module, and a result output module; The multimodal sensing acquisition module is used to collect multi-source dynamic monitoring data of the drug metabolism process of the target monitored object, including first biochemical response data, second physiological state data, and third environmental behavior correlation data. The first biochemical response data is used to characterize the changes in drug concentration and metabolites, the second physiological state data is used to characterize the vital sign response of the target monitored object after being treated with the drug, and the third environmental behavior correlation data is used to characterize the influence of drug administration behavior, activity status, and external environmental factors on the drug metabolism process.

[0047] The data synchronization and encoding module is used to perform time synchronization processing, data cleaning, and feature encoding processing on different types of data acquired by the multimodal sensing acquisition module. It converts sensing signals with different sampling frequencies and data dimensions into feature sequences in a unified format, thereby obtaining the first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C .

[0048] The metabolic state modeling module is used to model the first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C A drug metabolism state feature space is constructed, and through correlation analysis of multi-source sensor features, a state feature X describing the current drug metabolism state of the target monitored object is generated. S .

[0049] The state transition analysis module is used to establish a drug metabolism state transition relationship model based on historical drug metabolism monitoring data. By analyzing the change pattern of drug metabolism state between different time points, the state transition parameter M is obtained and used to describe the dynamic changes in the absorption, transformation and clearance process of the target monitored drug.

[0050] The multi-source data fusion module is used to evaluate the credibility of different sensor data sources based on the state transition parameter M, and adjust the data fusion weights according to the credibility results of various sensor data, and to integrate the first metabolic feature sequence X. A Second physiological characteristic sequence X B Third associated feature sequence X C The state transition parameter M is fused together to construct a dynamic metabolic joint feature vector Z.

[0051] The trend prediction module is used to input the dynamic metabolic joint feature vector Z into a pre-constructed drug metabolism trend prediction model. The prediction model performs joint analysis on the current drug metabolism state and historical change trends, and outputs the drug concentration change trend, metabolic rate change parameters, and future metabolic state prediction results of the target monitoring object within the future time window.

[0052] The adaptive optimization module is used to obtain the prediction deviation parameter E between the trend prediction result and the actual sensor feedback data, and dynamically adjust the state transition parameter M and data fusion weight in the state transition relationship model based on the comparison result between the prediction deviation parameter E and the preset error threshold T, so that the drug metabolism monitoring model can adapt to the individualized metabolic differences of different target objects.

[0053] The result output module is used to generate dynamic monitoring results based on drug metabolism status information and future metabolism trend prediction results, and outputs the current drug metabolism level of the target monitoring object, drug concentration change curve, metabolism trend prediction information, and abnormal risk marking results; when the drug metabolism status is detected to exceed the preset risk range, the result output module triggers an abnormal prompt and outputs the corresponding risk level information.

[0054] In this embodiment, the aforementioned monitoring system enables data acquisition, status analysis, trend prediction, and feedback optimization of the entire drug metabolism process. This allows the system to dynamically adjust model parameters based on real-time sensor data, improving the adaptability of drug metabolism monitoring results to different individuals and medication environments. Compared to traditional monitoring methods based on a single detection indicator, this system comprehensively considers changes in drug concentration, physiological responses, and external influencing factors, achieving continuous, dynamic, and individualized analysis of drug metabolism status.

[0055] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A sensor-based method for dynamic monitoring of drug metabolism, characterized in that, Includes the following steps: Step 1: Collect continuous dynamic monitoring data of drug metabolism process by deploying multimodal sensor nodes on the target monitoring object, classify and process the data according to the data type to obtain the first biochemical response data, the second physiological state data and the third environmental behavior correlation data; Step 2: Perform time synchronization processing and feature encoding processing on the first biochemical response data, the second physiological state data, and the third environmental behavior correlation data respectively to obtain the corresponding first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C ; Step 3: Based on the first metabolic feature sequence X A Second physiological characteristic sequence X B and the third associated feature sequence X C A multidimensional drug metabolism state feature space is constructed, and sensor data corresponding to different time points are mapped to a unified state representation space to obtain the current drug metabolism state feature X. S ; Step 4: Construct a state transition relationship model based on historically collected dynamic drug metabolism data, and analyze the current drug metabolism state characteristics X. S Perform time-series correlation analysis to obtain the drug metabolism state transition parameter M; Step 5: Based on the state transition parameter M, process the first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C A credibility assessment is performed, and the data weight coefficients corresponding to different sensor data sources are calculated respectively. Based on the weight coefficients, multi-source data fusion processing is performed to obtain the dynamic metabolic joint feature vector Z. Step 6: Input the dynamic metabolic joint feature vector Z into the pre-constructed drug metabolism trend prediction model, and perform nonlinear mapping analysis on the current drug metabolism state based on the drug metabolism trend prediction model to obtain the predicted drug concentration change trend Y and metabolic rate change parameter V within the future time window. Step 7: Obtain actual sensor feedback data, and calculate the error between the actual feedback data and the drug concentration change trend prediction result Y obtained in Step 6 to obtain the prediction deviation parameter E; Step 8: Determine whether the prediction deviation parameter E is greater than the preset error threshold T; if E > T, adjust the state transition parameter M and data weight coefficient in the state transition relationship model according to the prediction deviation parameter E; if E ≤ T, maintain the current model parameters. Step 9: Re-execute the drug metabolism state prediction process based on the adjusted state transition relationship model, and output the current drug metabolism state information of the target object, the drug concentration change curve, and the prediction results of future metabolism trends; Step 10: Generate dynamic monitoring results based on the drug metabolism status information and future metabolism trend prediction results, and mark abnormal drug metabolism statuses for risk.

2. The sensor-based drug metabolism dynamic monitoring method according to claim 1, characterized in that, The first biochemical response data includes drug concentration response signals, drug metabolite concentration change signals, and biochemical reaction parameters related to the drug metabolism process; the second physiological state data includes heart rate change parameters, blood oxygen change parameters, body temperature change parameters, and skin conductance response parameters; the third environmental behavior correlation data includes the target monitoring object's motion state parameters, drug administration time parameters, environmental temperature and humidity parameters, and behavioral activity parameters.

3. The sensor-based drug metabolism dynamic monitoring method according to claim 1, characterized in that, The first metabolic feature sequence X A It should include at least the characteristics of drug concentration change trends, the characteristics of metabolite change rates, and the characteristics of drug clearance; The second physiological characteristic sequence X B It includes at least the characteristics of vital sign fluctuations and the intensity of human response; the third associated feature sequence X C It includes at least the characteristics associated with drug administration time, the characteristics associated with behavioral state, and the characteristics associated with environmental disturbances.

4. The sensor-based drug metabolism dynamic monitoring method according to claim 1, characterized in that, The drug metabolic state characteristic X S These are comprehensive state characterization parameters constructed based on multi-source sensor data, used to describe the current drug metabolism state changes of the target monitored object; the drug metabolism state feature X S It should include at least the characteristics of drug concentration changes, metabolic rate, physiological response, environmental influence, and individual metabolic deviation.

5. The sensor-based drug metabolism dynamic monitoring method according to claim 1, characterized in that, The state transition relationship model is used to describe the changes in drug metabolism state between adjacent time points, and it is expressed as follows: ,in, This indicates the characteristics of the drug metabolism state at the current time point. Indicates the predicted state at the next time point. This refers to external input parameters that affect changes in the metabolic state of a drug.

6. The sensor-based drug metabolism dynamic monitoring method according to claim 1, characterized in that, In step 5, the data weighting coefficient is obtained through the calculation of sensor data reliability, and the calculation method is as follows: ,in, This represents the data fusion weight corresponding to the first-order sensor data at the current time point. This represents the reliability evaluation parameter for the type I sensor data.

7. The sensor-based drug metabolism dynamic monitoring method according to claim 1, characterized in that, The dynamic metabolic joint feature vector Z is represented as follows: Among them, X A X represents the first metabolic characteristic sequence. B The second physiological characteristic sequence, X C The third associated feature sequence is given, and M represents the state transition parameter. , , , Let represent the corresponding feature fusion weight coefficients, and satisfy the following: .

8. The sensor-based drug metabolism dynamic monitoring method according to claim 1, characterized in that, The drug metabolism trend prediction model in step 6 is represented as follows: ,in, This represents the dynamic metabolic joint feature vector. Represents the set of parameters for the prediction model. Represents a nonlinear mapping function; The prediction model uses: Obtain the probability distribution of future drug metabolism states, where W i Let b represent the weight matrix of the i-th layer network. i This represents the corresponding bias parameter. This represents a non-linear activation function.

9. The sensor-based drug metabolism dynamic monitoring method according to claim 1, characterized in that, The prediction deviation parameter daily in step 7 is expressed as follows: Among them, Y r Y represents the drug metabolism state parameter obtained from actual sensor feedback. p This represents the prediction parameters output by the drug metabolism trend prediction model.

10. A monitoring system for implementing the sensor-based dynamic monitoring method for drug metabolism according to claim 1, characterized in that, It includes a multimodal sensing acquisition module, a data synchronization and encoding module, a metabolic state modeling module, a state transition analysis module, a multi-source data fusion module, a trend prediction module, an adaptive optimization module, and a result output module; The multimodal sensing acquisition module is used to collect the first biochemical response data, the second physiological state data, and the third environmental behavior correlation data of the target monitored object during drug metabolism. The data synchronization encoding module is used to perform time synchronization processing on sensor data from different sources and with different sampling periods, and to generate a first metabolic feature sequence X. A Second physiological characteristic sequence X B and the third associated feature sequence X C ; The metabolic state modeling module is used to construct a drug metabolic state feature space based on multi-source sensing features and output drug metabolic state features X. S ; The state transition analysis module is used to establish a state transition relationship model based on historical drug metabolism data and to calculate the drug metabolism state transition parameter M. The multi-source data fusion module is used to evaluate the credibility of different sensor data based on the state transition parameter M, and to construct a dynamic metabolic joint feature vector Z. The trend prediction module is used to input the dynamic metabolic joint feature vector Z into the drug metabolism trend prediction model and output the future drug concentration change trend prediction result Y and the metabolic rate change parameter V. The adaptive optimization module is used to adjust the model parameters according to the prediction deviation parameter E between the prediction results and the actual feedback data. The result output module is used to output information on the drug metabolism status of the target monitoring object, concentration change curves, and abnormal risk labeling results.

Images